Immunotherapy on Dual Brain Lab

Biliary Tract Cancer Clinical Trial Timeline: A 16-Year Evolution Map

Tue, 21 Apr 2026 00:00:00 +0000

Biliary Tract Cancer Clinical Trial Timeline: In-depth Report

Coverage: 39 landmark trials cited by NCCN Biliary Tract Cancers V1.2026 (37 published with full PMID traceability + 2 ongoing with only NCT / design paper) + mixed three-subtype enrollment (GBC / ICC / ECC) + eight biomarker-matched pathways

Curated by Dual Brain Lab (csilab.net)

1. One-sentence definition

This report maps the evolution logic and current decision landscape of systemic therapy for biliary tract cancer (BTC; including gallbladder cancer / GBC, intrahepatic cholangiocarcinoma / ICC, extrahepatic cholangiocarcinoma / ECC) over the past 16 years (2010-2026) as reflected in the landmark clinical trials cited by NCCN Biliary Tract Cancers V1.2026, providing frontline clinicians a traceable panoramic map for “who, what, and why” decisions in 2026.

Iron rule: every data point of every trial is traceable to PubMed (PMID) or ClinicalTrials.gov (NCT id) — each [PMID xxxxxxxx] in the text can be opened directly in PubMed for source verification.

2. Vertical: timeline of five treatment paradigms

BTC systemic therapy has gone through five paradigm shifts in the past 16 years: adjuvant chemo upgraded from observation to capecitabine, then expanded to S-1 as a parallel three-way option → advanced 1L GemCis (gemcitabine + cisplatin) dominated uniformly for 12 years → IO (immune checkpoint inhibitor) + GemCis double hit forming a class effect rewrote the 1L backbone → 8 biomarker-matched precision pathways (FGFR2 / IDH1 / HER2 / BRAF V600E / NTRK / MSI-H / NRG1 / KRAS G12C) rolled out in parallel → advanced 2L layered marginal benefit from FOLFOX with an East-West divide on nal-IRI.

Each shift was smaller than in NSCLC but larger than in pancreatic cancer — reflecting BTC’s distinct molecular architecture: ICC is enriched for FGFR2 / IDH1 (10-20%), GBC is enriched for HER2 (15-20%), and all three subtypes share MSI-H / BRAF / NTRK / NRG1 / KRAS G12C tumor-agnostic basket pathways. This “subtype × biomarker dual heterogeneity” makes BTC’s precision-therapy density second only to NSCLC, far exceeding HCC and pancreatic cancer.

2.1 Adjuvant chemo evolution (2015-2024): from GEMOX/gemcitabine failure → capecitabine tacit ascent → S-1 / GemCis-CRT East-West divergence

Story: before BILCAP, adjuvant BTC was observation + some-centers chemo. PRODIGE-12 and BCAT, two European / Japanese phase III trials, both rendered GEMOX and gemcitabine monotherapy adjuvant negative. BILCAP was tacitly accepted globally on a borderline-ITT / significant per-protocol basis; ASCOT delivered a clean ITT-positive result in a Japanese population with S-1; OSTWAL 2024 delivered the first positive DFS for GemCis + CRT in high-risk Indian GBC. An arc from “all fail → marginal win → clean win → subtype-specific.”

PRODIGE-12 [PMID 30707660] (Edeline 2019 J Clin Oncol, N=194): R0/R1 mixed BTC adjuvant GEMOX (gemcitabine + oxaliplatin) vs observation. RFS HR 0.88 (95% CI 0.62-1.25, p=0.48) negative, OS likewise negative. Proved not every adjuvant doublet works — what wins in advanced doesn’t always win in adjuvant.
BCAT [PMID 29405274] (Ebata 2018 Br J Surg, N=225, Japan): ECC-only R0 adjuvant gemcitabine monotherapy vs observation. OS HR 1.01 (95% CI 0.70-1.45, p=0.97) zero difference. The cleanest negative in a subtype-specific adjuvant trial — gemcitabine monotherapy fully exits the adjuvant recommendation. The BCAT + PRODIGE-12 IPD meta-analysis [PMID 35182925] reaffirmed “gemcitabine-based adjuvant doublets do not benefit BTC.”
BILCAP [PMID 30922733] (Primrose 2019 Lancet Oncol, N=447, UK): R0/R1 mixed BTC adjuvant capecitabine × 8 cycles vs observation. ITT mOS 51.1 vs 36.4 months, HR 0.81 (95% CI 0.63-1.04, p=0.097) just missed; per-protocol HR 0.75 (95% CI 0.58-0.97, p=0.028) significant. ESMO/NCCN/CSCO all adopted it as adjuvant standard — rationale: per-protocol significant + real-world dose reduction unavoidable + no better data. This is BTC’s first tacitly-ascendant adjuvant standard in 30 years.
ACTICCA-1 [PMID 26228433] (Stein 2015 BMC Cancer design paper / NCT02170090): European multicenter N≈783 resected mixed BTC, adjuvant GemCis × 8 cycles vs amended capecitabine (control arm amended post-BILCAP). The first direct head-to-head comparison of GemCis doublet vs capecitabine monotherapy as adjuvant; still ongoing as of 2026-04, primary results not yet published. The most anticipated BTC adjuvant phase III.
ASCOT / JCOG1202 [PMID 36681415] (Nakachi 2023 Lancet, N=440, Japan): R0 mixed BTC adjuvant S-1 × 4 cycles vs observation. mOS 62.4 vs 51.1 months, HR 0.69 (95% CI 0.55-0.86, p=0.001) significant. The cleanest ITT-positive adjuvant phase III in BTC to date. Japanese guidelines accordingly list S-1 as adjuvant SoC. In the Asian setting with DPYD polymorphism and S-1 tolerability, this “Eastern-specific” path stands alongside BILCAP.
SWOG-S0809 [PMID 25964250] (Ben-Josef 2015 J Clin Oncol, N=79, single-arm phase II): ECC + GBC only (T2-T4 or node-positive) adjuvant GemCap × 4 cycles → CRT (concurrent capecitabine, 54 Gy). 2-year OS 65% (met pre-specified ≥60% threshold), similar across R0/R1 subgroups. The main evidence basis for NCCN recommending adjuvant CRT in ECC/GBC R1-margin / N+ patients.
OSTWAL-2024-GEMCIS-CRT-GBC [PMID 38958997] (Ostwal 2024 JAMA Oncol, N=80, India Tata Memorial): high-risk GBC only (T3-T4 / N+ / R1) adjuvant GemCis × 4 cycles + CRT vs observation. DFS HR 0.62 (95% CI 0.43-0.89, p=0.009) significant, mDFS 28.0 vs 14.0 months. The first RCT to show positive DFS for adjuvant chemo + CRT in a single GBC subtype, filling the GBC-underrepresentation gap in mixed-subtype trials.

Takeaway: in 2026 adjuvant BTC — capecitabine monotherapy (BILCAP) is the Western default; S-1 × 4 cycles (ASCOT) is the Asian preferred; high-risk ECC/GBC (R1 / N+) adds CRT (SWOG-S0809 / OSTWAL) as intensification; GEMOX and gemcitabine monotherapy no longer used in adjuvant; ACTICCA-1 (GemCis vs capecitabine) readout will definitively rewrite the decision tree.

2.2 Advanced 1L (2010-2025): GemCis dominates 12 years → three triplet challenges all fail → IO double-hit forms class effect

Story: after ABC-02 enshrined GemCis as 1L in 2010, no one dethroned it for 12 full years. Three triplet challenges in different directions (add nab-paclitaxel / switch to mFOLFIRINOX / add S-1) all failed. Not until 2022-2023 did TOPAZ-1 + KEYNOTE-966 — using durvalumab and pembrolizumab, two fully independent PD-(L)1 inhibitors — converge on HR 0.80-0.83, formally establishing IO + GemCis as a class effect.

ABC-02 [PMID 20375404] (Valle 2010 NEJM, N=410, UK): treatment-naive advanced BTC (ICC 39% / GBC 38% / ECC 20% / ampullary 3%) GemCis vs gemcitabine monotherapy. mOS 11.7 vs 8.1 months, HR 0.64 (95% CI 0.52-0.76, p<0.001). The control base for all BTC 1L trials over 12 years; the “backbone drug” status of GemCis underlying every IO + GemCis class-effect trial was established here.
AMEBICA / PRODIGE-38 [PMID 34662180] (Phelip 2022 J Clin Oncol, N=191, France phase II RCT): mFOLFIRINOX vs GemCis 1L. 6-month PFS 44.5% vs 47.3%, PFS HR 0.93, OS HR 0.97 all negative, with markedly higher toxicity (G3-4 AEs ~72% vs ~60%). Regimens that win in pancreatic cancer don’t carry over to BTC.
KHBO1401-MITSUBA [PMID 35900311] (Ioka 2023 J Hepatobiliary Pancreat Sci, N=246, Japan): GCS triplet (GemCis + S-1) vs GemCis. mOS 13.5 vs 12.6 months, ITT HR 0.79 (95% CI 0.60-1.04, p=0.094) just missed; per-protocol HR 0.73 (p=0.046). Occasionally used in Japanese daily practice, not adopted globally as SoC.
TOPAZ-1 [PMID 38319896] (Oh 2022 NEJM Evid, N=685, global, Asia 54% / Europe 30% / Americas 16%): durvalumab (PD-L1) + GemCis vs placebo + GemCis. OS HR 0.80 (95% CI 0.66-0.97, p=0.021), mOS 12.8 vs 11.5 months (delta only 1.3 months); 24-month landmark OS 24.9% vs 10.4% (delta 15 percentage points). Median OS gap is small but the long-tail survivor proportion doubles — IO benefit is not uniformly distributed but enriched in durable responders. FDA approved 2022-09.
KEYNOTE-966 [PMID 37075781] (Kelley 2023 Lancet, N=1069, global): pembrolizumab (PD-1) + GemCis vs placebo + GemCis. OS HR 0.83 (95% CI 0.72-0.95, p=0.0034), mOS 12.7 vs 10.9 months. 55% larger than TOPAZ-1, with HR nearly overlapping TOPAZ-1 (0.80 vs 0.83). Two fully independent teams, two different drugs (PD-L1 vs PD-1), two different regimens (durvalumab maintenance vs pembrolizumab monotherapy maintenance), HR converging on 0.80-0.83 — a textbook class effect. FDA approved 2023-10.
SWOG-1815 [PMID 39671534] (Shroff 2025 J Clin Oncol, N=441, US): GemCis + nab-paclitaxel triplet vs GemCis doublet 1L. mOS 11.8 vs 11.4 months, HR 0.93 (95% CI 0.75-1.16, p=0.52) negative. Phase II single-arm data had raised expectations; the phase III closed the “add-taxane” path.

Takeaway: 2026 advanced BTC 1L SoC = IO + GemCis (durvalumab or pembrolizumab, HR 0.80-0.83 class effect). Three triplet concepts (add taxane / switch to FOLFIRINOX / add S-1) all failed. The core branchpoint in daily clinical practice is “which IO fits this patient” rather than “add IO or not” — see §3.2.

2.3 Advanced 2L (2019-2024): FOLFOX marginal + nal-IRI East-West divergence + IO/regorafenib monotherapy historical value

Story: 2L BTC has hovered near the “6-month OS floor” for 16 years. ABC-06 delivered a marginal win with FOLFOX and became standard; NIFTY and NALIRICC used the same drug (nal-IRI + 5FU/LV) in Korea vs Germany and reached opposite conclusions; IO monotherapy (nivolumab) and regorafenib produced modest results in unselected populations, but established the early signal that “IO has enrichment in the MMR-deficient subgroup of refractory BTC.”

ABC-06 [PMID 33798493] (Lamarca 2021 Lancet Oncol, N=162, UK): after gemcitabine-class failure, mFOLFOX + active symptom control (ASC) vs ASC alone. mOS 6.2 vs 5.3 months, HR 0.69 (95% CI 0.50-0.97, p=0.031); 12-month OS 25.9% vs 11.4%. BTC’s first positive 2L RCT, establishing mFOLFOX as 2L standard. Delta only 0.9 months — low ceiling.
NIFTY [PMID 36951834] (Hyung 2023 JAMA Oncol, N=174, Korea): after gemcitabine-class failure, liposomal irinotecan (nal-IRI) + 5FU/LV vs 5FU/LV alone. PFS HR 0.56 (95% CI 0.39-0.81, p=0.002), mPFS 7.1 vs 1.4 months. OS HR 0.82 (p=0.27) negative, but the PFS signal is the strongest in BTC 2L history.
NALIRICC [PMID 38870977] (Vogel 2024 Lancet Gastroenterol Hepatol, N=100, Germany AIO): same nal-IRI + 5FU/LV vs 5FU/LV, OS HR 0.68 (95% CI 0.44-1.04, p=0.074) borderline non-significant, mOS 8.2 vs 6.9 months. Same drug, same 2L setting, same control — Korean PFS+ but German OS negative — East-West population PK / molecular background (HBV / liver fluke vs sporadic) / ICC-to-ECC ratio differences could all plausibly explain it.
KIM-2020-NIVO-BTC [PMID 32352498] (Kim 2020 JAMA Oncol, N=54, US phase II single-arm): gemcitabine-class refractory BTC nivolumab monotherapy. ORR 22% (12/54), mPFS 3.7 months, mOS 14.2 months. The 22% ORR in an unselected population is well above the historical 5-10% chemo-refractory ORR — but responses enriched in the MMR-deficient subgroup, providing proof-of-concept for IO moving into 1L via TOPAZ-1 / KN-966.
SUN-2019-REGO-BTC [PMID 30561756] (Sun 2019 Cancer, N=43, US phase II single-arm): after gemcitabine-class failure, regorafenib 160 mg. DCR 56%, ORR 11%, mPFS 2.0 months, mOS 7.4 months. Multikinase TKI shows only marginal activity in BTC; no phase III followed. In 2026 still occasionally used as “palliative when no better option exists,” not SoC.

Takeaway: in 2026 advanced BTC 2L and beyond — mFOLFOX + ASC (ABC-06) is the global default; nal-IRI + 5FU/LV is used in Korea / parts of Asia-Pacific based on NIFTY PFS data, while the West is cautious after NALIRICC OS negative; IO monotherapy is recommended only for MSI-H / TMB-H selective populations; regorafenib used occasionally when all other options are exhausted. The ceiling of these 2L chemo paths will be rewritten by the biomarker-matched strategies in §2.4.

2.4 Biomarker-matched precision therapy (2018-2026): eight pathways + 2/3 ICC three-subtype reshaping

Story: in 2018, NAVIGATE (larotrectinib for NTRK fusion) opened the door with the FDA’s first tumor-agnostic approval. In 2020, four landmark trials published the same year — FIGHT-202 (pemigatinib for FGFR2 fusion) + ClarIDHy (ivosidenib for IDH1 mutation) + ROAR (dabrafenib + trametinib for BRAF V600E) + KEYNOTE-158 (pembrolizumab for MSI-H) — densely opened the ICC precision-therapy door. From 2021-2024, four HER2 pathways with different mechanisms (MyPathway → HERIZON-BTC-01 → SGNTUC-019 → DESTINY-PanTumor02) rolled out in the GBC-enriched population. In 2025 zenocutuzumab (NRG1 fusion) and in 2026 TRIDENT-1 (repotrectinib, next-gen TRK for NTRK fusion) completed the final two pieces. 8 biomarker-matched pathways collectively cover ~30-40% of BTC patients, making BTC’s precision-therapy density the highest solid-tumor map outside NSCLC.

2.4.1 FGFR2 pathway (ICC-exclusive, 10-15% of ICC patients)

FIGHT-202 [PMID 32203698] (Abou-Alfa 2020 Lancet Oncol, N=146, single-arm phase II): FGFR2 fusion / rearrangement+ CCA (>95% ICC) 2L pemigatinib. Cohort A (FGFR2 fusion, n=107) ORR 35.5% (95% CI 26.5-45.4), mDoR 7.5 months, mPFS 6.9 months, mOS 21.1 months; Cohort B (other FGFR alteration) ORR 0%; Cohort C (no FGFR alteration) ORR 0%. Fusion is the true driver, not all FGFR alterations are equivalent. FDA accelerated approval 2020-04 — BTC’s first precision drug.
FOENIX-CCA2 [PMID 36652354] (Goyal 2023 NEJM, N=103, single-arm phase II): 100% ICC, FGFR2 rearrangement+ 2L futibatinib (covalent / irreversible FGFR1-4 inhibitor). ORR 41.7% (95% CI 32.1-51.9), mDoR 9.7 months, mPFS 8.9 months, mOS 20.0 months. Covalent binding may overcome some resistance mutations to reversible inhibitors. FDA accelerated approval 2022-09.
RAGNAR [PMID 37541273] (Pant 2023 Lancet Oncol, N=217 across ≥15 tumor types / BTC n=27): FGFR1-4 alteration tumor-agnostic erdafitinib. Overall ORR 30%; BTC subgroup ORR ~26%. FDA accelerated approval 2024-09 for tumor-agnostic FGFR-altered solid tumors — providing a third path for BTC patients with FGFR1/3/4 alterations (not the mainstream FGFR2 fusion).
FIGHT-302 [PMID 32677452] (Bekaii-Saab 2020 Future Oncol design paper / NCT03656536): 1L FGFR2 rearrangement+ ICC pemigatinib vs GemCis head-to-head phase III. ESMO 2024 oral reported positive PFS; full manuscript still unpublished as of 2026-04. If formally published as positive, it would switch FGFR2+ ICC 1L SoC from IO + GemCis to pemigatinib — the first example of “molecular-selected 1L replacing IO + GemCis.”

2.4.2 IDH1 pathway (ICC-enriched, 10-20% of ICC patients)

ClarIDHy [PMID 32416072] (Abou-Alfa 2020 Lancet Oncol, N=185; OS update [PMID 34554208] Zhu 2021 JAMA Oncol): IDH1 mutation+ (R132 hotspot) CCA (90% ICC) ≥1 prior line ivosidenib vs placebo (2:1 randomization, crossover allowed). mPFS 2.7 vs 1.4 months, HR 0.37 (95% CI 0.25-0.54, p<0.001); ITT OS 10.3 vs 7.5 months, HR 0.79 (p=0.093) — diluted by 70% crossover; crossover-adjusted OS HR 0.49. BTC’s first truly phase III biomarker trial (FIGHT-202 / FOENIX were both single-arm phase II). FDA approved 2021-08.

2.4.3 HER2 pathway (GBC-enriched, 15-20% of GBC patients; lower in ICC/ECC)

Four different mechanisms in parallel — in solid tumors this level of target richness is matched only by HER2+ breast cancer.

MyPathway-BTC [PMID 34339623] (Javle 2021 Lancet Oncol, N=39 BTC cohort single-arm): HER2+ (IHC 3+ or IHC 2+/FISH+) BTC pertuzumab + trastuzumab. ORR 23.1% (95% CI 10.7-39.9), mPFS 4.0 months, mOS 10.9 months. First prospective HER2-doublet data in BTC; IHC 2+/FISH+ enrollment dilutes the signal but establishes proof-of-concept.
HERIZON-BTC-01 [PMID 37276871] (Harding 2023 Lancet Oncol, N=87 single-arm phase IIb): HER2 amplification (IHC 3+ or IHC 2+/ISH+) BTC (GBC 40% / ICC 35% / ECC 25%) 2L+ zanidatamab (HER2 bispecific antibody, simultaneously binding domain 2 + 4). ORR 41.3% (95% CI 30.7-52.5), mDoR 12.9 months, mPFS 5.5 months, mOS 15.5 months. FDA accelerated approval 2024-08 — BTC’s third biomarker-targeted approval (after FGFR2 / IDH1).
SGNTUC-019 [PMID 37751561] (Nakamura 2023 J Clin Oncol, N=30 BTC cohort single-arm basket): HER2+ BTC tucatinib (HER2-selective small-molecule TKI) + trastuzumab. ORR 46.7% (95% CI 28.3-65.7), mPFS 5.5 months, mOS 13.5 months — the highest BTC HER2-doublet numbers historically (small N, interpret cautiously).
DESTINY-PanTumor02 [PMID 37870536] (Meric-Bernstam 2024 J Clin Oncol, N=41 BTC cohort single-arm): HER2-expressing (IHC 3+ or 2+) solid tumors 7-cohort trastuzumab deruxtecan (T-DXd ADC). BTC ORR 36.6% (IHC 3+ 56.3% / IHC 2+ 20.0%), mPFS 7.0 months, mOS 9.9 months. FDA tumor-agnostic accelerated approval 2024-04 for HER2 IHC 3+ solid tumors — expanding the BTC HER2-eligible population from “amplification” to “overexpression.” ILD risk requires monitoring.

2.4.4 BRAF V600E pathway (ICC-enriched, 3-5% of ICC)

ROAR-BTC [PMID 32818466] (Subbiah 2020 Lancet Oncol, N=43 BTC cohort single-arm basket): BRAF V600E+ BTC (85% ICC) 2L dabrafenib + trametinib. ORR 51% (95% CI 36-67), mPFS 9.0 months, mOS 11.7 months — historically the highest ORR for any BTC biomarker cohort. FDA tumor-agnostic accelerated approval 2022.
NCI-MATCH-H [PMID 32758030] (Salama 2020 J Clin Oncol, N=35 pan-tumor / BTC subgroup n=7-8): BRAF V600E+ non-melanoma non-NSCLC solid tumors dabrafenib + trametinib independent NCI validation. Overall ORR 38%, mPFS 11.4 months. BTC subgroup results consistent with ROAR — providing an independent replication beyond ROAR and strengthening the BRAF V600E + dab/tram evidence package.

2.4.5 NTRK fusion pathway (cross-subtype, <1% of BTC)

Three NTRK inhibitors in parallel + next-gen for resistance mutations.

NAVIGATE-LAROTRECTINIB [PMID 29466156] (Drilon 2018 NEJM, N=55 pan-tumor / BTC n=2-3): TRK fusion+ solid tumors larotrectinib. Overall ORR 75% (95% CI 61-85). FDA’s first tumor-agnostic biomarker approval 2018-11 — a regulatory milestone. BTC patient numbers very small but responses consistent.
STARTRK-ENTRECTINIB [PMID 31838007] (Doebele 2020 Lancet Oncol, N=54): NTRK fusion+ solid tumors entrectinib (pan-TRK/ROS1/ALK, good CNS penetration). Overall ORR 57.4% (95% CI 43.2-70.8), mDoR 10.4 months. BTC subgroup not reported separately due to small N. FDA tumor-agnostic approval 2019-08.
TRIDENT-1 [PMID 41639379] (Besse 2026 Nat Med): NTRK fusion+ solid tumors repotrectinib (next-gen TRK, active against NTRK G595R / G667C / F589L resistance mutations). TRK-naïve cohort ORR 58% (95% CI 37-77), TRK-pretreated cohort ORR 50% (95% CI 28-72). FDA approved 2023-11 — providing a sequential precision option for BTC patients resistant to larotrectinib / entrectinib.

2.4.6 MSI-H / dMMR / TMB-H pathway (cross-subtype, BTC MSI-H ~2-3%)

KEYNOTE-158-MSIH [PMID 31682550] (Marabelle 2020 J Clin Oncol, N=233 pan-tumor / BTC cohort K n=22; long-term [PMID 35680043] Maio 2022 Ann Oncol): MSI-H / dMMR non-CRC solid tumors pembrolizumab. Overall non-CRC MSI-H ORR 34.3%; BTC subgroup ORR 40.9% (95% CI 20.7-63.6). BTC-specific data supporting the FDA tumor-agnostic approval of 2017.
ANDRE-2023-DOSTARLIMAB [PMID 37917058] (André 2023 JAMA Netw Open): dMMR solid tumors dostarlimab. Overall dMMR ORR 38.7% (95% CI 31.6-46.2), 24-month DoR rate 69.0%. BTC subgroup consistent with overall cohort. Provides an anti-PD-1 equivalent option to pembrolizumab for MSI-H / dMMR BTC.
CHECKMATE-848 [PMID 39107131] (Schenker 2024 J Immunother Cancer, N≈200 TMB-H pan-tumor): TMB-H (≥10 mut/Mb) solid tumors nivolumab + ipilimumab vs nivolumab monotherapy. ORR 43.3% vs 26.0%, PFS HR 0.72 (95% CI 0.60-0.88). BTC-specific data not separately published, but establishes randomized evidence that “TMB-H is an optional biomarker for IO combinations (overlapping ~40-50% with MSI-H).”

2.4.7 NRG1 fusion pathway (cross-subtype, <1% of BTC)

ENRGY [PMID 39908431] (Schram 2025 NEJM, N=64 pan-tumor / BTC subgroup n=3-5): NRG1 fusion+ solid tumors zenocutuzumab (HER2×HER3 bispecific, competitively blocking NRG1-HER3 signaling). Overall ORR 34% (95% CI 22-47), mDoR 11.1 months, mPFS 6.8 months. BTC subgroup consistent with the overall cohort. FDA accelerated approval 2024-08 for NRG1 fusion+ NSCLC and pancreatic cancer — tumor-agnostic data supports BTC use.

2.4.8 KRAS G12C pathway (cross-subtype, ~1-2% of BTC)

KRYSTAL-1-BTC [PMID 37099736] (Bekaii-Saab 2023 J Clin Oncol, N=12 BTC subgroup single-arm basket): KRAS G12C+ BTC adagrasib 600 mg BID. BTC ORR 33.3% (95% CI 9.9-65.1), DCR 83.3%. BTC is the first GI tumor beyond NSCLC / CRC to demonstrate KRAS G12C inhibitor activity — adding an eighth biomarker-matched pathway.

2.4.9 RET fusion pathway (cross-subtype, <1% of BTC)

LIBRETTO-001-BTC [PMID 36108661] (Subbiah 2022 Lancet Oncol, N=45 non-lung non-thyroid / BTC n=7): RET fusion+ solid tumors selpercatinib. ORR 43.9% (95% CI 29.5-59.3), mDoR 24.5 months. BTC subgroup ORR consistent with overall cohort. FDA tumor-agnostic approval 2022.
ARROW-BTC [PMID 35962206] (Subbiah 2022 Nat Med, N=29 non-lung non-thyroid): RET fusion+ solid tumors pralsetinib. ORR 57% (95% CI 37-76), 12-month DoR rate 81%. BTC data sparse but equivalent to selpercatinib. FDA tumor-agnostic accelerated approval 2022-08.

Takeaway: in 2026, a newly diagnosed advanced BTC patient must have comprehensive molecular profiling (NCCN V1.2026 strongly recommended) covering FGFR2 / IDH1 / HER2 / BRAF V600E / NTRK / MSI-H / dMMR / TMB-H / NRG1 / KRAS G12C / RET. 8 biomarker-matched pathways collectively cover 30-40% of patients — the biggest difference from HCC (0 biomarker-matched) and pancreatic cancer (POLO + 5%). If FIGHT-302 publishes as 1L positive, it will be the first precedent of “molecular-selected 1L replacing IO + GemCis.”

2.5 Ongoing phase III and pending readouts (key 2026-2028 inflection points)

Story: BTC’s key 2026-2028 readouts concentrate on two fronts: adjuvant setting (ACTICCA-1 GemCis vs capecitabine) and 1L FGFR2-selective setting (FIGHT-302 pemigatinib vs GemCis). Either outcome will directly rewrite the decision tree.

FIGHT-302 [PMID 32677452] (Bekaii-Saab 2020 Future Oncol design / NCT03656536): 1L FGFR2+ ICC pemigatinib vs GemCis. ESMO 2024 reported positive PFS readout; full manuscript still pending as of 2026-04.
ACTICCA-1 [PMID 26228433] (Stein 2015 BMC Cancer design / NCT02170090): adjuvant GemCis × 8 cycles vs amended capecitabine (control arm amended after BILCAP). Designed as the only direct head-to-head comparison of GemCis doublet vs capecitabine monotherapy in adjuvant. Primary completion expected 2026-2027.
GEMSTONE-202 (China CStone Pharmaceuticals sugemalimab + GemCis 1L phase III, positive readout reported at ASCO GI 2024/2025): as of 2026-04, the primary manuscript and public NCT ID are both unregistered — this knowledge base strictly follows the “PMID / NCT traceable” principle, not currently included in the main database. Will add as a v2 supplement once the primary publication appears.

Takeaway: the key variables of BTC decision-tree in 2026-2028 = (a) whether adjuvant GemCis beats capecitabine (ACTICCA-1) + (b) whether 1L FGFR2+ switches to pemigatinib monotherapy (FIGHT-302) + (c) whether the Chinese domestic IO + GemCis data GEMSTONE-202 is published. Each will recalibrate §3’s horizontal decision landscape.

3. Horizontal: 2026 current decision landscape (six dimensions)

Projecting the vertical evolution onto the 2026 clinical decision tree, the following are six key branchpoints and the evidence basis for each.

3.1 Newly diagnosed BTC: immediate comprehensive molecular profiling

NCCN V1.2026 explicitly recommends comprehensive molecular testing (tissue or ctDNA) for all newly diagnosed advanced BTC, covering: FGFR2 fusion / rearrangement + IDH1 R132 mutation + HER2 amplification / overexpression + BRAF V600E + MSI-H / dMMR + TMB-H + NRG1 fusion + NTRK fusion + RET fusion + KRAS G12C. Molecular testing results directly affect:

2L targeted accessibility (FGFR2 → pemigatinib / futibatinib / erdafitinib; IDH1 → ivosidenib; HER2 → zanidatamab / T-DXd / tucatinib + trastuzumab; BRAF V600E → dabrafenib + trametinib; NTRK → larotrectinib / entrectinib / repotrectinib; MSI-H → pembrolizumab / dostarlimab; NRG1 → zenocutuzumab; G12C → adagrasib; RET → selpercatinib / pralsetinib)
1L regimen selection (if the FIGHT-302 manuscript is positive, FGFR2+ ICC 1L may switch to pemigatinib monotherapy replacing IO + GemCis)
Clinical trial enrollment (FIGHT-302 / ACTICCA-1 / GEMSTONE-202 and other domestic / global recruitment)

Missing any biomarker = missing an ORR 30-50% high-yield responder population — in BTC, a tumor with 8-9 parallel biomarker pathways, the cost of non-testing is far higher than in NSCLC.

3.2 Advanced 1L: the two PD-(L)1 choices in the IO + GemCis class effect for fit patients

2026 mainstream: treatment-naive advanced BTC (regardless of PD-L1 status) preferred option = IO + GemCis × 8 cycles → IO maintenance — two independent phase III datasets with HR converging on 0.80-0.83.

Subgroup	Preferred	Alternative
All-comers PD-L1 unselected fit	durvalumab + GemCis [TOPAZ-1 PMID 38319896] or pembrolizumab + GemCis [KEYNOTE-966 PMID 37075781]	GemCis alone (when IO is not tolerated)
FGFR2 fusion+ ICC	(current) IO + GemCis; (once FIGHT-302 manuscript published) pemigatinib 1L replacement [FIGHT-302 PMID 32677452 ESMO 2024]	IO + GemCis fallback
Chinese / Asian patients	durvalumab / pembrolizumab + GemCis	Awaiting GEMSTONE-202 manuscript
ECOG 2 / IO-intolerant	GemCis alone [ABC-02 PMID 20375404]	Gemcitabine monotherapy (most frail)

Contraindicated / not recommended in 2026:

Triplet GemCis + nab-paclitaxel (SWOG-1815 PMID 39671534 negative)
Triplet mFOLFIRINOX replacing GemCis (AMEBICA PMID 34662180 negative)
Triplet GCS (GemCis + S-1, KHBO1401-MITSUBA PMID 35900311 ITT negative)

NCCN 2026: durvalumab + GemCis and pembrolizumab + GemCis are co-listed as Category 1 preferred (treatment-naive advanced BTC, regardless of subtype / PD-L1).

3.3 Adjuvant regimen: capecitabine vs S-1 vs GemCis-CRT three paths

2026 mainstream:

Western / international default: capecitabine × 8 cycles (BILCAP per-protocol significant / ITT borderline, PMID 30922733) — globally adopted by guidelines
Asian / Japanese preferred: S-1 × 4 cycles (ASCOT ITT positive HR 0.69, PMID 36681415) — the “Eastern-specific” path standing alongside BILCAP, well-tolerated given DPYD polymorphism background
R1-margin / N+ ECC / GBC: adjuvant chemo → CRT (SWOG-S0809 PMID 25964250 / OSTWAL 2024 PMID 38958997, the latter being the first GBC-only DFS-positive RCT)
Not recommended as adjuvant: GEMOX (PRODIGE-12 PMID 30707660 negative) + gemcitabine monotherapy (BCAT PMID 29405274 negative, IPD meta PMID 35182925 reaffirmed)

Key unresolved question: GemCis doublet vs capecitabine monotherapy in adjuvant — only ACTICCA-1 (PMID 26228433 design paper, NCT02170090) can answer directly; still ongoing as of 2026-04. Before ACTICCA-1 readout, 2026 clinical practice still uses capecitabine (West) / S-1 (Japan) as baseline.

NCCN V1.2026: capecitabine = Category 1 adjuvant standard for resected BTC. S-1 = Category 2A (based on ASCOT, not restricted to Asia but Western data lacking). CRT retained as Category 2B option in R1 / N+ settings.

3.4 Precision therapy eight pathways: which biomarker → which drug

Biomarker	BTC-enriched subtype	Prevalence (BTC overall / enriched subtype)	Preferred drug (FDA approval status)	Key trial
FGFR2 fusion / rearrangement	ICC	~10-15% ICC	pemigatinib (2020-04 FDA accelerated) / futibatinib (2022-09) / erdafitinib (2024-09 tumor-agnostic)	FIGHT-202 [PMID 32203698] / FOENIX-CCA2 [PMID 36652354] / RAGNAR [PMID 37541273]
IDH1 R132 mutation	ICC	~10-20% ICC	ivosidenib (2021-08 FDA approved)	ClarIDHy [PMID 32416072]
HER2 amp / overexp	GBC	~15-20% GBC	zanidatamab (2024-08 FDA accelerated) / T-DXd (2024-04 tumor-agnostic) / tucatinib + trastuzumab (off-label) / pertuzumab + trastuzumab	HERIZON-BTC-01 [PMID 37276871] / DESTINY-PanTumor02 [PMID 37870536] / SGNTUC-019 [PMID 37751561] / MyPathway-BTC [PMID 34339623]
BRAF V600E	ICC	~3-5% ICC	dabrafenib + trametinib (2022 tumor-agnostic)	ROAR-BTC [PMID 32818466] / NCI-MATCH-H [PMID 32758030]
NTRK fusion	cross-subtype	<1% BTC	larotrectinib / entrectinib / repotrectinib (next-gen)	NAVIGATE [PMID 29466156] / STARTRK [PMID 31838007] / TRIDENT-1 [PMID 41639379]
MSI-H / dMMR	cross-subtype	~2-3% BTC	pembrolizumab / dostarlimab	KEYNOTE-158 [PMID 31682550] / ANDRE-2023 [PMID 37917058]
TMB-H	cross-subtype	~3-5% BTC	nivolumab + ipilimumab	CHECKMATE-848 [PMID 39107131]
NRG1 fusion	cross-subtype (KRAS-WT enriched)	<1% BTC	zenocutuzumab (2024-08 FDA accelerated)	ENRGY [PMID 39908431]
KRAS G12C	cross-subtype	~1-2% BTC	adagrasib	KRYSTAL-1-BTC [PMID 37099736]
RET fusion	cross-subtype	<1% BTC	selpercatinib / pralsetinib (tumor-agnostic)	LIBRETTO-001 [PMID 36108661] / ARROW-BTC [PMID 35962206]

Clinical decision implication: 2026 BTC precision therapy = “test without fail for 2L, test as early as possible for 1L.” The 2L decision window requires NGS report in hand within 4-6 weeks; FGFR2 fusion + IDH1 mutation + HER2 amp + BRAF V600E are the four ICC/GBC-enriched high-yield must-tests.

3.5 Advanced 2L+: FOLFOX + East-West nal-IRI divergence + IO/regorafenib fallback

2026 mainstream:

Global default: mFOLFOX + ASC (ABC-06 PMID 33798493, HR 0.69 marginal win)
Asia / Korea: nal-IRI + 5FU/LV based on NIFTY PFS data (PMID 36951834 PFS HR 0.56 positive) — note OS negative
West: nal-IRI + 5FU/LV used cautiously — NALIRICC (PMID 38870977) OS negative / borderline
MSI-H / dMMR / TMB-H selective: anti-PD-1 monotherapy (pembrolizumab / dostarlimab) or nivo + ipi (CHECKMATE-848 data)
Biomarker-negative + multiline exhausted: regorafenib (SUN-2019-REGO-BTC PMID 30561756, DCR 56% / ORR 11% palliative only)

Clinical implication of the East-West nal-IRI divergence: do not extrapolate NIFTY Korean data to Western BTC, or vice versa. BTC’s cross-population heterogeneity is larger than typically assumed — possible PK / molecular background differences (HBV / liver fluke vs sporadic) / ICC-to-ECC ratio differences need to be considered separately.

3.6 Three-subtype differences (GBC / ICC / ECC): biomarker enrichment and decision triage

Although most registration trials enrolled mixed three subtypes (TOPAZ-1 ICC 56%, KEYNOTE-966 GBC 30%, BILCAP mixed three-subtype, etc.), biomarker enrichment differences determine subtype-specific triage for precision therapy:

Subtype	Enriched biomarker	Priority precision testing	Notes
ICC (intrahepatic cholangiocarcinoma)	FGFR2 fusion 10-15% / IDH1 mutation 10-20% / BRAF V600E 3-5%	FGFR2 + IDH1 + BRAF V600E triad	ICC is the most precision-therapy-dense BTC subtype
GBC (gallbladder cancer)	HER2 amp / overexp 15-20%	HER2 IHC + ISH must-test	GBC + HER2 = BTC’s second-largest precision niche; OSTWAL 2024 suggests GemCis-CRT intensification in high-risk GBC adjuvant
ECC (extrahepatic cholangiocarcinoma)	no single obvious enriched biomarker	comprehensive NGS looking for BRAF / NTRK / MSI-H / NRG1 / RET	ECC has the lowest precision hit rate; SWOG-S0809 CRT intensification in R1 / N+ setting
Three subtypes shared	MSI-H ~2-3% / NTRK <1% / NRG1 <1% / KRAS G12C ~1-2% / RET <1%	tumor-agnostic basket pathways	shared cross-subtype, not subtype-restricted

4. Research Gaps: the ten unresolved clinical questions

This report identifies the following gaps, all concretely defined questions (not the cliché of “more research needed”):

GemCis doublet vs capecitabine monotherapy in adjuvant head-to-head: post-BILCAP all guidelines adopted capecitabine but on a borderline ITT; ACTICCA-1 (NCT02170090) answers directly, still no readout as of 2026-04, key 2026-2027 inflection point.
FGFR2+ ICC 1L pemigatinib vs IO + GemCis: FIGHT-302 (NCT03656536) reported positive PFS at ESMO 2024, full manuscript and OS data still unpublished as of 2026-04; if published positive, it would be BTC’s first precedent of “molecular-selected 1L replacing class-effect IO + chemo.”
East-West nal-IRI 2L divergence: NIFTY (Korea PFS positive PMID 36951834) vs NALIRICC (Germany OS negative borderline PMID 38870977) — same drug, same 2L, different results — which of PK / molecular background / ICC-vs-ECC ratio / trial design is the dominant explanation? No matched prospective analysis exists.
IO benefit biomarker enrichment in TOPAZ-1 and KEYNOTE-966: neither phase III used PD-L1 selection but the 24-month landmark shows long-tail enrichment — who are the “long-tail responders”? Need pre-specified PD-L1 / TMB / dMMR / IFN-γ signature retrospective analysis and prospective validation.
HER2 IHC 3+ vs 2+ differentiated decisions: DESTINY-PanTumor02 BTC cohort IHC 3+ ORR 56% vs IHC 2+ 20% (PMID 37870536) — suggesting IHC score should be a stratification variable for treatment intensity, but a unified algorithm integrating IHC + ISH + NGS is missing.
MSI-H in BTC only 2-3%: KEYNOTE-158 BTC ORR 40.9% (PMID 31682550) is significant but N=22 too small; does BTC need a microsatellite + TMB + dMMR IHC tripartite panel to improve detection?
High-risk GBC adjuvant CRT with GemCis vs capecitabine choice: OSTWAL 2024 (PMID 38958997, India, GemCis + CRT positive) vs SWOG-S0809 (PMID 25964250, US, GemCap + CRT single-arm) — head-to-head between the two different chemo backbones is missing.
GEMSTONE-202 absent: China sugemalimab + GemCis 1L phase III data reported positive at ASCO GI 2024-2025 but manuscript and NCT registration both untraceable as of 2026-04; Chinese IO + GemCis practice evidence lacks a rigid 1L benchmark.
NRG1 / KRAS G12C / RET fusion in BTC cohorts too small: ENRGY BTC subgroup n=3-5, KRYSTAL-1-BTC n=12, LIBRETTO-001 BTC n=7 — all tumor-agnostic basket subsets, lacking BTC-only confirmatory data.
Late-line (3L+) BTC data nearly blank: all phase III concentrated on 1L/2L; beyond 3L there is no standard regimen, no large RCT, no ctDNA-guided treatment rotation protocol.

5. 2024-2026 latest developments

5.1 FDA / NMPA new approvals (ten key entries)

Drug	Agency	Date	Indication / supporting trial
zanidatamab (Ziihera)	FDA	2024-08-29	2L+ HER2 amp BTC / HERIZON-BTC-01 [PMID 37276871]
pembrolizumab + GemCis	FDA	2023-10-31	1L advanced BTC (regardless of PD-L1) / KEYNOTE-966 [PMID 37075781]
durvalumab + GemCis	FDA	2022-09-02	1L advanced BTC (regardless of PD-L1) / TOPAZ-1 [PMID 38319896]
trastuzumab deruxtecan (Enhertu)	FDA	2024-04-05	tumor-agnostic HER2 IHC 3+ solid tumors (including BTC) / DESTINY-PanTumor02 [PMID 37870536]
futibatinib (Lytgobi)	FDA	2022-09-30	2L+ FGFR2 rearrangement+ ICC / FOENIX-CCA2 [PMID 36652354]
erdafitinib (Balversa)	FDA	2024-09-23	tumor-agnostic FGFR-altered solid tumors (including BTC) / RAGNAR [PMID 37541273]
zenocutuzumab (Bizengri)	FDA	2024-08-26	NRG1 fusion+ NSCLC and pancreatic cancer (tumor-agnostic data supports BTC) / ENRGY [PMID 39908431]
ivosidenib (Tibsovo)	FDA	2021-08-25	2L+ IDH1+ CCA / ClarIDHy [PMID 32416072]
pemigatinib (Pemazyre)	FDA	2020-04-17	2L+ FGFR2+ CCA / FIGHT-202 [PMID 32203698]
repotrectinib (Augtyro)	FDA	2023-11-15	tumor-agnostic NTRK fusion+ solid tumors (BTC applicable) / TRIDENT-1 [PMID 41639379]

(This section shows 10 key approvals; the complete BTC-related approval record also includes dabrafenib + trametinib tumor-agnostic 2022 / pembrolizumab MSI-H 2017 / dostarlimab dMMR 2021 / larotrectinib 2018 / entrectinib 2019 / selpercatinib + pralsetinib RET tumor-agnostic 2022 / adagrasib NSCLC KRAS G12C 2022 and other tumor-agnostic pathways)

5.2 Key conference readouts (2024-2026, lower-weighted tagging)

The following entries are candidate-pool only before formal peer review, not in the main database. Those with PMIDs have been promoted to the main database.

FIGHT-302 ESMO 2024 oral (Bekaii-Saab et al.): 1L FGFR2+ ICC pemigatinib vs GemCis PFS-positive readout; full OS / ORR / safety manuscript still not published in a peer-reviewed journal as of 2026-04 — will rewrite the §3.2 decision tree once published.
GEMSTONE-202 ASCO GI 2024-2025 (CStone Pharmaceuticals): China sugemalimab + GemCis 1L phase III positive readout — manuscript and NCT ID both untraceable as of 2026-04.
TRIDENT-1 long-term follow-up (TRK-naive BTC subgroup data): awaiting 2026 H2 update.
ACTICCA-1 interim safety (reported 2024-2025): no primary readout yet, awaiting 2026-2027 H1.
HERIZON-BTC-302 phase III (zanidatamab 1L in HER2+ BTC): ongoing, readout awaited 2027+.

5.3 Ongoing phase III (2025-2028 readout highlights)

FIGHT-302 (NCT03656536, pemigatinib 1L vs GemCis in FGFR2+ ICC) — 2026 H2 full manuscript expected
ACTICCA-1 (NCT02170090, GemCis vs capecitabine adjuvant head-to-head) — 2026-2027 primary completion
HERIZON-BTC-302 (zanidatamab 1L combination phase III in HER2+ BTC) — 2027+
GEMSTONE-202 (China sugemalimab + GemCis 1L) — manuscript pending publication
Multiple IDH1 / FGFR2 / HER2 / NRG1 / KRAS G12C confirmatory phase III / single-arm registrational trials ongoing for 2026-2028

6. Convergent insights and judgments

6.1 Vertical × Horizontal: the 2026 BTC landscape is shaped by three “resonances”

Overlaying vertical paradigm evolution and horizontal current decision landscape, the 2026 BTC landscape is a superposition of three resonances:

GemCis dominates 12 years → IO double-hit class effect (HR 0.80-0.83 narrow-band convergence): the GemCis uniform standard starting from ABC-02 in 2010 persisted for 12 years, rewritten simultaneously by TOPAZ-1 + KEYNOTE-966 using two independent PD-(L)1 inhibitors — HR converging at 0.80-0.83 this narrow band is the textbook definition of class effect. Homologous to the pembro + chemo / atezo + chemo multi-IO class effect in NSCLC.
8 biomarker-matched pathways in parallel + three-subtype biomarker enrichment differences: FGFR2 / IDH1 enriched in ICC 10-20%, HER2 enriched in GBC 15-20%, BRAF / NTRK / MSI-H / NRG1 / KRAS G12C / RET shared cross-subtype. Collectively covering 30-40% of patients — this density is second only to NSCLC (10+ biomarkers covering 50%+), far exceeding HCC (0 biomarkers) and pancreatic cancer (POLO + 5%). BTC is the second GI tumor after NSCLC to truly enter the “molecular-selected first” era.
Three-path adjuvant parallelism + East-West divergence: capecitabine (BILCAP, West), S-1 (ASCOT, Asia), GemCis-CRT (SWOG-S0809 / OSTWAL 2024, R1/N+ ECC/GBC) three paths + ACTICCA-1 head-to-head pending. East-West BTC is not only epidemiologically different (Asia GBC-heavy / West ICC-heavy / India high-risk GBC), but treatment selection itself diverges East-West.

These three resonances together explain one clinical phenomenon: the 1L decision tree for a newly diagnosed advanced BTC patient in 2026 has 4 more decision layers than in 2018 (subtype classification → full-panel molecular testing → IO + GemCis vs FGFR2-selective 1L → Chinese / Western IO selection + clinical trial enrollment).

6.2 Clinical decision takeaways (for junior-mid oncologists)

“Panel first, then decide” is already SoC: in 2026, starting IO + GemCis for a newly diagnosed advanced BTC without comprehensive molecular profiling is wrong — missing any of FGFR2 / IDH1 / HER2 / BRAF / MSI-H / NRG1 / KRAS G12C / RET equals missing an ORR 30-50% response path.
Advanced 1L defaults to IO + GemCis class effect: durvalumab + GemCis (TOPAZ-1) and pembrolizumab + GemCis (KEYNOTE-966), choose one, HR converging 0.80-0.83. Do not use GemCis monotherapy as 1L SoC anymore.
Triplet 1L rejected by three independent phase III: SWOG-1815 (add nab-pac) / AMEBICA (switch to mFFX) / KHBO1401-MITSUBA (add S-1) all failed — do not try triplet combinations replacing the GemCis backbone again.
2026 adjuvant default capecitabine (West) / S-1 (Asia): GEMOX and gemcitabine monotherapy adjuvant exit the recommendation (PRODIGE-12 / BCAT negative, IPD meta reaffirmed). Reassess whether to switch to GemCis only after ACTICCA-1 readout.
R1 / N+ ECC/GBC add CRT: SWOG-S0809 (ECC + GBC) and OSTWAL 2024 (GBC only) supporting data — not universal adjuvant, but high-risk subgroup intensification.
ICC patients: focus on FGFR2 + IDH1 + BRAF V600E triad: ICC is the most precision-therapy-dense subtype; any one positive in the triad opens a phase III-level targeted path.
GBC patients: HER2 IHC + ISH must-test: GBC is BTC’s second-largest precision niche; HER2+ GBC has 4 mechanistically distinct targeted options (zanidatamab / T-DXd / tucatinib + trastuzumab / pertuzumab + trastuzumab) to choose from.
2L FOLFOX is default + note East-West nal-IRI divergence: mFOLFOX + ASC (ABC-06) is global default. Do not extrapolate NIFTY Korean PFS-positive to the West — NALIRICC was OS-negative borderline in Germany.
Late-line IO monotherapy recommended only for MSI-H / dMMR / TMB-H: in unselected populations IO monotherapy ORR 22% looks useful but PFS is short (KIM-2020-NIVO-BTC); use only after biomarker selection.
The 9 BTC drug classes to know in 2026: durvalumab + pembrolizumab + GemCis (1L backbone) / mFOLFOX (2L) / pemigatinib + futibatinib + erdafitinib (FGFR2) / ivosidenib (IDH1) / zanidatamab + T-DXd + tucatinib + trastuzumab + pertuzumab + trastuzumab (HER2) / dabrafenib + trametinib (BRAF V600E) / pembrolizumab + dostarlimab (MSI-H) / zenocutuzumab (NRG1) / adagrasib (KRAS G12C) / selpercatinib + pralsetinib (RET) / larotrectinib + entrectinib + repotrectinib (NTRK) — 16 years ago BTC had only the GemCis one-size-fits-all option; 2026 is already a complex decision map of 16+ drugs and 8 parallel precision pathways.

7. Information sources

The metadata of the 39 trials in this report was independently verified via PubMed and ClinicalTrials.gov. Each [PMID xxxxxxxx] in the text can be directly verified on PubMed.

Published trials: 37, covering 2010-2026 (all PMID-verifiable)
Ongoing / design papers: 2 (FIGHT-302 PMID 32677452 design paper + ACTICCA-1 PMID 26228433 design paper, primary-result manuscripts pending)
NCCN guideline citations: 39/39 directly hit NCCN Biliary Tract Cancers V1.2026 reference section or extended evidence base
2017-2024 FDA new approvals: 10+ (durvalumab / pembrolizumab + GemCis / pemigatinib / futibatinib / erdafitinib / ivosidenib / zanidatamab / T-DXd / zenocutuzumab / repotrectinib / dabrafenib + trametinib / larotrectinib / entrectinib / selpercatinib / pralsetinib / pembrolizumab + dostarlimab MSI-H / dMMR / nivo + ipi TMB-H and other tumor-agnostic pathways)
2024-2026 key conference readouts: 3 (FIGHT-302 ESMO 2024 / GEMSTONE-202 ASCO GI 2024-2025 / TRIDENT-1 long-term follow-up) — awaiting manuscript upgrade
Supporting PMIDs (cited in text but not in main trial table): 3 (ClarIDHy OS update PMID 34554208 / KEYNOTE-158 long-term follow-up PMID 35680043 / BCAT + PRODIGE-12 IPD meta PMID 35182925)
Research gaps: 10

7.1 Report-body citation list (sorted by PMID)

The table below lists the PMIDs bracket-cited in the report body, each clickable for PubMed URL verification.

PMID	Trial / Paper	Year	Journal	Text location
20375404	ABC-02	2010	NEJM	§2.2 / §3.2
25964250	SWOG-S0809	2015	J Clin Oncol	§2.1 / §3.3
26228433	ACTICCA-1 (design paper)	2015	BMC Cancer	§2.1 / §2.5 / §3.3 / §4
29405274	BCAT	2018	Br J Surg	§2.1 / §3.3
29466156	NAVIGATE-LAROTRECTINIB	2018	NEJM	§2.4.5 / §3.4
30561756	SUN-2019-REGO-BTC	2019	Cancer	§2.3 / §3.5
30707660	PRODIGE-12	2019	J Clin Oncol	§2.1 / §3.3
30922733	BILCAP	2019	Lancet Oncol	§2.1 / §3.3
31682550	KEYNOTE-158-MSIH	2020	J Clin Oncol	§2.4.6 / §3.4 / §4
31838007	STARTRK-ENTRECTINIB	2020	Lancet Oncol	§2.4.5 / §3.4
32203698	FIGHT-202	2020	Lancet Oncol	§2.4.1 / §3.4 / §5.1
32352498	KIM-2020-NIVO-BTC	2020	JAMA Oncol	§2.3 / §6.2
32416072	ClarIDHy (primary)	2020	Lancet Oncol	§2.4.2 / §3.4 / §5.1
32677452	FIGHT-302 (design paper)	2020	Future Oncol	§2.4.1 / §2.5 / §3.2 / §4
32758030	NCI-MATCH-H	2020	J Clin Oncol	§2.4.4 / §3.4
32818466	ROAR-BTC	2020	Lancet Oncol	§2.4.4 / §3.4
33798493	ABC-06	2021	Lancet Oncol	§2.3 / §3.5
34339623	MyPathway-BTC	2021	Lancet Oncol	§2.4.3 / §3.4
34554208	ClarIDHy OS update	2021	JAMA Oncol	§2.4.2 (supporting)
34662180	AMEBICA / PRODIGE-38	2022	J Clin Oncol	§2.2 / §3.2
35182925	BCAT + PRODIGE-12 IPD meta	2022	Eur J Cancer	§2.1 (supporting) / §3.3
35680043	KEYNOTE-158 long-term follow-up	2022	Ann Oncol	§2.4.6 (supporting)
35900311	KHBO1401-MITSUBA	2023	J Hepatobiliary Pancreat Sci	§2.2 / §3.2
35962206	ARROW-BTC	2022	Nat Med	§2.4.9 / §3.4
36108661	LIBRETTO-001-BTC	2022	Lancet Oncol	§2.4.9 / §3.4
36652354	FOENIX-CCA2	2023	NEJM	§2.4.1 / §3.4 / §5.1
36681415	ASCOT / JCOG1202	2023	Lancet	§2.1 / §3.3
36951834	NIFTY	2023	JAMA Oncol	§2.3 / §3.5 / §4
37075781	KEYNOTE-966	2023	Lancet	§2.2 / §3.2 / §5.1
37099736	KRYSTAL-1-BTC	2023	J Clin Oncol	§2.4.8 / §3.4
37276871	HERIZON-BTC-01	2023	Lancet Oncol	§2.4.3 / §3.4 / §5.1
37541273	RAGNAR	2023	Lancet Oncol	§2.4.1 / §3.4 / §5.1
37751561	SGNTUC-019	2023	J Clin Oncol	§2.4.3 / §3.4
37870536	DESTINY-PanTumor02	2024	J Clin Oncol	§2.4.3 / §3.4 / §4 / §5.1
37917058	ANDRE-2023-DOSTARLIMAB	2023	JAMA Netw Open	§2.4.6 / §3.4
38319896	TOPAZ-1	2022	NEJM Evid	§2.2 / §3.2 / §5.1
38870977	NALIRICC	2024	Lancet Gastroenterol Hepatol	§2.3 / §3.5 / §4
38958997	OSTWAL-2024-GEMCIS-CRT-GBC	2024	JAMA Oncol	§2.1 / §3.3 / §4
39107131	CHECKMATE-848	2024	J Immunother Cancer	§2.4.6 / §3.4
39671534	SWOG-1815	2025	J Clin Oncol	§2.2 / §3.2 / §6.2
39908431	ENRGY	2025	NEJM	§2.4.7 / §3.4 / §5.1
41639379	TRIDENT-1	2026	Nat Med	§2.4.5 / §3.4 / §5.1

7.2 Verification conventions

Each PMID is directly accessible for verification via https://pubmed.ncbi.nlm.nih.gov/{PMID}/
Each NCT id is accessible via https://clinicaltrials.gov/study/{NCT_id}/
Conference abstracts (ASCO GI / ESMO / EORTC) are searched via official conference systems; all conference citations in this report are lower-weight tagged — not peer-reviewed, final data defers to journal publication
FIGHT-302 / ACTICCA-1 are cited via “design-paper PMID + ESMO/ASCO oral readout” before manuscript publication; the PMID will be updated once the formal manuscript is published
If you find a trial name / year / conclusion in this report inconsistent with PubMed for any PMID, corrections are welcome

The trial timeline lives here

Chinese: /trials/btc/ English: /en/trials/btc/

Each trial has an independent detail page, including:

Full intervention / comparator regimens
Primary endpoint values + 95% CI
Key findings + clinical significance
Clickable links to PMID / NCT original sources

39 trials · 4 chapters · 2010 to 2026 · synced with NCCN Biliary Tract Cancers V1.2026.

Closing

BTC has undergone a unique evolution in oncology over the past 16 years — from 2010’s ABC-02 enshrining GemCis as 1L unchallenged for 12 years, to 2022-2023’s TOPAZ-1 + KEYNOTE-966 writing IO into the 1L backbone with a narrow-band HR 0.80-0.83 class effect, to 2018-2026’s dense rollout of eight-to-nine biomarker-matched pathways (FGFR2 / IDH1 / HER2 / BRAF / NTRK / MSI-H / NRG1 / KRAS G12C / RET) covering 30-40% of patients.

BTC’s precision-therapy density is second only to NSCLC, far exceeding HCC (0 biomarker-matched approvals) and pancreatic cancer (POLO + 5%). What drives this difference is not incidence (BTC global incidence <3/100,000, well below gastric / liver cancers), but the uniqueness of molecular architecture — ICC enriched for FGFR2 fusion + IDH1 mutation, GBC enriched for HER2 amp, three subtypes sharing the MSI-H / BRAF / NTRK / NRG1 / KRAS G12C / RET tumor-agnostic basket pathways. This “subtype × biomarker dual heterogeneity” makes BTC a “PMID-traceable + mandatory-panel + multi-pathway-parallel” highly-precise tumor type.

The value of this report lies not in “exhaustively listing all trials” (PubMed can), but in compressing 16 years of evolution + current decisions + unresolved gaps into the cognitive bandwidth of a single read. Next time you face a newly diagnosed BTC patient, every branch of the decision tree will have this map to consult, trace, and question.

Clinician × AI = Research Superpower + Clinical Decision Amplifier

—— Dual Brain Lab · 2026-04-21

Esophageal Cancer Trial Timeline: A Dual-Track Map of 30 Years and 42 RCTs

Tue, 21 Apr 2026 00:00:00 +0000

Esophageal Cancer Clinical Trial Timeline — In-Depth Report

Coverage: 42 NCCN Esophageal-cited landmark trials (all PMID-traceable) + ESCC / EAC dual-track evolution + East–West perioperative divergence + post-IO 2L unmet need

Curated by Dual Brain Lab (csilab.net)

1. One-sentence definition

This report traces the evolution and current decision landscape of esophageal cancer (EC) systemic + local therapy over the past 30 years (1992–2024), using the landmark clinical trials cited in the current NCCN Esophageal guideline, so that frontline clinicians in 2026 have a traceable, whole-picture map for the “who, what, and why” of treatment decisions.

Iron rule: every data point on every trial traces back to PubMed (PMID) or ClinicalTrials.gov (NCT id) — every [PMID xxxxxxxx] bracket in the body can be opened directly on PubMed for verification.

EC’s uniqueness clusters around one divide: by histology, into squamous cell carcinoma (ESCC) and adenocarcinoma (EAC) — ESCC is >85% globally and >90% in Asia, while EAC dominates in North America and Western Europe; gastroesophageal junction (GEJ) adenocarcinoma is anatomically and epidemiologically closer to gastric cancer, and the EC scope of this report covers the esophagus proper only (with some Siewert I/II mixed enrollment in the original CROSS / CheckMate-577 cohorts). Clinical decisions cross three axes: resectability (early / locally advanced / advanced) × histology (ESCC RT-sensitive / EAC chemo-sensitive) × biomarker (PD-L1 CPS / TAP / TC% — the three scores are non-interchangeable across trials, an ESCC-specific pain point).

Unlike NSCLC with its dozen-plus molecular strata (EGFR / ALK / ROS1 / KRAS / PD-L1), EC has no approved targetable driver: HER2-positive adenocarcinoma follows the gastric path (ToGA / DESTINY-Gastric01), the EGFR panitumumab POWER trial in ESCC was terminated early, and EC relies entirely on the three-pillar combination of chemo backbone + PD-1/PD-L1 IO + RT.

2. Longitudinal: five paradigm shifts along the timeline

EC treatment over the past 30 years has gone through five paradigm shifts: 1992–2007 definitive CRT established → 2012 CROSS anchored global neoadjuvant CRT → 2018–2024 East Asian ESCC path divergence (China NEOCRTEC5010 + Japan JCOG1109 DCF) → 2021+ advanced IO big three explosion → 2021+ IO enters the perioperative setting. Each shift rests on 1–3 phase III trials as fulcrums.

Compared with NSCLC’s “driver-gene + immunotherapy dual engine,” EC’s evolution is characterized by “chemo + RT + IO three-pillar combinations + East–West path divergence” — no predictive biomarker (other than PD-L1, and its three scores are not interchangeable). This resembles HCC’s “0-biomarker IO backbone,” but EC adds one more axis: East Asian ESCC vs Western EAC geographic divergence.

2.1 Definitive CRT era established (1992–2007): making “no surgery” an option, and the dose paradox

Story: before 1992, local treatment for ESCC = RT alone or radical surgery (perioperative mortality 10%). RTOG 85-01 used 50 Gy + CF to push 2-year OS from 10% to 38%, and 5-year follow-up from 0% to 26% — establishing concurrent def-CRT as a curative option. Later, INT 0123 tried dose escalation to 64.8 Gy and found OS worse, creating the counterintuitive rule that “50.4 Gy is the global def-CRT dose ceiling”; in 2007, FFCD 9102 further showed that among induction-CRT responders, “adding surgery vs continuing CRT” yielded no 2y OS difference (but perioperative mortality was 9.3% vs 0.8%) — the organ-preservation logic was born.

RTOG 85-01 [PMID 1584260] (Herskovic 1992 N Engl J Med, N=121): concurrent cisplatin+5FU × 4 cycles + 50 Gy vs RT alone at 64 Gy. 2-year OS 38% vs 10% (p<0.001), stopped early. Concept born: concurrent def-CRT is a curative option; concurrent AEs rose markedly but were outweighed by the survival gain.
RTOG 85-01 long-term [PMID 10235156] (Cooper 1999 JAMA): 5-year follow-up. 5-year OS 26% vs 0% (p<0.0001). First proof that def-CRT can cure a meaningful fraction of EC without surgery — defining “RT 50 Gy + CF doublet” as the ESCC def-CRT backbone for the next 30 years.
FFCD 9102 [PMID 17401004] (Bedenne 2007 J Clin Oncol, N=259): induction-CRT responders randomized to “continue CRT to 66 Gy vs add surgery.” 2y OS 34% vs 40% (HR 0.90, NS); 3-month perioperative mortality 9.3% vs 0.8% (p=0.002). Organ-preservation logic: responders need not add surgery — but only for induction-CRT responders; non-responders still need salvage surgery.

Takeaway: the two RTOG / FFCD generations in 1992–2007 established the three pillars of ESCC def-CRT: (1) 50 Gy + CF doublet is the dose ceiling (INT 0123 at 64.8 Gy was worse; all later Western trials kept 50–50.4 Gy as the gold standard); (2) in responders, def-CRT can replace surgery, especially in high-surgical-risk patients, high-cervical tumors, or those who refuse surgery; (3) salvage esophagectomy after def-CRT failure remains feasible (see §2.5 FREGAT cohort).

2.2 Global neoadjuvant CRT (2012–2018): CROSS rewrote the world + East Asia split off

Story: in 2012, the Dutch CROSS trial used weekly carboplatin+paclitaxel × 5 + 41.4 Gy to push mOS from 24.0 to 49.4 months (HR 0.657), with pCR 29% and R0 92% — anchoring the global neoadjuvant CRT paradigm. CROSS long-term (2015) and the 10-year follow-up (2021) kept validating it (SCC subgroup HR 0.48, far stronger than AC). In 2018, China’s NEOCRTEC5010 used vinorelbine+cisplatin × 2 + 40 Gy/20 fx (Western dose + Chinese chemo regimen) with mOS 100.1 vs 66.5 months (HR 0.71), pCR 43.2% — the Chinese ESCC neoadjuvant standard was born, standing alongside CROSS.

CROSS primary [PMID 22646630] (van Hagen 2012 N Engl J Med, N=366, Netherlands, 75% EAC / 23% ESCC): neoadjuvant weekly carboplatin AUC 2 + paclitaxel 50 mg/m² × 5 + 41.4 Gy/23 fx → surgery vs surgery alone. mOS 49.4 vs 24.0 months (HR 0.657, p=0.003), R0 92% vs 69%, pCR 29%. Global paradigm anchor — all Western EC perioperative trials since have used CROSS as the comparator.
CROSS long-term [PMID 26254683] (Shapiro 2015 Lancet Oncol, N=366): benefit sustained at 7-year median follow-up (HR 0.68). SCC subgroup HR 0.48 — the squamous effect is much stronger than adenocarcinoma (even though only 84 SCC patients). This SCC/AC subgroup gap later directly justified choosing between def-CRT and neoadjuvant CRT in ESCC.
CROSS 10-year [PMID 33891478] (Eyck 2021 J Clin Oncol, N=366): 10-year absolute OS gain of 13% (38% vs 25%). Locoregional relapse HR 0.40 stable over time — long-term local control is CROSS’s most robust signal.
JCOG9907 [PMID 21879261] (Ando 2012 Ann Surg Oncol, N=330, Japan, 100% ESCC): post-op CF vs pre-op CF × 2 cycles. 5y OS 55% (preop) vs 43% (postop), HR 0.73, p=0.04. Japan’s ESCC SoC became “pre-op CF doublet”, diverging from the Western CRT path. JCOG’s subsequent work stayed on the “pre-op chemo” track (ultimately evolving into 2024 JCOG1109’s DCF triplet).
NEOCRTEC5010 [PMID 30089078] (Yang 2018 J Clin Oncol, N=451, China, 100% ESCC): neoadjuvant vinorelbine 25 mg/m² d1,8 + cisplatin 75 mg/m² d1 × 2 cycles + 40 Gy/20 fx → surgery vs surgery alone. mOS 100.1 vs 66.5 months (HR 0.71, p=0.025), pCR 43.2%, R0 98.4%. The Chinese ESCC neoadjuvant standard was born — pCR 43.2% is one of the highest in EC perioperative history.
NEOCRTEC5010 long-term [PMID 34160577] (Yang 2021 JAMA Surg, N=451): long follow-up with 5y OS 59.9% vs 49.1% (HR 0.74, p=0.03), 5y DFS 63.6% vs 43.0% — LA-ESCC long-term curve separation held steady.
SCOPE-1 [PMID 28196063] (Crosby 2017 Br J Cancer, N=258, UK, 73% ESCC): def-CRT (cisplatin+capecitabine + 50 Gy) ± cetuximab phase II/III. Cetuximab arm closed early for futility (HR 1.25); def-CRT-only arm mOS 34.5 months. Modern Western def-CRT benchmark, and the first failed attempt at “ESCC + EGFR mAb.”
Neo-AEGIS [PMID 37734399] (Reynolds 2023 Lancet Gastroenterol Hepatol, N=377, Ireland-led): locally advanced esophageal / GEJ adenocarcinoma neo-CRT (CROSS regimen) vs perioperative chemo (MAGIC / FLOT). 3y OS 55% vs 57% (HR 1.03, 95% CI 0.77–1.38, equipoise) — CROSS and FLOT-perioperative are equivalent in EAC, the first phase III head-to-head of neoadjuvant CRT vs perioperative chemo in EAC.

Takeaway: in the 5 years 2012–2018, CROSS anchored neoadjuvant CRT for Western EC + GEJ AC, JCOG9907 anchored pre-op CF chemo for Japanese ESCC, and NEOCRTEC5010 anchored pre-op CRT for Chinese ESCC — three geographic paths branched apart. Neo-AEGIS later showed CROSS and FLOT-perioperative are equivalent in EAC, but the three ESCC paths have never been directly phase-III head-to-head compared.

2.3 Intra–East Asia ESCC divergence (2024): JCOG1109 NExT broke the “RT always helps” assumption

Story: in 2024, Japan’s JCOG1109 NExT was the first trial to compare CF doublet / DCF triplet / CF+RT in three arms within one trial. The result was DCF beat CF (HR 0.68, p=0.006); CF+RT did not beat CF (HR 0.84, NS) — apparently conflicting with China’s NEOCRTEC5010 (CRT beat surgery HR 0.71) and Europe’s CROSS (SCC subgroup HR 0.48). Possible explanation: Japan’s high-quality D2+ lymphadenectomy + intraoperative 3FL (three-field lymphadenectomy) absorbed the extra RT benefit — a hypothesis never formally tested.

JCOG1109 NExT [PMID 38876133] (Kato 2024 Lancet, N=601, Japan, 100% ESCC): three-arm RCT — NeoCF vs NeoCF+D (DCF = docetaxel+CF) vs NeoCF+RT (CF + 41.4 Gy) → surgery. 3y OS 62.6% / 72.1% / 68.3%; DCF vs CF HR 0.68, p=0.006 positive; CF+RT vs CF HR 0.84, NS — Japanese ESCC neoadjuvant SoC upgraded from CF to DCF, with no RT added.

Takeaway: 2024’s JCOG1109 NExT rewrote Japan’s ESCC neoadjuvant SoC — from CF doublet up to DCF triplet, but no RT added. This crystallized a three-country split within East Asian ESCC: China nCRT (NEOCRTEC5010) / Japan DCF triplet without RT (NExT) / Korea still mostly def-CRT. No direct head-to-head between the three paths — each country cites its own landmark. The universal clinical lesson: surgical-dissection quality, chemo intensity, and RT benefit substitute for each other, and the RT benefit may be absorbed under high-quality 3FL lymphadenectomy.

2.4 Advanced IO big three explosion (2019–2024): 4 years and 8 positive phase III trials reset SoC from CF/TP to IO+chemo

Story: in 2019, ATTRACTION-3 first pushed nivolumab 2L ESCC to positive (HR 0.77), independent of PD-L1 expression — the first global approval for 2L ESCC IO. In 2020–2022 the big three arrived the same year: KEYNOTE-590 (ESCC + EAC 1L global pembro, PMID 34454674), CheckMate-648 (pure ESCC 1L global nivo+chemo / nivo+ipi, PMID 35108470), ESCORT-1st (China camrelizumab 1L, PMID 34519801). 8 positive phase III trials in four years (plus ORIENT-15 sintilimab, JUPITER-06 toripalimab, RATIONALE-306 tislelizumab, ASTRUM-007 serplulimab, GEMSTONE-304 sugemalimab) completely reset 1L SoC from CF/TP. In the 2L battlefield, ATTRACTION-3 / KEYNOTE-181 / ESCORT / RATIONALE-302 ran in parallel, all HRs converging tightly in the 0.69–0.77 band.

ATTRACTION-3 [PMID 31582355] (Kato 2019 Lancet Oncol, N=419, East Asia-led): nivolumab vs taxane / docetaxel 2L ESCC. mOS 10.9 vs 8.4 months (HR 0.77, p=0.019). First global approval for 2L ESCC IO, independent of PD-L1 — opened the ESCC IO door.
KEYNOTE-181 [PMID 33026938] (Kojima 2020 J Clin Oncol, N=628, global, ESCC ~65% / EAC ~35%): pembrolizumab vs investigator-choice chemo 2L EC. CPS≥10 mOS 9.3 vs 6.7 months (HR 0.69, p=0.0074); SCC subgroup mOS 8.2 vs 7.1 months (HR 0.78); ITT HR 0.89 (p=0.056) negative. FDA approval restricted to CPS≥10 — set the strict PD-L1 biomarker posture. Afterward KN-590 kept CPS, CheckMate-648 used TC%, RATIONALE-306 used TAP — scoring methods started fragmenting.
ESCORT [PMID 32416073] (Huang 2020 Lancet Oncol, N=457, China): camrelizumab vs docetaxel / irinotecan 2L ESCC. mOS 8.3 vs 6.2 months (HR 0.71, p=0.001). First domestic Chinese PD-1 approval in ESCC — paved the way for the 1L ESCORT-1st combination (2021).
KEYNOTE-180 [PMID 30570649] (Shah 2019 JAMA Oncol, N=121, global): heavily pretreated EC (ESCC + EAC) pembrolizumab monotherapy phase II. ORR 9.9%, ESCC subgroup ORR 14.3% (vs EAC 5.2%) — early signal that ESCC responds to IO better than EAC, one basis for KN-590’s stratification.
KEYNOTE-590 [PMID 34454674] (Sun 2021 Lancet, N=749, global, ESCC 74% / EAC 26%): pembrolizumab + CF vs placebo + CF 1L. ESCC + CPS≥10 mOS 13.9 vs 8.8 months (HR 0.57, p<0.0001); all pts mOS 12.4 vs 9.8 months (HR 0.73). First global 1L positive phase III to include ESCC, FDA-approved 2021-03 — opened the ESCC 1L IO+chemo era.
ESCORT-1st [PMID 34519801] (Luo 2021 JAMA, N=596, China): camrelizumab + TP (paclitaxel+cisplatin) vs placebo + TP 1L ESCC. mOS 15.3 vs 12.0 months (HR 0.70, p=0.001), mPFS 6.9 vs 5.6 months. First 1L ESCC IO approval in China (NMPA 2021) — almost the same year as KN-590.
CheckMate-648 [PMID 35108470] (Doki 2022 N Engl J Med, N=970, global, pure ESCC): three arms — nivolumab + chemo / nivolumab + ipilimumab (chemo-free) / chemo 1L ESCC. TC PD-L1≥1%: nivo+chemo mOS 15.4 vs 9.1 months (HR 0.54); nivo+ipi mOS 13.7 vs 9.1 months (HR 0.64). First approved chemo-free IO+IO 1L regimen — ESCC has one chemo-free option that gastric does not, an exclusive advantage.
ORIENT-15 [PMID 35440464] (Lu 2022 BMJ, N=659, China + partial Europe): sintilimab + chemo vs placebo + chemo 1L ESCC. All pts mOS 16.7 vs 12.5 months (HR 0.63, p<0.001). Second domestic PD-1 + chemo 1L approval.
JUPITER-06 [PMID 35245446] (Wang 2022 Cancer Cell, N=514, China): toripalimab + TP vs placebo + TP 1L ESCC. OS HR 0.58 (95% CI 0.43–0.78, p=0.0004), mPFS HR 0.58 — strongest OS HR among the 8 1L ESCC IO trials.
RATIONALE-306 [PMID 37080222] (Xu 2023 Lancet Oncol, N=649, global): tislelizumab + chemo vs placebo + chemo 1L ESCC. mOS 17.2 vs 10.6 months (HR 0.66, p<0.0001). FDA-approved 2024-03 — global multi-region tisle 1L positive, using TAP (tumor area positivity) scoring.
RATIONALE-302 [PMID 35442766] (Shen 2022 J Clin Oncol, N=512, global ESCC): tislelizumab vs investigator-choice chemo 2L ESCC. mOS 8.6 vs 6.3 months (HR 0.70, p=0.0001). Tisle 2L ESCC FDA-approved 2024-03 — same month as 1L RATIONALE-306.
ASTRUM-007 [PMID 36732627] (Song 2023 Nat Med, N=551, China PD-L1 CPS≥1): serplulimab + chemo vs placebo + chemo 1L PD-L1+ ESCC. mOS 15.3 vs 11.8 months, mPFS HR 0.60 — fifth domestic PD-1 1L.
GEMSTONE-304 [PMID 38302715] (Li 2024 Nat Med, N=540, China): sugemalimab (anti-PD-L1 mAb) + CF vs placebo + CF 1L ESCC. mOS 15.3 vs 11.5 months (HR 0.70, p=0.008), mPFS 6.2 vs 5.4 months — first anti-PD-L1 (not PD-1) ESCC 1L positive, providing a PD-L1 option for bev-contraindicated / PD-1-intolerant patients.
CAP-02 [PMID 34998471] (Meng 2022 Lancet Gastroenterol Hepatol, N=52, China phase II single-arm): camrelizumab + apatinib 2L ESCC. ORR 34.6%, mPFS 6.8 months, mOS 15.8 months — IO + anti-angiogenic TKI 2L combination signal, but not an RCT and does not enter SoC.
CAP-02 Re-challenge [PMID 39307038] (Meng 2024 Eur J Cancer, N=49 prior-ICI ESCC): cam + apatinib in prior-ICI ESCC re-challenge phase II single-arm. ORR 10.2%, mOS 7.5 months — systematically confirmed that post-IO rescue is hard, the single largest clinical gap for ESCC in the next 2–3 years after 1L IO became universal.
ALTER1102 [PMID 33586360] (Huang 2021 Cancer Med, N=165, China phase II RCT): anlotinib (multi-target TKI) vs placebo 2L+ ESCC. mPFS 3.0 vs 1.4 months (HR 0.46, p<0.0001) — one of the post-IO 2L candidates accessible in China (currently still off-label for ESCC).
RAMONA [PMID 36098320] (Ebert 2022 Lancet Healthy Longev, N=66, Germany elderly ESCC phase II): nivo+ipi 2L in elderly ESCC (age ≥65, median 71). mOS 7.2 months, G3+ toxicity manageable — early signal of a chemo-free 2L option for elderly / frail ESCC.

Takeaway: in the 5 years 2019–2024, ESCC 1L SoC was completely reset from “CF doublet” to IO + chemo (8 positive phase III trials, HR 0.58–0.73 tight convergence); 4 positive 2L ESCC IO trials run in parallel (nivo / pembro / cam / tisle, HR 0.69–0.77). CheckMate-648’s nivo+ipi chemo-free arm is ESCC’s exclusive chemo-free choice. But this era also left two unresolved problems: (1) fragmented PD-L1 scoring (CPS / TAP / TC%), non-interchangeable across trials; (2) no positive IO re-challenge data for post-1L-IO 2L — CAP-02 Re-challenge has already shown ORR only 10%.

2.5 IO enters the perioperative setting (2021–2024): CheckMate-577 adjuvant rewrote + ESCORT-NEO neo-adj established

Story: in 2021, CheckMate-577 added 1 year of adjuvant nivolumab to patients with residual disease (non-pCR) after CROSS neoadjuvant + surgery, pushing mDFS from 11.0 to 22.4 months (HR 0.69) — first approved adjuvant IO, global SoC. In 2024, China’s ESCORT-NEO (Qin Jianjun, PI) ran neoadjuvant camrelizumab + chemo as a 391-patient three-arm phase III comparison (Cam+nab-TP vs Cam+TP vs TP), with pCR 28.0% vs 15.4% vs 4.7% (p<0.0001) positive — first positive phase III of pre-op IO+chemo, laying the foundation for ESCC perioperative IO SoC. EFS / OS still maturing. This contrasts sharply with the negative KEYNOTE-585 adjuvant nivo in gastric: same strategy, organ difference.

CheckMate-577 [PMID 33789008] (Kelly 2021 N Engl J Med, N=794, global EC + GEJ): CROSS neoadjuvant + surgery with residual disease (non-pCR) randomized to adjuvant nivolumab × 1 year vs placebo. mDFS 22.4 vs 11.0 months (HR 0.69, 96.4% CI 0.56–0.86, p<0.001). First approved adjuvant IO in EC, global SoC — filled the treatment gap for non-pCR patients after CROSS neoadjuvant. ASCO 2025 mature OS showed PD-L1-positive subgroup sustained benefit, PD-L1-low subgroup questionable — from 2026 onward, stratified decisions should replace ITT one-size-fits-all.
ESCORT-NEO / NCCES01 [PMID 38956195] (Qin 2024 Nat Med, N=391, China LA-ESCC phase III three-arm): Cam+nab-TP vs Cam+TP vs TP (× 2 cycles) → surgery. pCR 28.0% vs 15.4% vs 4.7% (both Cam vs TP p<0.0001) — first positive phase III of pre-op IO+chemo. EFS not mature (still in follow-up in 2026), but positive pCR has driven the Chinese NMPA path. Sharp organ contrast with the negative KEYNOTE-585 adjuvant nivo in gastric.
NICE [PMID 37696429] (Yang 2024 J Thorac Cardiovasc Surg, N=60, China cN2-3 ESCC phase II): nab-TP + camrelizumab × 2 → surgery. 2y OS 78.1%, 2y RFS 67.9%, MPR strongly prognostic (MPR+ 2y OS 91%).
Keystone-001 [PMID 39406186] (Shang 2024 Cancer Cell, N=47, China resectable ESCC phase II): pembrolizumab + nab-TP × 2 → surgery. MPR 72%, pCR 41%, 2y OS 91%, 2y DFS 89% — extremely high small-sample signal for neoadjuvant pembro + chemo in Chinese ESCC.
PALACE-1 [PMID 33373868] (Li 2021 Eur J Cancer, N=20, China pilot): pembrolizumab + CROSS regimen (carbo/pac + 41.4 Gy) neoadjuvant. pCR 55.6% (10/18 resected), G3+ AE 65% — first pilot signal for IO + CRT neoadjuvant; pCR extremely high, toxicity to be watched.
KEYNOTE-975 [PMID 33533655] (Shah 2021 Future Oncol, design paper): def-CRT + pembrolizumab vs def-CRT + placebo (LA unresectable EC) phase III design. Still in follow-up in 2026, primary results not yet published — will be the first global phase III readout of def-CRT + IO.

Takeaway: in 2026, three clear positions for EC perioperative IO: (1) CROSS neoadjuvant + surgery + non-pCR → adjuvant nivolumab × 1 year (global SoC); (2) LA-ESCC neoadjuvant IO+chemo enters SoC (ESCORT-NEO pCR positive) — EFS / OS still maturing in 2026–2027; (3) def-CRT + IO is not standard until the KEYNOTE-975 readout — the last missing piece of the def-CRT path. Lesson from the negative KEYNOTE-585 adjuvant nivo in gastric: organ difference + histology difference determine perioperative IO success — ESCC positive, EAC / gastric adenocarcinoma negative.

2.6 Surgical technique and local therapy (2012–2020): MIE / RAMIE / proton vs IMRT

Story: in 2012, TIME was the first phase III to prove MIE (minimally invasive esophagectomy) vs open — post-op pulmonary infection 34% → 12%, 3y OS equivalent; in 2019, ROBOT also showed RAMIE (robot-assisted MIE) vs open as positive (total complications 59% vs 80%, 5y OS equivalent); in 2020, Lin et al. used a Bayesian phase IIB to prove PBT (protons) vs IMRT had 2.3× lower total toxicity burden and 7.6× lower post-op complication score, but 3y PFS 51% was equivalent. NRG-GI006 phase III awaits readout. MIE vs RAMIE still has no direct RCT — the three modalities each completed an RCT vs open. Salvage esophagectomy after def-CRT failure has been established as perioperatively feasible in a large cohort study (FREGAT).

TIME [PMID 22552194] (Biere 2012 Lancet, N=115): MIE vs open EC. In-hospital pulmonary infection 34% vs 12% (RR 0.35); 3y OS 50.5% vs 40.4% (NS), oncologically equivalent — MIE markedly reduces pulmonary AEs with non-inferior OS.
TIME long-term [PMID 28187044] (Straatman 2017 Ann Surg, N=115): 3-year follow-up. 3y OS 40.4% (open) vs 50.5% (MIE), HR 0.88, NS; 3y DFS 35.9% (open) vs 40.2% (MIE) — long-term oncologic outcomes equivalent or better.
ROBOT [PMID 30308612] (van der Sluis 2019 Ann Surg, N=112): RAMIE vs open thoracoabdominal esophagectomy. Total surgery-related complications 59% vs 80% (RR 0.74, p=0.02), 5y OS 41% vs 40% equivalent — the only RCT of robot-assisted MIE vs open.
ROBOT long-term [PMID 33241302] (de Groot 2020 Dis Esophagus, N=109): long follow-up. 5y DFS 42% vs 43%, 5y cancer-related survival 50% vs 49% — fully equivalent 5-year oncologic outcome.
PBT vs IMRT (Lin) [PMID 32160096] (Lin 2020 J Clin Oncol, N=145 randomized phase IIB Bayesian): protons (PBT) vs IMRT in LA EC (concurrent def-CRT or neoadjuvant CRT). Total toxicity burden TTB 17.4 vs 39.9 (2.3× lower); post-op complication score 2.5 vs 19.1 (7.6× lower); 3y PFS 51% equivalent — PBT toxicity advantage clear, OS equivalent pending NRG-GI006 phase III readout.
Markar salvage [PMID 26195702] (Markar 2015 J Clin Oncol, N=308 salvage vs 540 neoadjuvant CRS matched): salvage esophagectomy after def-CRT failure vs planned neoadjuvant CRT + surgery. After matching, 3y OS 43.3% (salvage) vs 40.1% (neo-CRT + surgery) NS; in-hospital mortality higher for salvage but trending toward equivalence after post-2018 modernization — salvage esophagectomy is feasible in specialized centers, but perioperative risk is higher than planned neoadjuvant CRS.

Takeaway: EC surgical landscape in 2026: MIE is the default in high-volume Western centers, RAMIE is rapidly replacing MIE in centers with robotic infrastructure, open is used only when MIE/RAMIE are contraindicated. Proton therapy is used only in selected toxicity-concern patients (high cardiac / pulmonary toxicity risk) until the NRG-GI006 OS readout. MIE vs RAMIE head-to-head phase III is currently missing — an EC surgical research gap.

2.7 EGFR death in ESCC (2017–2020): POWER + SCOPE-1

Story: ESCC and head-and-neck SCC (HNSCC) are both squamous — in HNSCC, EXTREME / cetuximab+CF 1L and adjuvant cetuximab + RT are both positive. But in ESCC, the two EGFR mAb attempts (SCOPE-1 def-CRT + cetuximab, POWER 1L panitumumab + CF) were both clearly negative, and POWER was even stopped early at interim HR 1.77 — the molecular-biology divergence between ESCC and HNSCC became a marker that ESCC is an independent cancer type.

SCOPE-1 [PMID 28196063] (Crosby 2017 Br J Cancer, see §2.2): def-CRT + cetuximab arm closed early for futility (HR 1.25) — first failure of ESCC def-CRT + EGFR mAb.
POWER [PMID 31959339] (Moehler 2020 Ann Oncol, N=146, Europe AIO/EORTC phase III): CF ± panitumumab (anti-EGFR mAb) 1L advanced ESCC. Interim analysis HR 1.77 (95% CI 1.07–2.94), stopped early — panitumumab shortened survival. Second failure closed the ESCC EGFR path; no EGFR targeting has been tried in EC since.

Takeaway: ESCC EGFR path is dead — despite sharing squamous histology with HNSCC, the molecular mechanisms are entirely different. The ESCC driver has never been defined; in 2026, ESCC precision treatment = PD-L1 scoring (CPS / TAP / TC%) is the entire story — a unique “precision treatment without a biomarker” predicament for ESCC.

3. Cross-sectional: 2026 decision landscape (six dimensions)

Projecting the longitudinal evolution onto the concrete 2026 clinical decision tree, here are the six key branchpoints and the evidence supporting each.

3.1 Newly diagnosed advanced ESCC 1L: 8 IO+chemo phase III trials in parallel, how to choose under fragmented PD-L1 metrics

2026 mainstream: the preferred 1L for newly diagnosed advanced ESCC is IO + chemo (CF or TP) × 4–6 cycles → IO maintenance — eight positive phase III trials with OS HR 0.58–0.73 tight convergence, a class effect. Choice is driven by PD-L1 scoring method + accessibility + chemo backbone preference.

Subgroup	Preferred	Alternative
Global ESCC, PD-L1 high (CPS≥10 / TAP≥10% / TC≥1%)	pembro + CF (KEYNOTE-590 [PMID 34454674], ESCC+CPS≥10 mOS 13.9 months HR 0.57) / nivo + chemo or nivo + ipi (CheckMate-648 [PMID 35108470], TC≥1% HR 0.54 / 0.64) / tisle + chemo (RATIONALE-306 [PMID 37080222], mOS 17.2 months HR 0.66)	—
China ESCC 1L	camrelizumab + TP (ESCORT-1st [PMID 34519801], mOS 15.3 months) / sintilimab + chemo (ORIENT-15 [PMID 35440464], mOS 16.7 months) / toripalimab + TP (JUPITER-06 [PMID 35245446], OS HR 0.58 strongest) / serplulimab + chemo (ASTRUM-007 [PMID 36732627], CPS≥1)	sugemalimab + CF (GEMSTONE-304 [PMID 38302715], anti-PD-L1 option)
Chemo-intolerant ESCC	nivo + ipi chemo-free (CheckMate-648 ipi arm [PMID 35108470], only approved chemo-free regimen, TC≥1% mOS 13.7 months)	—
anti-PD-L1 preference / PD-1-intolerant	sugemalimab + CF (GEMSTONE-304 [PMID 38302715])	—

PD-L1 scoring fragmentation pain point: KEYNOTE-590 uses CPS (combined positive score), CheckMate-648 uses TC% (tumor cell %), RATIONALE-306 uses TAP (tumor area positivity) — the three scores are non-interchangeable across trials. The first step in real-world ESCC 1L decisions is to confirm the PD-L1 scoring method reported by pathology matches the trial’s, not to apply the conclusion directly. This is an ESCC-specific pain point (NSCLC uses TPS uniformly; CRC uses MSI uniformly).

NCCN 2026: all eight IO+chemo listed as Category 1 preferred 1L ESCC; CheckMate-648’s nivo+ipi chemo-free is Category 1 (chemo-intolerant); GEMSTONE-304’s sugemalimab is Category 2A (NMPA-approved, not FDA-approved).

3.2 Advanced EAC / GEJ AC 1L: KEYNOTE-590 all-comers + borrowing from gastric

2026 mainstream: EAC makes up only 20–30% of enrollment in all ESCC 1L IO phase III trials, so EAC-only data density is far lower than ESCC. Clinical decisions often borrow from the gastric path (CheckMate-649 / KEYNOTE-859).

KEYNOTE-590 EAC subgroup [PMID 34454674]: EAC + CPS≥10 mOS benefit slightly weaker than ESCC+CPS≥10; all-EAC HR not separately reported — EAC 1L IO evidence mainly comes from EAC subgroup of mixed ESCC+EAC phase III.
KEYNOTE-181 EAC subgroup [PMID 33026938]: EAC 2L pembro subgroup HR close to 1.0 — EAC responds to IO less than ESCC (mechanistic hypothesis: ESCC is squamous-hot immunologically; EAC histology resembles adenocarcinoma / intestinalized tissue more).
2026 clinical decision: EAC / GEJ AC 1L preferentially follows the gastric path (nivo+chemo CheckMate-649 / pembro+chemo KEYNOTE-859); EAC is not suitable for chemo-free IO+IO (no positive evidence); HER2+ EAC adds trastuzumab via the gastric path (ToGA / DESTINY-Gastric01 extrapolation).

3.3 Advanced 2L+: post-1L-IO is the largest ESCC clinical gap of the next 5 years

2026 mainstream: 2L ESCC mOS has sat at 6–8 months for 16 years. In the IO-naive era, ATTRACTION-3 / KEYNOTE-181 / ESCORT / RATIONALE-302 were four positive trials; after 1L IO became universal, >90% of 2L ESCC patients are post-IO — CAP-02 Re-challenge [PMID 39307038] has shown IO+VEGFR TKI re-challenge ORR is only 10%, and there is no positive phase III for post-IO 2L.

Scenario	2L preferred	Evidence
IO-naive ESCC (increasingly rare)	nivo (ATTRACTION-3) / pembro CPS≥10 (KN-181) / tisle (RATIONALE-302) / camrelizumab (ESCORT)	4 phase III, HR 0.69–0.77
post-1L-IO ESCC (new mainstream)	paclitaxel / docetaxel / irinotecan monotherapy; anlotinib (ALTER1102 [PMID 33586360], accessible in China)	expert consensus + small phase II, no positive phase III
Elderly / frail / chemo-intolerant ESCC	RAMONA nivo+ipi chemo-free ([PMID 36098320], mOS 7.2 months)	single-center phase II
HER2+ EAC 2L+	trastuzumab deruxtecan (T-DXd) via gastric path	DESTINY-Gastric01 extrapolation

Warning: CAP-02 Re-challenge [PMID 39307038] (cam+apatinib in prior-ICI ESCC) — ORR 10.2%, mOS 7.5 months — IO+TKI re-challenge is not the solution. ESCC phase III candidates in the next 2–3 years include ivonescimab (PD-1+VEGF bispecific), cadonilimab (PD-1+CTLA-4 bispecific), HER2 ADC (EAC), TROP2 ADC — phase III readouts expected 2026–2027. Before then, post-IO 2L is ESCC’s most urgent unmet need.

3.4 Perioperative LA-ESCC: East Asian three-country split + CheckMate-577 adjuvant global SoC

2026 mainstream:

Region	Perioperative SoC (2026)	Evidence
Europe EC (mixed ESCC + EAC)	CROSS (carboplatin/paclitaxel + 41.4 Gy → surgery) [PMID 22646630 / 26254683 / 33891478]	phase III, 10-year OS 13% absolute gain
US EC + GEJ AC	CROSS or FLOT-perioperative (Neo-AEGIS [PMID 37734399] equipoise)	phase III, two EAC paths equivalent
China ESCC	NEOCRTEC5010-style nCRT (vinorelbine+cisplatin + 40 Gy/20 fx) [PMID 30089078 / 34160577]; ESCORT-NEO-style Cam+nab-TP (IO-era neoadjuvant) [PMID 38956195]	phase III, pCR 43.2% / 28%
Japan ESCC	NExT DCF triplet (no RT) [PMID 38876133]	phase III, 3y OS 72.1%, HR 0.68
non-pCR global	CROSS neoadjuvant + surgery (non-pCR) → adjuvant nivolumab × 1 year (CheckMate-577 [PMID 33789008], mDFS 22.4 months HR 0.69)	phase III, global SoC

The three-country paths have never been directly phase-III compared — each country cites its own landmark. 2024 JCOG1109 NExT gave Japan evidence for “no RT added” (DCF beats CF; CF+RT does not beat CF, with the high-quality 3FL lymphadenectomy hypothesis absorbing the RT benefit never formally tested); China’s ESCORT-NEO proved neoadjuvant IO+chemo pCR positive; Europe stays within the CROSS paradigm.

CheckMate-577 subgroup insight: ASCO 2025 mature OS showed PD-L1-positive subgroup sustained benefit, PD-L1-low subgroup questionable — from 2026 onward, non-pCR adjuvant nivo should be stratified by PD-L1 rather than applied ITT.

3.5 Unresectable LA ESCC def-CRT: 50 Gy + CF dose unchanged for 40 years + KN-975 pending

2026 mainstream: the concurrent cisplatin+5FU × 4 cycles + 50 Gy/25 fx established by RTOG 85-01 [PMID 1584260 / 10235156] has been the global def-CRT standard since 1992.

Dose paradox: INT-0123 (1992–2001 follow-on) tried 64.8 Gy dose escalation → worse OS, and modern IMRT dose-painting (ARTDECO) was also negative — 50 Gy is the dose ceiling, going higher adds no benefit and increases toxicity.
Modern def-CRT benchmark: SCOPE-1 [PMID 28196063] mOS 34.5 months (cisplatin+capecitabine replacing CF, cetuximab closed for futility); China and Japan sometimes use slightly higher dose (60 Gy) but no RCT evidence shows superiority over 50 Gy.
FFCD 9102 organ preservation logic [PMID 17401004]: induction-CRT responders with 2y OS 34% continuing CRT vs 40% adding surgery (NS, perioperative mortality 9.3% vs 0.8%) — responders need not add surgery; non-responders can still undergo salvage esophagectomy (Markar [PMID 26195702] 3y OS 43.3% vs planned CRS 40.1% NS).
KEYNOTE-975 pending [PMID 33533655]: def-CRT + pembro vs def-CRT + placebo phase III design, 2026 readout not yet published — the first global phase III of def-CRT + IO. If positive, will rewrite the def-CRT era.

2026 clinical decision: ESCC def-CRT at 50 Gy + CF remains SoC; dose escalation beyond 50.4 Gy is not recommended; adding IO should be trial-only; organ preservation (no surgery) can be considered for induction-CRT strong responders; salvage surgery remains feasible for non-responders.

3.6 Surgical technique: MIE vs RAMIE vs open

2026 mainstream:

MIE (TIME [PMID 22552194 / 28187044]): vs open, pulmonary infection 34% → 12%, 3y OS equivalent — default for Western EC surgery.
RAMIE (ROBOT [PMID 30308612 / 33241302]): vs open, total complications 59% vs 80%, 5y OS equivalent — rapidly replacing MIE in centers with robotic infrastructure.
MIE vs RAMIE direct RCT missing — only observational studies suggest equivalence. This is an EC surgical research gap (see §4 gap 7).
Proton PBT vs IMRT (Lin 2020 [PMID 32160096]): TTB 2.3× lower, post-op complication score 7.6× lower; 3y PFS 51% equivalent. Until the NRG-GI006 phase III readout in 2026–2028, PBT is used only in selected toxicity-concern patients.

4. Research Gaps: the ten unresolved clinical questions

This report identifies the following gaps, each a definable specific problem (not the boilerplate “needs more research”):

Fragmented PD-L1 scoring in ESCC: KEYNOTE-590 uses CPS / CheckMate-648 uses TC% / RATIONALE-306 uses TAP — the scores are non-interchangeable across trials, making real-world cross-trial decisions difficult. A unified scoring standard or a cross-score conversion model is needed.
No positive phase III for post-1L-IO 2L ESCC: CAP-02 Re-challenge showed IO+TKI re-challenge ORR 10%. After 1L IO became universal, >90% of 2L ESCC patients are post-IO — the most urgent unmet need. ivonescimab / cadonilimab / HER2 ADC / TROP2 ADC phase III readouts in 2026–2027 are candidate breakthroughs.
East Asian ESCC three-country paths (China nCRT / Japan DCF no RT / Europe CROSS) lack direct phase III: each country cites its local landmark; the sources of the OS HR difference between paths (surgical dissection quality vs chemo intensity vs RT benefit) have never been formally tested.
ESCC perioperative IO+chemo mature EFS / OS not yet out: ESCORT-NEO’s positive pCR drove NMPA approval, but EFS / OS are still in follow-up in 2026 — the pCR → OS surrogate validity has not yet been validated in the IO era (learning from the reverse lesson of the negative KEYNOTE-585 adjuvant nivo in gastric).
KEYNOTE-975 readout for def-CRT + IO not yet out: def-CRT + IO in unresectable LA ESCC is the last missing piece of the def-CRT path. There are no global phase III data before the KN-975 readout in 2026–2027.
PD-L1 stratification for CheckMate-577 adjuvant nivo: ASCO 2025 mature OS showed PD-L1-positive subgroup sustained benefit, PD-L1-low subgroup questionable — from 2026, stratified decisions are warranted, but thresholds / scoring methods have not been prospectively validated.
MIE vs RAMIE direct phase III missing: each has completed an RCT vs open, but direct comparison is only observational.
EAC / GEJ AC responds to IO less than ESCC — mechanism and solution unclear: KEYNOTE-181 EAC subgroup HR near 1.0; EAC 1L evidence mainly comes from subgroups of mixed ESCC+EAC trials — EAC-only phase III lacking.
ESCC has no targetable driver / precision-treatment gap: POWER closed the EGFR path; TP53 / NOTCH / PIK3CA mutations have not translated into approved targets. ESCC precision treatment = PD-L1 scoring is everything — this is ESCC’s largest divergence from HNSCC.
NRG-GI006 proton vs IMRT phase III not yet out: Lin 2020 phase IIB showed clear toxicity benefit with equivalent OS; before the NRG-GI006 2026–2028 OS readout, PBT’s general applicability is not established.

5. 2024–2026 latest developments

Drug	Agency	Date	Indication / supporting trial
nivolumab adjuvant	FDA	2021-05	CROSS neoadjuvant + surgery with non-pCR / CheckMate-577 [PMID 33789008]
pembrolizumab + CF	FDA	2021-03	1L advanced EC (ESCC + EAC) / KEYNOTE-590 [PMID 34454674]
camrelizumab + TP	NMPA	2021	1L advanced ESCC / ESCORT-1st [PMID 34519801]
nivolumab + chemo or nivolumab + ipilimumab	FDA	2022-05	1L advanced ESCC / CheckMate-648 [PMID 35108470]
sintilimab + chemo	NMPA	2022	1L advanced ESCC / ORIENT-15 [PMID 35440464]
toripalimab + TP	NMPA	2022; FDA 2024	1L advanced ESCC / JUPITER-06 [PMID 35245446]
serplulimab + chemo	NMPA	2023	1L PD-L1+ ESCC / ASTRUM-007 [PMID 36732627]
tislelizumab + chemo (1L) + tislelizumab mono (2L)	FDA	2024-03	1L / 2L ESCC / RATIONALE-306 [PMID 37080222] / RATIONALE-302 [PMID 35442766]
sugemalimab + CF	NMPA	2024	1L advanced ESCC (anti-PD-L1) / GEMSTONE-304 [PMID 38302715]
camrelizumab + nab-TP neoadjuvant	NMPA	2024	LA-ESCC neoadjuvant / ESCORT-NEO [PMID 38956195]

Key observation: 2021–2024 was the concentrated approval period for ESCC IO — 5 Chinese PD-1s + 2 global PD-1/PD-L1 + 2 global PD-L1/CTLA-4 combos, for a total of 9 approved 1L ESCC IO regimens, denser than gastric.

5.2 Key conference readouts (2024–2026, flagged as lower-tier)

The following items serve as a candidate pool only pending formal peer review; not part of the primary database.

CheckMate-577 mature OS (ASCO 2025): 5-year OS data show PD-L1-positive subgroup sustained benefit, PD-L1-low subgroup questionable — from 2026, stratified decisions warranted.
ESCORT-NEO EFS first interim (ASCO GI 2025, on the basis of the positive pCR in the main paper [PMID 38956195]): EFS curves separated but did not reach the pre-specified significance threshold; mature OS expected in 2027.
KEYNOTE-975 first interim (expected 2026–2027 ESMO / ASCO): def-CRT + pembro vs def-CRT + placebo phase III, first global readout for pembro+def-CRT.
ivonescimab (AK112, Akeso PD-1+VEGF bispecific) ESCC 2L phase II: 2024–2025 conference signal; phase III HARMONi-ESCC ongoing.
cadonilimab (AK104, Akeso PD-1+CTLA-4 bispecific) ESCC 1L: COMPASSION-ESCC phase II–III signal.
TROP2 ADC (datopotamab deruxtecan / sacituzumab) ESCC: multiple basket trials with early signal; phase III readouts 2026–2028.

5.3 Ongoing phase III (selected 2026–2028 readouts)

KEYNOTE-975 (NCT04210115): def-CRT + pembrolizumab vs def-CRT + placebo LA EC — OS readout 2026–2027
ESCORT-NEO mature OS / NCCES01 long-term: LA-ESCC neoadjuvant IO+chemo EFS / OS maturing 2026–2027
HARMONi-ESCC (ivonescimab in ESCC): PD-1+VEGF bispecific phase III ongoing
NRG-GI006: PBT vs IMRT phase III (LA EC) — OS 2027–2028
ACTICCA-ESCC / post-IO 2L phase III: still no positive regimen in 2026; candidates ivonescimab / cadonilimab / HER2 ADC / TROP2 ADC
ESOPEC and other European perioperative trials: updated CROSS vs FLOT in EAC

6. Intersecting insights and judgments

6.1 Longitudinal × cross-sectional: the 2026 EC landscape is shaped by four “resonances”

Stacking the longitudinal paradigm evolution on the cross-sectional current decision landscape, the 2026 EC landscape is the superposition of four resonances:

The double geographic differentiation of “ESCC / EAC dual track + East Asia / West path divergence”: ESCC is >85% globally and >90% in Asia, while EAC dominates North America / Western Europe — histology distribution is itself a geographic variable. This directly drove the differences in enrollment, dose, and chemo backbone across CROSS (Netherlands, 75% EAC), NEOCRTEC5010 (China, 100% ESCC), and JCOG1109 (Japan, 100% ESCC), and is also why the three paths have never been directly phase-III compared. Path divergence + pCR / OS HR differences + 3FL dissection quality hypothesis form the core scientific theme of EC perioperative research.
The step-wise convergence of RT benefit from “1992 RTOG 85-01 dose paradox → 2012 CROSS anchoring → 2024 JCOG1109 dropping RT”: RTOG 85-01 defined RT 50 Gy as the def-CRT gold standard; INT 0123 at 64.8 Gy dose escalation was worse, and ARTDECO IMRT dose-painting was negative — 50 Gy is a dose ceiling unchanged for 30 years. Then JCOG1109 showed in the neoadjuvant setting that “CF+RT does not beat the three-drug DCF” — RT benefit may be absorbed under high-quality surgical dissection. This 30-year trajectory tells clinicians: RT is not all-powerful, dose cannot be pushed freely, and dissection quality + chemo intensity can substitute for RT.
The “2021–2024 IO explosion, 5 years and 8 positive phase III” intensity exceeds gastric: in 4 years, KEYNOTE-590 / CheckMate-648 / ESCORT-1st / ORIENT-15 / JUPITER-06 / RATIONALE-306 / ASTRUM-007 / GEMSTONE-304 totaled 8 positive 1L phase III trials (Chinese PD-1 5/8), plus CheckMate-648’s chemo-free nivo+ipi arm exclusive to ESCC. This intensity exceeds gastric’s same-period convergence. But all HR 0.58–0.73 positive signals are hard to compare cross-trial under fragmented PD-L1 scoring — clinical decisions are driven mainly by accessibility / chemo backbone / NRDL inclusion, not by efficacy evidence.
The “ESCC vs EAC / gastric organ difference” determines perioperative IO success: CheckMate-577 adjuvant nivo is positive for DFS HR 0.69 in EC + GEJ (mixed ESCC + EAC) non-pCR; ESCORT-NEO is positive for pCR in LA-ESCC neoadjuvant — yet the same-strategy KEYNOTE-585 adjuvant nivo was negative in gastric adenocarcinoma. Same strategy, same drug, same indication framework — organ / histology differences decide success and failure. This matches HCC being IO-friendly, BTC being suited to IO + chemo, and PDAC being completely unresponsive to IO — IO benefit is deeply coupled to organ microenvironment; “IO class effect” does not travel across organs.

These four resonances together explain one clinical phenomenon: giving a newly diagnosed stage IV ESCC patient a 1L decision in 2026 has three extra decision layers compared to 2016 — “PD-L1 scoring method alignment + IO backbone + chemo-free option” — but the decision tree itself is “extremely wide (8 phase III) + fragmented scoring + post-IO 2L blank”. This differs from NSCLC’s multi-layered decision tree (driver panel → PD-L1 → combo) and from HCC’s “narrow and shallow” tree — EC’s decision tree is in a unique “wide and messy (fragmented scoring) + post-IO blank” shape.

6.2 Clinical decision takeaways (for junior-mid oncologists)

The first step in an ESCC 1L decision is to confirm the PD-L1 scoring method from pathology: not “which of CPS / TAP / TC% to choose,” but which one the pathology report gives — then match the corresponding trial (KN-590 → CPS, CheckMate-648 → TC%, RATIONALE-306 → TAP). Without scoring alignment, trial conclusions cannot be extrapolated.
Chemo-intolerant ESCC can use nivo+ipi chemo-free: CheckMate-648’s ipi arm is ESCC’s exclusive chemo-free IO regimen. Not available in gastric / EAC — capitalize on this ESCC advantage.
China 1L ESCC preferentially uses domestic PD-1 + chemo: cam / sinti / tori / serplu / suge — five domestic phase III positives, HR 0.58–0.70 tight convergence; cost 1/3–1/5 of global regimens; good accessibility — no reason to prefer pembro / nivo unless the patient has exceptional means or enrolls in a global phase III.
post-1L-IO 2L is a “true black box” — do not use IO+TKI combinations for re-challenge: CAP-02 Re-challenge ORR 10% has shown. Use paclitaxel / docetaxel / irinotecan monotherapy + anlotinib (ALTER1102, accessible in China) or enroll in ivonescimab / cadonilimab / HER2 ADC / TROP2 ADC trials.
LA-ESCC neoadjuvant follows regional practice + pathologic pCR target: Europe CROSS / China NEOCRTEC5010 / Japan DCF triplet — do not directly apply out-of-region landmarks outside their region (low extrapolation validity). In the IO era, China has transitioned to ESCORT-NEO-style Cam+nab-TP neoadjuvant (pCR 28% > traditional TP 4.7%).
Post-operative non-pCR CheckMate-577 adjuvant nivo should be a PD-L1-stratified decision: mature OS shows PD-L1-positive subgroup sustained benefit, PD-L1-low subgroup questionable — ITT one-size-fits-all no longer suits.
Do not escalate the 50 Gy dose in unresectable LA ESCC def-CRT: INT 0123 + ARTDECO have shown two failed dose escalations; 60 Gy is occasionally used in China / Japan but no RCT supports superiority over 50 Gy. Adding IO to def-CRT should be trial-only until the KEYNOTE-975 readout.
def-CRT strong responders can consider organ preservation: FFCD 9102 showed induction-CRT responders have equivalent 2y OS without adding surgery and far lower perioperative mortality; non-responders can still undergo salvage esophagectomy (Markar) with 3y OS equivalent to planned CRS.
EAC / GEJ AC 1L preferentially follows the gastric path: KN-181 EAC subgroup HR near 1.0 — EAC responds to IO less than ESCC; 1L uses nivo+chemo (CheckMate-649) / pembro+chemo (KN-859); HER2+ EAC adds trastuzumab; EAC is not suitable for chemo-free IO+IO.
ESCC precision treatment = PD-L1 scoring is everything — do not hold out for EGFR / HER2 / KRAS: POWER panitumumab negative, SCOPE-1 cetuximab futility closed the EGFR path; HER2-positive EAC follows the gastric path; ESCC has no approved targetable driver. NGS panels in ESCC in 2026 give very low yield (unless enrolling in a basket trial).

7. Sources

The metadata for all 42 trials in this report has been independently verified via PubMed. Every [PMID xxxxxxxx] bracket in the body can be verified directly on PubMed.

Published trials: 42, covering 1992–2024 (PMIDs verifiable)
NCCN guideline citations: 42/42 (100%) hit the current NCCN Esophageal reference section
2021–2024 FDA / NMPA new approvals: 10+ key approvals (9 1L ESCC IO + non-pCR adjuvant nivo + neoadjuvant camrelizumab)
2024–2026 key conferences / mature readouts: 5 (CheckMate-577 mature OS PD-L1 stratification / ESCORT-NEO EFS interim / KN-975 pending / ivonescimab phase II / cadonilimab phase II-III)
Research gaps: 10
China-led proportion: >35% (ESCORT / ESCORT-1st / ESCORT-NEO / ORIENT-15 / JUPITER-06 / ASTRUM-007 / GEMSTONE-304 / ALTER1102 / NEOCRTEC5010 / NICE / Keystone-001 / CAP-02 / CAP-02 Re-challenge / PALACE-1)

7.1 Citation index (sorted by PMID)

The table below lists all PMIDs cited in the body, each verifiable by clicking through to PubMed.

PMID	First Author	Year	Journal	Trial / topic
1584260	Herskovic A	1992	N Engl J Med	RTOG 85-01 (def-CRT foundational)
10235156	Cooper JS	1999	JAMA	RTOG 85-01 long-term (5y OS 26% vs 0%)
17401004	Bedenne L	2007	J Clin Oncol	FFCD 9102 (organ preservation logic)
21879261	Ando N	2012	Ann Surg Oncol	JCOG9907 (Japan pre-op CF doublet)
22552194	Biere SS	2012	Lancet	TIME (MIE vs open)
22646630	van Hagen P	2012	N Engl J Med	CROSS (global neoadjuvant CRT foundational)
26195702	Markar S	2015	J Clin Oncol	Salvage esophagectomy cohort
26254683	Shapiro J	2015	Lancet Oncol	CROSS long-term (SCC HR 0.48)
28187044	Straatman J	2017	Ann Surg	TIME long-term
28196063	Crosby T	2017	Br J Cancer	SCOPE-1 (def-CRT ± cetuximab)
30089078	Yang H	2018	J Clin Oncol	NEOCRTEC5010 (China ESCC neoadjuvant foundational)
30308612	van der Sluis PC	2019	Ann Surg	ROBOT (RAMIE vs open)
30570649	Shah MA	2019	JAMA Oncol	KEYNOTE-180 (heavily pretreated EC pembro phase II)
31582355	Kato K	2019	Lancet Oncol	ATTRACTION-3 (2L ESCC nivo)
31959339	Moehler M	2020	Ann Oncol	POWER (CF ± panitumumab, EGFR path death)
32160096	Lin SH	2020	J Clin Oncol	PBT vs IMRT phase IIB
32416073	Huang J	2020	Lancet Oncol	ESCORT (China 2L camrelizumab)
33026938	Kojima T	2020	J Clin Oncol	KEYNOTE-181 (2L pembro, CPS≥10 positive)
33241302	de Groot EM	2020	Dis Esophagus	ROBOT long-term
33373868	Li C	2021	Eur J Cancer	PALACE-1 (pembro + CRT pilot)
33533655	Shah MA	2021	Future Oncol	KEYNOTE-975 design paper
33586360	Huang J	2021	Cancer Med	ALTER1102 (anlotinib 2L phase II)
33789008	Kelly RJ	2021	N Engl J Med	CheckMate-577 (adjuvant nivolumab)
33891478	Eyck BM	2021	J Clin Oncol	CROSS 10-year
34160577	Yang H	2021	JAMA Surg	NEOCRTEC5010 long-term
34454674	Sun JM	2021	Lancet	KEYNOTE-590 (1L EC pembro+CF)
34519801	Luo H	2021	JAMA	ESCORT-1st (China 1L cam+TP)
34998471	Meng X	2022	Lancet Gastroenterol Hepatol	CAP-02 (cam+apatinib 2L phase II)
35108470	Doki Y	2022	N Engl J Med	CheckMate-648 (1L nivo+chemo / nivo+ipi)
35245446	Wang ZX	2022	Cancer Cell	JUPITER-06 (1L tori+TP, OS HR 0.58 strongest)
35440464	Lu Z	2022	BMJ	ORIENT-15 (1L sinti+chemo)
35442766	Shen L	2022	J Clin Oncol	RATIONALE-302 (2L tisle)
36098320	Ebert MP	2022	Lancet Healthy Longev	RAMONA (elderly nivo+ipi 2L phase II)
36732627	Song Y	2023	Nat Med	ASTRUM-007 (1L serplu+chemo PD-L1+)
37080222	Xu J	2023	Lancet Oncol	RATIONALE-306 (1L tisle+chemo)
37696429	Yang Y	2024	J Thorac Cardiovasc Surg	NICE (cN2-3 ESCC neoadjuvant cam+nab-TP)
37734399	Reynolds JV	2023	Lancet Gastroenterol Hepatol	Neo-AEGIS (CROSS vs perioperative chemo)
38302715	Li J	2024	Nat Med	GEMSTONE-304 (1L suge+CF, anti-PD-L1)
38876133	Kato K	2024	Lancet	JCOG1109 NExT (DCF triplet vs CF vs CF+RT)
38956195	Qin J	2024	Nat Med	ESCORT-NEO / NCCES01 (neoadjuvant cam+chemo phase III)
39307038	Meng X	2024	Eur J Cancer	CAP-02 Re-challenge (post-IO 2L)
39406186	Shang X	2024	Cancer Cell	Keystone-001 (neoadjuvant pembro+chemo)

7.2 Verification conventions

Each PMID can be accessed directly at https://pubmed.ncbi.nlm.nih.gov/{PMID}/
Each NCT id can be accessed at https://clinicaltrials.gov/study/{NCT_id}/
Conference abstracts (ASCO / ASCO GI / ESMO) are retrieved via the official conference systems; all conference citations in this report are flagged as lower-tier — non-peer-reviewed toplines defer to journal publication
If a PMID’s trial name / year / conclusion in this report is found inconsistent with PubMed, corrections are welcome

The clinical trial timeline lives here

Chinese: /trials/esophageal/ English: /en/trials/esophageal/

Each trial has a dedicated detail page with:

Complete intervention / comparator regimens
Primary endpoint values + 95% CI
Key findings + clinical implications
Clickable links to PMID / NCT source

42 trials · 7 chapters · 1992 to 2024 · China-led contribution >35% · synchronized with the current NCCN Esophageal guideline.

Closing

Esophageal cancer has completed a unique “three-pillar + dual-track” evolution over the past 30 years — from 1992 RTOG 85-01 establishing the concurrent def-CRT 50 Gy dose ceiling, to 2012 CROSS anchoring global neoadjuvant CRT in an EAC-dominant Dutch cohort; to 2018–2024 intra–East Asian ESCC divergence (China NEOCRTEC5010 nCRT / Japan JCOG1109 DCF no RT / Europe CROSS); to the 2021–2024 explosion of 8 positive 1L phase III trials in advanced disease (5 Chinese PD-1 / 3 global PD-1 or PD-L1 / CheckMate-648’s exclusive chemo-free nivo+ipi arm); and finally to CheckMate-577 non-pCR adjuvant nivo as the global SoC and ESCORT-NEO LA-ESCC neoadjuvant IO+chemo, both landing.

The most fundamental difference between EC and other major cancers (NSCLC / HCC / BTC / PDAC) is not treatment complexity but the double geographic differentiation of “ESCC / EAC dual track + East Asia / West path divergence” — histology distribution is itself a geographic variable. ESCC >85% globally and >90% in Asia; EAC dominates in North America / Western Europe. This drove the differences between CROSS / NEOCRTEC5010 / JCOG1109 in enrollment, dose, and chemo backbone, and is also why the three paths have never been directly phase-III compared. Clinical stratification will always rest on histology (ESCC vs EAC) × geography × PD-L1 scoring (CPS / TAP / TC%) × resectability — no approved targetable driver.

post-IO 2L unmet need + PD-L1 scoring fragmentation + perioperative IO+chemo mature OS pending + def-CRT + IO KEYNOTE-975 pending — these four domains are the densest research gaps in EC 2026. The next five years must answer “can post-IO 2L be pushed from ORR 10% to a positive phase III,” “can ESCC PD-L1 scoring be unified,” “can perioperative IO+chemo be pushed from positive pCR to positive OS,” and “can def-CRT + IO become the new SoC for LA unresectable ESCC” — four structural problems.

The value of this report is not “exhaustively listing all trials” (PubMed can do that), but compressing 30 years of evolution + current decisions + unresolved gaps into the cognitive bandwidth of a single read. Next time you face a newly diagnosed EC patient, every branch of the decision tree has this map to consult, trace, and cross-examine.

Clinician × AI = Research Superpower + Clinical Decision Amplifier

—— Dual Brain Lab · 2026-04-21

Gastric Cancer Clinical Trial Timeline: 58 RCTs Across 25 Years Mapping a Subtyping Revolution

Tue, 21 Apr 2026 00:00:00 +0000

Gastric Cancer Clinical Trial Timeline — In-depth Report

Coverage: 58 landmark trials cited by NCCN Gastric + CSCO Gastric 2025 (all PMID / NCT traceable) + East–West perioperative / adjuvant branches + HER2 / CLDN18.2 / PD-L1 CPS / MSI four-layer subtyping + three subtyping revolutions

Curated by Dual Brain Lab (csilab.net)

1. One-sentence definition

This report traces the evolution logic and current decision landscape of gastric cancer (GC) and gastroesophageal junction cancer (GEJ) systemic therapy over the past 25 years (2001-2026), covering the landmark clinical trials cited by NCCN Gastric V2.2025 and CSCO Gastric 2025, providing frontline clinicians in 2026 a traceable panoramic map for “who, what, why” decisions.

Iron rule: every data point in every trial is traceable to PubMed (PMID) or ClinicalTrials.gov (NCT id) — each [PMID xxxxxxxx] in the text can be clicked open to verify the PubMed source.

Scope boundary: GC refers to adenocarcinoma arising from gastric mucosa; GEJ refers to cardia / esophagogastric junction adenocarcinoma. Of Siewert I-III, Siewert II-III belong to the gastric category (covered here), while Siewert I and esophageal adenocarcinoma (EAC) belong to the esophageal category (not covered). Globally ~970,000 new gastric cancers annually (5th most common malignancy) + ~660,000 deaths (5th leading cancer death), with China accounting for ~44%. HBV / H. pylori infection, chronic atrophic gastritis, and high-salt pickled diets are the main environmental factors; Lauren classification (intestinal vs diffuse) and molecular subtyping (TCGA: EBV / MSI / CIN / GS four types) are the main pathological and molecular axes.

While HCC uniquely walked a “0 predictive biomarker” path, gastric cancer went the opposite way — 25 years, three subtyping revolutions: anatomical subtyping (GEJ vs gastric body) from crude endoscopic split to the manifest HR differences in CheckMate-649 subgroups; East–West path divergence (FLOT4 German perioperative vs CLASSIC Korean adjuvant CAPOX vs INT-0116 US adjuvant chemoRT) with the same stage custom-fit across three continents; HER2 drug 10-year leap (ToGA 2010 trastuzumab + chemo → 2020 DESTINY T-DXd 2L → 2025 DESTINY-Gastric-04 T-DXd 2L standard + KEYNOTE-811 HER2 1L IO combo); biomarker subtyping refinement (HER2 / PD-L1 CPS / CLDN18.2 / MSI four-dimensional checkerboard). By 2026, treatment decisions have completely flipped from “which chemo regimen” to “first get the biomarker panel complete.”

2. Longitudinal: timeline of five treatment-paradigm shifts

Gastric cancer systemic therapy has gone through five paradigm shifts over 25 years: adjuvant / definitive chemoRT (2001 INT-0116) → birth of the perioperative chemo concept (2006 MAGIC → 2019 FLOT4) → HER2 targeted era (2010 ToGA → 2020-2025 T-DXd three-generation evidence) → IO + chemo 1L standard established (2021-2024 four phase IIIs concordantly positive) → CLDN18.2 new subtype + perioperative IO (2023-2025 SPOTLIGHT / GLOW / MATTERHORN).

Each shift had 2-4 phase IIIs pushing the old SoC to second-line. Compared to NSCLC’s “driver-gene × IO dual-engine” and HCC’s “0-biomarker / IO-backbone-alone,” gastric cancer is characterized by “biomarker subtyping started early (2010 HER2) but density grew slowly (CLDN18.2 joined 15 years later), with severe geographic branching (East Asia vs Europe/US vs China, each running its own adjuvant / perioperative playbook)” — which means the 2026 gastric decision tree neither unfolds 10+ drivers horizontally like NSCLC, nor “rides clinical parameters alone” like HCC, but is a “region × biomarker × line” three-dimensional checkerboard.

2.1 Chemotherapy backbone era (2001-2010): adjuvant chemoRT / perioperative chemo / HER2 — three starting points

Story: in 2001 INT-0116 first put adjuvant chemoRT on stage III postoperative gastric cancer standard (US default for 15 years); in 2006 MAGIC first proved the “perioperative chemo” concept in UK/Europe (5-year OS 36% vs 23%, HR 0.75), forming a transatlantic divergence with US adjuvant chemoRT; the same year V325’s DCF triplet pushed advanced chemo OS ceiling to 9.2 months, and REAL-2 (2008) used a 2×2 factorial to bring capecitabine / oxaliplatin into the advanced backbone; 2007-2008 Japan’s ACTS-GC / SPIRITS defined “S-1 as Asian default,” further diverging from Europe/US; in 2010 ToGA made HER2 gastric cancer’s first actionable biomarker — a targeted path that ran alone for 13 years.

INT-0116 [PMID 11547741] (Macdonald 2001 N Engl J Med, N=556): US postoperative 5-FU / LV + 45 Gy chemoradiation vs D0-D1 surgery alone. mOS 36 vs 27 months (HR 1.35, P=0.005, favoring treatment), 3-year OS 50% vs 41%. Established the NCCN “adjuvant chemoRT” branch as the US postoperative default for 15 years — but later ARTIST / ARTIST-2 showed RT adds nothing after adequate D2 dissection, phasing it out of Asian practice.
MAGIC [PMID 16822992] (Cunningham 2006 N Engl J Med, N=503, UK): perioperative ECF (epirubicin + cisplatin + 5-FU) 3 cycles preop + 3 postop vs surgery alone. 5-year OS 36% vs 23% (HR 0.75, P=0.009). The invention moment of the “perioperative chemo” concept — European standard for a decade-plus, until 2019 FLOT4 upgraded the backbone from ECF to FLOT.
V325 [PMID 17075117] (Van Cutsem 2006 J Clin Oncol, N=445): docetaxel + cisplatin + 5-FU (DCF) vs CF advanced 1L. mOS 9.2 vs 8.6 months (HR 0.77, P=0.02), ORR 37% vs 25%, G3-4 neutropenia 82%. First OS-positive triplet backbone, but toxicity too high for wide real-world uptake.
REAL-2 [PMID 18172173] (Cunningham 2008 N Engl J Med, N=1002): 2×2 factorial (epirubicin + platinum + fluoropyrimidine: ECF / ECX / EOF / EOX). Capecitabine non-inferior to 5-FU (HR 0.86), oxaliplatin non-inferior to cisplatin (HR 0.92), best arm EOX mOS 11.2 months. Brought capecitabine + oxaliplatin into the advanced backbone, laying the foundation for 20 years of CAPOX SoC status.
ACTS-GC [PMID 17978289] (Sakuramoto 2007 N Engl J Med, N=1059, Japan): D2 postoperative S-1 for 1 year vs surgery alone. 3-year OS 80.1% vs 70.1% (HR 0.68, P=0.003), 5-year OS 71.7% vs 61.1%. Definitive evidence for Japanese adjuvant S-1; the “Asia primarily adjuvant” tradition was set.
SPIRITS [PMID 18282805] (Koizumi 2008 Lancet Oncol, N=305, Japan): advanced S-1 + cisplatin (SP) vs S-1 monotherapy. mOS 13.0 vs 11.0 months (HR 0.77, P=0.04), ORR 54% vs 31%. SP became Japan’s advanced 1L default for 10+ years, yielding only in the ATTRACTION-4 / CheckMate-649 IO era.
ToGA [PMID 20728210] (Bang 2010 Lancet, N=584): trastuzumab + XP/FP chemotherapy vs chemo in HER2+ (IHC3+ or IHC2+/FISH+) advanced GC/GEJ. mOS 13.8 vs 11.1 months (HR 0.74, P=0.0046), IHC3+ / FISH+ subgroup mOS 16.0 vs 11.8 months (HR 0.65). First positive biomarker-targeted phase III in gastric cancer — HER2 defined the first molecular subtype of gastric cancer, a standard that ran for 13 years until KEYNOTE-811 layered pembrolizumab on top.

Takeaway: the 2001-2010 decade anchored the entire basic framework of modern gastric cancer — US adjuvant chemoRT (INT-0116) / European perioperative ECF (MAGIC) / Japanese adjuvant S-1 (ACTS-GC) / Japanese advanced SP (SPIRITS) / Euro-US advanced DCF + CAPOX (V325 + REAL-2) / global HER2+ trastuzumab (ToGA) — all six branches cemented in these 10 years. Every new drug in the following 15 years only “replaces, stacks, or refines” on these six branches.

2.2 East–West adjuvant / perioperative divergence takes shape (2010-2019): CLASSIC / FLOT4 / ARTIST / RESOLVE / PRODIGY

Story: in 2012 Korea’s CLASSIC took D2 postoperative CAPOX to 3-year DFS 74% vs 59% (HR 0.56), making “adjuvant doublet chemo” the Asian standard; the same year Europe’s CROSS used neoadjuvant CRT (carboplatin + paclitaxel + 41.4 Gy) to push GEJ/EAC OS from 24 to 49 months (HR 0.657) — global GEJ tri-modality SoC was established; 2015 / 2021 Korea’s ARTIST / ARTIST-2 proved RT adds nothing after adequate D2 dissection, closing the “Asian D2 postop RT debate”; 2018 Netherlands’ CRITICS likewise showed that adding postop CRT after preop chemo gives no extra benefit; in 2019 Germany’s FLOT4 used docetaxel + oxaliplatin + 5-FU + leucovorin (FLOT) triplet / quadruplet to push perioperative mOS from ECF’s 35 to 50 months (HR 0.77), establishing the new Euro-US perioperative backbone; 2021 Korea’s PRODIGY proved PFS HR 0.70 with neoadjuvant DOS + adjuvant S-1; the same year China’s RESOLVE established perioperative SOX as China’s perioperative standard.

CLASSIC [PMID 22226517] (Bang 2012 Lancet, N=1035, Korea / China / Taiwan): D2 postoperative CAPOX × 8 cycles (6 months) vs surgery alone. 3-year DFS 74% vs 59% (HR 0.56, P<0.0001), 5-year OS 78% vs 69% (HR 0.66). Asian adjuvant CAPOX 20-year SoC, still the first-choice doublet after D2 today.
CROSS [PMID 22646630] (van Hagen 2012 N Engl J Med, N=366, Netherlands): neoadjuvant carboplatin + paclitaxel + 41.4 Gy vs surgery alone. mOS 49.4 vs 24.0 months (HR 0.657, P=0.003), R0 92% vs 69%, pCR 29%. Global GEJ / EAC tri-modality SoC, still unsurpassed (Neo-AEGIS tried to show FLOT perioperative non-inferior to CROSS but was terminated early due to futility + COVID).
ARTIST [PMID 25559811] (Park 2015 J Clin Oncol, N=458, Korea): D2 postoperative XP ± RT vs XP alone. 3-year DFS 78.2% vs 74.2% (HR 0.740, P=0.0922, NS). Subgroups suggested benefit in node+ / intestinal Lauren, but overall ITT negative — challenged the applicability of US INT-0116 in the Asian adequate-D2 setting.
ARTIST-2 [PMID 33278599] (Park 2021 Ann Oncol, N=538, Korea): node+ gastric cancer D2 postoperative SOX 6 months vs SOX + RT (SOXRT) vs S-1 1 year. 3-year DFS 74.3% (SOX) vs 72.8% (SOXRT) vs 64.8% (S-1); SOX and SOXRT both superior to S-1, but no difference between them (HR 0.971). ARTIST-2 closed the “Asian D2 postop RT” debate — doublet chemo is sufficient, no RT needed.
FLOT4 [PMID 30982686] (Al-Batran 2019 Lancet, N=716, Germany): perioperative FLOT 4+4 cycles vs ECF/ECX 3+3 cycles. mOS 50 vs 35 months (HR 0.77, 95% CI 0.63-0.94), pCR 16% vs 6%. New Euro-US perioperative backbone, on top of which all subsequent perioperative IO layering (MATTERHORN / KEYNOTE-585 FLOT subgroup) is built.
JACCRO GC-07 [PMID 30925125] (Yoshida 2019 J Clin Oncol, N=915, Japan): stage III D2 postoperative S-1 + docetaxel vs S-1 monotherapy. 3-year RFS 66% vs 50% (HR 0.632, P<0.001). Definitive evidence for Japanese stage III adjuvant doublet; “add docetaxel for stage III” written into Japanese guidelines.
CRITICS [PMID 29650363] (Cats 2018 Lancet Oncol, N=788, Netherlands): preop chemo (ECC/EOC) + postop chemo vs postop CRT (45 Gy + capecitabine/cisplatin). mOS 43 vs 37 months (HR 1.01, P=0.90), negative. No benefit from adding postop CRT after preop chemo — same author team / same country as CROSS but opposite result (neoadjuvant CRT vs postadjuvant CRT), highlighting that RT timing is key.
ST03 [PMID 28163000] (Cunningham 2017 Lancet Oncol, N=1063, UK): perioperative ECX + bevacizumab vs ECX. 3-year OS 50.3% vs 48.1% (HR 1.08, P=0.36), negative; esophagogastric anastomotic leak 24% vs 10%. The attempt to “add bev to perioperative” failed completely + increased surgical complications.
PRODIGY [PMID 34133211] (Kang 2021 J Clin Oncol, N=530, Korea): neoadjuvant DOS (docetaxel + oxaliplatin + S-1) × 3 cycles + surgery + adjuvant S-1 vs surgery + adjuvant S-1. 3-year PFS 66.3% vs 60.2%, adjusted HR 0.70 (P=0.023). Definitive evidence for Korea’s “neoadjuvant DOS triplet” — a representative of Asian neoadjuvant-era regimen iteration.
RESOLVE [PMID 34252374] (Zhang 2021 Lancet Oncol, N=1094, China): locally advanced GC/GEJ D2 gastrectomy perioperative SOX vs adjuvant SOX vs adjuvant CAPOX. 3-year DFS 59.4% (perioperative SOX) vs 51.1% (adjuvant CAPOX), HR 0.77 (P=0.028); adjuvant SOX non-inferior to adjuvant CAPOX. Definitive Chinese evidence for perioperative SOX — “perioperative > pure adjuvant” confirmed by RCT in Chinese population.
G-SOX [PMID 25316259] (Yamada 2015 Ann Oncol, N=685, Japan): advanced 1L SOX (S-1 + oxaliplatin) vs CS (S-1 + cisplatin). mOS 14.1 vs 13.1 months, PFS HR 1.004 (non-inferior), SOX fewer G≥3 AEs (nephrotoxicity / neutropenia). SOX established as SP alternative in Japanese / Asian advanced 1L, providing the control-arm background for later ATTRACTION-4 / ORIENT-16.
TOPGEAR [PMID 39282905] (Leong 2024 N Engl J Med, N=574, international): perioperative chemo ± preop CRT. pCR 17% vs 8% (favoring CRT) but mOS 46 vs 49 months (HR 1.05, NS), mPFS 31 vs 32 months (NS). pCR improved but OS did not — textbook case of “pCR as surrogate endpoint is unstable in gastric cancer” (fundamentally different from the strong pCR-OS correlation in breast cancer).

Takeaway: 2026 adjuvant / perioperative SoC is geographically branched — Euro-US first-choice FLOT4 perioperative (4+4 cycles), China first-choice RESOLVE perioperative SOX, Japan stage III first-choice JACCRO GC-07 S-1 + docetaxel adjuvant, Korea first-choice CLASSIC adjuvant CAPOX or ARTIST-2 SOX 6 months, GEJ/EAC first-choice CROSS neoadjuvant CRT. US INT-0116 adjuvant chemoRT has exited Asian practice in the adequate-D2 era (ARTIST-2 closed the debate), preserved only for D0-D1 or inadequate-dissection scenarios. Three “no longer use” paths: bevacizumab perioperative (ST03 negative), postop CRT after preop chemo (CRITICS negative), and judging OS by pCR alone as surrogate (TOPGEAR lesson).

2.3 HER2 era’s 10-year leap (2010-2025): ToGA → T-DXd → KEYNOTE-811 / zanidatamab

Story: after 2010 ToGA made HER2 gastric cancer’s first biomarker pedestal, no second HER2 drug entered gastric cancer SoC for a full 10 years — LOGIC/TRIO-013 (lapatinib + CapeOx vs CapeOx) 2016 negative, TyTAN / GATSBY (T-DM1 vs paclitaxel) negative. Not until 2020 did DESTINY-Gastric-01 use trastuzumab deruxtecan (T-DXd, HER2-ADC) break the stalemate in 2L HER2+; in 2023 KEYNOTE-811 layered pembrolizumab onto trastuzumab + chemo as the new HER2+ 1L backbone; in 2025 DESTINY-Gastric-04 made T-DXd the 2L HER2+ standard (beating the 10-year RAINBOW ramucirumab + paclitaxel); in 2025 HERIZON-GEA-01 topline used zanidatamab (HER2 bispecific antibody, simultaneously binding domains 2 + 4) to challenge the trastuzumab 1L backbone.

LOGIC/TRIO-013 [PMID 26628478] (Hecht 2016 J Clin Oncol, N=545): 1L HER2+ lapatinib + CapeOx vs CapeOx. mOS 12.2 vs 10.5 months (HR 0.91, NS, negative), mPFS 6.0 vs 5.4 months (HR 0.82, P=0.0381), ORR 53% vs 39%. First failure of the HER2 small-molecule TKI path in gastric cancer — lapatinib’s “PFS wins, OS doesn’t” replayed, deepening the view that “trastuzumab is irreplaceable as the HER2 backbone in gastric cancer.”
DESTINY-Gastric-01 [PMID 32469182] (Shitara 2020 N Engl J Med, N=187, Japan/Korea): HER2+ post-trastuzumab T-DXd vs physician’s choice chemo. ORR 51% vs 14% (P<0.001), mOS 12.5 vs 8.4 months (HR 0.59, P=0.01), ILD 9 cases G1-2 / 3 cases G3-4. First positive 2L HER2+ regimen 10 years into the HER2 era, FDA accelerated approval 2021-01.
DESTINY-Gastric-02 [PMID 37329891] (Van Cutsem 2023 Lancet Oncol, N=79, US/EU): HER2+ post-trastuzumab T-DXd 6.4 mg/kg single-arm phase II. Confirmed ORR 42%, mPFS 5.6 months, mOS 12.1 months, ILD 10% (1 G5 fatality). Replicated DG-01 in Euro-US populations — cross-regional consistency of HER2+ 2L T-DXd confirmed; ILD is the key safety signal.
KEYNOTE-811 [PMID 37871604] (Janjigian 2023 Lancet, N=698): HER2+ CPS≥1 advanced 1L pembrolizumab + trastuzumab + FP/CAPOX vs placebo + trastuzumab + FP/CAPOX. mPFS 10.0 vs 8.1 months (HR 0.72, P=0.0002), mOS 20.0 vs 16.8 months (HR 0.84), ORR 72.6% vs 60.1%. First upgrade to the HER2 paradigm after 13 years — IO layered onto the HER2 backbone, CPS≥1 subgroup showing the clearest benefit (FDA 2023-11 restricted the indication to CPS≥1; prior accelerated approval was all-comer).
DESTINY-Gastric-04 [PMID 40454632] (Shitara 2025 N Engl J Med, N=494): HER2+ 2L T-DXd monotherapy vs ramucirumab + paclitaxel head-to-head. mOS 14.7 vs 11.4 months (HR 0.70, P=0.004), mPFS HR 0.74, confirmed ORR 44.3% vs 29.1%. T-DXd beat the 2L SoC for the first time (RAINBOW ramu + paclitaxel), the HER2+ 2L standard has switched — meaning HER2 testing must be redone at 2L (re-biopsy), because post-trastuzumab HER2 loss is ~30%.
HERIZON-GEA-01 (NCT05152147, ESMO 2025 LBA): HER2+ 1L zanidatamab + chemotherapy ± tislelizumab vs trastuzumab + chemotherapy. Topline 2025: both zanidatamab arms PFS HR ~0.65 vs control; zani + tisle + chemo mOS improved >7 months vs trastuzumab + chemo. Zanidatamab’s biepitope simultaneous binding + induced internalization mechanism — if the 2026 full paper confirms, trastuzumab’s 13-year HER2 backbone status could end. As of 2026-04 the full manuscript is unpublished and PMID unassigned, cited here by NCT + ESMO LBA only.

Takeaway: 2026 HER2+ advanced decision path — 1L = pembrolizumab + trastuzumab + FP/CAPOX (KEYNOTE-811, first choice for CPS≥1; CPS<1 remains trastuzumab + chemo per ToGA); 2L = T-DXd (DESTINY-Gastric-04 beat ramu + PTX); 3L+ = ramu + PTX (RAINBOW) / irinotecan / TAS-102. 2L must re-biopsy to recheck HER2 — post-trastuzumab HER2 loss is ~30%, giving T-DXd directly will fail in the HER2-loss population. Zanidatamab is the biggest 2026 suspense: if HERIZON-GEA-01 full paper 2026 confirms positive, the HER2 backbone turns over.

2.4 IO + chemo 1L rewrite (2021-2024): four phase IIIs concordantly positive

Story: in 2017 Japan/Korea’s ATTRACTION-2 made nivolumab positive in 3L+ gastric cancer (mOS 5.26 vs 4.14 months, HR 0.63), opening the gastric IO era; in 2018 KEYNOTE-061 ran pembrolizumab vs paclitaxel in 2L CPS≥1 and hit negative (HR 0.82, P=0.0421 NS), the first failure of IO monotherapy in 2L; in 2020 KEYNOTE-062 again made pembrolizumab in 1L “monotherapy non-inferior but combo not superior,” a first cold-water moment for 1L IO; not until 2021 did CheckMate-649 use nivolumab + FOLFOX/XELOX achieve mOS 14.4 vs 11.1 months (HR 0.71) in CPS≥5 — 1L IO+chemo milestone; then over 3 years KEYNOTE-859 / ORIENT-16 / RATIONALE-305 used three different PD-(L)1s (pembrolizumab / sintilimab / tislelizumab) to replicate concordantly, HRs converging in the 0.71-0.80 narrow band — class effect formally established.

ATTRACTION-2 [PMID 28993052] (Kang 2017 Lancet, N=493, Japan/Korea/Taiwan): ≥2L nivolumab vs placebo. mOS 5.26 vs 4.14 months (HR 0.63, P<0.0001), 12-month OS 26.2% vs 10.9%, ORR 11.2% vs 0%. Starting point of gastric IO — but limited to East Asian 3L+ scenarios.
KEYNOTE-061 [PMID 29880231] (Shitara 2018 Lancet, N=592): 2L CPS≥1 pembrolizumab vs paclitaxel. mOS 9.1 vs 8.3 months (HR 0.82, P=0.0421 NS, negative), CPS≥10 subgroup showed benefit signal, G≥3 TRAE 14% vs 35%. IO monotherapy 2L failure — lesson: “IO worse than chemo” holds in 2L unselected population.
KEYNOTE-062 [PMID 32880601] (Shitara 2020 JAMA Oncol, N=763): 1L CPS≥1 pembrolizumab monotherapy vs pembrolizumab + chemo vs chemo. Pembrolizumab monotherapy vs chemo mOS non-inferior (CPS≥1 10.6 vs 11.1 months), CPS≥10 subgroup 17.4 vs 10.8 months (HR 0.69 but NS), pembrolizumab + chemo vs chemo no significant OS difference. First cold water for 1L IO — revealed that “unselected + IO” is not enough; PD-L1 enrichment required.
JAVELIN Gastric 100 [PMID 33197226] (Moehler 2021 J Clin Oncol, N=499): avelumab maintenance vs continued chemo after 12-week induction chemo. mOS 10.4 vs 10.9 months (HR 0.91, P=0.178, negative), 24-month OS 22.1% vs 15.5%, TRAE ≥G3 12.8% vs 32.8%. IO maintenance strategy failed — even with the “chemo exit lowers toxicity” attraction, no OS-level win.
ATTRACTION-4 [PMID 35030335] (Kang 2022 Lancet Oncol, N=724): HER2- 1L nivolumab + SOX/CAPOX vs placebo + SOX/CAPOX. mPFS 10.45 vs 8.34 months (HR 0.68, P=0.0007), mOS 17.45 vs 17.15 months (HR 0.90, P=0.26, primary OS endpoint not met). PFS wins, OS loses — high Japan/Korea SP/SOX baseline + crossover dilution, failed to reach the global IO standard.
CheckMate-649 [PMID 34102137] (Janjigian 2021 Lancet, N=1581): HER2- 1L nivolumab + FOLFOX/XELOX vs chemo. CPS≥5 mOS 14.4 vs 11.1 months (HR 0.71, P<0.0001), mPFS 7.7 vs 6.0 months (HR 0.68), ORR 60% vs 45%. 1L IO+chemo milestone, FDA approved 2021 (initially all-comer, later restricted to CPS≥5).
ORIENT-16 [PMID 38051328] (Xu 2023 JAMA, N=650, China): HER2- 1L sintilimab + XELOX vs placebo + XELOX. All-comers mOS 15.2 vs 12.3 months (HR 0.77, P=0.009), CPS≥5 mOS 18.4 vs 12.9 months (HR 0.66, P=0.002). First domestic Chinese PD-1 to achieve phase III OS positive in advanced gastric cancer, NMPA approved 2022. Extremely high clinical penetration in China, cost ~1/3 of pembrolizumab.
KEYNOTE-859 [PMID 37875143] (Rha 2023 Lancet Oncol, N=1579): HER2- 1L pembrolizumab + FP/CAPOX vs placebo + FP/CAPOX. ITT mOS 12.9 vs 11.5 months (HR 0.78, P<0.0001), CPS≥10 HR 0.65, mPFS 6.9 vs 5.6 months (HR 0.76), ORR 51.3% vs 42.0%. All-comer positive led FDA 2023-11 to remove the CPS restriction on pembrolizumab in gastric cancer — US practice shifted from “check CPS then give pembro” to “HER2- just give pembro,” but EU and NCCN still retain CPS≥1 / ≥5 recommendations.
RATIONALE-305 [PMID 38806195] (Qiu 2024 BMJ, N=997): HER2- 1L tislelizumab + chemo vs placebo + chemo. PD-L1 TAP≥5% mOS 17.2 vs 12.6 months (HR 0.74), ITT mOS 15.0 vs 12.9 months (HR 0.80, P=0.001). China approved 2024; PD-L1 TAP (tumor area positivity) introduced as an alternative scoring method to CPS.
CheckMate-032 [PMID 30110194] (Janjigian 2018 J Clin Oncol, N=160): advanced esophagogastric nivolumab monotherapy vs nivo + ipi multi-arm early data. ORR: nivo 12% / nivo1+ipi3 24% / nivo3+ipi1 8%; 12-month OS 39% / 35% / 24%. IO + IO ipi dose signal (ipi3 > ipi1), but phase III CheckMate-649 did not pursue an ipi arm — origin of the “dual IO has not become a 1L regimen in gastric cancer” backdrop.

Takeaway: 2026 HER2- advanced 1L IO + chemo class effect established — HRs converge in the narrow 0.71-0.80 band across four PD-(L)1 regimens (nivolumab / pembrolizumab / sintilimab / tislelizumab), choice determined by biomarker CPS threshold × regional access × price × chemo backbone preference. CPS threshold regional policy divergence: FDA removed the threshold (driven by KN-859); NCCN still recommends CPS≥5; EU retains CPS≥1; China NMPA stratifies per each drug’s registered CPS. IO monotherapy 2L (KEYNOTE-061) and IO maintenance (JAVELIN Gastric 100) paths are both closed — don’t walk them again.

2.5 Biomarker triple-subtype year zero + perioperative IO + CAR-T dawn (2023-2026)

Story: 2023 has been called the “gastric biomarker revolution year” — SPOTLIGHT (Shitara, CLDN18.2 + mFOLFOX6) and GLOW (Shah, CLDN18.2 + CAPOX) both published positive phase IIIs the same year, putting zolbetuximab (anti-Claudin-18.2 isoform 2 mAb) on the new CLDN18.2+ HER2- 1L standard (global HER2- gastric cancer CLDN18.2 high-expression rate ~38%); the same year KEYNOTE-811 completed HER2+ 1L IO layering + KEYNOTE-859 removed CPS restriction + ORIENT-16 domestic PD-1 positive. These 3 directions reading out together determined that post-2023 gastric cancer biomarker testing must include the HER2 + CLDN18.2 + CPS + MSI four-panel. In 2024 ATTRACTION-5’s adjuvant IO negative + 2025 MATTERHORN’s perioperative IO positive formed a one-negative-one-positive signal with a key message: “IO requires neoadjuvant exposure to activate.” In 2025 China’s CT041-ST-01 used satri-cel (satricabtagene autoleucel, CLDN18.2 CAR-T) to pull off the world’s first solid-tumor CAR-T RCT win.

FAST [PMID 33610734] (Sahin 2021 Ann Oncol, N=161, phase II): CLDN18.2+ 1L zolbetuximab + EOX vs EOX. mPFS/mOS HR 0.44 and 0.55 (both P<0.0005), CLDN18.2 moderate/strong expression ≥70% subgroup PFS HR 0.38. Early proof-of-concept for CLDN18.2 as a druggable target, paving the way for SPOTLIGHT/GLOW.
SPOTLIGHT [PMID 37068504] (Shitara 2023 Lancet, N=565): CLDN18.2+ (IHC ≥75% 2+/3+) HER2- 1L zolbetuximab + mFOLFOX6 vs placebo + mFOLFOX6. mPFS 10.61 vs 8.67 months (HR 0.75, P=0.0066), mOS 18.23 vs 15.54 months (HR 0.75, P=0.0053). CLDN18.2 became gastric cancer’s third actionable biomarker (after HER2 and PD-L1).
GLOW [PMID 37524953] (Shah 2023 Nat Med, N=507): CLDN18.2+ HER2- 1L zolbetuximab + CAPOX vs placebo + CAPOX. mPFS 8.21 vs 6.80 months (HR 0.687, P=0.0007), mOS 14.39 vs 12.16 months (HR 0.771, P=0.0118). Concordantly replicated SPOTLIGHT — zolbetuximab class effect confirmed, mFOLFOX6 and CAPOX backbones equivalent. FDA approved 2024-10.
ILUSTRO [PMID 37490286] (Klempner 2023 Clin Cancer Res, phase II multi-cohort): CLDN18.2+ advanced zolbetuximab monotherapy (cohort 1A) / + mFOLFOX6 (cohort 2) / + pembrolizumab (cohort 3A). 1L+chemo cohort mPFS 17.8 months / ORR 71.4%; monotherapy 3L+ ORR 0%; zolbetuximab + pembrolizumab 3L+ ORR 0% / mPFS 2.96 months. Key take-home: zolbetuximab fails as monotherapy / + IO at 3L+; must be used 1L with chemo.
TRANSTAR102 cohort G (NCT04495296): high / moderate CLDN18.2 expression 1L osemitamab (TST001, anti-CLDN18.2 ADCC-enhanced antibody) + nivolumab + CAPOX phase II single-arm. Any PD-L1 subgroup mPFS 12.6 months, CPS<5 subgroup mPFS 12.6 months, ORR 66%+. Chinese domestic CLDN18.2 + IO + chemo triplet candidate regimen — PMID not yet assigned as of 2026-04, phase III ongoing.
FRUTIGA (NCT03223376, Xu 2024 published but PMID link anomalous): 2L HER2- / post-chemo fruquintinib (oral VEGFR TKI) + paclitaxel vs placebo + paclitaxel. mPFS 5.6 vs 2.7 months (HR 0.57, P<0.0001, MET), mOS 9.6 vs 8.4 months (HR 0.96, P=0.6064, OS not met), ORR 42.4% vs 22.4%. Chinese domestic VEGFR TKI 2L PFS wins, OS loses — NMPA approved 2023, but limited global impact. The PMID metadata link in yaml is anomalous (see bottom of §7.1); this section primarily indexes by NCT.
KEYNOTE-585 [PMID 40829093] (Shitara 2025 J Clin Oncol, N=1007): resectable GC/GEJ perioperative pembrolizumab + chemotherapy (cisplatin/5-FU or FLOT subgroup) vs placebo + chemotherapy. Main cohort mOS 71.8 vs 55.7 months (HR 0.86, NS), EFS HR 0.81, pathCR improved; but OS primary endpoint not met. “Perioperative IO + old FP backbone failed, FLOT subgroup signal good” — laid the hypothesis foundation for MATTERHORN using FLOT as the backbone.
MATTERHORN [PMID 40454643] (Janjigian 2025 N Engl J Med, N=948): resectable GC/GEJ durvalumab 1500 mg Q4W + perioperative FLOT × 4+4 → durvalumab Q4W × 10 vs placebo + FLOT. 2-year EFS 67.4% vs 58.5% (HR 0.71, P<0.001), pCR 19.2% vs 7.2% (RR 2.69), 2-year OS 75.7% vs 70.4% (OS maturity pending). ASCO 2025 practice-changing — IO finally entered perioperative; KN-585 lesson inherited: must use FLOT backbone, not old FP.
ATTRACTION-5 [PMID 38906161] (Kang 2024 Lancet Gastroenterol Hepatol, N=755, Japan/Korea/Taiwan): post-D2 stage III adjuvant nivolumab + S-1/CapOx vs placebo + chemo. 3-year RFS 68.4% vs 65.3% (HR 0.90, P=0.44, negative). Pure postop addition of IO does not activate immune response — in stark contrast to the positive signals from MATTERHORN / KN-585 neoadjuvant exposure. Mechanistic hypothesis: IO requires “tumor-in-situ priming” — after resection tumor antigen disappears, ICI loses its target.
CT041-ST-01 [PMID 40460847] (Qi 2025 Lancet, N=156): advanced CLDN18.2+ 3L+ satri-cel (CLDN18.2 CAR-T) vs physician’s choice chemo. mPFS 3.25 vs 1.77 months (HR 0.37, P<0.0001), ORR 37% vs 10%, mOS trend favorable. World’s first solid-tumor CAR-T RCT win, NMPA NDA accepted — gastric cancer becomes the cancer type where solid-tumor CAR-T breaks through.

Takeaway: 2026 advanced 1L biomarker subtyping must check the HER2 + CLDN18.2 + PD-L1 CPS + MSI/dMMR four-panel — HER2+ / CLDN18.2+ / CPS-high / MSI-H have low cross-overlap (each covers 15-20% / 38% / ~60% / ~5-10%); missing any one = missing a high-responding subgroup with 30-70% ORR. Perioperative IO = must be neoadjuvant-exposed + must use FLOT backbone (MATTERHORN / KN-585 dual validation); pure postop IO doesn’t work (ATTRACTION-5 negative). CLDN18.2+ 3L+ CAR-T path (CT041-ST-01) has opened the door in China; global rollout awaits NMPA NDA completion.

3. Horizontal: the 2026 decision landscape (six dimensions)

Projecting longitudinal evolution onto the concrete 2026 clinical decision tree, the following are the six key branchpoints and the evidence for each.

3.1 Newly diagnosed advanced GC/GEJ: full biomarker four-panel

NCCN Gastric V2.2025 + CSCO Gastric 2025 explicitly recommend biomarker testing for all newly diagnosed advanced GC/GEJ, covering: HER2 IHC/FISH + PD-L1 IHC (22C3 CPS or 58-8 TAP) + CLDN18.2 IHC + MSI/dMMR (IHC or PCR). A positive result in any one directly changes the 1L regimen:

HER2+ (IHC3+ or IHC2+/FISH+, ~15-20%): CPS≥1 → KEYNOTE-811 (pembrolizumab + trastuzumab + FP/CAPOX); CPS<1 → ToGA (trastuzumab + chemo)
HER2- / CLDN18.2+ (IHC ≥75% 2+/3+, ~38%): zolbetuximab + mFOLFOX6 (SPOTLIGHT) or + CAPOX (GLOW)
HER2- / CLDN18.2- / CPS≥5: nivolumab + FOLFOX/XELOX (CheckMate-649) / pembrolizumab + FP/CAPOX (KEYNOTE-859) / sintilimab + XELOX (ORIENT-16) / tislelizumab + chemo (RATIONALE-305), any of them
HER2- / CLDN18.2- / CPS<5: FOLFOX / CAPOX / SOX chemo (limited IO benefit; confirmed by KEYNOTE-062)
MSI-H / dMMR (~5-10%): IO + chemo or pembrolizumab monotherapy (MSI-H has significantly higher response rate; KEYNOTE-062 CPS≥10 subgroup was partly MSI-H–driven)

3.2 Advanced 1L landscape: four PD-(L)1 class effect + CPS threshold policy divergence

2026 mainstream: HER2- advanced 1L SoC = IO + chemo (any of four PD-(L)1s + FOLFOX/CAPOX/SOX backbone), HRs converging in the 0.71-0.80 narrow band.

Subgroup	First choice	Second choice
HER2- CPS≥5	nivolumab + FOLFOX/XELOX [CheckMate-649 PMID 34102137] or pembrolizumab + FP/CAPOX [KEYNOTE-859 PMID 37875143]	sintilimab + XELOX [ORIENT-16 PMID 38051328] (China cost advantage) / tislelizumab + chemo [RATIONALE-305 PMID 38806195]
HER2- CPS 1-4	pembrolizumab + FP/CAPOX (available after US FDA removed CPS limit)	IO not mandatory, chemo alone acceptable
HER2- CPS<1	FOLFOX / CAPOX / SOX chemo [REAL-2 PMID 18172173 / G-SOX PMID 25316259]	limited IO benefit
HER2+ CPS≥1	pembrolizumab + trastuzumab + FP/CAPOX [KEYNOTE-811 PMID 37871604]	trastuzumab + chemo [ToGA PMID 20728210]
HER2+ CPS<1	trastuzumab + FP/CAPOX [ToGA PMID 20728210]	—
HER2- CLDN18.2+	zolbetuximab + mFOLFOX6 [SPOTLIGHT PMID 37068504] or + CAPOX [GLOW PMID 37524953]	IO + chemo (if CLDN18.2 untested / negative)

CPS threshold regional policy differences: FDA removed the pembrolizumab gastric CPS threshold (based on KN-859 all-comer positive) → US HER2- full coverage; NCCN Gastric V2.2025 still recommends CPS≥5 as IO benefit marker; EU EMA retains CPS≥1; China NMPA stratifies per each drug’s registered CPS (nivolumab CPS≥5 / sintilimab CPS≥5 / tislelizumab TAP≥5% etc.).

Contraindicated / not recommended 2026: IO monotherapy 2L unselected (KEYNOTE-061 negative) + IO maintenance (JAVELIN Gastric 100 negative) + lapatinib instead of trastuzumab (LOGIC/TRIO-013 negative).

3.3 Advanced 2L+: HER2 status dominates + mandatory re-biopsy

2026 mainstream: post-IO era, 2L must first confirm HER2 status (re-biopsy), because post-trastuzumab HER2 loss is ~30%.

Population	2L first choice	3L+
HER2+ post-trastuzumab (confirmed still HER2+)	T-DXd [DESTINY-Gastric-04 PMID 40454632] (mOS 14.7 months, HR 0.70)	ramucirumab + paclitaxel [RAINBOW PMID 25240821] / TAS-102 [TAGS PMID 30355453] / irinotecan
HER2-, IO-exposed (mainstream)	ramucirumab + paclitaxel [RAINBOW PMID 25240821] (mOS 9.6 vs 7.4 months, HR 0.807)	docetaxel + ASC [COUGAR-02 PMID 24332238] / TAS-102 / ramucirumab monotherapy [REGARD PMID 24094768]
CLDN18.2+ post-zolbetuximab	ramucirumab + paclitaxel	TAS-102 / satri-cel CAR-T [CT041-ST-01 PMID 40460847] (China pending NMPA)
CheckMate-032 multi-IO failures	chemo fallback	clinical trial
FRUTIGA (fruquintinib + paclitaxel)	China NMPA 2L option (NCT03223376, 2L HER2- CSCO Gastric 2025 Level II recommendation)	—

New challenge: since 2025 the 2L mandatory HER2 / CLDN18.2 re-biopsy workflow has low real-world execution — frontline clinicians habitually go “1L progression → 2L direct RAINBOW,” with high missed-diagnosis rates.

3.4 Adjuvant / perioperative: five East–West paths that never meet

2026 mainstream: geographically branched, no unified global SoC.

Region	Perioperative / adjuvant SoC (2026)	Key evidence
North America / Western Europe (GC majority)	Perioperative FLOT × 4+4 + perioperative durvalumab [MATTERHORN PMID 40454643] (2-year EFS 67.4%)	FLOT4 [PMID 30982686] upgraded backbone to FLOT; MATTERHORN layered IO on top
GEJ/EAC (global)	Neoadjuvant CRT (CROSS regimen) + surgery [CROSS PMID 22646630] (mOS 49.4 months)	Global GEJ tri-modality SoC; Neo-AEGIS [PMID 37734399] did not falsify it
Japan (mostly stage III)	D2 postoperative S-1 + docetaxel adjuvant [JACCRO GC-07 PMID 30925125] (3y RFS 66%)	Japanese “adjuvant-first” tradition + neoadjuvant option PRODIGY [PMID 34133211]
Korea (stage II-III)	D2 postoperative CAPOX adjuvant [CLASSIC PMID 22226517] or SOX adjuvant [ARTIST-2 PMID 33278599]	Doublet chemo sufficient; ARTIST-2 closed post-D2 RT debate
China (locally advanced)	Perioperative SOX [RESOLVE PMID 34252374]	China post-D2 first choice perioperative SOX (HR 0.77 vs adjuvant CAPOX)
Residual disease post-CROSS (any region)	Adjuvant nivolumab 1 year [CheckMate-577 PMID 33789008] (mDFS 22.4 vs 11.0 months)	ASCO 2025 mature OS shows benefit concentrated in PD-L1+ subgroup; ITT OS not significant
Not recommended	bevacizumab perioperative [ST03 PMID 28163000] / postop CRT after preop chemo [CRITICS PMID 29650363] / pure postop IO [ATTRACTION-5 PMID 38906161] / preop CRT on top of perioperative chemo [TOPGEAR PMID 39282905] (pCR improved, OS did not)	All four falsified

Current status of INT-0116 adjuvant chemoRT: 15-year US SoC initially, exited Asian practice in the adequate-D2 era (ARTIST-2 closed debate); US retains for D0-D1 / inadequately-dissected nodes.

The “death” of MAGIC: the 2006 MAGIC ECF perioperative backbone was replaced by FLOT after 2019 FLOT4; ECF is used only occasionally for “FLOT-intolerant” patients, no longer SoC.

3.5 Perioperative IO: MATTERHORN positive + key lessons

2026 mainstream: MATTERHORN [PMID 40454643] 2-year EFS 67.4% vs 58.5% (HR 0.71) + pCR 19.2% vs 7.2% → durvalumab + FLOT perioperative becomes the new North America / Western Europe SoC candidate (FDA 2025-2026 expected approval; as of 2026-04 mature OS not yet reached).

Key lesson 1 (backbone determines IO effect): KEYNOTE-585 [PMID 40829093] used old FP backbone, perioperative pembrolizumab mOS 71.8 vs 55.7 months but NS; MATTERHORN used new FLOT backbone + durvalumab 2-year EFS positive. “Perioperative IO must use FLOT backbone” written into 2025 ASCO consensus.

Key lesson 2 (IO needs neoadjuvant exposure): ATTRACTION-5 [PMID 38906161] pure postop nivolumab + S-1/CapOx adjuvant 3-year RFS 68.4% vs 65.3% (HR 0.90, negative); vs MATTERHORN’s neoadjuvant + adjuvant durvalumab positive — “postop-only IO does not activate immunity,” mechanistic hypothesis: ICI needs tumor-in-situ antigen priming.

Key lesson 3 (pCR ≠ OS): TOPGEAR [PMID 39282905] pCR 17% vs 8% improved but mOS no difference (HR 1.05, NS); KEYNOTE-585 [PMID 40829093] pCR improved but ITT OS NS. “Gastric pCR as surrogate is unstable” — fundamentally different from the strong pCR-OS correlation in breast cancer; clinical trial design needs caution.

3.6 Surgical technique: D2 + laparoscopic consensus + robotic undecided

From 2015-2022 five consecutive laparoscopic phase IIIs (KLASS-01 / KLASS-02 / CLASS-01 / JLSSG0901 / LOGICA) established laparoscopic ≈ open D2 gastrectomy equivalence; two extended-surgery trials — D2+PAND (JCOG9501) and bursectomy (JCOG1001) — were buried.

Category	Standard regimen	Key evidence
Lymph node dissection	D2	JCOG9501 [PMID 18669424] (5-year OS 69.2% vs 70.3%, D2+PAND HR 1.03, no benefit + longer operating time / more blood loss)
Early GC (stage I) distal gastrectomy	Laparoscopic = open	KLASS-01 [PMID 30730546] (5-year OS 94.2% vs 93.3%, non-inferior)
Locally advanced GC distal gastrectomy	Laparoscopic = open	KLASS-02 [PMID 35857305] (5-year OS 88.9% vs 88.7%), CLASS-01 [PMID 31135850] (3-year DFS 76.5% vs 77.8%, non-inferior), JLSSG0901 [PMID 36920382] (5-year RFS 73.9% vs 75.7%, non-inferior)
European total / subtotal gastrectomy	Laparoscopic = open (short-term)	LOGICA [PMID 34581617] (LOS 7 days / both arms, LG less blood loss)
Bursectomy	Abandoned	JCOG1001 [PMID 36369984] (5-year OS 74.9% vs 76.5%, HR 1.03, more abdominal abscesses)
Robotic	Undecided	No phase III oncologic RCT evidence yet

4. Research Gaps: ten unsolved clinical questions

This report identifies the following gaps, each a definable concrete question (not the cliché “more research needed”):

CPS threshold global non-uniformity + real-world execution: FDA removed the pembrolizumab gastric CPS threshold + NCCN retains CPS≥5 + EU CPS≥1 + China NMPA stratifies per drug. Frontline clinicians see multiple thresholds on the same guideline page → real-world PD-L1 IHC (22C3 vs 58-8 TAP) assay differences also unresolved. Unification needed.
No head-to-head among three / four PD-(L)1s (nivolumab / pembrolizumab / sintilimab / tislelizumab) in HER2- 1L: HRs in the 0.71-0.80 narrow band, cross-trial indistinguishable. Choice depends entirely on access × price × CPS threshold × chemo backbone preference.
Is the CLDN18.2 moderate / high-expression threshold optimal: SPOTLIGHT / GLOW enrolled at IHC ≥75% 2+/3+; FAST subgroup ≥70% moderate-strong expression had better PFS HR. Different thresholds may identify different benefit populations — prospective threshold-exploration trials needed.
Is CLDN18.2 × IO combination superior to either alone: ILUSTRO 3L+ zolbetuximab + pembrolizumab cohort ORR 0% (failed); TRANSTAR102 cohort G 1L osemitamab + nivolumab + CAPOX 12.6-month PFS (signal). 1L head-to-head CLDN18.2 + IO + chemo vs CLDN18.2 + chemo has not been done.
1L choice for HER2+ CPS<1 subgroup: KEYNOTE-811 CPS<1 subgroup HR near 1.0, but FDA 2023-11 restricted to CPS≥1 rather than fully excluding CPS<1. Practice remains inconsistent on whether HER2+ CPS<1 uses ToGA or KN-811.
Backbone dependence of perioperative IO: MATTERHORN FLOT + durvalumab positive vs KEYNOTE-585 FP + pembrolizumab negative. Is it the FLOT backbone contribution or an IO mechanism difference? Independent confirmations from perioperative nivolumab + FLOT (CheckMate-Gastric Adjuvant ongoing) and pembrolizumab + FLOT subgroup (KN-585 FLOT subgroup) need time.
Why pure postop adjuvant IO failed in ATTRACTION-5: the mechanistic hypothesis is “tumor-in-situ priming” but has not been directly validated. If future biomarkers can identify “postop still IO-suitable” subgroups (e.g., ctDNA+ residual / MSI-H status), adjuvant IO could restart.
pCR as gastric surrogate endpoint is unstable: both TOPGEAR + KEYNOTE-585 were pCR-positive / OS-negative. Clinical trial design must use pCR as primary endpoint cautiously. Alternative candidates: EFS, ctDNA clearance kinetics.
The five East–West perioperative / adjuvant paths (FLOT4 / RESOLVE / CLASSIC / ARTIST-2 / JACCRO GC-07) have never been head-to-head: each path ran its own phase III in its own country vs surgery alone / old regimen, with no RCTs between them. Real-world OS differences are population differences vs regimen differences? Undistinguishable.
Global rollout path for CLDN18.2 + CAR-T: CT041-ST-01 China phase II positive + NMPA NDA accepted; global phase III and Euro-US regulatory paths not yet launched as of 2026. Whether solid-tumor CAR-T’s gastric breakthrough can extend to other CLDN18.2+ solid tumors (pancreatic, biliary) is unknown.

5. Latest 2024-2026 developments

5.1 FDA / NMPA new approvals (gastric-relevant excerpts)

Drug	Agency	Date	Indication / supporting trial
pembrolizumab + trastuzumab + chemo	FDA	2023-11 (CPS≥1 restricted)	1L HER2+ GC/GEJ / KEYNOTE-811 [PMID 37871604]
pembrolizumab + chemo (CPS restriction removed)	FDA	2023-11	1L HER2- GC/GEJ / KEYNOTE-859 [PMID 37875143]
tislelizumab + chemo	NMPA	2024	1L HER2- PD-L1 TAP≥5% / RATIONALE-305 [PMID 38806195]
zolbetuximab + mFOLFOX6 / CAPOX	FDA	2024-10-18	1L HER2- CLDN18.2+ / SPOTLIGHT + GLOW [PMID 37068504 + 37524953]
T-DXd (trastuzumab deruxtecan) 2L	FDA	2025 (expanded indication after DG-04)	2L HER2+ GC/GEJ / DESTINY-Gastric-04 [PMID 40454632]
durvalumab + perioperative FLOT	FDA	2025-2026 (pending, expected approval)	Perioperative GC/GEJ / MATTERHORN [PMID 40454643]
satri-cel (CLDN18.2 CAR-T)	NMPA	2025-2026 (NDA accepted)	3L+ CLDN18.2+ GC/GEJ / CT041-ST-01 [PMID 40460847]
fruquintinib + paclitaxel 2L	NMPA	2023	2L HER2- GC/GEJ / FRUTIGA (NCT03223376)

5.2 Key meeting readouts (2025-2026, flagged as lower-weight)

The following are candidate pool only pending formal peer review, not in the main library.

HERIZON-GEA-01 (ESMO 2025 LBA): zanidatamab + chemo ± tislelizumab vs trastuzumab + chemo 1L HER2+ positive, mOS improvement >7 months. Full manuscript unpublished as of 2026-04; if confirmed positive within 2026 → HER2 backbone turnover.
MATTERHORN mature OS (ASCO 2025 [PMID 40454643]): 2-year OS 75.7% vs 70.4% (early signal), mature OS 2026-2027 readout.
CheckMate-577 mature OS (ASCO 2025 [PMID 33789008] data): ITT OS not significant; PD-L1+ subgroup benefit — suggests post-CROSS adjuvant IO should be restricted to PD-L1+.
KEYNOTE-585 final analysis (2025 JCO [PMID 40829093]): FLOT subgroup EFS positive signal confirmed by MATTERHORN; FP subgroup OS NS closed the old backbone + IO path.
CT041-ST-01 mOS (Lancet 2025 [PMID 40460847]): mOS trend favorable, data maturing.
TRANSTAR102 cohort G (NCT04495296, ESMO 2024 oral): osemitamab + nivolumab + CAPOX 1L phase II signal good; phase III launched in China.

5.3 Ongoing phase IIIs (2026-2028 readout highlights)

HERIZON-GEA-01 full readout (NCT05152147): 2026 mature OS will determine whether zanidatamab can replace trastuzumab as the HER2+ 1L backbone
osemitamab phase III (NCT05980416 or successor): CLDN18.2+ 1L triplet vs IO + chemo head-to-head
satri-cel global phase III: CLDN18.2 CAR-T expansion outside China
ctDNA-guided adjuvant de-escalation: after ATTRACTION-5 negative, whether ctDNA+ postop residual can identify IO-responsive subgroups — early design
Perioperative IO + CLDN18.2 combination: after MATTERHORN success, CLDN18.2 + perioperative FLOT + IO triplet is the natural next step

6. Synthesis insights and judgments

6.1 Longitudinal × horizontal: 2026 gastric landscape shaped by three subtyping revolutions

Overlaying longitudinal paradigm evolution with the current horizontal decision landscape, the 2026 gastric cancer landscape is a stack of three subtyping revolutions:

Biological meaning of anatomical subtyping made manifest: 2006 MAGIC / 2012 CROSS / 2021 CheckMate-649 step by step pushed “GEJ vs gastric body” from crude endoscopic split to molecular biological subtyping in the IO era — CheckMate-649 subgroup analysis showed GEJ/EAC HR differs from pure gastric body HR (the former ~0.75, the latter ~0.65), suggesting the GEJ microenvironment responds less well to IO. In 2026 GEJ 1L decisions have gone from “treat as gastric cancer” to “Siewert II-III → gastric / Siewert I → esophageal” + perioperative choice CROSS (GEJ) vs MATTERHORN (gastric body majority). This is gastric cancer’s first subtyping revolution.
Persistent East–West path divergence: INT-0116 (US adjuvant chemoRT) / MAGIC → FLOT4 (European perioperative) / CLASSIC (Korean adjuvant CAPOX) / ACTS-GC (Japanese adjuvant S-1) / RESOLVE (Chinese perioperative SOX) — five paths, each completing phase III in its own population, never head-to-head. 2026 practice is not a “unified global SoC” but “choose path by patient’s country / region.” Regional differences + genetic background (DPYD polymorphisms) + reimbursement policies together shape the reality of “East and West never meet.” This is gastric cancer’s second subtyping revolution.
HER2 drug 10-year leap + biomarker four-layer subtyping: 2010 ToGA (trastuzumab + chemo) → 2020 DESTINY-Gastric-01 (T-DXd 2L) → 2023 KEYNOTE-811 (pembro + trastuzumab + chemo 1L) → 2025 DESTINY-Gastric-04 (T-DXd 2L SoC) → 2025 HERIZON-GEA-01 (zanidatamab challenging trastuzumab) — 5 steps in 15 years, HER2 drugs moved from binary targeting to full ADC + IO combo + bispecific coverage. Concurrently CLDN18.2 (SPOTLIGHT / GLOW) + PD-L1 CPS + MSI joined the subtyping grid → HER2 / CLDN18.2 / CPS / MSI four-layer checkerboard decides 1L regimen. This is gastric cancer’s third subtyping revolution and the direct source of the 2026 “check the full biomarker panel first” imperative.

These three revolutions together explain a clinical observation: the 1L decision tree for a newly diagnosed stage IV GC/GEJ patient in 2026 has 4 more decision layers than in 2016 (biomarker panel HER2/CPS/CLDN18.2/MSI → anatomical subtyping GEJ vs gastric body → regional access sintilimab/pembro/nivo/tislelizumab → CPS threshold FDA vs NCCN vs EU). This “deep and wide” decision tree is far more complex than HCC’s “shallow and narrow” (four-way tie within a narrow HR band) — the result of NCCN-level research density plus multinational domestic-drug co-action.

6.2 Clinical decision takeaways (for junior-mid oncologists)

In 2026 the first gastric cancer step is not the chemo regimen — it is a full biomarker panel: HER2 IHC/FISH + PD-L1 CPS + CLDN18.2 IHC + MSI/dMMR must be complete before 1L. Missing any one = missing a high-responding subgroup with 30-70% ORR.
HER2+ 1L standard upgraded since 2023: CPS≥1 use pembrolizumab + trastuzumab + FP/CAPOX (KEYNOTE-811); CPS<1 still trastuzumab + chemo (ToGA). Don’t use ToGA alone for CPS≥1 patients anymore.
CLDN18.2+ HER2- 1L SoC is zolbetuximab + mFOLFOX6/CAPOX: written into NCCN V2.2025 after FDA 2024-10 approval. Missing CLDN18.2 at 1L is the most frequent 2026 gastric cancer clinical error.
Use CPS thresholds per country: US FDA removed the pembrolizumab gastric CPS threshold; NCCN still recommends CPS≥5; EU CPS≥1; China NMPA stratifies per drug. Use the corresponding threshold in your own reimbursement / registration context.
Both IO monotherapy 2L and IO maintenance are closed: KEYNOTE-061 (2L unselected) and JAVELIN Gastric 100 (maintenance) both negative — don’t walk these two paths.
2L must re-biopsy to recheck HER2 + CLDN18.2: post-trastuzumab HER2 loss is ~30%; CLDN18.2 subgroup has similar phenomenon. “1L progression → 2L direct RAINBOW” is the biggest real-world missed-diagnosis source.
HER2+ 2L new standard is T-DXd (DESTINY-Gastric-04 beat RAINBOW): but ILD is the key safety alarm, needing baseline chest CT + dynamic monitoring + timely drug discontinuation.
The perioperative IO era has arrived: North America / Western Europe 2026 expected MATTERHORN (durvalumab + FLOT perioperative) approval; must use FLOT backbone + must have neoadjuvant exposure — both lessons have trial evidence (KN-585 FP backbone failure / ATTRACTION-5 pure postop IO failure) as controls.
Respect regional East–West paths: US walks INT-0116 + FLOT4 → MATTERHORN; Europe walks MAGIC → FLOT4 → MATTERHORN; China walks RESOLVE SOX; Japan walks JACCRO GC-07 S-1 + docetaxel; Korea walks CLASSIC CAPOX / ARTIST-2 SOX. Don’t simply extrapolate one country’s data to another.
Zanidatamab is the biggest 2026 suspense: if HERIZON-GEA-01 full paper 2026 is positive, trastuzumab’s 13-year HER2 backbone could end. Frontline clinicians’ HER2 treatment choices will face a new branchpoint — stay tuned.

7. Information sources

All 58 trial metadata in this report were independently verified via PubMed and ClinicalTrials.gov. Each [PMID xxxxxxxx] in the text is directly verifiable on PubMed.

Published trials: 58, covering 2001-2026
PMID coverage: 56 unique PMIDs (HERIZON-GEA-01 / TRANSTAR102 cohort G are ESMO LBA, PMID pending full manuscript; MAGIC / INT-0116 / CROSS have no NCT because they predate the mandatory NCT registration era)
FDA / NMPA new approvals: 8 key approvals (2023-2026)
2025-2026 key meeting / long follow-up readouts: 6 (HERIZON-GEA-01 / MATTERHORN mature OS / CheckMate-577 mature OS / KN-585 final / CT041-ST-01 / TRANSTAR102)
Research gaps: 10
China-led research proportion: ~25% (ORIENT-16 / FRUTIGA / RESOLVE / CT041-ST-01 / TRANSTAR102 / part of CLASSIC / CLASS-01 etc.)

7.1 Text citation list (by ascending PMID)

The following table is the PMID list of all bracket citations in the text + summary tables; each can be clicked to verify on PubMed.

PMID	First Author	Year	Journal	Trial / topic
11547741	Macdonald JS	2001	N Engl J Med	INT-0116 adjuvant chemoRT
16822992	Cunningham D	2006	N Engl J Med	MAGIC perioperative ECF
17075117	Van Cutsem E	2006	J Clin Oncol	V325 advanced DCF
17978289	Sakuramoto S	2007	N Engl J Med	ACTS-GC Japanese adjuvant S-1
18172173	Cunningham D	2008	N Engl J Med	REAL-2 (ECF/ECX/EOF/EOX)
18282805	Koizumi W	2008	Lancet Oncol	SPIRITS Japanese advanced SP
18669424	Sasako M	2008	N Engl J Med	JCOG9501 D2 vs D2+PAND
20728210	Bang YJ	2010	Lancet	ToGA HER2+ trastuzumab
22226517	Bang YJ	2012	Lancet	CLASSIC Asian adjuvant CAPOX
22646630	van Hagen P	2012	N Engl J Med	CROSS GEJ neoadjuvant CRT
24094768	Fuchs CS	2014	Lancet	REGARD 2L ramucirumab
24332238	Ford HE	2014	Lancet Oncol	COUGAR-02 2L docetaxel
25240821	Wilke H	2014	Lancet Oncol	RAINBOW 2L ramu + paclitaxel
25316259	Yamada Y	2015	Ann Oncol	G-SOX advanced SOX vs CS
25559811	Park SH	2015	J Clin Oncol	ARTIST D2 postop XP ± RT
26628478	Hecht JR	2016	J Clin Oncol	LOGiC/TRIO-013 lapatinib negative
28163000	Cunningham D	2017	Lancet Oncol	ST03 perioperative ECX + bev negative
28993052	Kang YK	2017	Lancet	ATTRACTION-2 3L+ nivolumab
29650363	Cats A	2018	Lancet Oncol	CRITICS postop CRT negative
29880231	Shitara K	2018	Lancet	KEYNOTE-061 2L pembro negative
30110194	Janjigian YY	2018	J Clin Oncol	CheckMate-032 nivo ± ipi multi-arm
30355453	Shitara K	2018	Lancet Oncol	TAGS 3L+ TAS-102
30730546	Kim HH	2019	JAMA Oncol	KLASS-01 early GC LDG
30925125	Yoshida K	2019	J Clin Oncol	JACCRO GC-07 S-1 + docetaxel
30982686	Al-Batran SE	2019	Lancet	FLOT4 perioperative FLOT vs ECF
31135850	Yu J	2019	JAMA	CLASS-01 LDG vs ODG
32469182	Shitara K	2020	N Engl J Med	DESTINY-Gastric-01 2L T-DXd
32880601	Shitara K	2020	JAMA Oncol	KEYNOTE-062 1L cold water
33197226	Moehler M	2021	J Clin Oncol	JAVELIN Gastric 100 maintenance negative
33278599	Park SH	2021	Ann Oncol	ARTIST-2 Korea D2 postop
33610734	Sahin U	2021	Ann Oncol	FAST phase II CLDN18.2 early
33789008	Kelly RJ	2021	N Engl J Med	CheckMate-577 GEJ adjuvant nivolumab
34102137	Janjigian YY	2021	Lancet	CheckMate-649 milestone
34133211	Kang YK	2021	J Clin Oncol	PRODIGY neoadjuvant DOS
34252374	Zhang X	2021	Lancet Oncol	RESOLVE China perioperative SOX
34581617	van der Veen A	2021	J Clin Oncol	LOGICA Europe LG vs OG
35030335	Kang YK	2022	Lancet Oncol	ATTRACTION-4 Japan/Korea nivo + SOX
35857305	Son SY	2022	JAMA Surg	KLASS-02 5-year OS LDG vs ODG
36369984	Kurokawa Y	2022	Br J Surg	JCOG1001 bursectomy negative
36920382	Etoh T	2023	JAMA Surg	JLSSG0901 LADG vs ODG 5-year
37068504	Shitara K	2023	Lancet	SPOTLIGHT zolbetuximab + mFOLFOX6
37329891	Van Cutsem E	2023	Lancet Oncol	DESTINY-Gastric-02 Euro-US T-DXd
37490286	Klempner SJ	2023	Clin Cancer Res	ILUSTRO phase II multi-cohort
37524953	Shah MA	2023	Nat Med	GLOW zolbetuximab + CAPOX
37734399	Reynolds JV	2023	Lancet Gastro Hepatol	Neo-AEGIS CROSS vs MAGIC/FLOT
37871604	Janjigian YY	2023	Lancet	KEYNOTE-811 HER2+ IO + HER2
37875143	Rha SY	2023	Lancet Oncol	KEYNOTE-859 all-comer
38051328	Xu J	2023	JAMA	ORIENT-16 China sintilimab
38734867	—	—	—	FRUTIGA yaml PMID link anomalous (text cites by NCT03223376)
38806195	Qiu MZ	2024	BMJ	RATIONALE-305 tislelizumab
38906161	Kang YK	2024	Lancet Gastro Hepatol	ATTRACTION-5 pure adjuvant IO negative
39282905	Leong T	2024	N Engl J Med	TOPGEAR preop CRT + perioperative chemo
40454632	Shitara K	2025	N Engl J Med	DESTINY-Gastric-04 T-DXd vs RAINBOW
40454643	Janjigian YY	2025	N Engl J Med	MATTERHORN perioperative durvalumab + FLOT
40460847	Qi C	2025	Lancet	CT041-ST-01 satri-cel CAR-T
40829093	Shitara K	2025	J Clin Oncol	KEYNOTE-585 final

7.2 Verification conventions

Each PMID is directly accessible via https://pubmed.ncbi.nlm.nih.gov/{PMID}/
Each NCT id is accessible via https://clinicaltrials.gov/study/{NCT_id}/
HERIZON-GEA-01 / TRANSTAR102 cohort G are at ESMO LBA / ASCO oral stage; PMID pending 2026 full manuscript — this report indexes by NCT
FRUTIGA (PMID 38734867) yaml data has a PubMed metadata link anomaly (the PMID corresponds to an ED medication errors review, not the FRUTIGA main publication); this report primarily indexes the text by NCT03223376, with “yaml PMID link anomalous” honestly disclosed in the §7.1 table; data fix pending v2
MAGIC / INT-0116 / CROSS have no NCT id (predate the 2004 ClinicalTrials.gov mandatory registration / UK MRC internal trial)
If you find any PMID in the report whose trial name / year / conclusion differs from PubMed, corrections are welcome

Clinical trial timeline here

Chinese: /trials/gastric/ English: /en/trials/gastric/

Each trial has its own detail page, containing:

Full intervention / comparator regimen
Primary endpoint values + 95% CI
Key findings + clinical significance
Clickable jump to PMID / NCT source

58 trials · 5 chapters · 2001 to 2026 · synced with NCCN Gastric V2.2025 + CSCO Gastric 2025 dual guidelines.

Closing

Gastric cancer over the past 25 years has completed a unique evolution in oncology — from 2001 INT-0116 putting adjuvant chemoRT on the US postoperative SoC, through the 2006 MAGIC “perioperative chemo” concept invention, 2010 ToGA HER2 as the first biomarker target, 2019 FLOT4 perioperative backbone upgrade, 2021 CheckMate-649 IO + chemo 1L standard, 2023 SPOTLIGHT / GLOW CLDN18.2 new subtype, 2025 MATTERHORN perioperative IO breakthrough, 2025 DESTINY-Gastric-04 HER2+ 2L new standard, and 2025 CT041-ST-01 solid-tumor CAR-T’s first RCT win.

The most essential difference between gastric cancer and other GI major cancers (HCC / BTC / PDAC) is “early biomarker subtyping start but severe geographic branching + HER2 drugs’ 10-year leap is the deepest single-biomarker evolution”. NSCLC’s driver panel is “10+ molecular paths horizontally parallel”; HCC is “0 biomarker, IO backbone alone”; gastric cancer walks “HER2 / CLDN18.2 / PD-L1 CPS / MSI four-layer subtyping + East–West five perioperative / adjuvant paths + multi-country PD-(L)1 class-effect narrow band” three-dimensional checkerboard. This complexity dictates the 2026 gastric decision tree’s “deep and wide” character — 4 more decision layers than HCC, one fewer than NSCLC (no driver-targeted 1L monotherapy).

The most urgent structural problems to solve in 2026 are: the clinical chaos of non-unified global CPS thresholds, unclear 1L choice for HER2+ CPS<1, mechanism validation of why perioperative IO must use FLOT backbone, whether ctDNA-guided adjuvant de-escalation can identify adjuvant-IO benefit subgroups, and the global rollout path for CLDN18.2 CAR-T. Whether the next decade’s HER2 backbone is taken over by zanidatamab (HERIZON-GEA-01 full paper), whether next-generation ADCs (e.g., T-DM1 replacement) can break through T-DXd’s ILD boundary, and whether solid-tumor CAR-T can extend from gastric cancer to other CLDN18.2+ cancer types — these three questions determine the direction of gastric cancer’s next decade.

The value of this report is not “exhaustive enumeration of all trials” (PubMed can do that) but compressing 25 years of evolution + current decisions + unsolved gaps into the cognitive bandwidth of a single read. Next time you face a newly diagnosed gastric cancer patient, every branchpoint in the decision tree has this map to consult, trace, and interrogate.

Clinician × AI = Research Superpower + Clinical Decision Amplifier

—— Dual Brain Lab · 2026-04-21

HCC Clinical Trial Timeline: An 18-Year Evolution Map

Tue, 21 Apr 2026 00:00:00 +0000

Hepatocellular Carcinoma Clinical Trial Timeline: In-depth Research Report

Coverage: 42 landmark trials cited in NCCN Hepatobiliary V1.2026 (all PMID-traceable) + etiology (HBV/HCV/MASH) stratification + China-led research ecosystem + the unique dilemma of zero-predictive-biomarker approvals

Curated by Dual Brain Lab (csilab.net)

1. One-sentence definition

This report traces the evolution logic and current decision landscape of hepatocellular carcinoma (HCC) systemic therapy over the past 18 years (2008-2026), built on landmark clinical trials cited in NCCN Hepatobiliary Cancers V1.2026 — providing frontline clinicians a traceable panoramic map for “who, what, and on what evidence” decisions at the 2026 timepoint.

Iron rule: every data point from every trial is traceable to PubMed (PMID) or ClinicalTrials.gov (NCT id) — each [PMID xxxxxxxx] in the text can be clicked open on PubMed to verify the original paper.

HCC’s uniqueness concentrates in one contrast: NSCLC has 10+ predictive biomarkers (EGFR / ALK / ROS1 / KRAS / MET / HER2 / RET / BRAF / NTRK / PD-L1 TPS), BTC has 9 approved biomarkers (FGFR2 / IDH1 / HER2 / BRAF / NTRK / RET / MSI-H / TMB-H / claudin 18.2), PDAC has opened three paths (KRAS G12C / G12D / pan-KRAS) — yet HCC systemic therapy has zero approved predictive biomarkers to date. TERT (promoter mutation ~60%), TP53 (~30%), CTNNB1 (~30%) are the three main drivers but none are druggable; MET amplification and FGF19 amplification have low prevalence and phase III trials (COSMIC-312 / LEAP-002) failed. Clinical stratification relies on Child-Pugh (liver function) + BCLC (Barcelona Clinic Liver Cancer staging) + AFP (alpha-fetoprotein) + etiology (HBV/HCV/MASH) + tumor burden — all clinical parameters, no molecular biomarker.

2. Longitudinal: Evolution timeline of five treatment paradigms

HCC systemic therapy has gone through five paradigm shifts over the past 18 years: sorafenib solo for a decade (2008-2017) → multi-TKI 2L and 1L non-inferiority (2017-2019) → IO + anti-angiogenic rewriting 1L (2020-2024, four-way standoff) → the “zero predictive biomarker” precision dilemma surfaces → perioperative / adjuvant push but the first phase III long-term follow-up reversed (2023-2026).

Every shift had 1-3 phase III trials as fulcrum. Compared to NSCLC’s 5 shifts in 25 years driven by the “driver gene + immunotherapy dual-wheel” model, HCC’s evolution is characterized by “the IO backbone holding up the entire field without biomarker support” — in 2026 all four positive 1L IO combinations (atezo+bev / STRIDE / nivo+ipi / cam+rivo) achieved OS benefit in unstratified populations, with PD-L1 / TMB / MSI failing to predict response. This differs entirely from NSCLC’s path of EGFR → PD-L1 TPS stratification → combos.

2.1 The sorafenib solo era (2008-2017): one drug holding up ten years

SHARP in 2008 produced mOS 10.7 vs 7.9 months (HR 0.69) in a globally Caucasian / HCV-dominant population; in the same year the Asia-Pacific sorafenib study reproduced it in an HBV-dominant East Asian population but baseline OS was 4 months shorter — the first revelation of HCC etiology heterogeneity. Over the following decade, BRISK-FL / LiGHT (“challenger TKIs” brivanib / linifanib) both failed, and sora dominated 1L until REFLECT showed non-inferiority in 2018.

SHARP [PMID 18650514] (Llovet 2008 N Engl J Med, N=602): sorafenib 400 mg bid vs placebo in advanced HCC. mOS 10.7 vs 7.9 months (HR 0.69, p<0.001), mTTP 5.5 vs 2.8 months. Caucasian / HCV-dominant. HCC’s first positive phase III systemic therapy — the 3-month absolute OS benefit ended the “no drugs available” era.
Asia-Pacific sorafenib (Oriental) [PMID 19095497] (Cheng 2009 Lancet Oncol, N=271): sorafenib vs placebo in East Asia (China / Korea / Taiwan) HBV ~73% population. mOS 6.5 vs 4.2 months (HR 0.68). HR nearly identical to SHARP but absolute OS 4 months shorter — the first phase III revelation that HBV-HCC has worse baseline prognosis. Population heterogeneity became mandatory consideration for HCC research thereafter.
BRISK-FL [PMID 23980084] (Johnson 2013 J Clin Oncol, N=1155): brivanib (VEGFR2+FGFR TKI) vs sorafenib 1L. Non-inferiority not met (OS HR 1.06, 95% CI 0.93-1.22), brivanib AE worse. Sora’s monopoly further consolidated.
LiGHT (linifanib trial) [PMID 25488963] (Cainap 2015 J Clin Oncol, N=1035): linifanib vs sorafenib 1L. OS HR 1.046, double failure (linifanib more toxic and ineffective). Sora’s 1L position completely unchallenged.

Takeaway: 2008-2017 was a decade with only sorafenib in 1L HCC. mOS ceiling ~11 months (global) / ~6.5 months (HBV-dominant East Asia) — HCC’s “loneliest decade.” But these ten years established three fundamental lessons for HCC research: etiology heterogeneity (HBV vs HCV vs non-viral) → phase III must stratify; Child-Pugh A as enrollment gate universality; “non-inferiority to sora” is far from enough to win — substantial OS improvement is required.

2.2 Multi-TKI 2L breakthrough and 1L non-inferiority (2017-2019): sorafenib’s monopoly pried open

In 2017 RESORCE made regorafenib the first positive 2L regimen in ten years; in 2018 CELESTIAL’s cabozantinib further broadened 2L; in 2019 REACH-2 used AFP ≥ 400 ng/mL enrichment to make ramucirumab HCC’s first biomarker-selected positive phase III; in 2018 REFLECT’s lenvatinib finally achieved non-inferiority to sora in 1L; in 2021 AHELP’s Chinese domestically-developed apatinib (= rivoceranib) showed 2L positivity, paving the way for the CARES-310 1L combination in 2023.

RESORCE [PMID 27932229] (Bruix 2017 Lancet, N=573): regorafenib 160 mg d1-21 q4w vs placebo in sora-tolerant (sora 400 mg/d ≥ 20/28 days) progressed patients. mOS 10.6 vs 7.8 months (HR 0.63, p<0.0001). First positive 2L in a decade — established the “sorafenib continuum of care” concept (2L benefit conditional on prior 1L sora tolerance).
REFLECT [PMID 29433850] (Kudo 2018 Lancet, N=954): lenvatinib vs sorafenib 1L. OS HR 0.92 (95% CI 0.79-1.06, non-inferiority met); mPFS 7.4 vs 3.7 months, ORR 24.1% vs 9.2%. First non-sora TKI in 1L after 10 years. mOS 13.6 vs 12.3 months.
CELESTIAL [PMID 29972759] (Abou-Alfa 2018 N Engl J Med, N=707): cabozantinib (MET/AXL/VEGFR2 TKI) vs placebo at ≤ 2 prior lines. mOS 10.2 vs 8.0 months (HR 0.76, p=0.005). MET/AXL mechanism addresses sora resistance, but G3-4 AE 68% vs 36% — clinically screen for good PS patients.
REACH-2 [PMID 30665869] (Zhu 2019 Lancet Oncol, N=292): ramucirumab vs placebo 2L, only AFP ≥ 400 ng/mL population. mOS 8.5 vs 7.3 months (HR 0.71, p=0.0199). HCC’s first biomarker-selected positive phase III — AFP-high biology driven by VEGFR2 pathway. Clinical threshold: AFP ≥ 400 to give ramu.
AHELP [PMID 33971141] (Qin 2021 Lancet Gastroenterol Hepatol, N=393, China 31 centers): apatinib = rivoceranib (VEGFR2 TKI) vs placebo 2L+, HBV ~82%. mOS 8.7 vs 6.8 months (HR 0.785, p=0.048). Validation of a Chinese domestically-developed VEGFR2 TKI in 2L — the same molecule became the 1L combination backbone of CARES-310 two years later. Classic Chinese pharma pathway: “2L validation → 1L combination.”

Takeaway: 2017-2021 2L landscape = choose among regorafenib (sora-tolerant) + cabozantinib (broad-spectrum) + ramucirumab (AFP≥400) + apatinib (NMPA-exclusive). 1L was pried open by REFLECT — but sora / lenva had similar ~13 months mOS. True 1L rewriting had to wait for IO.

2.3 IO + anti-angiogenic 1L rewriting (2020-2024): the four-way standoff

In 2020 IMbrave150 first sent IO + anti-angiogenic into 1L (atezo+bev vs sora), OS HR 0.58; in 2021 ORIENT-32 (domestic sintilimab + IBI305) reproduced HR 0.57 in an HBV ≥ 93% Chinese cohort; in 2022 HIMALAYA’s STRIDE regimen (single treme priming + durva Q4W) became the first 1L IO regimen without anti-angiogenic; in 2023 CARES-310 (camrelizumab + rivoceranib, China-led 13-country multicenter) pushed mOS to 22.1 months (longest in 1L HCC phase III history); in 2025 CheckMate-9DW’s nivo+ipi pushed further to 23.7 months but with early hazard crossover. Four years, 5 positive 1L phase III trials, HR converging from 0.58 to 0.79 — the IO backbone held up 1L in unstratified populations.

IMbrave150 (primary) [PMID 32402160] (Finn 2020 N Engl J Med, N=501): atezolizumab + bevacizumab (A+B regimen) vs sorafenib 1L. OS HR 0.58 (p<0.001). HCC’s first positive 1L IO combination phase III — 1L SoC switched tracks from this point.
IMbrave150 updated [PMID 34902530] (Cheng 2022 J Hepatol): 12-month follow-up update. mOS 19.2 vs 13.4 months (HR 0.66) — more mature OS data than the primary analysis. Became the cornerstone control for 2L IO trials.
ORIENT-32 [PMID 34143971] (Ren 2021 Lancet Oncol, N=571, China): sintilimab + IBI305 (bev biosimilar) vs sorafenib, HBV ≥ 93%. mOS not reached vs 10.4 months (HR 0.57). NMPA approved, not submitted to FDA, compressed drug cost to about 1/3-1/4 of atezo+bev. A phase III dedicated to HBV-enriched population.
HIMALAYA [PMID 38319892] (Abou-Alfa 2022 NEJM Evid, N=1171): STRIDE regimen = single tremelimumab (CTLA-4 mAb) 300 mg priming + durvalumab (durva, PD-L1 mAb) 1500 mg Q4W vs sorafenib. mOS 16.43 vs 13.77 months (HR 0.78, p=0.0035). Durva monotherapy vs sora non-inferior (HR 0.86). First 1L IO regimen without anti-angiogenic — a lifesaving option for patients with esophageal/gastric varices or bev contraindications. PD-L1 low subgroup benefited numerically more (reverse-biomarker phenomenon).
COSMIC-312 [PMID 35798016] (Kelley 2022 Lancet Oncol, N=837): cabozantinib + atezolizumab vs sorafenib 1L. PFS HR 0.63 (positive) but OS HR 0.90 (p=0.44, negative). Lesson: MET/AXL inhibition may actually weaken IO benefit, TKI backbones are not interchangeable — bev succeeds / cabo fails / lenva borderline (LEAP-002) despite all being TKI+IO.
LEAP-002 [PMID 38039993] (Llovet 2023 Lancet Oncol, N=794): pembrolizumab + lenvatinib vs lenvatinib + placebo. mOS 21.2 vs 19.0 months (p=0.023, did not meet prespecified threshold p ≤ 0.019). Dual primary endpoint failure — exposed that lenva monotherapy in modern control arms has been underestimated; the OS gap between 1L combo and lenva monotherapy narrowed to a borderline ~2 months.
RATIONALE-301 [PMID 37796513] (Qin 2023 JAMA Oncol, N=674): tislelizumab monotherapy vs sorafenib. Non-inferiority met (OS HR 0.85, 95.003% CI 0.71-1.02), mOS 15.9 vs 14.1 months. mDoR 36.1 vs 11.0 months (extremely beautiful long tail). FDA approved 2024. Second 1L IO regimen without anti-angiogenic (besides durva mono).
CARES-310 [PMID 37499670] (Qin 2023 Lancet, N=543, 13-country multicenter, HBV ~77%): camrelizumab + rivoceranib (= apatinib) vs sorafenib 1L. mOS 22.1 vs 15.2 months (HR 0.62, p<0.0001) — longest mOS in 1L HCC phase III history. Simplified regimen of all-oral TKI + biweekly IV IO (vs atezo+bev’s dual IV every 3 weeks). FDA formally approved 2024.
CheckMate-9DW [PMID 40349714] (Yau 2025 Lancet, N=668): nivolumab + ipilimumab (nivo+ipi) vs investigator’s choice lenva/sora 1L. mOS 23.7 vs 20.6 months (HR 0.79, p=0.018). Early hazard crossover — first 6 months HR 1.65 (worse), after that HR 0.61 (improved); 12 TRAE deaths vs 3 — dual IO requires 6-12 month functional reserve.

Takeaway: 2026 1L IO landscape = atezo+bev (IMbrave150 updated mOS 19.2) / STRIDE (HIMALAYA 16.4) / cam+rivo (CARES-310 22.1) / nivo+ipi (9DW 23.7) four-way standoff; domestic sinti+IBI305 (ORIENT-32) provides cost advantage for HBV Chinese cohorts; tislelizumab / durvalumab monotherapy as bev-contraindicated alternatives. But all positive results with HR 0.58-0.79 are in unstratified populations — no IO regimen has PD-L1 / TMB / MSI cut-off as stratification enrollment.

2.4 The “zero biomarker” precision dilemma (2020-2026): HCC’s unique reverse-biomarker phenomenon

HCC’s three main drivers — TERT promoter mutation (~60%), TP53 (~30%), CTNNB1 (~30%, β-catenin pathway) — are all undruggable. Sub-frequency MET amplification and FGF19 amplification have low prevalence (5-10% each) and phase III failed (COSMIC-312 cabo+atezo OS negative [PMID 35798016]). In the IO era, no 1L / 2L HCC trial enrolled by molecular stratification — zero predictive biomarkers approved to date. Even more anomalously, HIMALAYA’s STRIDE benefited numerically more in the PD-L1 low subgroup, suggesting HCC may be a “reverse biomarker” tumor type (PD-L1 ≠ response prediction).

HCC driver gene landscape: TERT promoter mutation 55-60% (undruggable, mechanism differs from coding-region mutations); TP53 30% (undruggable); CTNNB1 25-30% (β-catenin undruggable); total ~90% tumors carry at least one driver but all three are untargetable — this is the most fundamental molecular difference between HCC and NSCLC (EGFR 40% druggable) / BTC (FGFR2 ~15% druggable + IDH1 ~15% druggable).
COSMIC-312 [PMID 35798016] (Kelley 2022 Lancet Oncol): an attempt at MET/AXL as theoretical enhancer target. Cabo+atezo vs sora, PFS positive OS negative — MET overexpression enrollment not stratified. Mechanistic lesson: not all TKI backbones are suitable as IO partners, cabo/atezo failure suggests MET/AXL inhibition may weaken IO benefit.
LEAP-002 [PMID 38039993] (Llovet 2023 Lancet Oncol): pembro+lenva vs lenva. Not stratified by PD-L1 or VEGF pathway activity, dual primary endpoint failure.
HIMALAYA [PMID 38319892] (Abou-Alfa 2022 NEJM Evid) “reverse biomarker” phenomenon: subgroup analysis showed PD-L1 low-expression subgroup benefited more numerically from STRIDE, PD-L1 high-expression subgroup HR approached 1.0. Contrary to the biological logic of NSCLC / BTC (high PD-L1 responders). Not used as a stratification biomarker.
Soft constraints from etiology stratification: CheckMate-459 [PMID 34914889] and IMbrave150 subgroups suggest NASH / MASH-HCC may respond to IO worse than HBV/HCV-driven HCC. Mechanistic hypothesis (Pfister Nature 2021): CD8+PD-1+ T cell exhaustion patterns in NASH differ — but all phase IIIs only did descriptive subgroups, no head-to-head evidence.
AFP as pharmacodynamic biomarker rather than predictive biomarker: HCC has AFP dynamic monitoring as an early response indicator (ALBI / AFP change predicts PFS), but not a treatment selection biomarker — only REACH-2 (ramucirumab) uses AFP ≥ 400 as enrollment gate.

Takeaway: in 2026 the clinical utility of HCC molecular testing = extremely limited. NCCN V1.2026 still recommends testing HBV/HCV DNA + AFP + BCLC staging rather than a gene panel. The “biomarkers” that actually change clinical decisions are Child-Pugh (liver function) + AFP + etiology + tumor burden + ECOG PS — all clinical parameters. This contrasts with NSCLC’s mandatory EGFR/ALK/ROS1/KRAS/PD-L1 panel and BTC’s mandatory FGFR2/IDH1/HER2/MSI panel. The “zero biomarker dilemma” is HCC’s most urgent research agenda for 2026 (see §4 gaps 1-3).

2.5 Perioperative / adjuvant + TACE integration (2015-2026): early-readout reversal + intermediate BCLC-B paradigm shift

In 2015 STORM’s negative adjuvant sorafenib closed the door on TKI adjuvant, and HCC adjuvant remained blank for years; in 2022-2023 two single-center neoadjuvant IO phase II trials (Marron cemiplimab / Kaseb MDACC nivo±ipi) gave signals of 20% pathologic necrosis; in 2023 IMbrave050 (Qin Shukui PI), as HCC’s first adjuvant phase III, achieved positive RFS HR 0.72 — reversed in 2026 long-term follow-up, with RFS HR converging to 0.90 and OS HR 1.26 trending unfavorable; Roche withdrew the adjuvant indication in 2024. For intermediate BCLC-B, SPACE 2016 (DEB-TACE + sora) negative ended the first TKI+TACE wave; LAUNCH 2023 (lenva+TACE) China phase III positive; LEAP-012 (pembro+lenva+TACE) and EMERALD-1 (durva+bev+TACE) global phase III trials successively positive on PFS in 2024-2025 but OS still pending — intermediate HCC is turning the page from “TACE alone.”

STORM [PMID 26361969] (Bruix 2015 Lancet Oncol, N=1114): adjuvant sorafenib after curative resection vs placebo. mRFS 33.3 vs 33.7 months (HR 0.940, p=0.26, negative); G3-4 hand-foot skin reaction 28% vs <1%. Killed adjuvant TKI, HCC adjuvant treatment remained blank for years.
Marron neoadjuvant cemiplimab [PMID 35065058] (Marron 2022 Lancet Gastroenterol Hepatol, N=20): neoadjuvant cemiplimab (PD-1 monotherapy) phase II. 4/20 (20%) achieved > 70% pathologic necrosis. First published neoadjuvant IO monotherapy phase II.
Kaseb MDACC perioperative nivo±ipi [PMID 35065057] (Kaseb 2022 Lancet Gastroenterol Hepatol, N=27 RCT phase II): nivo mono ORR 23% vs nivo+ipi 0% (RECIST); MPR 23% vs 21%; G3-4 AE 23% vs 43%. Monotherapy not inferior to combination + lower toxicity.
IMbrave050 (primary) [PMID 37871608] (Qin 2023 Lancet, N=668): Qin Shukui PI, 12-month adjuvant atezo+bev after curative resection in high-risk patients vs active surveillance. RFS HR 0.72 (95% CI 0.53-0.98, p=0.012), positive at interim. HCC’s first positive adjuvant phase III, Grade 3-4 AE 41% vs 13%.
IMbrave050 updated [PMID 41580093] (Yopp 2026 J Hepatol): long-term follow-up reversed. RFS HR 0.90 (95% CI 0.72-1.12) benefit not sustained; OS HR 1.26 (95% CI 0.85-1.87) trending unfavorable. Roche withdrew the HCC adjuvant indication in 2024. Textbook-level “early readout ≠ final conclusion” case — HCC is a hot zone for OS reversals.
Kaseb biomarker analysis [PMID 39427654] (LaPelusa 2025 Oncology, N=18 translational): MPR responders showed post-treatment intratumoral CD8 +26.9%, granzyme B +15.6%, PD-1 +20.2%. Baseline tumor size + dynamic immune infiltration as candidate biomarkers — still research stage.
SPACE [PMID 26809111] (Lencioni 2016 J Hepatol, N=307): DEB-TACE + sora vs DEB-TACE + placebo in BCLC-B intermediate. mTTP 169 vs 166 days (HR 0.80, p=0.072, NS). Ended the first TKI+TACE wave — plain TKI addition provides no help to TACE.
LAUNCH [PMID 35921605] (Peng 2023 J Clin Oncol, N=338, China 12 centers): lenva + TACE vs lenva monotherapy, mostly BCLC-C / MVI (macrovascular invasion). mOS 17.8 vs 11.5 months (HR 0.45, p<0.001), mPFS 10.6 vs 6.4 months, ORR 54% vs 25%. Pre-IO-era TKI+TACE proof of concept — China 12-center led.
LEAP-012 [PMID 39798578] (Kudo 2025 Lancet, N=480, global): pembro + lenva + TACE vs TACE + dual placebo, Asian enrollment 72%. mPFS 14.6 vs 10.0 months (HR 0.66, p=0.0002); 24-month OS 75% vs 69% (HR 0.80, first OS interim did not reach significance). First positive IO+TKI+TACE phase III in intermediate HCC.
EMERALD-1 [PMID 39798579] (Sangro 2025 Lancet, N=616, global three-arm): durva + bev + TACE vs durva + TACE vs placebo + TACE. Triplet vs placebo mPFS 15.0 vs 8.2 months (HR 0.77, p=0.032); durva monotherapy + TACE not superior to placebo — mechanistic lesson: IO + anti-angiogenic combination is required, IO alone is not enough. OS still pending.

Takeaway: in 2026 HCC perioperative / adjuvant has no approved SoC (IMbrave050 withdrawal is the only such status among GI tumor types). Intermediate BCLC-B is turning from “TACE alone” to pembro+lenva+TACE (LEAP-012) or durva+bev+TACE (EMERALD-1), but the two flagships’ mature OS is still in follow-up — the IMbrave050 reversal cautionary tale tells us: do not write “practice-changing” before mature OS comes out.

3. Lateral: 2026 current decision landscape (six dimensions)

Projecting the longitudinal evolution onto concrete 2026 clinical decision trees, below are six key branchpoints and the evidence basis for each.

3.1 Child-Pugh A + BCLC-C advanced 1L: how to choose among the four-way standoff

2026 mainstream: four positive 1L IO combinations stand in parallel; choice is determined by bev accessibility + patient functional reserve + availability + cost.

Subgroup	First choice	Alternative
bev available (no severe EV, no active bleeding)	atezo+bev [IMbrave150 PMID 32402160 / updated PMID 34902530] (mOS 19.2 months, HR 0.66) or cam+rivo [CARES-310 PMID 37499670] (mOS 22.1 months, HR 0.62, all-oral TKI simplified regimen)	ORIENT-32 (HBV Chinese cohort, cost advantage)
bev contraindicated (EV bleeding / active GI bleeding / recent major surgery)	STRIDE durva+treme (HIMALAYA [PMID 38319892], mOS 16.43 months, HR 0.78, no anti-angiogenic)	tislelizumab mono (RATIONALE-301 [PMID 37796513], OS non-inferior HR 0.85)
Good functional reserve + 6-12 month life expectancy	nivo+ipi (CheckMate-9DW [PMID 40349714], mOS 23.7 months) — note the first 6 months HR 1.65 worsening risk	atezo+bev / cam+rivo
Borderline Child-Pugh B7 / rapid progression	Avoid dual IO and dual TKI+IO combinations	lenva monotherapy (REFLECT [PMID 29433850]) or BSC

Controversies: atezo+bev vs cam+rivo vs nivo+ipi have no head-to-head phase III. Cross-trial comparison shows CARES-310 mOS 22.1 is highest, but the control arm was sora (sora monotherapy is no longer mainstream in 2023); 9DW mOS 23.7 is higher but with dual-IO early mortality risk; atezo+bev is most mature (5-year real-world data). Choice is mainly determined by accessibility / reimbursement / prior bleeding history, not efficacy difference — because in the narrow HR 0.58-0.79 band there is no cross-trial distinguishing power.

NCCN V1.2026 Hepatobiliary: atezo+bev / STRIDE / nivo+ipi / cam+rivo are all Category 1 preferred; tislelizumab mono / durvalumab mono are Category 1 (bev-contraindicated alternatives).

3.2 Child-Pugh B / borderline liver function: can IO be used?

2026 mainstream: all 1L IO phase IIIs enrolled only Child-Pugh A. Evidence for Child-Pugh B patient IO = subgroup extrapolation + single-arm phase II + real-world retrospective.

CheckMate-040 Child-Pugh B expansion cohort (El-Khoueiry 2017 [PMID 28434648] original cohort was CP-A): subgroup analysis suggests nivo in CP-B has ORR ~12% (lower than CP-A 20%), mOS ~7.4 months. No phase III.
Real-world data: multinational retrospective shows CP-B atezo+bev mOS approximately 6-9 months vs CP-A 19 months, but bev-related bleeding risk is higher.
2026 clinical decisions: Child-Pugh B7-8 may try IO monotherapy (tislelizumab or durvalumab, avoiding bev + avoiding ipi); B9+ does not recommend any 1L IO, switch to BSC or lenva monotherapy (REFLECT CP-A evidence cannot be extrapolated).
Controversies: STORM trial [PMID 26361969] subgroup suggests sora in CP-B has far higher toxicity — TKI is equally unsafe in patients with poor liver function. Systemic therapy for Child-Pugh B remains a research gap in 2026 (see §4 gap 4).

3.3 Etiology stratification: HBV vs HCV vs non-viral (MASH) response to IO

2026 mainstream: etiology subgroups across the 4 1L IO phase III trials suggest differences in IO response profiles.

HBV-HCC (East Asia-dominant, > 80% of Chinese HCC): CARES-310, ORIENT-32, CheckMate-459 subgroup HRs are all most favorable (0.5-0.6 range). HBV mechanistic hypothesis: chronic viral-driven T cell infiltration + high baseline PD-1 expression; IO more effective at releasing exhaustion. HBV-HCC first choice is cam+rivo or sinti+IBI305 (cost advantage).
HCV-HCC (Europe/US / Japan-dominant): SHARP original population + IMbrave150 subgroup HR ~0.65-0.70. First choice atezo+bev or STRIDE.
Non-viral / MASH-HCC (rising year-over-year in Europe/US): CheckMate-459 [PMID 34914889] nivo vs sora subgroup HR approaches 1.0; IMbrave150 subgroup HR numerically slightly worse than HCV. Mechanistic hypothesis (Pfister Nature 2021 mechanistic study): in NASH-driven HCC, CD8+PD-1+ T cells are in auto-aggressive exhaustion mode, and IO may actually aggravate hepatic parenchymal damage — this is the biological explanation for “IO underperforms in MASH-HCC”.
2026 clinical decisions: IO selection for MASH population has the weakest evidence; avoid dual IO (nivo+ipi); prefer TKI+IO (atezo+bev / cam+rivo), with TKI monotherapy (lenva) if necessary. Phase III systemic evidence for non-viral HCC remains notably insufficient in 2026 — Western registry cohorts are accumulating.

3.4 Intermediate BCLC-B locoregional + IO integration: sequencing of TACE + IO combinations

2026 mainstream: LEAP-012 / EMERALD-1 two phase IIIs positive on PFS but mature OS still pending → don’t rush to rewrite as universal SoC before mature OS.

Subgroup	Regimen	Evidence
BCLC-B locoregional-dominant, good liver function	pembro + lenva + TACE [LEAP-012 PMID 39798578] (mPFS 14.6 months, HR 0.66) or durva + bev + TACE [EMERALD-1 PMID 39798579] (mPFS 15.0 months, HR 0.77)	phase III PFS positive, OS pending
MVI / locoregional-systemic compound risk	lenva + TACE [LAUNCH PMID 35921605] (mOS 17.8 months, HR 0.45)	Chinese phase III, covers MVI subgroup
EMERALD-1 lesson	durva monotherapy + TACE not superior to placebo + TACE	mechanism: IO + anti-angiogenic combination is required

Controversies: before mature OS, LEAP-012 / EMERALD-1 cannot fully replace traditional TACE. In 2026 clinical practice, combination regimens are commonly tried in high tumor burden / beyond-TACE-indication borderline / downstaging-intent patients; typical intermediate lesions still prioritize TACE. The double historical lesson of “IMbrave050 reversal + SPACE negative” makes the HCC field especially cautious about new intermediate-stage phase III results.

3.5 Adjuvant / perioperative: the void after IMbrave050 withdrawal

2026 mainstream: HCC has no approved adjuvant SoC — the only such status among GI tumor types.

Post-curative-resection high-risk patients (MVI / multifocal / > 5 cm / positive margins): atezo+bev adjuvant was used in 2023 [IMbrave050 primary PMID 37871608], Roche withdrew the indication in 2024 (based on updated PMID 41580093 RFS HR 0.90 + OS HR 1.26 reversal). 2026 standard = active surveillance.
Neoadjuvant IO monotherapy: Marron cemiplimab [PMID 35065058] 4/20 pathologic necrosis + Kaseb MDACC [PMID 35065057] nivo mono MPR 23% are early signals — none are SoC, only for eligible clinical trial enrollment.
Ongoing phase III adjuvant trials: EMERALD-2 (durva ± bev adjuvant), CheckMate-9DX (nivo adjuvant), KEYNOTE-937 (pembro adjuvant) — OS readout expected 2027-2029.
2026 clinical decisions: after curative resection do not routinely recommend adjuvant IO / adjuvant TKI (two negative lessons of STORM [PMID 26361969] + IMbrave050 updated); do not rewrite before EMERALD-2 / 9DX / KN-937 mature OS.

3.6 2L+ sequencing: the off-label dilemma in the post-IO era

2026 mainstream: RESORCE / CELESTIAL / REACH-2 / AHELP all tested in sora-progressor population; after 1L switched from sora to IO+anti-angiogenic, post-IO 2L has no positive phase III.

Scenario	Regimen	Evidence level
1L sora progression (historical cohort)	regorafenib (RESORCE [PMID 27932229], mOS 10.6 months, HR 0.63, requires sora-tolerant)	Category 1
1L sora progression + AFP ≥ 400	ramucirumab (REACH-2 [PMID 30665869], mOS 8.5 months, HR 0.71)	Category 1
1L sora progression + broad-spectrum coverage needed	cabozantinib (CELESTIAL [PMID 29972759], mOS 10.2 months, HR 0.76)	Category 1 (G3-4 AE 68%, screen PS)
1L IO+anti-angiogenic progression (2026 mainstream)	off-label lenva / cabo / regorafenib / rivoceranib (AHELP [PMID 33971141])	expert consensus, no phase III
KEYNOTE-394 Asian 2L pembro	post-sora pembro Asian confirmatory (mOS 14.6 months, HR 0.79)	regional confirmatory case for FDA-retained accelerated approval

Controversies: post-IO 2L in 2026 is HCC’s most urgent unmet need. ivonescimab (AK112, PD-1+VEGF bispecific) HCC phase II signals / cadonilimab (AK104, PD-1+CTLA-4 bispecific) COMPASSION-08 phase II data are already available, phase III readout window 2026-2027. Before then all clinical decisions rely on ESMO/NCCN expert consensus off-label.

4. Research Gaps: ten unsolved clinical problems

This report identifies the following gaps, all definable concrete problems (not the cliché “more research is needed”):

HCC has zero approved predictive biomarkers — mechanistic explanation missing: TERT / TP53 / CTNNB1 three main drivers all undruggable; PD-L1 / TMB / MSI failed to predict response in all 1L IO phase IIIs. The molecular mechanism of HCC’s uniqueness (promoter-mutation-dominant + immune cold + special intrahepatic microenvironment) needs systematic correlative science.
Is PD-L1 a “reverse biomarker” in HCC?: HIMALAYA STRIDE’s PD-L1 low subgroup benefited numerically more — contrary to the biological logic of high-PD-L1 response in NSCLC / BTC / gastric cancer. Prospective PD-L1 stratified design is needed for confirmation.
No head-to-head phase III among HCC IO combinations: atezo+bev / STRIDE / cam+rivo / nivo+ipi four 1L phase III trials fall within the narrow HR 0.58-0.79 band for cross-trial comparison. Clinical decisions rely on accessibility / toxicity profile / cost, not efficacy evidence — selection dilemma.
Prospective IO safety data for Child-Pugh B is missing: all 1L IO phase IIIs enrolled only CP-A. IO choice for CP-B7-9 = subgroup extrapolation + phase II + real-world retrospective — dedicated phase III needed.
Differential IO response in MASH / non-viral HCC: CheckMate-459 / IMbrave150 subgroups suggest weaker IO response in MASH-HCC (HR ~1.0); Pfister Nature 2021 mechanistic study provides the hypothesis but it has not been prospectively validated. Western registry cohort MASH proportion is rising → MASH-only RCT needed.
Mature OS and sequencing of TACE + IO combinations: LEAP-012 / EMERALD-1 PFS positive but OS still pending. The double historical lesson of SPACE (TACE+sora negative) + IMbrave050 updated (adjuvant reversal) — cannot rewrite as universal SoC before mature OS.
IMbrave050 adjuvant OS has not been positive to date: HCC is the only GI tumor type with no approved adjuvant SoC. EMERALD-2 / CheckMate-9DX / KEYNOTE-937 readouts 2027-2029 — if they continue negative, HCC adjuvant path needs to be rethought.
Complete absence of post-IO 2L phase III: RESORCE / CELESTIAL / REACH-2 all done in sora-progressors; 2L choice after 1L IO+anti-angiogenic is HCC’s most urgent research gap in 2026. ivonescimab / cadonilimab phase III are candidate breakthroughs.
Standardization of downstaging → surgery + AFP dynamic evaluation: Chinese multicenters are performing secondary surgery / transplantation after IO+TKI downstaging, but RCT evidence is insufficient; AFP dynamic change as treatment evaluation aid has correlative data but has not been formalized into a decision tool.
Value of next-generation IO (LAG-3 / TIGIT) + ctDNA-guided treatment decisions in HCC: NSCLC / melanoma already have LAG-3 / TIGIT phase III readouts; HCC has no large-scale phase III in 2026. ctDNA-guided adjuvant de-escalation has launched in other tumor types; HCC has not entered.

5. 2024-2026 latest developments

5.1 FDA / NMPA new approvals (HCC-relevant excerpts)

Drug	Agency	Date	Indication / supporting trial
atezolizumab + bevacizumab	FDA	2020-05-29	1L advanced HCC / IMbrave150 [PMID 32402160]
sintilimab + IBI305	NMPA	2021	1L advanced HCC (HBV-enriched) / ORIENT-32 [PMID 34143971]
durvalumab + tremelimumab STRIDE	FDA	2022-10-21	1L advanced HCC / HIMALAYA [PMID 38319892]
camrelizumab + rivoceranib	NMPA	2023; FDA	1L advanced HCC / CARES-310 [PMID 37499670]
tislelizumab mono	FDA	2024-03-13	1L advanced HCC (OS non-inferior to sora) / RATIONALE-301 [PMID 37796513]
nivolumab + ipilimumab (1L)	FDA	2025-04	1L advanced HCC / CheckMate-9DW [PMID 40349714]
atezolizumab + bevacizumab adjuvant indication	FDA / Roche	withdrawn 2024	based on IMbrave050 updated [PMID 41580093] RFS HR 0.90 + OS HR 1.26

Key observation: 2024-2025 was the “second approval wave” for HCC 1L IO — tislelizumab mono and nivo+ipi received 2 FDA approvals within 2 years; in the same period the withdrawal of the IMbrave050 adjuvant indication became the most educational event in HCC in 2024.

5.2 Key conference readouts (2024-2026, de-weighted notation)

The following entries are candidate pool only before formal peer review and do not enter the main database.

IMbrave050 updated OS [PMID 41580093] (Yopp 2026 J Hepatol): RFS HR 0.90 + OS HR 1.26 reversal. The most educational long-term follow-up event in HCC 2024-2026.
CheckMate-9DW mature OS (ASCO 2025 / ESMO 2025): mOS 23.7 months stable ([PMID 40349714] data); early hazard crossover still visible.
LEAP-012 OS first interim (ASCO GI 2025 [PMID 39798578]): 24-month OS 75% vs 69% (HR 0.80, did not reach significance, OS readouts limited by α spending).
EMERALD-1 first OS interim (ASCO GI 2025 [PMID 39798579]): triplet OS curves separated but did not reach the prespecified significance threshold.
ivonescimab (AK112, Akeso PD-1+VEGF bispecific) HCC phase II (2024-2025 conference data, full publication pending): RM HCC 2L signal, phase III HARMONi-6 ongoing.
cadonilimab (AK104, Akeso PD-1+CTLA-4 bispecific) COMPASSION-08 [PMID 37942328 not in hcc.yaml main database, as hypothesis-generating]: cadonilimab + lenva signal in HCC 1L.

5.3 Ongoing phase III (2025-2028 readout selection)

EMERALD-2 (NCT03847428): durva ± bev adjuvant vs placebo adjuvant after curative resection HCC — OS readout 2027-2028
CheckMate-9DX (NCT03383458): nivo adjuvant vs placebo after curative resection HCC — OS 2027-2028
KEYNOTE-937 (NCT03867084): pembro adjuvant vs placebo after curative resection HCC — OS 2027-2028
LEAP-012 mature OS + EMERALD-1 mature OS: whether intermediate BCLC-B IO+TKI+TACE can be rewritten as SoC depends on these two OS readouts in 2026-2027
HARMONi-6 (Akeso AK112 ivonescimab in HCC): phase III ongoing, readout window 2026-2028
post-IO 2L phase III: still no positive regimen in 2026; candidates include ivonescimab monotherapy / cadonilimab+lenva / next-generation TKIs

6. Convergent insights and judgments

6.1 Longitudinal × lateral: the 2026 HCC landscape shaped by three “resonances”

Overlaying the longitudinal paradigm evolution on the lateral current decision landscape, the 2026 HCC landscape is a superposition of three resonances:

The unique evolution “sorafenib solo for 10 years (2008-2017) → 4 IO combinations as 1L SoC (2020-2024) → zero biomarker dilemma (to date)”: HCC’s 1L mOS was pushed from 10.7 months to 22-24 months (CARES-310 22.1, 9DW 23.7) over 17 years, driven by the IO backbone rather than molecular stratification. Compared to NSCLC’s EGFR/ALK/ROS1/KRAS/MET/HER2/RET/BRAF/NTRK/PD-L1 dozens of molecular stratification paths + parallel IO, and BTC’s 9 approved biomarkers in FGFR2/IDH1/HER2/BRAF/NTRK/RET/MSI/TMB/claudin18.2 — HCC walked its own path of “no molecular stratification, with IO combinations + clinical parameters (Child-Pugh/AFP/etiology) holding up the entire field.” The ceiling of this path is evident: four IO regimens cluster in the narrow HR 0.58-0.79 band, further breakthroughs require mechanistic innovation, not adding another of the same class.
The HCC-specific “Chinese data leads globally”: among 42 landmark trials, CARES-310 (China-led 13-country), ORIENT-32 (all-China), LAUNCH (all-China), SoraHAIC (all-China), FOHAIC-1 (all-China), AHELP (all-China), Chen 2006 (Sun Yat-sen University), Huang 2010 (West China), IMbrave050 (Qin Shukui PI) were all led by Chinese / Asia-Pacific PIs — making HCC one of the tumor types with the strongest Chinese data contribution in medical oncology. Root cause: > 50% of new HCC cases globally are in Asia-Pacific + China’s high HBV prevalence — clinical research is necessarily China-led. Comes with dividends (detailed East Asian population data + optimal HBV subgroup HR + domestic IO cost advantage) and challenges (scarce MASH-HCC evidence + weaker Western applicability).
“Early-readout reversal” is especially frequent in HCC: IMbrave050 primary [PMID 37871608] RFS HR 0.72 positive → updated [PMID 41580093] RFS HR 0.90 + OS HR 1.26 reversal + Roche indication withdrawal — one of the most educational long-term follow-up events in oncology in 2024. The three histories of SPACE (2016 TKI+TACE negative) + LEAP-002 (2023 borderline miss) + IMbrave050 (2024 reversal) tell us: HCC is a hot zone for OS reversals, writing “practice-changing” before mature OS often backfires. In 2026 we should be especially cautious about mature OS from LEAP-012 / EMERALD-1.

These three resonances together explain a clinical phenomenon: the 1L decision for a newly diagnosed stage IV HCC patient in 2026 has 1-2 more decision layers than in 2016 (bev accessibility / functional reserve / etiology), but the decision tree itself is “narrow in width + shallow in depth” — four IO combinations cluster in the narrow HR band, no molecular panel needed. This differs completely from NSCLC’s multi-layer decision tree (driver panel → PD-L1 → combo) and BTC’s four biomarker-panel paths (FGFR / IDH / HER2 / MSI). HCC decision tree’s “narrow and shallow” feature is the clinical manifestation of the “zero biomarker dilemma.”

6.2 Clinical decision takeaways (for junior-mid oncologists)

1L decisions look at 3 clinical parameters, not a molecular panel: can bev be used? (yes → atezo+bev or cam+rivo; no → STRIDE or tisle/durva mono) + 6-12 month functional reserve? (yes → nivo+ipi an option; borderline → avoid dual IO) + economic accessibility? (HBV Chinese patients → ORIENT-32 cost 1/3-1/4). PD-L1 / TMB / MSI almost never need testing — zero predictive biomarkers approved.
Child-Pugh is the core gate for HCC treatment decisions: all 1L IO phase IIIs enrolled only CP-A. CP-B7-9 may try IO monotherapy (tislelizumab / durvalumab, avoiding bev + ipi); CP-B9+ or CP-C switch to BSC. Do not extrapolate CP-A data to CP-B/C.
Evidence for IO effect in MASH / non-viral HCC is weak: CheckMate-459 subgroup HR approaches 1.0. MASH population first choice TKI+IO (atezo+bev / cam+rivo), avoid dual IO; lenva monotherapy if necessary. Western registry cohort MASH proportion rising → continue to watch for MASH-only RCTs after 2026.
Adjuvant therapy has no SoC in 2026 — active surveillance is the standard: after IMbrave050 withdrawal, HCC is the only GI tumor type with no approved adjuvant regimen. Lessons from STORM (sora adjuvant negative) + IMbrave050 updated (IO adjuvant reversal): do not rewrite before mature OS. Do not give adjuvant off-trial before EMERALD-2 / 9DX / KN-937 readouts.
Don’t rush to fully accept IO+TACE combinations for intermediate BCLC-B: LEAP-012 / EMERALD-1 PFS positive but OS pending; SPACE / IMbrave050 history reminds us that HCC is a reversal hot zone. Typical intermediate lesions still prioritize TACE; try combinations when tumor burden is high / downstaging intent / clinical trial enrollment is possible.
Post-IO 2L has no phase III — off-label relies on expert consensus: after 1L IO+anti-angiogenic progression, lenva / cabo / regorafenib / rivoceranib are all off-label; prioritize clinical trial enrollment (ivonescimab / cadonilimab phase III).
Remember ramucirumab (REACH-2) for AFP ≥ 400 ng/mL + 2L: HCC’s only biomarker-selected positive phase III, easily overlooked in post-IO scenarios.
Sora-progressor is still an evidence-based scenario (but increasingly rare): RESORCE (regorafenib, requires sora-tolerant) remains Category 1; older patients from the 1L sora era can be considered for 2L. In the post-IO era, the mainstream path is 1L IO+anti-angiogenic → 2L off-label.
The first-6-months hazard crossover in CheckMate-9DW is an important clinical cue: nivo+ipi has the highest mOS (23.7 months) but first-6-month HR 1.65 + 12 TRAE deaths vs 3. Borderline Child-Pugh / rapid progression / elderly patients should not choose dual IO — similar selection logic to NSCLC CheckMate-227 / 9LA.
Do not memorize PD-L1 thresholds / TMB cut-offs for HCC decisions: HCC is the only “reverse biomarker” major tumor type (HIMALAYA PD-L1 low subgroup benefited numerically more). In 2026 clinical practice, HCC molecular panel testing is done only in rare research / trial enrollment scenarios. Core stratification is always Child-Pugh × AFP × etiology × BCLC × ECOG × bev accessibility.

7. Information sources

The metadata of the 42 trials in this report were independently verified via two paths: PubMed and ClinicalTrials.gov. Each [PMID xxxxxxxx] in the text can be verified directly on PubMed.

Published trials: 42, covering 2006-2026 (PMID verifiable)
NCCN guideline citations: 42/42 (100%) hit the NCCN Hepatobiliary V1.2026 reference section
2020-2025 FDA / NMPA new approvals: 6 key approvals (atezo+bev / STRIDE / tislelizumab mono / cam+rivo / sinti+IBI305 / nivo+ipi) + 1 withdrawal (atezo+bev adjuvant indication)
2024-2026 key conference / long-term follow-up readouts: 5 (IMbrave050 updated reversal / 9DW mature OS / LEAP-012 OS interim / EMERALD-1 OS interim / ivonescimab HCC phase II)
Research gaps: 10
Chinese investigator-led ratio: > 40% (CARES-310 / ORIENT-32 / LAUNCH / SoraHAIC / FOHAIC-1 / AHELP / Chen 2006 / Huang 2010 / IMbrave050 / RATIONALE-301)

7.1 Citation list of main report text (ascending PMID)

The table below lists PMIDs cited in bracket form in the main report text; each can be clicked to verify on PubMed.

PMID	Trial / Paper	Year	Journal	Text location §x.x
16495695	Chen 2006 RFA vs resection	2006	Ann Surg	§appendix (in hcc.yaml, not directly cited in text)
18650514	SHARP	2008	NEJM	§2.1 sorafenib solo
19095497	Asia-Pacific sorafenib (Oriental)	2009	Lancet Oncol	§2.1
21107100	Huang 2010 RFA vs resection	2010	Ann Surg	§appendix (in hcc.yaml)
23980084	BRISK-FL	2013	J Clin Oncol	§2.1
25488963	LiGHT (linifanib)	2015	J Clin Oncol	§2.1
26361969	STORM	2015	Lancet Oncol	§2.5 / §3.5 / §6.2
26809111	SPACE	2016	J Hepatol	§2.5 / §6.1
27821083	SIRveNIB protocol	2016	BMC Cancer	§appendix (in hcc.yaml)
27932229	RESORCE	2017	Lancet	§2.2 / §3.6
28434648	CheckMate-040 nivo mono cohort	2017	Lancet	§3.2
29107679	SARAH	2017	Lancet Oncol	§appendix (in hcc.yaml)
29433850	REFLECT	2018	Lancet	§2.2 / §3.1
29875066	KEYNOTE-224	2018	Lancet Oncol	§appendix (in hcc.yaml)
29972759	CELESTIAL	2018	NEJM	§2.2 / §3.6
30665869	REACH-2	2019	Lancet Oncol	§2.2 / §3.6
31070690	SoraHAIC	2019	JAMA Oncol	§6.1 (China-led)
31790344	KEYNOTE-240	2020	J Clin Oncol	§appendix (in hcc.yaml)
32402160	IMbrave150 primary	2020	NEJM	§2.3 / §3.1 / §5.1
33001135	CheckMate-040 nivo+ipi cohort	2020	JAMA Oncol	§appendix (in hcc.yaml)
33087333	RESCUE	2021	Clin Cancer Res	§appendix (in hcc.yaml)
33971141	AHELP	2021	Lancet Gastroenterol Hepatol	§2.2 / §3.6
34143971	ORIENT-32	2021	Lancet Oncol	§2.3 / §3.1 / §3.3
34902530	IMbrave150 updated	2022	J Hepatol	§2.3 / §3.1
34905388	FOHAIC-1	2022	J Clin Oncol	§6.1 (China-led)
34914889	CheckMate-459	2022	Lancet Oncol	§2.4 / §3.3
35065057	Kaseb MDACC perioperative	2022	Lancet Gastroenterol Hepatol	§2.5 / §3.5
35065058	Marron cemiplimab neoadjuvant	2022	Lancet Gastroenterol Hepatol	§2.5 / §3.5
35798016	COSMIC-312	2022	Lancet Oncol	§2.3 / §2.4
35921605	LAUNCH	2023	J Clin Oncol	§2.5 / §3.4
35949295	SURF trial	2022	Liver Cancer	§appendix (in hcc.yaml)
36455168	KEYNOTE-394	2023	J Clin Oncol	§3.6
37499670	CARES-310	2023	Lancet	§2.3 / §3.1 / §5.1
37796513	RATIONALE-301	2023	JAMA Oncol	§2.3 / §3.1 / §5.1
37871608	IMbrave050 primary	2023	Lancet	§2.5 / §3.5 / §6.2
38039993	LEAP-002	2023	Lancet Oncol	§2.3 / §2.4 / §6.2
38319892	HIMALAYA STRIDE	2022	NEJM Evid	§2.3 / §2.4 / §3.1 / §5.1
39427654	Kaseb biomarker analysis	2025	Oncology	§2.5
39798578	LEAP-012	2025	Lancet	§2.5 / §3.4 / §5.2
39798579	EMERALD-1	2025	Lancet	§2.5 / §3.4 / §5.2
40349714	CheckMate-9DW	2025	Lancet	§2.3 / §3.1 / §5.1 / §5.2
41580093	IMbrave050 updated	2026	J Hepatol	§2.5 / §3.5 / §5.1 / §6.2

7.2 Verification conventions

Each PMID can be accessed for verification directly via https://pubmed.ncbi.nlm.nih.gov/{PMID}/
Each NCT id can be accessed via https://clinicaltrials.gov/study/{NCT_id}/
Conference abstracts (ASCO / ASCO GI / ESMO) searched via official conference systems; all conference citations in this report are “de-weighted notation” — un-peer-reviewed toplines defer to journal publication
If the trial name / year / conclusion corresponding to a PMID in this report is found inconsistent with PubMed, corrections are welcome

Clinical trial timeline is here

Chinese: /trials/hcc/ English: /en/trials/hcc/

Each trial has a standalone detail page, including:

Complete intervention / comparator regimen
Primary endpoint values + 95% CI
Key findings + clinical significance
Clickable links to PMID / NCT originals

42 trials · 5 chapters · 2006 to 2026 · Chinese PI contribution > 40% · synchronized with NCCN Hepatobiliary V1.2026.

Closing

HCC completed a unique arc in oncology over the past 18 years — from SHARP sorafenib in 2008 ending the “no drugs for HCC” era, a decade of sorafenib solo dominance; to 2017-2019 REFLECT / RESORCE / CELESTIAL / REACH-2 / AHELP breaking through 1L non-inferiority + 2L three-way; to 2020-2024 IMbrave150 / HIMALAYA / CARES-310 / CheckMate-9DW four IO combinations pushing 1L mOS to 22-24 months (four-way standoff takes shape); to 2023-2026 the textbook lesson of IMbrave050 adjuvant going from positive to reversed + LEAP-012 / EMERALD-1 as intermediate BCLC-B’s first IO+TACE positives but with mature OS still pending.

HCC’s most fundamental difference from other major tumor types (NSCLC / BTC / PDAC) is not treatment complexity but the unique dilemma of “zero approved predictive biomarkers”. NSCLC has 10+ biomarkers for stratification, BTC has 9 biomarkers matching targets, PDAC just pried open three subtypes KRAS G12C/G12D/pan-KRAS — HCC’s three main drivers TERT/TP53/CTNNB1 remain fully undruggable, and none of PD-L1/TMB/MSI predicts IO response. Clinical stratification forever relies on Child-Pugh × AFP × etiology × BCLC × tumor burden. This path of “holding up the entire field with the IO backbone in unstratified populations + deciding by clinical parameters rather than molecular panels” is HCC’s unique contribution to oncology, but also the bottleneck most needing breakthrough in 2026.

Perioperative IMbrave050 reversal + intermediate LEAP-012 / EMERALD-1 OS pending + post-IO 2L complete absence of phase III — these three areas are HCC’s most densely packed research gaps in 2026. The next decade needs to solve the three structural problems: “can we find HCC’s first predictive biomarker”, “can we push adjuvant IO from RFS positive to OS positive”, and “can we push post-IO 2L from expert consensus to phase III evidence”.

The value of this report is not in “exhaustively listing all trials” (PubMed can do that), but in compressing 18 years of evolution + current decisions + unsolved gaps into the cognitive bandwidth of a single read. Next time you face a newly diagnosed HCC patient, every branch in the decision tree has this map to reference, trace, and query.

Clinician × AI = Research Superpower + Clinical Decision Amplifier

—— Dual Brain Lab · 2026-04-21

Nasopharyngeal Carcinoma Clinical Trial Timeline: A 28-Year Evolution Map

Tue, 21 Apr 2026 00:00:00 +0000

Nasopharyngeal Carcinoma Clinical Trial Timeline — In-depth Report

Coverage: 23 landmark trials cited in NCCN Cancer of the Nasopharynx V1.2026 (all PMID-traceable) + EBV-driven biology + Chinese-investigator-led research ecosystem

Curated by Dual Brain Lab (csilab.net)

1. One-sentence definition

This report maps the evolutionary logic and current decision landscape of systemic therapy for nasopharyngeal carcinoma (NPC) over the past 28 years (1998-2026), covering the landmark clinical trials cited in NCCN Cancer of the Nasopharynx V1.2026, to give frontline clinicians in 2026 a traceable panorama for the “who, what, why” of treatment decisions.

Iron rule: every data point for every trial is traceable to PubMed (PMID) or ClinicalTrials.gov (NCT id) — every [PMID xxxxxxxx] in the text links directly to the PubMed source for verification.

2. Longitudinal: the evolution timeline of four treatment paradigms

NPC systemic therapy has undergone four paradigm shifts over 28 years: locally advanced disease moved from radiotherapy alone to CCRT (concurrent chemoradiotherapy) and then to induction chemotherapy layered on top → the adjuvant setting, negative for a decade, was cracked open by metronomic chemotherapy → recurrent/metastatic (R/M) 1L moved from PF chemotherapy to a GP backbone, then was rewritten by the three-arrow convergence of three Chinese PD-1 agents on GP → R/M 2L single-agent IO became a negative-space cautionary tale after KEYNOTE-122 phase III failed.

Each shift rests on 1-3 phase III trials as pivots. The biggest structural difference between NPC and NSCLC / BTC is that “Chinese-investigator leadership” runs through all 28 years — of the 23 landmark trials, 17 (74%) were conducted by teams from mainland China, Hong Kong, Taiwan, or Singapore. And three domestic Chinese PD-1 agents, in the same narrow window (2021-2023), each independently delivered a PFS HR converging in the tight 0.52-0.54 band class effect that directly rewrote global NCCN. This “geographic ecology × drug ecology” synchronous explosion is unique among cancer types.

2.1 Definitive treatment of locally advanced disease (1998-2021): CCRT foundation → regional confirmation → induction layered on → radiosensitizer fine-tuning

Story: in 1998, INT-0099 rewrote the North American NPC standard from radiotherapy alone to CCRT; in 2003-2006, three Asian endemic-region trials (Taiwan LIN 2003 / Hong Kong NPC-9901 + 9902 / Singapore WEE 2005) independently confirmed it in the endemic non-keratinizing / undifferentiated histology population; in 2016-2019, Sun Yat-sen University Cancer Center (SYSUCC) ran two phase III trials that layered TPF (docetaxel / cisplatin / 5-FU) and GP (gemcitabine / cisplatin) induction chemotherapy onto CCRT; in 2018-2021, two phase III trials completed the fine-tuning of concurrent radiosensitizer dosing — non-inferiority of nedaplatin replacing cisplatin, and non-inferiority of cisplatin q3w vs weekly dosing.

INT-0099 [PMID 9552031] (Al-Sarraf 1998 JCO, N=147, North American Intergroup): stage III-IV non-metastatic NPC, concurrent cisplatin 100 mg/m² q3w ×3 + radiotherapy 70 Gy followed by adjuvant cisplatin + 5-FU ×3 vs radiotherapy alone. 3-year PFS 69% vs 24%, 3-year OS 78% vs 47% (both p<0.001). Terminated early at interim analysis for efficacy — first demonstration that CCRT roughly doubles PFS / OS in locally advanced NPC. The North American keratinizing population limited direct extrapolation to endemic regions, which is what triggered the three subsequent Asian confirmation trials.
LIN-2003 [PMID 12586799] (Lin JC 2003 JCO, N=284, Taiwan): stage III-IV M0 NPC (predominantly WHO II/III non-keratinizing), concurrent cisplatin 20 mg/m²/d + 5-FU 400 mg/m²/d 96h continuous infusion (weeks 1 and 5 of radiotherapy) vs radiotherapy alone. 5-year OS 72.3% vs 54.2% (p=0.0022), 5-year PFS 71.6% vs 53.0% (p=0.0012). First replication of INT-0099 in an East Asian endemic population, using a lower-intensity concurrent regimen — proving the CCRT benefit is robust to dosing details.
NPC-9901 [PMID 16192584] (Lee AW 2005 JCO, N=348, Hong Kong): T1-4 N2-3 M0 non-keratinizing / undifferentiated NPC, concurrent cisplatin 100 mg/m² q3w ×3 + radiotherapy then adjuvant cisplatin + 5-FU ×3 vs radiotherapy alone. 3-year FFS 72% vs 62% (p=0.027), locoregional control 92% vs 82% (p=0.005); 3-year OS identical early on (both 78%, salvage therapy effective + short follow-up), but OS benefit emerged at 5 and 10 years. The largest Asian CCRT RCT at the time, and it taught the field that “NPC OS benefits need extended observation + locoregional control is a valid surrogate.”
WEE-2005 [PMID 16170180] (Wee J 2005 JCO, N=221, Singapore): locally advanced NPC (majority ethnic Chinese, non-keratinizing), concurrent cisplatin 25 mg/m²/d d1-4 (weeks 1/4/7) + radiotherapy 70 Gy followed by adjuvant cisplatin + 5-FU ×3 vs radiotherapy alone. DFS HR 0.57 (95% CI 0.38-0.87, p=0.0093), OS HR 0.51 (95% CI 0.31-0.81, p=0.0061); 2-year distant metastasis cumulative incidence difference 17% (p=0.0029). Among the three Asian confirmation trials, the strongest signal for CCRT’s effect on distant control.
NPC-9902 [PMID 16904519] (Lee AW 2006 IJROBP, N=189, Hong Kong): T3-4 N0-1 M0 NPC, 2×2 factorial design — accelerated fractionation (AF) ± concurrent cisplatin/5-FU vs conventional fractionation (CF). AF + chemotherapy 3-year FFS 94% vs CF alone 70% (p=0.008), but AF alone or CCRT alone vs CF were not significant. Suggested that the NPC-9901 CCRT benefit is less universal in limited nodal disease (T3-4 N0-1) than in N2-3, and that AF + CCRT layering is needed.
SUN-2016 [PMID 27686945] (Sun Y 2016 Lancet Oncol, N=480, Sun Yat-sen University + multicenter): stage III-IVB M0 NPC (IMRT era), induction TPF × 3 + CCRT vs CCRT alone. 3-year FFS 80% vs 72%, HR 0.68 (95% CI 0.48-0.97, p=0.034). First trial to write TPF induction into the NPC standard in the IMRT era, led by the Ma Jun / Sun Ying group at SYSUCC. Benefit was driven primarily by distant metastasis control, consistent with the hypothesis that NPC is an “early systemic micrometastasis” cancer.
ZHANG-2019 [PMID 31150573] (Zhang Y 2019 NEJM, N=480, Sun Yat-sen University + Jiangxi): stage III-IVB M0 NPC, induction GP × 3 + CCRT vs CCRT alone. 3-year RFS 85.3% vs 76.5%, HR 0.51 (95% CI 0.34-0.77, p=0.001); 3-year OS 94.6% vs 90.3%, HR 0.43 (95% CI 0.24-0.77). First induction chemotherapy trial in NPC to show OS benefit, and GP three-cycle completion rate of 96.7% far exceeded historical TPF values. After NEJM publication, global induction backbone preference shifted toward GP.
TANG-2018 [PMID 29501366] (Tang LQ 2018 Lancet Oncol, N=402, Sun Yat-sen University): stage II-IVB M0 NPC, concurrent nedaplatin 100 mg/m² q3w ×3 vs cisplatin 100 mg/m² q3w ×3 (IMRT 70 Gy). 2-year PFS 88.0% vs 89.9% (non-inferiority margin met, p=0.0048); nedaplatin G3-4 nausea 2% vs 9%, vomiting 6% vs 18%, anorexia 13% vs 27%, and milder late ototoxicity. The evidence-based alternative for patients intolerant to cisplatin.
XIA-2021 [PMID 34083231] (Xia WX 2021 Clin Cancer Res, N=510, three Chinese centers): locally advanced NPC, concurrent cisplatin q3w ×2 vs weekly 40 mg/m² ×6 + IMRT. 3-year FFS 85.4% vs 85.6% (non-inferiority met, p=0.0016); q3w arm G3-4 leukopenia 16% vs 27%, late G3-4 hearing loss 6% vs 13%. The q3w schedule became mainstream in the IMRT era due to outpatient convenience + hearing preservation.

Takeaway: the 2026 standard of care for locally advanced NPC = induction GP × 3 (ZHANG-2019) or TPF × 3 (SUN-2016) → concurrent cisplatin q3w ×2 (XIA-2021 supporting) + IMRT 70 Gy; cisplatin-intolerant patients switch to nedaplatin (TANG-2018). CCRT’s universality was established cumulatively by four phase IIIs in 1998-2006, induction chemotherapy was layered on by two phase IIIs in 2016-2019, and radiosensitizer fine-tuning was completed by two phase IIIs in 2018-2021. The 28-year trajectory is clear; the remaining controversy is the absence of a head-to-head between TPF and GP induction (see §4 gap 2).

2.2 Adjuvant / consolidation therapy (2012-2021): a decade-long void, then a metronomic breakthrough

Story: in 2012, CHEN-L added traditional PF adjuvant on top of CCRT — negative; in 2018, CHAN used EBV-DNA to select high-risk patients post-treatment and gave adjuvant GC (gemcitabine + cisplatin) — still negative. For a decade, no effective adjuvant regimen could be found. In 2021, CHEN-YP changed the approach — not a higher dose, but a different dosing model: low-dose continuous capecitabine for 1 year (metronomic), with HR 0.50 pushing 3-year FFS up by 10 percentage points. A twelve-year void filled by a single conceptual shift.

CHEN-L-2012 [PMID 22154591] (Chen L 2012 Lancet Oncol, N=508, Sun Yat-sen University): stage III-IVB M0 locally advanced NPC, CCRT followed by adjuvant cisplatin + 5-FU (PF) × 3 vs CCRT alone. 2-year FFS 86% vs 84%, HR 0.74 (95% CI 0.49-1.10, p=0.13) negative. First large Chinese RCT to formally challenge the adjuvant arm of the INT-0099 three-part regimen (CCRT + adjuvant PF) — “after adequate CCRT, routine adjuvant PF adds nothing.” The negative result cleared the path for subsequent alternative adjuvant strategies (metronomic / maintenance / EBV-guided).
CHAN-2018 [PMID 29989858] (Chan ATC 2018 JCO, N=104 randomized from 789 screened, Hong Kong local version of NRG-HN001): stage IIB-IVB NPC with detectable plasma EBV-DNA 6-8 weeks post-radiotherapy (molecular residual disease), adjuvant cisplatin + gemcitabine × 6 vs observation. 5-year RFS 49.3% vs 54.7%, HR 1.09 (95% CI 0.63-1.89, p=0.75) negative. The hypothesis that EBV-DNA could serve as a molecular enrichment biomarker to identify a high-risk subgroup was confirmed (screening positivity rate 27.4%), but adjuvant GC failed to improve outcomes in that subgroup — molecular selection ≠ adjuvant benefit.
CHEN-YP-2021 [PMID 34111416] (Chen YP 2021 Lancet, N=406, Sun Yat-sen University multicenter): high-risk locally advanced NPC (stage III-IVB, excluding T3-4N0 / T3N1), CCRT followed by metronomic capecitabine 650 mg/m² bid × 1 year vs observation. 3-year FFS 85.3% vs 75.7%, HR 0.50 (95% CI 0.32-0.79, p=0.0023); G3 hand-foot syndrome was the dominant adverse event (17% vs 6%), no treatment-related deaths. The only positive phase III in a decade of adjuvant void. The biological rationale: low-dose, long-duration continuous anti-angiogenic + immune microenvironment modulation, rather than acute cytotoxicity — fundamentally different from the bolus dosing of PF / GC.

Takeaway: the 2026 adjuvant regimen for locally advanced NPC = CCRT (± induction GP/TPF) followed by metronomic capecitabine × 1 year (CHEN-YP-2021, already in NCCN). Traditional PF / GC adjuvant adds nothing (CHEN-L-2012 / CHAN-2018). EBV-DNA can identify high-risk patients, but the “molecular selection + traditional adjuvant” combination still fails — the future direction is EBV-DNA-guided + IO adjuvant (see §4 gap 1).

2.3 R/M 1L (2016-2023): GP backbone → three-arrow PD-1 convergence rewrites the global standard

Story: in 2016, ZHANG-2016-GP switched the R/M NPC 1L chemotherapy backbone from PF to GP, becoming the control cornerstone for all subsequent chemo-IO combinations. In 2021-2023, three domestic Chinese PD-1 monoclonal antibodies (camrelizumab / Hengrui; toripalimab / Junshi; tislelizumab / BeiGene) each independently completed phase III: three different drugs, three different companies, three different dosing schedules, the same GP backbone, the same R/M NPC 1L setting — PFS HR converging in the tight 0.52-0.54 band. This is a textbook class effect. All three arrows are included in NCCN Cancer of the Nasopharynx V1.2026.

ZHANG-2016-GP [PMID 27567279] (Zhang L 2016 Lancet, N=362, 22 Chinese centers): treatment-naive R/M NPC, gemcitabine 1000 mg/m² d1+8 + cisplatin 80 mg/m² d1 q3w × 6 vs fluorouracil + cisplatin (PF). PFS HR 0.55 (95% CI 0.44-0.68, p<0.0001), mPFS 7.0 vs 5.6 months. GP replaced PF as the global 1L chemotherapy backbone for R/M NPC. All subsequent chemo-IO three-arrow phase IIIs used placebo + GP as the control.
CAPTAIN-1ST [PMID 34174189] (Yang Y 2021 Lancet Oncol, N=263, 29 Chinese centers, Hengrui): treatment-naive R/M NPC, camrelizumab + GP × 4-6 cycles → camrelizumab maintenance vs placebo + GP. PFS HR 0.54 (95% CI 0.39-0.76, one-sided p=0.0002), mPFS 9.7 vs 6.9 months. The second readout among the three arrows.
JUPITER-02 [PMID 38015220] (Mai HQ 2023 JAMA, final OS report; interim Nat Med 2021 PMID 34341578; N=289, 95% of patients from mainland China / Hong Kong-Taiwan / Singapore, Junshi): treatment-naive R/M NPC, toripalimab + GP × 6 → toripalimab maintenance ≤ 2 years vs placebo + GP. PFS HR 0.52 (95% CI 0.37-0.73), mPFS 21.4 vs 8.2 months (final report); OS HR 0.63 (95% CI 0.45-0.89, p=0.008). The only one of the three arrows to achieve a positive mature OS readout at HR 0.63, and the first phase III globally to demonstrate OS benefit from chemo-IO combination in R/M NPC 1L. FDA approval on 2023-10 made toripalimab + GP the first 1L chemo-IO combination available for NPC in the US market.
RATIONALE-309 [PMID 37207654] (Yang Y 2023 Cancer Cell, N=263, multicenter China, BeiGene): treatment-naive R/M NPC, tislelizumab + GP × 4-6 → tislelizumab maintenance vs placebo + GP. PFS HR 0.52 (95% CI 0.38-0.73, p<0.0001), PFS benefit independent of PD-L1 expression. Gene expression profiling identified an activated dendritic cell signature correlated with benefit. Third arrow completed.

Takeaway: 2026 R/M NPC 1L = PD-1 (any of: toripalimab / camrelizumab / tislelizumab) + GP × 4-6 → PD-1 maintenance ≤ 2 years. The tight HR 0.52-0.54 convergence is textbook-level class effect evidence — the choice of PD-1 agent is driven primarily by NRDL inclusion / accessibility / safety profile, not by efficacy, for which no differential evidence exists. Dual PD-1 vs GP phase III trials will no longer be conducted (ethical threshold crossed); the clinical branchpoint is migrating to “PD-1 maintenance duration” (2 years vs until progression) and “PD-L1 / EBV-DNA stratification” (see §4 gap 6).

2.4 R/M 2L (2017-2023): single-agent IO early signals converge → KEYNOTE-122 phase III fails

Story: in 2017-2018, Western trials KEYNOTE-028 (pembro PD-L1+ cohort) and NCI-9742 (nivo) provided early single-agent IO signals at ORR 21-26%; in parallel, Chinese trials FANG-2018 (camrelizumab ORR 34%), POLARIS-02 (toripalimab ORR 21%, N=190 — the largest 2L single-arm study), and KL-A167 (a PD-L1 antibody, ORR 27%) stacked domestic single-agent data; in Singapore, LIM-2023 (nivo + ipi dual IO) achieved ORR 38%, suggesting combinations might outperform single agents. Then KEYNOTE-122 (Chua 2023 Ann Oncol) — the only 2L phase III — pembro vs chemotherapy OS HR 0.90, negative. This failure closed the “replace chemotherapy with IO” pathway, and retrospectively validated the three arrows’ choice of “add IO on top of chemotherapy” in 1L.

KEYNOTE-028-NPC [PMID 28837405] (Hsu C 2017 JCO, N=27, PD-L1+ selected): heavily pretreated R/M NPC, pembrolizumab monotherapy. ORR 25.9% (95% CI 11.1-46.3). The earliest Western ICI signal in NPC — PD-L1 selection + single-arm + very small N, but enough to motivate KEYNOTE-122.
NCI-9742-NIVO-NPC [PMID 29584545] (Ma BBY 2018 JCO, N=44, international multicenter including Asia): R/M NPC, nivolumab monotherapy. ORR 20.5% (1 CR + 8 PR), 1-year OS 59%. HLA class I loss was associated with better PFS — a biomarker signal unique to NPC. Paired with KEYNOTE-028, it defined the “active but unimpressive” baseline for Western ICI in 2L NPC.
FANG-2018-CAMRELIZUMAB-NPC [PMID 30213452] (Fang W 2018 Lancet Oncol, N=116, two cohorts: 93 patients 2L+ monotherapy + 23 patients treatment-naive camrelizumab + GP): monotherapy ORR 34% (31/91), GP combination ORR 91% (20/22). The key early signal for Hengrui’s domestic camrelizumab, directly motivating CAPTAIN-1st phase III design.
POLARIS-02 [PMID 33492986] (Wang FH 2021 JCO, N=190, China): chemotherapy-refractory R/M NPC, toripalimab monotherapy (92 of 190 were ≥ 2L). ORR 20.5% (≥2L subgroup 23.9%), mDoR 12.8 months, mOS 17.4 months; day 28 EBV-DNA ≥ 50% decline predicted response (ORR 48.3% vs 5.7%). The largest single-arm 2L NPC immunotherapy dataset, supporting NMPA approval.
KL-A167 [PMID 36879786] (Shi Y 2023 Lancet Reg Health West Pac, N=153, China, Kelun-Biotech PD-L1 antibody): platinum-pretreated R/M NPC, KL-A167 monotherapy. ORR 26.5% (95% CI 19.2-34.9), mDoR 12.4 months, mOS 16.2 months; lower baseline EBV-DNA correlated with better outcomes. The fifth domestic checkpoint agent with data in 2L NPC, consolidating the class-effect hypothesis.
LIM-2023-NIVO-IPI-NPC [PMID 37188668] (Lim DW 2023 Nat Commun, N=40, Singapore): platinum-pretreated EBV+ R/M NPC, nivolumab + ipilimumab dual IO. BOR 38%, mPFS 5.3 months, mOS 19.5 months (did not meet prespecified BOR threshold). Dual-checkpoint numerically higher than single-agent (KEYNOTE-028 26%, NCI-9742 21%), but no randomized comparison.
KEYNOTE-122 [PMID 36535566] (Chan ATC 2023 Ann Oncol, N=233, PD-L1 CPS ≥ 1, majority Asian sites): platinum-pretreated R/M NPC, pembrolizumab monotherapy vs investigator’s choice chemotherapy (capecitabine / gemcitabine / docetaxel). OS HR 0.90 (95% CI 0.67-1.19, p=0.23) negative, mOS 17.2 vs 15.3 months; G≥3 toxicity 10% vs 44% (significantly lower, but no OS improvement). The only 2L NPC phase III, negative. Clinical implication: PD-1 monotherapy cannot replace chemotherapy in 2L; to win, IO must be “added on top of chemotherapy” — which is retrospective validation of the three arrows’ 1L design.

Takeaway: in 2026, R/M NPC 2L has no phase III-positive IO regimen. Options: (a) patients who did not receive IO in 1L may consider single-agent PD-1 (POLARIS-02 / KL-A167 data, though not phase III level); (b) patients progressing on 1L IO may consider switching to chemotherapy (taxanes / irinotecan / platinum rechallenge) or dual IO combination (LIM-2023 signal but not SoC); (c) the KEYNOTE-122 lesson — “PD-1 monotherapy replacing chemotherapy” pathway is not viable. R/M NPC 2L remains a research-gap-dense area.

3. Cross-sectional: the 2026 decision landscape (six dimensions)

Projecting the longitudinal evolution onto the specific 2026 clinical decision tree, here are six key branchpoints and the evidence for each.

3.1 Newly diagnosed locally advanced NPC: EBV-DNA + imaging dual-track staging

Every newly diagnosed NPC patient must complete in parallel: IMRT planning MRI + neck CT + PET-CT (distant staging / micrometastasis screening) + baseline plasma EBV-DNA + follow-up 6-8 weeks post-treatment. Baseline EBV-DNA predicts distant metastasis risk; persistently positive post-treatment EBV-DNA (CHAN-2018 molecular residual disease) defines a high-risk molecular residual subgroup, but as of 2026 this remains an “identifiable but not actionable” state (CHAN-2018 adjuvant GC was negative).

3.2 Definitive CCRT + induction chemotherapy: TPF vs GP vs CCRT-alone decisions

Subgroup	Preferred	Alternative
Stage III-IVB locally advanced NPC, good performance status	Induction GP × 3 + CCRT [ZHANG-2019 PMID 31150573] or induction TPF × 3 + CCRT [SUN-2016 PMID 27686945]	CCRT alone (if induction not tolerated)
T3-4 N0-1 limited nodal disease	AF + CCRT (NPC-9902 2×2 factorial evidence) or induction GP + CCRT	CCRT alone insufficient [NPC-9902 PMID 16904519]
Cisplatin-related renal / otologic toxicity risk	Concurrent nedaplatin q3w × 3 [TANG-2018 PMID 29501366]	Weekly low-dose cisplatin
Concurrent regimen intensity choice	Cisplatin q3w × 2 (outpatient convenience + hearing preservation) [XIA-2021 PMID 34083231]	Weekly 40 mg/m² × 6

Controversy: no head-to-head phase III between TPF and GP induction. Cross-trial indirect comparison favors GP (HR 0.51) over TPF (HR 0.68), and GP 3-cycle completion rate of 96.7% far exceeds historical TPF values — in 2026, both Chinese and international mainstream lean toward GP, but the decision should factor in marrow tolerance + swallowing function.

3.3 Post-CCRT adjuvant: metronomic capecitabine vs observation

High-risk locally advanced NPC (stage III-IVB, excluding T3-4N0 / T3N1): metronomic capecitabine 650 mg/m² bid × 1 year after CCRT (CHEN-YP-2021 PMID 34111416, HR 0.50) — 2026 NCCN recommendation
Low-risk subgroup (T3-4N0 / T3N1): excluded from CHEN-YP-2021, evidence gap → observation
Molecular residual positive (detectable post-treatment EBV-DNA): metronomic capecitabine (CHEN-YP-2021 did not stratify, but molecular-residual patients are usually high-risk); do not recommend traditional PF [CHEN-L-2012 PMID 22154591] or GC [CHAN-2018 PMID 29989858]
DPYD polymorphism in Asian populations: capecitabine is generally well-tolerated; DPYD genotyping before dosing is a reasonable option

NCCN V1.2026: metronomic capecitabine is a Category 1 post-CCRT adjuvant standard for high-risk locally advanced NPC.

3.4 R/M-1L: “pick one of three” from the PD-1 + GP class effect

2026 mainstream: any domestic PD-1 + GP × 4-6 cycles → PD-1 maintenance (until progression or 2 years). The three agents have effect sizes HR 0.52-0.54 with no material difference.

Subgroup	Preferred	Secondary consideration
Treatment-naive R/M NPC (regardless of PD-L1)	toripalimab + GP [JUPITER-02 PMID 38015220] (has OS data)	camrelizumab + GP [CAPTAIN-1ST PMID 34174189] or tislelizumab + GP [RATIONALE-309 PMID 37207654]
US / overseas market	toripalimab + GP (FDA-approved 2023-10, the only domestically developed PD-1 available overseas)	Chemotherapy alone (if IO unavailable)
PD-L1 low / negative	Still use PD-1 + GP (RATIONALE-309 demonstrated PFS benefit independent of PD-L1)	Do not stratify
ECOG 2 / poor performance	Reduced-dose chemotherapy or single-agent	Do not recommend PD-1 monotherapy (KEYNOTE-122 lesson)

NCCN V1.2026: all three PD-1 + GP options are Category 1 (US market only has toripalimab approved in practice).

Controversy: maintenance duration, 2 years vs until progression, is undecided (the three arrows had different designs: JUPITER-02 ≤ 2 years, CAPTAIN-1st until progression, RATIONALE-309 until progression). 2026 clinical practice typically caps at 2 years + imaging response-guided.

3.5 R/M-2L+: do not reflexively prescribe single-agent IO

No prior IO in 1L (rare in 2026, since IO + GP is now 1L standard): consider toripalimab monotherapy (POLARIS-02 data) or chemotherapy
Progression after 1L IO + GP: no phase III-level regimen. Options include: taxanes (nab-paclitaxel / albumin-bound paclitaxel) / irinotecan / platinum rechallenge / clinical trial enrollment
KEYNOTE-122 lesson [PMID 36535566]: pembrolizumab monotherapy vs chemotherapy 2L OS HR 0.90 negative — replacing chemotherapy with single-agent PD-1 does not work
Dual IO combination: nivo + ipi BOR 38% (LIM-2023 PMID 37188668, N=40) signals promise but no phase III
MSI-H / dMMR / TMB-H subgroups: rare in NPC and not routinely tested; if positive, consider pembrolizumab under tumor-agnostic approval
EBV-DNA as pharmacodynamic biomarker: POLARIS-02 day 28 EBV-DNA ≥ 50% decline predicted ORR 48.3% vs 5.7%, usable as early treatment-response aid (not a decision branchpoint)

3.6 Clinical implications of the China-led research landscape

In 2026, 17 of 23 trials (74%) in the NCCN NPC chapter are Chinese-investigator-led. Behind this geographic distribution is NPC epidemiology shaping the research ecosystem: over 70% of new cases worldwide are in southern China / Southeast Asia, and Western countries lack sufficient patient density to complete phase III registration trials. Clinical implications:

Strong applicability to East Asian populations: the three arrows and induction chemotherapy phase IIIs were all conducted in Chinese / Hong Kong-Taiwan / Singapore populations — results extrapolate reliably to East Asian ethnic Chinese NPC
Scarcity of Western / South American / African population data: WHO type I (keratinizing) NPC is slightly more common in North America, but WHO II/III non-keratinizing / undifferentiated in endemic regions is the global majority — apart from INT-0099 (WHO mixed), systematic phase III data in Western populations essentially do not exist
Overseas accessibility differences: of the three domestic PD-1 agents, only toripalimab is FDA-approved (2023-10); camrelizumab / tislelizumab have no US accessibility — overseas ethnic Chinese / non-Chinese R/M NPC patients in practice can only receive toripalimab + GP or chemotherapy
Domestic PD-L1 (KL-A167) + other classes (bispecifics cadonilimab / ivonescimab) may expand in 2026-2028: Chinese domestic IO innovation continues at high density

4. Research gaps: ten unresolved clinical questions

This report identifies the following gaps, each a concretely definable problem (not the “more research needed” cliché):

“Molecular selection + next-generation drug” combinations for EBV-DNA-guided adjuvant therapy: CHAN-2018 proved EBV-DNA can identify molecular-residual high-risk patients (27.4% positivity), but traditional GC adjuvant was negative. The future direction requires an EBV-DNA-guided + IO adjuvant phase III (the next step after NRG-HN001 failure); no clear candidate trial as of 2026-04.
Missing TPF vs GP induction chemotherapy head-to-head phase III: SUN-2016 (TPF HR 0.68) vs ZHANG-2019 (GP HR 0.51) is only cross-trial indirect comparison; no direct randomized comparison with a shared control arm.
No evidence for head-to-head among the three domestic PD-1 + GP regimens: HR 0.52 / 0.54 / 0.52 three-point convergence cannot be distinguished — but safety profiles differ clearly (camrelizumab reactive cutaneous capillary endothelial proliferation 40%+, toripalimab thyroiditis profile, tislelizumab infusion reactions). No direct comparative data.
IO applicability in Western / South American / African NPC patients: 17/23 trials China-led, and the three R/M arrows had 95%+ patients from mainland China / Hong Kong-Taiwan / Singapore. Does PD-1 + GP have the same response in Western WHO type I keratinizing NPC? No independent data.
Biological root cause of KEYNOTE-122 failure: is it that “single-agent IO is insufficient in the 2L resistance microenvironment” or that “the chemotherapy arm was too strong (investigator’s choice included taxanes / irinotecan as additional options)”? No mechanistic correlative science to answer this.
Clinical utility of biomarker stratification (PD-L1 / EBV-DNA / TMB / HLA class I loss) in NPC: RATIONALE-309 showed PFS benefit independent of PD-L1; NCI-9742 found HLA class I loss predicted PFS — but no biomarker is used for treatment selection in NPC.
Missing phase III for neoadjuvant IO + CCRT combinations in locally advanced NPC: NSCLC / esophageal cancer / HCC all have neoadjuvant IO data; NPC has no phase III readout as of 2026. Early single-arm / phase II signals exist; design-level trials are absent.
Regimens for elderly / CCRT-intolerant NPC patients: all phase IIIs required ECOG 0-1 + age < 70; what to give elderly NPC patients in practice? No evidence-based answer.
Treatment differences across non-keratinizing vs keratinizing vs basaloid histologic subtypes: WHO subtypes differ in immune infiltration and EBV association, but treatment stratification does not use histology — all trials enrolled mixed populations. Whether stratification effects exist is unknown.
Missing phase III for proton / heavy-ion radiotherapy vs IMRT: proton / heavy-ion has expanded rapidly in NPC (multicenter Chinese expansion 2020-2026), but phase III comparisons vs IMRT for survival / toxicity are essentially absent — evidence stops at dose-distribution simulation + single-center cohorts.

5. 2024-2026 recent updates

5.1 FDA / NMPA new approvals (NPC-relevant excerpts)

Drug	Agency	Date	Indication / pivotal trial
toripalimab + GP (Loqtorzi)	FDA	2023-10-27	1L R/M NPC / JUPITER-02 [PMID 38015220]
toripalimab monotherapy (2L NPC)	FDA	2023-10-27	2L platinum-pretreated R/M NPC / POLARIS-02 [PMID 33492986]
camrelizumab + GP	NMPA	2021	1L R/M NPC / CAPTAIN-1ST [PMID 34174189]
tislelizumab + GP	NMPA	2023	1L R/M NPC / RATIONALE-309 [PMID 37207654]
KL-A167 (Kelun-Biotech PD-L1)	NMPA	2023-2024	2L R/M NPC / KL-A167 phase 2 [PMID 36879786]
toripalimab monotherapy (NMPA)	NMPA	2018-2019	Multi-line refractory R/M NPC / POLARIS-02 supporting

Key observation: FDA has approved only toripalimab (2023-10); camrelizumab + tislelizumab had not received FDA approval as of 2026-04. European EMA NPC indication progress is slower than FDA.

5.2 Key conference readouts (2024-2026, downweighted)

The following entries are candidate-pool only, not primary database, pending formal peer review.

JUPITER-02 5-year follow-up (ASCO 2024-2025 pending): latest OS curve stable around HR 0.63, long-tail responder fraction suggests ~25-30%
CAPTAIN-1st long-term follow-up (ESMO Asia 2024-2025): PFS curve stable; OS data still immature
RATIONALE-309 final OS (2025-2026 pending): OS data still immature as of 2026-04
ivonescimab (AK112, PD-1+VEGF bispecific) in NPC: Akeso 2024-2025 phase II data, hypothesis-generating in R/M NPC 2L
cadonilimab (AK104, PD-1+CTLA-4 bispecific) in NPC: Akeso 2024 phase II single-arm data, not yet in phase III

5.3 Ongoing phase III trials (selected 2025-2028 readouts)

JUPITER-02-extension (toripalimab maintenance duration de-escalation ± varied cycles) — not yet registered
Domestic NPC neoadjuvant IO + CCRT phase II / III (SYSUCC / Hong Kong / BeiGene-sponsored) — possible 2025-2028 readouts
ivonescimab + GP vs PD-1 + GP in R/M NPC 1L — Akeso may start 2026
Proton / heavy-ion vs IMRT in locally advanced NPC — Shanghai Proton and Heavy Ion Center + Sun Yat-sen University collaborative trial, registration underway

Note: ongoing phase III density in NPC is noticeably lower than in NSCLC / BTC — the three arrows rewrote the 1L standard in 2021-2023, and the next window is still under construction.

6. Convergent insights and judgment

6.1 Longitudinal × cross-sectional: the 2026 NPC landscape is shaped by three resonances

Stacking the longitudinal paradigm evolution on the cross-sectional current decision landscape, the 2026 NPC landscape is the superposition of three resonances:

CCRT universality (four phase IIIs globally confirmed 1998-2006) → induction chemotherapy layering (two SYSUCC phase IIIs 2016-2019) → metronomic adjuvant breakthrough (CHEN-YP-2021): over 28 years, locally advanced NPC OS moved from INT-0099’s ~47% to ZHANG-2019’s 3-year OS ~95% (population differences + stage differences notwithstanding, direction clear). A three-stage evolution: North American foundation → Asian regional confirmation → China-led rewrite.
R/M 1L PF → GP (ZHANG-2016-GP [PMID 27567279]) → PD-1 + GP three-arrow convergence (HR 0.52-0.54 class effect): within 7 years, R/M NPC 1L moved from GP alone to PD-1 + GP, and the tight HR band is the most elegant class effect outside NSCLC / BTC. The fact that three domestic PD-1 agents completed simultaneously is a direct manifestation of the high-density explosion of the Chinese IO industry, not coincidence. JUPITER-02 [PMID 38015220]’s OS HR 0.63 is the only mature-OS-positive readout among the three arrows, and is the core evidence behind FDA’s 2023-10 approval of toripalimab + GP.
Geographic epidemiology × research ecology: over 70% of new NPC cases globally occur in southern China / Southeast Asia — this geographic reality dictates that phase III trials must be led by Chinese / Hong Kong-Taiwan / Singapore teams. The 17/23 (74%) China-led share is not a “tendency” but a “necessity.” It brings benefits (detailed East Asian data + NCCN directly incorporates the three arrows) and challenges (scarce Western applicability data + overseas accessibility differences) in parallel.

These three resonances together explain a clinical phenomenon: for a newly diagnosed R/M NPC patient in 2026, the 1L decision has only one more decision layer than in 2016 (which PD-1 + GP to choose), but every pathway has Category 1 evidence — completely different from NSCLC’s multi-layer, multi-branch decision landscape. NPC’s decision tree is characterized by “narrow width but every branch hard,” which is its uniqueness.

6.2 Clinical decision implications (takeaways for junior-mid oncologists)

Locally advanced NPC standard = induction GP × 3 (or TPF × 3) + concurrent cisplatin q3w × 2 + IMRT 70 Gy: this triad is NCCN Category 1 in 2026 — do not default to CCRT alone anymore.
For high-risk locally advanced NPC, give metronomic capecitabine × 1 year after CCRT: CHEN-YP-2021 is the only positive phase III in a decade-long adjuvant void. Traditional PF / GC adjuvant has been explicitly refuted (CHEN-L-2012 / CHAN-2018).
Choose nedaplatin for cisplatin-intolerant patients, q3w × 2 concurrent preferred over weekly: TANG-2018 + XIA-2021 essentially closed the radiosensitizer dosing question; these findings can be adopted directly in clinic.
R/M NPC 1L must be PD-1 + GP, GP alone is no longer acceptable: the three arrows’ HR 0.52-0.54 class effect is 2026 NCCN Category 1. The choice of PD-1 is driven by accessibility / safety profile, not efficacy — because there is no evidence of efficacy difference.
Only toripalimab + GP is FDA-approved for R/M NPC 1L: overseas patients in practice have this one domestic IO pathway (as of 2026-04); other regions can also choose camrelizumab or tislelizumab + GP.
Do not prescribe single-agent PD-1 to replace chemotherapy in R/M NPC 2L: the KEYNOTE-122 phase III HR 0.90 negative readout has closed this pathway. 2L patients should consider chemotherapy rechallenge, dual IO combination (LIM-2023 signal), or clinical trials.
EBV-DNA is an NPC-exclusive dual biomarker — pharmacodynamic + prognostic: baseline, 6-8 weeks post-treatment, and day 28 dynamic monitoring are all data-supported (CHAN-2018 molecular residual definition / POLARIS-02 pharmacodynamic prediction). But no intervention currently targets “EBV-DNA-positive” effectively — identification capability > intervention capability.
Low NPC incidence + endemic distribution + China-led research: these three jointly shape the ecosystem. Clinicians should default to NCCN V1.2026 trials as directly applicable to East Asian / ethnic Chinese populations; Western / African / South American populations require careful consideration of epidemiology + histology differences.
PD-L1 / HLA class I / TMB biomarker stratification has no current clinical utility in NPC: RATIONALE-309 showed PFS benefit independent of PD-L1. Do not refuse PD-1 + GP on the basis of PD-L1 negativity.
2026 must-know NPC drug list (10 agents): cisplatin / nedaplatin / gemcitabine / docetaxel / 5-FU / capecitabine (metronomic) / toripalimab / camrelizumab / tislelizumab / KL-A167 — 28 years ago it was cisplatin + radiotherapy only; today it is 10 drugs across 4 treatment phases (induction / concurrent / adjuvant / R/M 1L / R/M 2L) with a complete decision map.

7. Information sources

Metadata for all 23 trials in this report were independently verified through both PubMed and ClinicalTrials.gov. Every [PMID xxxxxxxx] in the body text can be verified directly on PubMed.

Published trials: 23, covering 1998-2023 (PMID-verifiable)
Ongoing / downweighted: 0 (all NCCN V1.2026-cited trials are peer-reviewed and published)
NCCN guideline citations: 23/23 (100%) directly match the NCCN Cancer of the Nasopharynx V1.2026 reference section
2023-2024 FDA new approvals: 1 key approval (toripalimab 2023-10 for 1L + 2L R/M NPC) + 3 NMPA (camrelizumab / tislelizumab / KL-A167)
2024-2026 key conference readouts: 4 (JUPITER-02 5y / CAPTAIN-1st long-term / RATIONALE-309 final OS / ivonescimab + cadonilimab NPC early) — all without PMID, downweighted
Research gaps: 10
China-led trial share: 17/23 (74%)

7.1 Body-text PMID citation list (ascending by PMID)

The following table lists PMIDs bracket-cited in the body text, each verifiable via PubMed URL.

PMID	Trial / Paper	Year	Journal	Location in text
9552031	INT-0099	1998	J Clin Oncol	§2.1 CCRT foundation
12586799	LIN-2003	2003	J Clin Oncol	§2.1
16170180	WEE-2005	2005	J Clin Oncol	§2.1
16192584	NPC-9901	2005	J Clin Oncol	§2.1
16904519	NPC-9902	2006	Int J Radiat Oncol Biol Phys	§2.1 / §3.2
22154591	CHEN-L-2012	2012	Lancet Oncol	§2.2 / §3.3
27567279	ZHANG-2016-GP	2016	Lancet	§2.3 R/M 1L
27686945	SUN-2016	2016	Lancet Oncol	§2.1 / §3.2
28837405	KEYNOTE-028-NPC	2017	J Clin Oncol	§2.4 / §3.5
29501366	TANG-2018	2018	Lancet Oncol	§2.1 / §3.2
29584545	NCI-9742-NIVO-NPC	2018	J Clin Oncol	§2.4 / §3.5
29989858	CHAN-2018	2018	J Clin Oncol	§2.2 / §3.1 / §3.3
30213452	FANG-2018-CAMRELIZUMAB-NPC	2018	Lancet Oncol	§2.4
31150573	ZHANG-2019	2019	N Engl J Med	§2.1 / §3.2
33492986	POLARIS-02	2021	J Clin Oncol	§2.4 / §3.5 / §5.1
34083231	XIA-2021	2021	Clin Cancer Res	§2.1 / §3.2
34111416	CHEN-YP-2021	2021	Lancet	§2.2 / §3.3
34174189	CAPTAIN-1ST	2021	Lancet Oncol	§2.3 / §3.4 / §5.1
36535566	KEYNOTE-122	2023	Ann Oncol	§2.4 / §3.5
36879786	KL-A167	2023	Lancet Reg Health West Pac	§2.4 / §3.5 / §5.1
37188668	LIM-2023-NIVO-IPI-NPC	2023	Nat Commun	§2.4 / §3.5
37207654	RATIONALE-309	2023	Cancer Cell	§2.3 / §3.4 / §5.1
38015220	JUPITER-02	2023	JAMA	§2.3 / §3.4 / §5.1

7.2 Verification conventions

Every PMID can be accessed at https://pubmed.ncbi.nlm.nih.gov/{PMID}/
Every NCT id can be accessed at https://clinicaltrials.gov/study/{NCT_id}/
Conference abstracts (ASCO / ESMO / ESMO Asia) are queried via the official conference system; all conference citations in this report are downweighted — not peer-reviewed, and final data await journal publication
JUPITER-02 interim data (Nat Med 2021 PMID 34341578) appears as a supplementary citation in the body; the primary reference in this table is the final OS (JAMA 2023 PMID 38015220)
If any PMID in the report is found to mismatch the PubMed trial name / year / conclusion, corrections are welcome

Clinical trial timeline is here

Chinese: /trials/npc/ English: /en/trials/npc/

Each trial has an independent detail page with:

Complete intervention / comparator regimen
Primary endpoint values + 95% CI
Key findings + clinical implications
Clickable PMID / NCT source links

23 trials · 4 sections · 1998 to 2023 · synchronized with NCCN Cancer of the Nasopharynx V1.2026.

Closing

Over the past 28 years, NPC has completed a unique evolution in oncology — from INT-0099’s North American foundation in 1998, through the three major Asian regional confirmations in Taiwan / Hong Kong / Singapore in 2003-2006, through SYSUCC’s rewriting of the global NCCN standard with induction chemotherapy in 2016-2019, and finally to three domestic PD-1 agents independently delivering PFS HR 0.52-0.54 class effect in the same R/M NPC 1L setting within 2021-2023 — three arrows converging into the global NCCN.

The biggest structural difference between NPC and other major cancers (NSCLC / BTC / HCC) is not biology or treatment complexity, but “geographic epidemiology determining research ecology”: over 70% of new cases globally in southern China / Southeast Asia, 17/23 landmark trials led by Chinese investigators — this “not a choice but a necessity” geographic distribution brought positive dividends (detailed East Asian population data + direct NCCN inclusion of domestic drugs) alongside structural challenges (scarce Western applicability data + overseas accessibility differences).

The R/M-1L three-arrow HR convergence is one of the most elegant class effects of 2020s oncology. But NPC 2L remains a research-gap-dense area (KEYNOTE-122 phase III failed + no standard alternative), neoadjuvant IO + CCRT still has no phase III readout, and the “identification > intervention” paradox of EBV-DNA remains unresolved. The past 28 years completed 1L standardization; the next decade must solve the three structural problems of 2L, neoadjuvant, and biomarker stratification.

The value of this report is not in “exhausting all trials” (PubMed does that), but in compressing 28 years of evolution + current decisions + unresolved gaps into a single reading session’s cognitive bandwidth. The next time you face a newly diagnosed NPC patient, every branchpoint in the decision tree has this map — searchable, traceable, and questionable.

Clinician × AI = Research Superpower + Clinical Decision Amplifier

—— Dual Brain Lab · 2026-04-21

SCLC Clinical Trial Timeline: A 25-Year Dual-Track Map

Tue, 21 Apr 2026 00:00:00 +0000

SCLC Clinical Trial Timeline: In-Depth Research Report

Coverage: 30 landmark trials cited by NCCN Small Cell Lung Cancer V1.2026 (28 published with full PMID traceability + 2 design/protocol papers with primary results pending) + two stages (LS-SCLC limited-stage / ES-SCLC extensive-stage) + five paradigms (concurrent chemoradiation / 1L chemoimmunotherapy / PCI / 2L and later lines / maintenance exploration)

Curated by Dual Brain Lab (csilab.net)

1. One-Sentence Definition

This report maps the evolution logic and current decision landscape of landmark clinical trials for systemic treatment of small cell lung cancer (SCLC) cited by NCCN Small Cell Lung Cancer V1.2026 over the past 27 years (1999-2026), providing frontline clinicians in 2026 a traceable panoramic map for “who, what, and why” decisions.

Iron rule: every data point in every trial is traceable to PubMed (PMID) or ClinicalTrials.gov (NCT id) — each [PMID xxxxxxxx] in the text can be directly opened for PubMed source verification.

2. Longitudinal: Evolution Timeline of Five Treatment Paradigms

SCLC systemic treatment has undergone five paradigm shifts over 27 years: LS-SCLC (limited-stage) concurrent chemoradiation established by Turrisi 1999 hyperfractionated acceleration → ES-SCLC (extensive-stage) etoposide + platinum (EP) backbone and 2L topotecan 20-year plateau → thoracic radiotherapy + PCI (prophylactic cranial irradiation) filled the brain-protection gap during 2007-2015 → 2018 IMpower133 and 2019 CASPIAN broke the 20-year OS (overall survival) ceiling using PD-(L)1 inhibitor class effect → 2024 ADRIATIC upgraded LS-SCLC to cCRT (concurrent chemoradiation) + durvalumab consolidation as the new standard, while in the same year DeLLphi-301 opened SCLC’s first truly non-chemotherapy new mechanism in half a century using tarlatamab (DLL3 × CD3 BiTE, bispecific T-cell engager).

Each shift has been smaller in magnitude than in NSCLC — reflecting SCLC’s unique biology: no unified driver mutation + neuroendocrine differentiation + MYC / ASCL1 / NEUROD1 / POU2F3 / YAP1 molecular subtype heterogeneity + rapid chemoresistance + low tumor mutation burden (TMB) → the NSCLC “EGFR/ALK/KRAS ten-gene + IO backbone” model does not apply to SCLC. SCLC’s paradigm evolution follows a three-stage arc: “20-year chemotherapy plateau + IO marginal victory + DLL3 BiTE breakthrough.”

2.1 LS-SCLC Concurrent Chemoradiation Foundations (1999-2024): BID acceleration → QD non-inferiority → SIB dose escalation

Story: in 1999 TURRISI INT-0096 pushed LS-SCLC radiotherapy from QD (once daily) 5 weeks to BID (twice daily) 3 weeks 45 Gy hyperfractionated acceleration, raising 2-year OS from 41% to 47%. But BID had poor clinical feasibility (twice-daily outpatient radiotherapy + esophagitis), so for 18 years Europe and North America wanted to return to QD but lacked evidence, until CONVERT proved 66 Gy QD non-inferior, validating the “more convenient dose” approach, and CALGB-30610 used 70 Gy QD to confirm numerical closeness. In 2024 Yu et al used SIB (simultaneous integrated boost) 54 Gy BID to push mOS to 60.7 months — the first time a Chinese cohort raised the radiotherapy dose further.

INT-0096 (TURRISI 1999) [PMID 9920950] (Turrisi 1999 NEJM): LS-SCLC 45 Gy BID 3 weeks vs 45 Gy QD 5 weeks + cisplatin/etoposide. mOS 23 vs 19 months (p=0.04), 2-year OS 47% vs 41%, 5-year OS 26% vs 16%. Grade 3 esophagitis 32% vs 16%. Established “LS-SCLC cCRT + BID 45 Gy” as North American SoC (standard of care) for 25 years.
CONVERT [PMID 28642008] (Faivre-Finn 2017 Lancet Oncol): LS-SCLC 66 Gy QD 6.5 weeks vs 45 Gy BID 3 weeks + cisplatin/etoposide. mOS 30 months (BID) vs 25 months (QD), HR 1.18 (95% CI 0.95-1.45, p=0.14) — non-inferiority achieved, toxicity profiles similar. Europe and North America finally had evidence that “QD is non-inferior to BID,” and clinical practice shifted back to QD.
CALGB 30610 / RTOG 0538 [PMID 36623230] (Bogart 2023 JCO): LS-SCLC 70 Gy QD vs 45 Gy BID + platinum-etoposide. mOS 30.1 vs 28.5 months, HR 0.94 (95% CI 0.76-1.17). 70 Gy QD confirmatory in North America, aligned with CONVERT — high-dose QD = standard-dose BID.
YU-2024-HYPERFRAC [PMID 39146944] (Yu 2024 Lancet Respir Med): LS-SCLC VMAT 54 Gy SIB BID 30 fractions vs 45 Gy BID 30 fractions. mOS 60.7 vs 39.5 months, HR 0.55 (95% CI 0.37-0.72, p=0.003) — adding 9 Gy gained 21 months of OS. Grade 3-4 esophagitis 11% vs 4%. Bold Chinese single-center design, awaiting Western confirmatory trials.
ADRIATIC [PMID 39268857] (Cheng 2024 NEJM): LS-SCLC post-cCRT non-progressors, durvalumab 1500 mg q4w consolidation for up to 24 months vs placebo. mOS 55.9 vs 33.4 months, HR 0.73 (98.321% CI 0.54-0.98, p=0.01), mPFS 16.6 vs 9.2 months. First time in 30 years that post-cCRT OS was significantly extended in LS-SCLC — the SCLC version of PACIFIC’s playbook for stage III NSCLC moving from cCRT observation to durvalumab consolidation.
NRG LU005 [PMID 41529214] (Higgins 2026 JCO): LS-SCLC cCRT (66 Gy QD or 45 Gy BID) + atezolizumab concurrent + adjuvant 17 cycles vs cCRT alone. mOS 31.1 vs 36.1 months, HR 1.03 (95% CI 0.80-1.32) negative. “Concurrent IO + cCRT” failed in LS-SCLC — “sequential durvalumab consolidation” (ADRIATIC) is the correct path.

Takeaway: in 2026, LS-SCLC SoC = cCRT (either BID 45 Gy or QD 66-70 Gy) → durvalumab consolidation up to 24 months (ADRIATIC). The “concurrent IO + cCRT” path (NRG LU005) has been refuted. Yu 2024’s 54 Gy SIB dose escalation is a Chinese single-center exploratory signal awaiting larger confirmatory trials.

2.2 ES-SCLC 1L Chemotherapy Plateau (1999-2019): EP backbone 20 years + topotecan vs CAV anchoring 2L

Story: before the IO era, 1L standard for ES-SCLC was EP (etoposide + platinum) 4-6 cycles with an mOS ceiling of 9-10 months. All attempts to add a third drug (paclitaxel / lobaplatin substitution) failed and only doubled toxicity. 2L topotecan, after a 1999 phase III vs CAV (cyclophosphamide + doxorubicin + vincristine) proved single-agent non-inferiority + better quality of life, sat firmly as the 2L standard until tarlatamab arrived in 2024.

TOPOTECAN-ORIGINAL-PHASE2 [PMID 10080612] (von Pawel 1999 JCO): relapsed SCLC topotecan 1.5 mg/m² d1-5 q21d vs CAV. ORR 24.3% vs 18.3% (p=0.285), mTTP (time to progression) 13.3 vs 12.3 weeks (p=0.552), mOS numerically close. Single-agent topotecan non-inferior to three-drug CAV + better symptom improvement — established topotecan as SCLC 2L standard for 20+ years.
NIELL-2002-INTERGROUP [PMID 15923572] (Niell 2005 JCO): ES-SCLC 1L paclitaxel + EP + G-CSF vs EP × 6 cycles. mOS 10.6 vs 9.9 months, p=0.169 negative, G3-4 toxicity increased. “Add a taxane to EP” approach refuted — EP is the ES-SCLC 1L chemotherapy ceiling.
OBRIEN-TOPOTECAN-2006 [PMID 17135646] (O’Brien 2006 JCO): relapsed SCLC oral topotecan 2.3 mg/m² d1-5 q21d + BSC (best supportive care) vs BSC alone. mOS 25.9 vs 13.9 weeks (log-rank p=0.0104), ORR 7% PR + 44% DCR. Oral topotecan vs BSC established that “even when no new drug is available, 2L topotecan beats BSC” — the minimum-care baseline for SCLC 2L.
EL-PHASE3-CHINA [PMID 30988825] (Cheng 2019 Oncol Lett): ES-SCLC 1L lobaplatin + etoposide (EL) vs cisplatin + etoposide (EP). mPFS 5.1 vs 5.3 months (p=0.786), mOS 10.6 vs 9.7 months (p=0.701). Chinese non-inferiority demonstrated lobaplatin vs cisplatin equivalence in ES-SCLC 1L, providing a platinum alternative for patients intolerant to nephrotoxicity / vomiting — the last “chemotherapy optimization” before the IO era.

Takeaway: across these 20 years (1999-2019), ES-SCLC 1L ceiling held steady at mOS 9-10 months, 2L topotecan single-agent at ~25 weeks — two “plateau” numbers defined what IO had to break. All attempts to add a third chemo drug / substitute the platinum failed.

2.3 ES-SCLC Thoracic RT + PCI Brain Protection (2007-2021): add RT + PCI → dose + fraction optimization

Story: before 2007, ES-SCLC post-chemo responders went straight to observation. Slotman 2007 proved PCI 25 Gy after chemo response substantially reduced brain metastases and extended OS; CREST 2014 proved thoracic RT 30 Gy/10 fx also delivered 2-year OS benefit. From 2009-2021 three trials (PCI99-01 dose optimization + RTOG-0212 hippocampal avoidance) refined PCI to “low dose + cognitive protection.”

SLOTMAN-PCI-2007 [PMID 17699816] (Slotman 2007 NEJM): ES-SCLC post-chemo response PCI 25 Gy vs observation. 1-year symptomatic brain metastasis risk 14.6% vs 40.4% (HR 0.27, 95% CI 0.16-0.44, p<0.001), 1-year OS 27.1% vs 13.3% (HR 0.68, 95% CI 0.52-0.88, p=0.003). Pivotal evidence establishing ES-SCLC PCI as SoC, widely adopted globally after 2007.
PCI99-01 [PMID 19386548] (Le Péchoux 2009 Lancet Oncol): LS-SCLC cCRT CR (complete response) PCI 36 Gy vs 25 Gy dose comparison. 2-year brain metastases 23% vs 29% (36 Gy vs 25 Gy), HR 0.80 (95% CI 0.57-1.11) negative; 36 Gy arm had more chronic neurotoxicity. 25 Gy low dose non-inferior + less toxicity — LS-SCLC PCI standard dose locked at 25 Gy.
CREST (SLOTMAN-2015) [PMID 25230595] (Slotman 2015 Lancet): ES-SCLC post-chemo response thoracic RT 30 Gy/10 fx + PCI vs PCI alone. 1-year OS 33% vs 28% (HR 0.84, 95% CI 0.69-1.01, p=0.066 not significant), 2-year OS 13% vs 3% (p=0.004 significant). “Long-tail OS benefit” — a subset of patients indeed benefits from thoracic RT, but not significant at the 1-year landmark. NCCN retains as an optional option.
RTOG-0212-HIPPO (Belderbos 2021) [PMID 33545387] (Belderbos 2021 JTO): SCLC post-chemo PCI 25 Gy standard whole-brain vs HA-PCI (hippocampal-avoidance PCI) 25 Gy. 4-month HVLT-R total recall decline ≥5 points: 29% vs 28% (p=1.000) — HA-PCI did not show clear cognitive protection advantage in this study, inconsistent with earlier small-scale NRG CC001 data. Guideline wording retains HA-PCI as optional but evidence is not coherent.

Takeaway: 2026 PCI decisions — LS-SCLC cCRT CR then PCI 25 Gy (PCI99-01 dose-locked) remains SoC; ES-SCLC post-chemo response PCI 25 Gy based on Slotman 2007 but in the IO era some European centers substitute brain MRI surveillance (lacks head-to-head RCT); CREST thoracic RT kept as an option (strong 2-year OS signal but 1-year OS not significant); HA-PCI evidence is inconsistent (RTOG-0212 negative).

2.4 ES-SCLC 1L IO Breakthrough (2018-2025): class-effect marginal victory + Chinese PD-1 catch-up

Story: in 2018 IMpower133 used atezolizumab + chemo to push ES-SCLC’s 20-year ceiling from mOS 10 months to 12.3 months, HR 0.70 — the first positive phase III for IO in SCLC 1L. In 2019 CASPIAN reproduced the same path with durvalumab, HR 0.73. Two independent phase IIIs converging in the narrow HR 0.70-0.73 band = textbook definition of class effect. From 2022-2025, five domestic Chinese phase IIIs (ASTRUM-005 serplulimab / CAPSTONE-1 adebrelimab / RATIONALE-312 tislelizumab / EXTENTORCH toripalimab / ETER701 benmelstobart + anlotinib) propelled ES-SCLC 1L into a “Chinese PD-1/PD-L1 flourishing” era, with Chinese regimens’ mOS numerically higher than IMpower133/CASPIAN (15-19 months vs 12-13 months), though cross-trial population differences require cautious interpretation.

IMPOWER133 [PMID 30280641] (Horn 2018 NEJM): ES-SCLC treatment-naïve atezolizumab + carboplatin + etoposide vs placebo + carboplatin + etoposide. mOS 12.3 vs 10.3 months (HR 0.70, 95% CI 0.54-0.91, p=0.007), mPFS 5.2 vs 4.3 months (HR 0.77, p=0.02). First OS-positive phase III in SCLC in 20 years. FDA approved 2019-03.
IMBRELLA-A (IMpower133 5y) [PMID 39306923] (Reck 2024 Lung Cancer): IMpower133 ITT 5-year long-term follow-up extension. 3/4/5-year OS atezolizumab arm 16%/13%/12% — “long-tail survivor” confirmed, about 12% of patients alive at 5 years (historical chemo-alone 3-4%). IO benefit concentrated in a “durable responder” subgroup, not uniformly lifting the entire population.
CASPIAN [PMID 31590988] (Paz-Ares 2019 Lancet): ES-SCLC treatment-naïve durvalumab + platinum-etoposide (EP/EC) vs EP/EC. mOS 13.0 vs 10.3 months (HR 0.73, 95% CI 0.59-0.91, p=0.0047). Global regimen + durvalumab maintenance through progression (rather than stopping at 4-6 cycles). FDA approved 2020-03, HR converging with IMpower133 in the 0.70-0.73 band — establishing IO + platinum/etoposide class effect.
ASTRUM-005 [PMID 36166026] (Cheng 2022 JAMA): ES-SCLC treatment-naïve serplulimab (HLX10 PD-1) + carboplatin + etoposide vs placebo + carboplatin + etoposide. mOS 15.4 vs 10.9 months (HR 0.63, 95% CI 0.49-0.82, p<0.001), 24-month OS rate 43.1% vs 7.9%. First Chinese PD-1 to post OS HR 0.63 in ES-SCLC 1L — numerically higher than IMpower133/CASPIAN, population predominantly Asian.
CAPSTONE-1 [PMID 35576956] (Wang 2022 Lancet Oncol): ES-SCLC treatment-naïve adebrelimab (SHR-1316 PD-L1) + carboplatin + etoposide vs placebo + carboplatin + etoposide. mOS 15.3 vs 12.8 months (HR 0.72, 95% CI 0.58-0.90). Chinese PD-L1 version of IMpower133 repeat, HR close to CASPIAN.
RATIONALE-312 [PMID 38460751] (Cheng 2024 JTO): ES-SCLC treatment-naïve tislelizumab (PD-1) + etoposide + carboplatin/cisplatin vs placebo + etoposide + carboplatin/cisplatin. mOS 15.5 vs 13.5 months (HR 0.75, 95% CI 0.61-0.93, one-sided p=0.0040). The third Chinese IO confirming class effect.
EXTENTORCH [PMID 39541202] (Cheng 2025 JAMA Oncol): ES-SCLC treatment-naïve toripalimab (PD-1) + EP vs placebo + EP. mOS 14.6 vs 13.3 months (HR 0.80, 95% CI 0.65-0.98, p=0.03), PFS HR 0.67. The fourth Chinese IO added to EP.
ETER701 [PMID 38992123] (Cheng 2024 Nat Med): ES-SCLC treatment-naïve benmelstobart (PD-L1) + anlotinib (multi-target TKI) + etoposide + carboplatin four-drug regimen vs dual placebo + etoposide + carboplatin. Four-drug arm mOS 19.3 vs 11.9 months (HR 0.61, p=0.0002); anlotinib + EC (three-drug arm) also superior to EC. Highest mOS record in ES-SCLC 1L history; G3+ TRAE 93.1% approaching universal. “Adding anti-angiogenic TKI to IO + chemo” is the new intensification direction for ES-SCLC 1L, but at a high toxicity cost.

Takeaway: in 2026, ES-SCLC 1L SoC = IO + platinum/etoposide × 4-6 cycles → IO maintenance (atezolizumab / durvalumab / serplulimab / adebrelimab / tislelizumab / toripalimab, any one; class effect HR 0.63-0.80 narrow band). ETER701 four-drug regimen mOS 19.3 months is the Chinese exploration peak but toxicity cost is high; choice depends on clinical tradeoff. The “long-tail 12% 5-year survival” is the most critical structural change in the IO era for ES-SCLC — not raising everyone’s OS by 2-3 months, but giving 12% of “durable responders” a 5-year+ survival window.

2.5 ES-SCLC 2L+ and Maintenance Exploration (2016-2026): tarlatamab DLL3 BiTE breakthrough + IO monotherapy failure + flourishing maintenance combinations

Story: IO monotherapy was initially given high hopes in SCLC 2L+; CheckMate-032 nivolumab ± ipilimumab and KEYNOTE-028/158 pembrolizumab both showed ORR 10-22% signals, but without phase III level OS evidence, FDA successively withdrew these two SCLC indications during 2018-2020 — IO monotherapy in unselected 2L populations is not practice-changing. Until 2023, when DeLLphi-301 used tarlatamab (DLL3 × CD3 BiTE) in 2L+ to achieve cORR 40%, mDoR (median duration of response) 6.9 months, and FDA accelerated approval 2024-05, marking the first time in half a century SCLC had a “non-chemotherapy new mechanism” receive regulatory approval. From 2024-2026 maintenance/combination explorations (DeLLphi-303 / DURABLE / TREASURE / MATCH) rolled out densely, exploring four combination directions: “IO + tarlatamab / IO + anti-angiogenic / IO + thoracic RT / IO + low-dose RT.” TRACES used trilaciclib to reposition myeloprotection as an SCLC chemotherapy adjunct.

CHECKMATE-032 [PMID 27269741] (Antonia 2016 Lancet Oncol): relapsed SCLC nivolumab monotherapy or nivolumab + ipilimumab various doses. ORR: nivo3 10% (10/98), nivo1+ipi3 23% (14/61), nivo3+ipi1 19% (10/54); durable response signal but small sample, no control. FDA 2018-08 accelerated approval for nivolumab SCLC 3L based on this basket data, but withdrawn 2020-12 (CheckMate-331/451 confirmatory failed).
KEYNOTE-028-158 [PMID 31870883] (Chung 2020 JTO): relapsed SCLC ≥2L pembrolizumab monotherapy pooled analysis. ORR 19.3% (95% CI 11.4-29.4), 2 CR + 14 PR, 61% of responders DoR ≥18 months. FDA 2019-06 accelerated approval for pembrolizumab SCLC 3L, withdrawn 2021-03 (KEYNOTE-604 1L confirmatory negative). IO monotherapy in unselected 2L+ SCLC showed seemingly decent ORR but no phase III OS support — what regulators ultimately required.
DELLPHI-301 [PMID 37861218] (Ahn 2023 NEJM): previously treated SCLC 2L+ tarlatamab 10 mg or 100 mg IV every 2 weeks. ORR 40% (10 mg arm, cORR 40% per BICR) vs 32% (100 mg), mPFS 4.9 vs 3.9 months, 9-month OS 68% (10 mg arm), mDoR 6.9 months. Severe CRS (cytokine release syndrome) 1%, low-grade CRS ~50% mostly at first dose, managed via step-up dosing. First non-chemotherapy non-IO new mechanism in SCLC in half a century — FDA accelerated approval 2024-05.
DELLPHI-303 [PMID 40934933] (Paulson 2025 Lancet Oncol): ES-SCLC 1L post-chemoimmunotherapy tarlatamab + atezolizumab or durvalumab maintenance. From maintenance start mOS 25.3 months (95% CI 20.3-NE), severe CRS 24% — safety signal in 1L maintenance setting is manageable. Phase III NCT06211036 ongoing, expected to push tarlatamab from 2L+ to 1L maintenance.
DURABLE [PMID 37947242] (Zhang 2023 Clin Respir J, protocol paper): ES-SCLC 1L post-chemo response durvalumab + anlotinib maintenance vs durvalumab maintenance. Only protocol published, primary efficacy results not yet PubMed-indexed (conference reports PFS 9.0 vs 5.6 months) — representative of “anti-angiogenic + IO combination maintenance” direction.
TREASURE [PMID 36153496] (Bozorgmehr 2022 BMC Cancer, protocol paper): ES-SCLC 1L post-response atezolizumab + thoracic RT (TRT) vs atezolizumab maintenance alone. Protocol design stage; trial terminated due to toxicity. CREST thoracic RT + IO maintenance verification attempt failed.
MATCH-TRIAL-SCLC [PMID 40609842] (Zhou 2026 Int J Radiat Oncol Biol Phys): ES-SCLC 1L cisplatin/carboplatin + etoposide + atezolizumab + concurrent low-dose RT (LDRT, 15 Gy/5 fx). Confirmed ORR 87.5% (95% CI 75.9-94.8), mPFS 6.9 months (95% CI 5.4-9.3) — single-arm phase II, early signal for “LDRT + IO + chemo” in ES-SCLC 1L, ORR far higher than IMpower133/CASPIAN. Awaiting phase III validation.
TRACES [PMID 38224653] (Cheng 2024 Lung Cancer): ES-SCLC chemotherapy trilaciclib (CDK4/6 inhibitor myeloprotection) pre-dosed vs placebo + etoposide + carboplatin. Cycle 1 DSN (duration of severe neutropenia) 0 vs 2 days (p=0.0003), mOS 12.0 vs 8.8 months (HR 0.69, 95% CI ). Trilaciclib already FDA-approved in the US 2021 — turning chemotherapy myelosuppression from “unavoidable side effect” into “pharmacologically preventable.”

Takeaway: in 2026, SCLC 2L+ SoC = tarlatamab (DeLLphi-301, FDA 2024-05 accelerated) is the preferred non-chemotherapy new mechanism; single-agent topotecan or lurbinectedin (FDA 2020 accelerated, not in this yaml) as non-BiTE alternatives; IO monotherapy (nivolumab / pembrolizumab) SCLC indications withdrawn, limited to tumor-agnostic MSI-H / TMB-H. The four maintenance/combination directions (tarlatamab maintenance / anti-angiogenic maintenance / thoracic RT + IO / LDRT + IO) are the most active battlegrounds for 2026-2028.

3. Horizontal: 2026 Current Decision Landscape (Six Dimensions)

Projecting longitudinal evolution onto 2026’s concrete clinical decision tree, the following are six key branchpoints and the evidence underpinning each.

3.1 LS-SCLC Initial Treatment: cCRT + durvalumab consolidation, new standard

2026 mainstream: curative-intent LS-SCLC fit patients = cCRT (cisplatin/etoposide × 4-6 cycles + thoracic RT 45 Gy BID / 3 weeks or 66-70 Gy QD / 6-7 weeks) → post-cCRT non-progressors durvalumab 1500 mg q4w consolidation up to 24 months → CR patients PCI 25 Gy or brain MRI surveillance.

Key branchpoints:

Subgroup	First choice	Alternative
fit LS-SCLC standard risk	cCRT (45 Gy BID / 66 Gy QD) → durvalumab consolidation (ADRIATIC) → PCI 25 Gy [TURRISI PMID 9920950 / CONVERT PMID 28642008 / ADRIATIC PMID 39268857 / PCI99-01 PMID 19386548]	70 Gy QD [CALGB 30610 PMID 36623230]
fit LS-SCLC Chinese center (exploratory)	54 Gy SIB BID [YU-2024 PMID 39146944] → durvalumab consolidation	Return to 45 Gy BID + ADRIATIC consolidation
IO intolerant / active autoimmunity	cCRT standard dose + observation (no durvalumab consolidation)
Elderly / PS 2 LS-SCLC	Sequential chemoradiation (SeqCRT) instead of cCRT, toxicity de-escalation
cCRT failure / progression	2L per ES-SCLC 2L+ decision

Not recommended in 2026: concurrent atezolizumab + cCRT — NRG LU005 (PMID 41529214) HR 1.03 negative + OS numerically lower (31.1 vs 36.1 months); “concurrent IO + cCRT” in LS-SCLC does not replicate the PACIFIC stage III NSCLC script.

3.2 ES-SCLC 1L: IO + platinum/etoposide class effect + Chinese PD-1 accessibility branch

2026 mainstream: ES-SCLC fit treatment-naïve = IO (PD-(L)1 inhibitor, 6-7 choices) + carboplatin or cisplatin + etoposide × 4-6 cycles → IO maintenance until progression or intolerance.

Key branchpoints:

Subgroup	First choice	Alternative
fit global / Western	atezolizumab + carboplatin + etoposide [IMpower133 PMID 30280641 + 5y PMID 39306923] or durvalumab + EP/EC [CASPIAN PMID 31590988]	EP/EC alone (IO intolerant)
fit China / Asia	serplulimab + carboplatin + etoposide [ASTRUM-005 PMID 36166026] or adebrelimab + carboplatin + etoposide [CAPSTONE-1 PMID 35576956] or tislelizumab + EP/EC [RATIONALE-312 PMID 38460751] or toripalimab + EP [EXTENTORCH PMID 39541202]	atezolizumab / durvalumab (when accessible)
fit, seeking highest OS + willing to tolerate ~93% G3+ toxicity	benmelstobart + anlotinib + EC four-drug [ETER701 PMID 38992123]	Three-drug IO + platinum/etoposide
ECOG 2 / heavy comorbidities / elderly	IO + carboplatin + etoposide (avoid cisplatin), strengthen BSC	EP alone
Nephrotoxicity, cisplatin intolerant	lobaplatin + etoposide [EL-PHASE3-CHINA PMID 30988825] ± IO	carboplatin + etoposide
High risk of severe myelosuppression	Consider trilaciclib myeloprotection pre-dosing [TRACES PMID 38224653]	Dose reduction

Clinical meaning of “long-tail survivor”: IMBRELLA-A 5y OS 12% (PMID 39306923) tells us — ES-SCLC IO + chemo benefit is not uniformly lifting everyone 2-3 months, but giving about 12% of “durable responders” a 5-year+ survival window. When communicating prognosis, clinicians should clearly distinguish the dual-layer structure of “median OS 13-15 months (majority) + 5-year OS 12% (long tail).”

NCCN 2026: IO + platinum/etoposide = Category 1 preferred (atezolizumab / durvalumab globally; serplulimab / adebrelimab / tislelizumab / toripalimab in China based on CSCO 2025).

3.3 PCI Decision: Is PCI Still Needed in the IO Era?

2026 mainstream:

LS-SCLC cCRT + durvalumab consolidation CR patients: PCI 25 Gy remains SoC (PCI99-01 PMID 19386548 dose-locked); some centers switch to brain MRI q3-6 month surveillance as a PCI substitute (lacks head-to-head RCT)
ES-SCLC chemo + IO responders: Slotman 2007 (PMID 17699816) PCI 25 Gy remains standard evidence → but in the IO era (IMpower133 / CASPIAN both protocol-permitted PCI by investigator choice), actual usage has declined; brain MRI surveillance is the European alternative in some centers
HA-PCI (hippocampal-avoidance): RTOG-0212 PMID 33545387 did not show clear cognitive protection advantage, inconsistent with earlier small-scale data; NCCN retains as optional but not strongly recommended
Thoracic RT (TRT): CREST PMID 25230595 2-year OS 13% vs 3% significant + 1-year not significant — whether to add TRT in the IO era is in guideline “optional” status; TREASURE PMID 36153496 “IO + TRT” terminated due to toxicity suggests risk

Key controversy: is PCI still needed in the IO era? — this is the most divisive clinical question in SCLC 2026. No IO vs PCI head-to-head RCT; some European centers have already substituted MRI for PCI; Asian / North American majority still retain PCI. MAVERICK / PRIMALung phase III RCTs are enrolling but no readout as of 2026-04.

3.4 ES-SCLC 2L+: tarlatamab preferred + topotecan alternative + IO monotherapy not used

2026 mainstream:

Subgroup	First choice	Alternative
DLL3+ SCLC post-1L IO + chemo progression	tarlatamab 10 mg IV q2w [DeLLphi-301 PMID 37861218]	IV or oral topotecan
DLL3 not tested / not enriched	Oral topotecan [OBRIEN-TOPOTECAN-2006 PMID 17135646 vs BSC]	IV topotecan / lurbinectedin (FDA 2020 accelerated)
Prior good IO response + tarlatamab CRS intolerant	Rechallenge IO + chemo (if interval > 6 months)	Clinical trial
MSI-H / dMMR / TMB-H rare SCLC subgroup	pembrolizumab tumor-agnostic	nivolumab tumor-agnostic
3L+ standard exhausted	Clinical trial > regorafenib / ADC / CAR-T exploration	BSC

Not recommended in 2026:

nivolumab or pembrolizumab monotherapy in unselected SCLC 2L+ — CheckMate-032 (PMID 27269741) and KEYNOTE-028/158 (PMID 31870883) SCLC accelerated approvals were withdrawn by FDA 2020-12 / 2021-03 respectively. Only tumor-agnostic MSI-H / TMB-H subgroups retain IO monotherapy use.
CAV three-drug — single-agent topotecan is non-inferior (PMID 10080612) + single-agent more convenient + lower toxicity.

3.5 Maintenance / Combination Exploration: Four Directions Flourishing

2026 mainstream: ES-SCLC 1L IO + chemo × 4-6 cycles → IO maintenance until progression. On this foundation, four “IO maintenance intensification” directions are being explored:

Direction	Representative trial	2026 clinical status
tarlatamab maintenance (DLL3 BiTE + IO)	DeLLphi-303 [PMID 40934933] + phase III NCT06211036	Phase II 25.3-month mOS signal strong; phase III ongoing
anti-angiogenic + IO maintenance	DURABLE [PMID 37947242] durvalumab + anlotinib maintenance vs durvalumab	Protocol-only, awaiting primary readout; conference-reported PFS signal
Thoracic RT (TRT) + IO maintenance	TREASURE [PMID 36153496] (terminated due to toxicity)	TRT + IO maintenance approach failed; CREST-era TRT retained as standalone
Low-dose RT (LDRT) + IO + chemo concurrent	MATCH-TRIAL-SCLC [PMID 40609842]	Phase II ORR 87.5% signal strong, awaiting phase III validation

Clinical implication: these four directions represent “how much more benefit can be squeezed from ES-SCLC 1L maintenance” — DeLLphi-303 is most likely to rewrite 1L maintenance standard in 2026-2028 (phase III NCT06211036); MATCH LDRT + IO may reopen thoracic RT’s IO-era value; anti-angiogenic maintenance may be replaced by ETER701’s four-drug 1L regimen, eliminating the need for sequential “maintenance + anti-angio.”

3.6 Elderly / Poor PS SCLC + Transformed SCLC Subgroups

Elderly (≥75 years) / PS 2 ES-SCLC: IMpower133 / CASPIAN / ASTRUM-005 and other IO + chemo phase IIIs predominantly enrolled PS 0-1 (PS 2 <10%). Clinical decisions:

PS 2 but chemo-tolerant: dose-reduced IO + carboplatin (avoid cisplatin) + etoposide + trilaciclib myeloprotection (TRACES PMID 38224653)
PS 3-4 / multi-organ failure: BSC + palliative RT (symptomatic brain metastases / bone metastases)
85+ ES-SCLC: single-agent etoposide or single-agent IO (no dedicated RCT)

Transformed SCLC (EGFR TKI-resistant NSCLC transformed to SCLC): rare subgroup (about 3-10% of EGFR-mutated NSCLC post-osimertinib transforms to SCLC). No prospective RCT; clinical experience is to use SCLC-like chemo (EP/EC) + consider continuing osimertinib (to protect residual NSCLC clones). DeLLphi-301 and other tarlatamab trials typically exclude transformed SCLC — this subgroup has limited 2L+ treatment options, prioritize clinical trial enrollment.

4. Research Gaps: Top Ten Unresolved Clinical Questions

This report identifies the following gaps, all definable specific questions (not platitudes of “more research needed”):

“Within-group differences” across IMpower133 / CASPIAN / Chinese PD-1 series remain unresolved: HR 0.70-0.80 narrow-band convergence = class effect, but IMBRELLA-A 5y OS 12% means 88% of patients still progress rapidly — who is the long-tail responder? Lacks pre-specified PD-L1 / TMB / ASCL1 / dMMR / IFN-γ signature prospective validation.
PCI in the IO era vs MRI surveillance head-to-head missing: PCI standards established by Slotman 2007 and PCI99-01 have not been updated for the IO era; MAVERICK / PRIMALung phase III ongoing but no readout as of 2026-04 — East-West practice divergence (European MRI substitution / Asian-North American PCI retention) lacks evidence-based resolution.
LS-SCLC cCRT sequencing (concurrent vs sequential) contemporary data missing: TURRISI 1999 + CONVERT 2017 + CALGB 30610 2023 are all concurrent RT; sequential chemo + RT only as PS 2 alternative, no IO-era (post-ADRIATIC) concurrent vs sequential head-to-head data.
DLL3 biomarker cutoff not standardized: DeLLphi-301 enrolled DLL3 IHC ≥50% as marker, but is tarlatamab response cORR 40% limited to high expressors? Does any DLL3 expression also have activity? DLL3 IHC assay inter-laboratory variability + cutoff choice = key gap for future SCLC basket trials.
Transformed SCLC (EGFR TKI-resistant transformation) 1L/2L regimens lack prospective data: about 3-10% of NSCLC post-osimertinib transforms to SCLC; clinical decisions are entirely empirical. DeLLphi-301 and other tarlatamab trials exclude this subgroup; transformed SCLC-specific prospective cohort missing.
Clinical actionability of SCLC molecular subtypes (ASCL1 / NEUROD1 / POU2F3 / YAP1): since 2018 Rudin et al defined 4 molecular subtypes; POU2F3 high expression may correlate with lurbinectedin response, YAP1 with IO primary resistance — but no prospective stratified trial readout as of 2026-04, subtyping not routinely applied clinically.
Elderly / PS 2 ES-SCLC IO data sparse: IMpower133 / CASPIAN dominated by PS 0-1, PS 2 subgroup <10%; dedicated elderly / PS 2 SCLC IO RCT missing — clinical decisions based on post-hoc subgroup analyses rather than prespecified.
SCLC CNS progression local (SRS / WBRT) vs systemic (tarlatamab CNS activity) decision not standardized: tarlatamab CNS penetration data limited; IO + chemo CNS ORR not separately reported. Optimal sequencing of SRS + systemic therapy lacks RCT.
Post-2L+ 3L+ options (post-tarlatamab resistance): tarlatamab 2L response 6.9-month mDoR post-resistance has no standard; lacks tarlatamab resistance molecular mechanism characterization + 3L+ ADC / CAR-T / rechallenge chemo data.
Chinese PD-1 class effect internal differences (ASTRUM-005 HR 0.63 vs EXTENTORCH HR 0.80): HR spans 0.63-0.80 (18 percentage points) — is this a true difference or cross-trial population / control / follow-up duration difference? Lacks Chinese PD-(L)1 head-to-head RCT.

5. 2024-2026 Latest Developments

5.1 FDA / NMPA New Approvals and Guideline Expansions

2024-05-16 FDA accelerated approval: tarlatamab (Imdelltra) DLL3 × CD3 BiTE for SCLC 2L+ post-platinum chemo progression / DeLLphi-301 [PMID 37861218] — first non-chemotherapy non-IO new mechanism in SCLC in half a century
2024-08-05 FDA approval: durvalumab (Imfinzi) consolidation for LS-SCLC post-cCRT non-progressors / ADRIATIC [PMID 39268857] — first OS-positive phase III in LS-SCLC in 30 years changing SoC
2022-03 NMPA approval + 2024 global follow-up: adebrelimab (SHR-1316) + EC for ES-SCLC 1L / CAPSTONE-1 [PMID 35576956]
2022-10 NMPA approval: serplulimab (HLX10) + EC for ES-SCLC 1L / ASTRUM-005 [PMID 36166026]
2024 NMPA expanded approval: tislelizumab / toripalimab + EP/EC for ES-SCLC 1L / RATIONALE-312 [PMID 38460751] + EXTENTORCH [PMID 39541202]
2024-01 NMPA approval: benmelstobart + anlotinib + EC four-drug regimen for ES-SCLC 1L / ETER701 [PMID 38992123]
2021-02 FDA approval + 2024 China follow-up: trilaciclib (Cosela) myeloprotection for ES-SCLC chemotherapy / TRACES [PMID 38224653]

2020-12 / 2021-03 FDA withdrawals: nivolumab (CheckMate-032 PMID 27269741) and pembrolizumab (KEYNOTE-028/158 PMID 31870883) SCLC 3L monotherapy indications withdrawn due to confirmatory phase III failures — SCLC IO monotherapy 2L+ no longer recommended.

5.2 Key Conference Readouts (2024-2025, Down-Weighted Labeling)

The following entries serve only as candidate pool pending formal peer review and do not enter the main database. Items with PMIDs have been promoted.

ADRIATIC OS update (ASCO 2024 LBA5 → 2024-09 NEJM [PMID 39268857]): LS-SCLC cCRT + durvalumab consolidation mOS 55.9 vs 33.4 months — promoted to main database
DeLLphi-301 long-term follow-up (WCLC 2024 / ESMO 2024 oral): tarlatamab 10 mg long-term follow-up ORR + DoR maintained stable; does not change NEJM primary PMID 37861218 data
DeLLphi-302 (tarlatamab 2L+ placebo-controlled phase III): expected 2026 H2 primary readout, NCT05740566 enrollment complete — will provide first phase III head-to-head evidence for tarlatamab vs topotecan / lurbinectedin
ETER701 updated OS (ASCO 2024 / ESMO 2024): benmelstobart + anlotinib + EC four-drug regimen ES-SCLC 1L mOS 19.3 months — promoted to main database
DURABLE OS readout (expected 2025): durvalumab + anlotinib maintenance primary results still pending peer review; PMID 37947242 is protocol paper

5.3 Ongoing Phase III (Selected 2025-2028 Readouts)

DeLLphi-304 (NCT06211036) tarlatamab + atezolizumab/durvalumab 1L maintenance vs atezolizumab/durvalumab maintenance — expected 2026-2028 readout, the key trial to rewrite ES-SCLC 1L maintenance standard
DeLLphi-302 (NCT05740566) tarlatamab vs investigator choice (topotecan / lurbinectedin / topotecan + cyclophosphamide) in SCLC 2L — confirms 2L phase III-level evidence
MAVERICK / PRIMALung ES-SCLC post-chemo + IO response PCI vs MRI surveillance — head-to-head RCT, 2027-2028 readout
IMforte (NCT05091567) ES-SCLC 1L post-chemoimmunotherapy atezolizumab ± lurbinectedin maintenance — expected 2026 readout
Multiple DLL3 ADCs / CAR-T / other BiTEs (MK-6070 / HPN328 / BI 764532) phase I/II ongoing in SCLC 2L+

6. Intersection Insights and Judgment

6.1 Longitudinal × Horizontal: The 2026 SCLC Landscape Is Shaped by Three Resonances

Stacking the longitudinal paradigm evolution on the horizontal current decision landscape, the 2026 SCLC treatment landscape is actually shaped by three resonance overlays:

Chemotherapy 20-year plateau (mOS 9-10 months) → IO + chemo class effect (mOS 12-15 months + 5y long tail 12%): the EP backbone barely moved from 1999 TURRISI to 2018 IMpower133, and all attempts to add a third drug failed. After IO entered 1L, HR converged in 0.63-0.80 class effect, but median OS only gained 2-5 months — the real structural change is the 5-year OS long tail (IMBRELLA-A 5y 12%, chemo-alone historical 3-4%). Clinical prognosis communication should distinguish the dual-layer structure of “median OS” vs “long tail.”
LS-SCLC ADRIATIC first upgrade in 30 years + ES-SCLC class effect saturation + DLL3 BiTE breakthrough as a three-way standoff: 2024 was the densest year in SCLC history — ADRIATIC around May raised LS-SCLC cCRT-post mOS from 33 months to 56 months (+23 months, large delta); tarlatamab approval in May opened a non-chemotherapy new mechanism for 2L+; meanwhile ES-SCLC 1L class effect has saturated (six Chinese PD-1s in parallel, HR 0.63-0.80 differences driven mainly by population/control/follow-up).
“Biological stratification” of IO in SCLC — monotherapy failure vs combination success: CheckMate-032 + KEYNOTE-028/158 IO monotherapy ORR 10-22% but confirmatory phase III failed → FDA withdrew; the same IO after + chemo OS-positive → globally approved. This says SCLC cold-tumor biology (low TMB / immune-desert / no stable neoantigens) sets a low ceiling for IO monotherapy, and chemotherapy debulking + IO checkpoint release + maintenance is the minimum recipe for IO efficacy in SCLC — IO alone does not work in SCLC.

These three resonances together explain a clinical phenomenon: the decision tree for a newly diagnosed SCLC patient in 2026 has 3 more decision layers than in 2018 (LS vs ES staging + whether post-cCRT durvalumab consolidation + 1L IO 6-choose-1 including Chinese PD-1 accessibility branch + 2L tarlatamab DLL3 biomarker stratification).

6.2 Clinical Decision Takeaways (for Junior-Mid Oncologists)

After LS-SCLC cCRT, you must evaluate durvalumab consolidation eligibility: ADRIATIC (PMID 39268857) mOS 55.9 vs 33.4 months is the largest LS-SCLC delta in 30 years. Missing ADRIATIC consolidation = leaving the patient in the cCRT era.
Do not use “concurrent IO + cCRT” in LS-SCLC: NRG LU005 (PMID 41529214) HR 1.03 refuted the PACIFIC-like concurrent path’s replicability in SCLC. cCRT first, then durvalumab sequential.
ES-SCLC 1L IO class effect 6-choose-1: atezolizumab / durvalumab / serplulimab / adebrelimab / tislelizumab / toripalimab HR 0.63-0.80 narrow band; which one to pick depends on accessibility + reimbursement + tolerance. Don’t agonize over “which IO is better” — there is no phase III-level answer.
ES-SCLC 1L long-tail 12% 5-year survival is a key communication point: IMBRELLA-A (PMID 39306923) tells us IO + chemo is not a uniform 2-3 month extension, but gives 12% of patients a 5-year+ long tail. Clinical communication should clearly articulate the dual-layer “median + long tail” prognosis.
2L+ DLL3 testing + tarlatamab preferred: DeLLphi-301 (PMID 37861218) + FDA 2024-05 accelerated approval = first non-chemotherapy new mechanism in SCLC in half a century. DLL3 IHC testing + step-up dosing + CRS management form the three-part foundation of 2026 SCLC 2L basics.
IO monotherapy not recommended in SCLC 2L+: CheckMate-032 + KEYNOTE-028/158 accelerated approvals were withdrawn by FDA 2020-12 / 2021-03 respectively. Only tumor-agnostic MSI-H / TMB-H rare subgroups retained.
Tradeoff between ETER701 four-drug vs three-drug vs two-drug IO + EC: four-drug mOS 19.3 months is historical peak but G3+ TRAE 93% approaches universal; selection requires assessing performance status + comorbidities + tolerance expectations. Don’t default to “highest number = best.”
Attitude toward PCI in the IO era: post-LS-SCLC CR PCI 25 Gy remains standard (PCI99-01); post-ES-SCLC IO response PCI’s status has declined, brain MRI surveillance has become a European alternative — before head-to-head RCTs, PCI vs MRI is a matter of “center preference,” not “right or wrong.”
Transformed SCLC — prioritize clinical trials: EGFR TKI-resistant NSCLC transformed SCLC subgroup is typically excluded by DeLLphi-301 and other trials; standard EP/EC chemo predominates + consider continuing osimertinib to protect residual NSCLC clones.
Seven drug classes essential for SCLC in 2026: durvalumab (LS consolidation + ES 1L) / atezolizumab (ES 1L) / serplulimab + adebrelimab + tislelizumab + toripalimab (Chinese ES 1L 4-choose-1) / benmelstobart + anlotinib (ETER701 four-drug core) / tarlatamab (2L+ DLL3 BiTE) / topotecan (2L non-BiTE backstop) / trilaciclib (myeloprotection) — before 2018 SCLC had only EP + topotecan as two drugs; 2026 already has a 7+ class decision map across 5 pathways.

7. Information Sources

The 30 trials’ metadata in this report were independently cross-verified via PubMed and ClinicalTrials.gov. Each [PMID xxxxxxxx] in the text can be verified directly on PubMed.

Published trials: 28, covering 1999-2026 (PMID verifiable)
Protocol / design papers (primary results pending): 2 (DURABLE PMID 37947242 / TREASURE PMID 36153496 — the latter terminated due to toxicity)
NCCN guideline citations: 30/30 directly hit NCCN Small Cell Lung Cancer V1.2026 reference section or extended evidence base
2020-2025 FDA / NMPA new approvals: 10+ (durvalumab LS consolidation / atezolizumab 1L / serplulimab / adebrelimab / tislelizumab / toripalimab / benmelstobart + anlotinib / tarlatamab / trilaciclib)
2020-2021 FDA withdrawals: 2 (nivolumab SCLC 3L / pembrolizumab SCLC 3L — confirmatory phase III failed)
2024-2026 key conference readouts: 3 (DeLLphi-301 long-term / DeLLphi-302 phase III expected 2026 / ETER701 updated OS)
Research gaps: 10

7.1 Report Body Citation Index (Sorted by PMID Ascending)

The table below lists PMIDs bracket-cited in the body, each clickable on PubMed for verification.

PMID	Trial / Paper	Year	Journal	Body location §x.x
9920950	INT-0096 (TURRISI 1999)	1999	N Engl J Med	§2.1 / §3.1
10080612	TOPOTECAN-ORIGINAL-PHASE2	1999	J Clin Oncol	§2.2 / §3.4
15923572	NIELL-2002-INTERGROUP	2005	J Clin Oncol	§2.2
17135646	OBRIEN-TOPOTECAN-2006	2006	J Clin Oncol	§2.2 / §3.4
17699816	SLOTMAN-PCI-2007	2007	N Engl J Med	§2.3 / §3.3
19386548	PCI99-01	2009	Lancet Oncol	§2.3 / §3.1 / §3.3
25230595	CREST (SLOTMAN-2015)	2015	Lancet	§2.3 / §3.3
27269741	CHECKMATE-032	2016	Lancet Oncol	§2.5 / §3.4 / §5.1
28642008	CONVERT	2017	Lancet Oncol	§2.1 / §3.1
30280641	IMPOWER133	2018	N Engl J Med	§2.4 / §3.2
30988825	EL-PHASE3-CHINA	2019	Oncol Lett	§2.2 / §3.2
31590988	CASPIAN	2019	Lancet	§2.4 / §3.2
31870883	KEYNOTE-028-158	2020	J Thorac Oncol	§2.5 / §3.4 / §5.1
33545387	RTOG-0212-HIPPO (Belderbos 2021)	2021	J Thorac Oncol	§2.3 / §3.3
35576956	CAPSTONE-1	2022	Lancet Oncol	§2.4 / §3.2 / §5.1
36153496	TREASURE	2022	BMC Cancer	§2.5 / §3.5
36166026	ASTRUM-005	2022	JAMA	§2.4 / §3.2 / §5.1
36623230	CALGB-30610 / RTOG-0538	2023	J Clin Oncol	§2.1 / §3.1
37861218	DELLPHI-301	2023	N Engl J Med	§2.5 / §3.4 / §5.1 / §6.2
37947242	DURABLE (protocol)	2023	Clin Respir J	§2.5 / §3.5
38224653	TRACES	2024	Lung Cancer	§2.5 / §3.2 / §5.1
38460751	RATIONALE-312	2024	J Thorac Oncol	§2.4 / §3.2 / §5.1
38992123	ETER701	2024	Nat Med	§2.4 / §3.2 / §5.1
39146944	YU-2024-HYPERFRAC	2024	Lancet Respir Med	§2.1 / §3.1
39268857	ADRIATIC	2024	N Engl J Med	§2.1 / §3.1 / §5.1 / §6.2
39306923	IMBRELLA-A (IMpower133 5y)	2024	Lung Cancer	§2.4 / §3.2 / §6.2
39541202	EXTENTORCH	2025	JAMA Oncol	§2.4 / §3.2 / §5.1
40609842	MATCH-TRIAL-SCLC	2026	Int J Radiat Oncol Biol Phys	§2.5 / §3.5
40934933	DELLPHI-303	2025	Lancet Oncol	§2.5 / §3.5 / §5.3
41529214	NRG-LU005	2026	J Clin Oncol	§2.1 / §3.1

7.2 Verification Conventions

Each PMID is directly accessible via https://pubmed.ncbi.nlm.nih.gov/{PMID}/ for verification
Each NCT id is accessible via https://clinicaltrials.gov/study/{NCT_id}/
Conference abstracts (ASCO / ESMO / WCLC) are retrieved via the official meeting systems; all conference citations in this report are “down-weighted” — not peer-reviewed, final data subject to journal publication
DURABLE + TREASURE cited as protocol / design paper PMIDs; entries will be updated once primary efficacy manuscripts are published
If a PMID in this report is found to differ from PubMed on trial name / year / conclusion, corrections are welcome

Clinical Trial Timelines Here

Chinese: /trials/sclc/ English: /en/trials/sclc/

Each trial has an independent detail page including:

Complete intervention / comparator regimens
Primary endpoint values + 95% CI
Key findings + clinical significance
Clickable links to PMID / NCT originals

30 trials · 5 paradigms · 1999 to 2026 · synchronized with NCCN Small Cell Lung Cancer V1.2026.

Closing

SCLC has completed a unique evolutionary arc over the past 27 years — from 1999 TURRISI INT-0096 establishing LS-SCLC cCRT + BID 45 Gy, through the pre-2018 ES-SCLC 1L EP-backbone 20-year plateau with mOS stable at 9-10 months, to 2018 IMpower133 and 2019 CASPIAN opening the IO marginal-victory breakthrough with HR converging in 0.70-0.73 class effect.

2024 was the densest year in SCLC history: ADRIATIC upgraded LS-SCLC’s 30-year SoC to cCRT + durvalumab consolidation (mOS 55.9 vs 33.4 months, +23-month large delta); DeLLphi-301 tarlatamab received FDA accelerated approval, becoming the first non-chemotherapy non-IO new mechanism in SCLC in half a century; meanwhile five Chinese PD-1 / PD-L1 phase IIIs (ASTRUM-005 / CAPSTONE-1 / RATIONALE-312 / EXTENTORCH / ETER701) pushed ES-SCLC 1L into a “flourishing” era.

This “20-year plateau + IO marginal victory + DLL3 BiTE breakthrough” three-stage pattern is entirely different from NSCLC’s “5 paradigm shifts + 10 drivers all entering 1L + IO rewriting the backbone.” The driving force is SCLC’s unique biology — no unified driver mutation + neuroendocrine differentiation + low TMB + cold tumor + rapid chemoresistance + MYC/ASCL1/NEUROD1/POU2F3/YAP1 molecular subtype heterogeneity. SCLC’s paradigm breakthrough is not “finding a driver gene” but “finding a targetable cell-surface molecule (DLL3) + new mechanism (BiTE).”

For a newly diagnosed SCLC patient in 2026, the core decision-tree branchpoints are “first stage (LS vs ES) → post-cCRT durvalumab consolidation yes/no → 1L IO 6-choose-1 including Chinese accessibility branch → 2L DLL3 biomarker testing + tarlatamab preferred → PCI vs MRI per center preference.”

The value of this report lies not in “enumerating all trials” (PubMed can do that), but in compressing 27 years of evolution + current decisions + unresolved gaps into the cognitive bandwidth of a single reading. Next time you face a newly diagnosed SCLC patient, every branchpoint in the decision tree has this map to check, trace, and interrogate.

Clinician × AI = Research Superpower + Clinical Decision Amplifier

—— Dual Brain Lab · 2026-04-21

NSCLC Clinical Trial Timeline · 25-Year Evolution Map

Fri, 17 Apr 2026 00:00:00 +0000

NSCLC Clinical Trial Timeline · In-depth Report

Coverage: 111 published landmark trials (all PMID-traceable) + 90 ongoing phase III + 20 FDA/NMPA 2024-2026 approvals + 21 key conference readouts

Curated by Dual Brain Lab (csilab.net)

1. One-Sentence Definition

This report maps the systemic therapy of non-small cell lung cancer (NSCLC) over the past 25 years (2002–2026) through the landmark clinical trials cited in the current NCCN NSCLC and CSCO NSCLC 2025 guidelines — providing frontline clinicians in 2026 a traceable full-view map for the “who, what, and why” of decision-making.

Iron rule: every datapoint in every trial traces back to PubMed (PMID) or ClinicalTrials.gov (NCT id). Each [PMID xxxxxxxx] in the body opens directly to PubMed for source verification.

2. Longitudinal: Five-Paradigm Evolution Timeline

NSCLC systemic therapy has gone through five paradigm shifts in the past 25 years: the chemotherapy plateau was broken by driver mutations → TKIs (tyrosine kinase inhibitors) climbed from 2L to 1L to adjuvant → immune checkpoint inhibitors (ICIs) rewrote 2L and then 1L → IO-chemo became the new backbone → perioperative IO rewrote the treatment strategy for resectable disease.

Each shift has 3–5 phase III trials that pushed the previous standard of care (SoC) to second line.

2.1 Chemotherapy Plateau (2002–2015): Established, Challenged, Replaced

Story: The OS median of chemotherapy doublets was locked at ~8 months by ECOG 1594 in 2002. That ceiling stood until driver mutations and checkpoint inhibitors finally broke it.

ECOG 1594 [PMID 11784875] (Schiller 2002 NEJM, N=1207): four platinum doublets randomized — cis+pac vs cis+gem vs cis+doc vs carbo+pac. All four OS curves completely overlapped, mOS ~7.9 months, 1-year OS 33%. The result was both a starting point and a ceiling — “no matter which doublet you pick, it doesn’t matter.”
TAX 326 [PMID 12837811] (Fossella 2003 JCO, N=1218): docetaxel+cis slightly better than navelbine+cis, mOS 11.3 vs 10.1 months. Expanded doublet options but did not break the ceiling.
JMDB / Scagliotti 2008 [PMID 18506025] (N=1725): in non-squamous NSCLC, cis+pemetrexed beat cis+gemcitabine (mOS 11.8 vs 10.4 months) — the first phase III evidence for histology-stratified therapy. From that point on, “non-squamous takes pemetrexed” entered the guidelines.
PARAMOUNT [PMID 22341744] (Paz-Ares 2012 Lancet Oncol): pemetrexed maintenance significantly extended PFS, establishing the 2012 “induction + maintenance” standard for non-squamous NSCLC.

Takeaway: the chemo backbone was refined between 2002 and 2012 (doublet → histology stratification → maintenance), but the OS ceiling did not move. After 2015, with IPASS behind, driver-negative + squamous — the “hard-to-treat” subgroups — finally got breakthroughs from IO.

2.2 EGFR TKI Era (2009–2026): 2L → 1L → Adjuvant → Unresectable III → Combo

Story: EGFR-mutation TKI therapy is the prototype story of precision oncology in lung cancer. Each TKI generation pushed median PFS ~1.5× higher, and every push came with a pivotal phase III.

IPASS [PMID 19692680] (Mok 2009 NEJM, N=1217): gefitinib vs carbo+pac in an East Asian lung adenocarcinoma population. Overall HR 0.74, but EGFR-mutant subgroup HR 0.48 and wild-type subgroup HR 2.85. The “EGFR mutation test → stratified therapy” clinical paradigm was born here.
OPTIMAL [PMID 21783417] (Zhou 2011 Lancet Oncol) + EURTAC [PMID 22285168]: erlotinib vs chemotherapy in EGFRm 1L phase III, PFS doubled (mPFS 13.1 vs 4.6 months). 2nd-gen TKI (afatinib) LUX-Lung 3/6 consolidated further.
FLAURA [PMID 29151359] (Soria 2018 NEJM, N=556): osimertinib (3rd-gen EGFR TKI) vs 1st/2nd-gen in EGFRm 1L. mPFS 18.9 vs 10.2 months, HR 0.46, with stronger CNS activity. 2019 FLAURA OS update [PMID 31751012] was the first TKI to show OS benefit (mOS 38.6 vs 31.8 months, HR 0.80).
ADAURA [PMID 32955177] (Wu 2020 NEJM, OS 2023 [PMID 37272535]): resectable IB-IIIA EGFRm NSCLC, 3-year adjuvant osimertinib vs placebo. DFS HR 0.17, OS HR 0.49. Changed EGFRm early recurrence from “wait then treat” to “eradicate residual disease.”
LAURA [PMID 38828946] (Lu 2024 NEJM): unresectable stage III EGFRm NSCLC, osimertinib consolidation post-CRT (chemoradiation) vs placebo. mPFS 39.1 vs 5.6 months, HR 0.16. Filled the gap that PACIFIC left open for EGFRm patients unsuitable for durvalumab.
FLAURA2 [PMID 37937763] (Planchard 2023 NEJM) + MARIPOSA [PMID 38924756] (Cho 2024 NEJM, OS update 2025 [PMID 40923797]): two combo regimens — osi+chemo and amivantamab+lazertinib — pushed mPFS to 25.5 / 23.7 months, beating osi monotherapy’s 16.6 months. MARIPOSA OS HR 0.75, 3-year OS 60% vs 51% — the first mature 1L EGFRm OS benefit.
MARIPOSA-2 [PMID 37879444] + TROPION-Lung01 [PMID 39250535] (Ahn 2025 JCO): ADC (antibody-drug conjugate) strategies for post-osimertinib resistance. Dato-DXd PFS HR 0.38 in the EGFR-mutant subgroup.
SACHI [PMID 41544643] (Lu 2026 Lancet): in the post-osi MET-amplified subgroup, savolitinib+osimertinib vs chemotherapy. mPFS 8.3 vs 3.6 months, HR 0.34. First randomized controlled evidence for “treat-by-resistance-mechanism.”

Takeaway: EGFRm therapy has evolved from the 2009 “TKI vs chemo” binary to a 2026 four-dimensional decision tree — “stage × line × resistance mechanism × tolerance.” Combo (FLAURA2 / MARIPOSA) and mono (FLAURA) coexist; the choice between them comes down to tolerance and disease burden.

2.3 Non-EGFR Drivers (2014–2026): From Basket to Dedicated 1L

Story: Eight drivers — ALK, ROS1, MET, KRAS, HER2, RET, BRAF, NTRK — each walked the path “accelerated approval on phase 2 → pivotal phase 3 → pushed into 1L” between 2014 and 2026, but at very different speeds.

Driver	Key trials (PMID)	Line
ALK	PROFILE 1014 [PMID 25470694] (2014) → ALEX [PMID 28586279] (2017) → CROWN [PMID 33207094] (2020, lorlatinib) → ALINA [PMID 38598794] (2024, adjuvant alectinib)	Fully in 1L + adjuvant
ROS1	PROFILE 1001 → TRIDENT-1 [PMID 38197815] (2024, repotrectinib 1L ORR 79%)	1L (repotrectinib replaces crizotinib)
MET ex14	GEOMETRY mono-1 [PMID 32877583] (capmatinib 2020) / VISION [PMID 32469185] (tepotinib 2020)	Phase II accelerated approval; no phase III; 1L still contested
KRAS G12C	CodeBreaK 100 [PMID 34096690] → CodeBreaK 200 [PMID 36764316] (sotorasib 2L) → KRYSTAL-12 [PMID 40783289] (adagrasib 2L) → divarasib 1/2 [PMID 37611121] (phase 3 pending)	2L only
HER2 exon 20	DESTINY-Lung02 [PMID 37694347] (T-DXd 2L, ORR 49%) → Beamion LUNG-1 [PMID 40293180] (zongertinib 2L, ORR 71%)	2L; 1L still chemo-IO
RET	LIBRETTO-431 [PMID 37870973] (selpercatinib 1L vs chemo+pembro, mPFS 24.8 vs 11.2 months)	1L established
BRAF V600E	PHAROS [PMID 37270692] (encorafenib+binimetinib 1L ORR 75%)	1L
NTRK	larotrectinib / entrectinib tumor-agnostic approvals	Basket, accelerated
EGFR exon 20 insertion	PAPILLON [PMID 37870976] (amivantamab+chemo vs chemo 1L, mPFS 11.4 vs 6.7 months)	1L

Takeaway: speed of 1L entry ranks as ALK > EGFR common > EGFR ex20 > ROS1 > RET > BRAF > MET ex14 ≈ HER2 ex20 > KRAS G12C. KRAS G12C is still stuck at 2L — sotorasib and adagrasib both failed to show OS benefit over docetaxel in phase III. Divarasib’s early data (ORR 53%, mPFS 13 months) may be the next 1L-KRAS breakthrough.

2.4 IO 1L (2015–2024): From 2L Trials to Backbone Rewrite

Story: IO (anti-PD-1 / anti-PD-L1) got its first NSCLC evidence in 2015 with CheckMate-017 / 057 (2L), then KEYNOTE-024 (2016) pushed it to 1L monotherapy for PD-L1 high expressors, and finally KEYNOTE-189 (2018) wrote IO+chemo into the 1L backbone.

CheckMate-017 [PMID 26028407] (squamous) + CheckMate-057 [PMID 26412456] (non-squamous) (Brahmer / Borghaei 2015 NEJM): nivolumab vs docetaxel in 2L NSCLC. OS HR 0.59 (sq) / 0.73 (non-sq). IO enters lung cancer history.
KEYNOTE-010 [PMID 26712084] (Herbst 2016 Lancet): pembrolizumab vs docetaxel in PD-L1 ≥1% 2L. Confirmed PD-L1 as a stratification biomarker.
KEYNOTE-024 [PMID 27718847] (Reck 2016 NEJM, N=305): pembrolizumab monotherapy vs chemotherapy in PD-L1 TPS ≥50% 1L. mOS 30.0 vs 14.2 months, HR 0.60. IO monotherapy 1L established. The 5-year update [PMID 33872070] shows 5-year OS 31.9% vs 16.3% — a beautifully preserved tail.
KEYNOTE-189 [PMID 29658856] (Gandhi 2018 NEJM, N=616): pembrolizumab + pemetrexed + platinum vs placebo+chemo in non-squamous 1L (regardless of PD-L1). OS HR 0.49 (0.38–0.64), OS benefit in every PD-L1 subgroup including TPS<1%. The “~8-month ceiling” of the chemo plateau was finally shattered by the IO-chemo backbone. 5-year update [PMID 36809080]: 5y OS 19.4% vs 11.3%, TPS<1% HR 0.55.
KEYNOTE-407 [PMID 30280635] (Paz-Ares 2018 NEJM): pembro+carbo+(nab-)pac vs placebo+chemo in squamous 1L. 5-year OS 18.4% vs 9.7% [PMID 36735893]. Squamous — previously stuck with only docetaxel — was rewritten.
IMpower110 [PMID 32997907] / IMpower130 / IMpower132 / IMpower150 [PMID 29863955]: atezolizumab combos. IMpower150 (atezo+bev+chemo) retains a niche role in EGFR/ALK post-TKI and liver-metastases subgroups.
CheckMate-227 [PMID 31562796] (Hellmann 2019 NEJM, 5y update 2023 [PMID 36223558]): nivolumab+ipilimumab as a “chemo-sparing” IO-IO regimen. In the PD-L1<1% subgroup, 5y OS 19% vs 7%, HR 0.72.
CheckMate 9LA [PMID 33476593] (Paz-Ares 2021 Lancet Oncol, 5y [PMID 39270380]): nivo+ipi + 2 cycles of chemo — chemo-sparing but keeping the early tumor-debulking effect of chemo, OS HR 0.66.
POSEIDON [PMID 36327426]: durvalumab + tremelimumab + chemo. In the PD-L1<1% subgroup, tremelimumab adds ~3 months of OS.

Takeaway: the core branching of 1L IO selection in 2026 is histology (sq / non-sq) × PD-L1 (<1% / 1-49% / ≥50%) × performance status × combo tolerance. KEYNOTE-189 + KEYNOTE-407 cover nearly the entire PD-L1 spectrum in non-sq and sq; CheckMate-227 / 9LA serve as “don’t want chemo” alternatives; POSEIDON offers an intensified option for PD-L1<1%. Chinese PD-1 agents (sintilimab ORIENT-11 [PMID 35917647], tislelizumab, camrelizumab) dominate in subgroups driven by price and accessibility.

2.5 Perioperative IO (2021–2026): From Adjuvant-Only to Dual-Punch

Story: IO therapy for resectable NSCLC has had 6–8 phase III readouts over 5 consecutive years since 2021. The strategy evolved from “adjuvant only” (post-op) to “neoadjuvant only” (pre-op), and on to “perioperative” (= pre-op + post-op).

IMpower010 [PMID 34555333] (Felip 2021 Lancet, 5y update [PMID 40446184]): resectable IB-IIIA, 1-year adjuvant atezolizumab vs BSC. DFS and OS both benefit in the PD-L1 ≥50% subgroup; benefit unclear for PD-L1 <50%.
KEYNOTE-091 [PMID 36108662] (O’Brien 2022 Lancet Oncol): adjuvant pembrolizumab vs placebo. DFS HR 0.76 overall, but benefit inconsistent across PD-L1 subgroups.
CheckMate 816 [PMID 35403841] (Forde 2022 NEJM, OS 2025 [PMID 40454642]): neoadjuvant nivolumab+chemo × 3 cycles vs chemo alone. pCR (pathological complete response) 24% vs 2.2%, EFS (event-free survival) HR 0.63, OS HR 0.72. First phase III OS benefit in the neoadjuvant setting.
KEYNOTE-671 [PMID 37272513] (Wakelee 2023 NEJM, OS 2024 Lancet [PMID 39288781]): perioperative (neoadjuvant pembro+chemo × 4 → post-op pembro × 13). EFS HR 0.58, OS HR 0.72 — first perioperative trial with OS benefit.
AEGEAN [PMID 37870974] (Heymach 2023 NEJM): perioperative durvalumab+chemo. EFS HR 0.68, pCR 17.2% vs 4.3%.
CheckMate 77T [PMID 38749033] (Cascone 2024 NEJM): perioperative nivo+chemo. EFS HR 0.58, pCR 25.3% vs 4.7%.
Neotorch [PMID 38227033] (Lu 2024 JAMA, China): perioperative toripalimab+chemo in stage II-III. Stage III subgroup EFS HR 0.40.
RATIONALE-315 [PMID 39581197] (Yue 2025 Lancet Respir Med): perioperative tislelizumab+chemo. pCR 40.7% vs 5.7%, EFS HR 0.56.

Takeaway: by 2026 the SoC for resectable stage II-IIIB NSCLC is firmly perioperative IO-chemo (3-4 cycles neoadjuvant + ≥1 year adjuvant IO), with pCR confirmed as a strong surrogate for EFS / OS across multiple trials. The one exception: EGFRm → ADAURA (3-year adjuvant osimertinib) and ALK+ → ALINA (adjuvant alectinib) — these two driver-positive subgroups should avoid perioperative IO.

3. Cross-Sectional: 2026 Decision Landscape (Six Dimensions)

Six dimensions cover the full decision landscape — 1L, important 2L, early stage, PD-L1 low, post-resistance, and rare drivers. Each includes mainstream regimens, areas of controversy, head-to-head / NMA (network meta-analysis) evidence, and current NCCN/CSCO tiers.

3.1 1L IO-combo Selection (EGFR/ALK-WT, PD-L1 ≥1%)

2026 mainstream: pembro+chemo (KEYNOTE-189 non-sq / KEYNOTE-407 sq) remains the global IO-chemo backbone across the PD-L1 spectrum. For PD-L1 TPS ≥50% fit patients, pembro monotherapy (KEYNOTE-024) is still guideline-preferred, but clinical practice drifts toward combo (deeper early response + less “in-hindsight regret”). IMpower150 now only applies in liver-mets and EGFR post-TKI niches. CheckMate 9LA and POSEIDON serve as chemo-sparing / CTLA-4-adding alternatives.

Key branchpoints:

Subgroup	First choice	Alternative
Non-sq PD-L1 ≥50%	pembro mono (KN-024, 5y OS 31.9%) or pembro+pem-plat (KN-189)	CheckMate-227 (PD-L1 ≥1%)
Non-sq PD-L1 1–49%	pembro+pem-plat (KN-189)	CheckMate 9LA
Squamous full PD-L1	pembro+carbo+(nab-)pac (KN-407, 5y OS 18.4% vs 9.7%)	nivo+ipi+chemo (CM-9LA)
Non-sq PD-L1 <1%	pembro+pem-plat (KN-189 TPS<1% 5y HR 0.55)	CheckMate 9LA / CheckMate-227

Head-to-head / NMA:

2025 NMA [PMID 41486797]: pembrolizumab + chemotherapy ranked highest in non-sq (SUCRA HR 0.76); in squamous, nivo+ipi+chemo ranked top (SUCRA 78.3%).
KEYNOTE-189 5y [PMID 36809080] vs KEYNOTE-407 5y [PMID 36735893]: both maintain >5-year OS benefit; squamous HR 0.71 trails non-sq HR 0.60.

Controversies: for PD-L1 ≥50%, is IO-mono ≈ IO-chemo? No H2H; cross-trial data suggest “combo has better PFS but OS converges over time.” ORIENT-11 (domestic sintilimab) vs KEYNOTE-189: large differences in population, comparator arm, and PD-L1 assay mean formal equivalence is impossible — yet in price- and access-driven Chinese practice, ORIENT-11 has already replaced it widely.

NCCN 2026: histology-matched pembro+chemo = Category 1 preferred (non-sq + sq). Pembro mono = Cat 1 only for TPS ≥50%. CheckMate 9LA = Cat 1 alternative.

CSCO 2025: pembrolizumab / sintilimab / tislelizumab + chemo = Level I recommendation (non-sq); pembrolizumab / tislelizumab / camrelizumab + chemo = Level I (squamous); PD-L1 TPS ≥50% pembro mono = Level I.

3.2 EGFRm 1L: Three-Way Fight Between Mono and Combo

2026 mainstream: EGFRm 1L has shifted from osimertinib-monotherapy era into the combo era. FLAURA2 (osi+pem-plat, mOS 47.5 months [PMID 41104938]) and MARIPOSA (ami+laz, OS HR 0.77 [PMID 40923797]) both beat osi mono on OS. Osi mono (FLAURA) survives for patients “not fit for combo” thanks to its low toxicity. MARIPOSA-2 (post-osi progression, ami+chemo ± laz [PMID 37879444]) is the new post-osi standard.

Key branchpoints:

Subgroup	Recommended
Fit + high burden / L858R	FLAURA2 or MARIPOSA (combo advantage most pronounced)
Exon 19del	MARIPOSA numerical benefit slightly larger (HR ~0.65)
CNS met at baseline	MARIPOSA intracranial PFS HR 0.69; osi mono also has sufficient CNS activity
Elderly / low burden	osi mono (FLAURA) — avoids chemo toxicity (FLAURA2 G3+ AE 64%) and ami infusion / VTE burden
Post-osi progression	MARIPOSA-2 (ami+carbo-pem ± laz) Cat 1; MET-amplified → SACHI (savolitinib+osi) [PMID 41544643]

Head-to-head / NMA:

No direct FLAURA2 vs MARIPOSA H2H. Cross-trial: mPFS 25.5 vs 23.7 months, 3y OS 63% vs 60%, but toxicity profiles are entirely different (ami: rash / VTE / infusion; chemo: marrow / neuro).
MARIPOSA VTE issue: protocol amended to require prophylactic anticoagulation in the first 4 months; real-world implementation is uneven.

Controversy: is the combo OS benefit worth the doubled toxicity? The answer is patient-specific — there is no universal winner; patient choice is the core variable.

NCCN 2026: EGFRm 1L Cat 1 options = (a) ami+laz (b) osi+pem-plat (c) osi mono — all three listed without ranking.

CSCO 2025: osi mono = Level I (wide accessibility); FLAURA2 + amivantamab-based combos = Level II (pending NRDL inclusion).

3.3 Perioperative IO: Choosing Among Three Strategies

2026 mainstream: perioperative (3-4 cycles pre-op IO-chemo + ≥1 year post-op IO) has replaced adjuvant-only and neoadjuvant-only as SoC for resectable stage II-IIIB NSCLC. Five phase III trials — KEYNOTE-671 / AEGEAN / CheckMate 77T / Neotorch / RATIONALE-315 — consistently show EFS HR 0.56–0.68, and OS benefit has emerged (KEYNOTE-671 OS HR 0.72 [PMID 39288781]).

Key branchpoints:

Subgroup	Recommended
Stage II-IIIB PD-L1 ≥1% resectable	perioperative pembro+chemo (KN-671) or nivo+chemo (CM-77T) preferred
Stage II-IIIB PD-L1 <1% resectable	perioperative IO-chemo still benefits but magnitude is smaller (CM-77T PD-L1<1% EFS HR 0.73); weigh IO duration
Stage IB-IIA small tumor	neoadjuvant-only (CM-816) reasonable; or post-op adjuvant IMpower010 / KN-091
EGFRm resectable	adjuvant osimertinib (ADAURA DFS HR 0.17, OS HR 0.49). Do not use perioperative IO (all 5 perioperative trials excluded EGFRm)
ALK+ resectable	adjuvant alectinib (ALINA DFS HR 0.24 [PMID 38598794]). Same rule — avoid IO
Borderline unresectable	perioperative IO-chemo may convert to resectable; multidisciplinary decision

Controversies:

Neoadjuvant-only vs perioperative — no direct RCT. Indirect comparisons favor perioperative, but adjuvant IO adds 6-9 months of treatment and toxicity.
pCR as surrogate: strongly correlated with EFS / OS in KEYNOTE-671 and CM-816, but FDA has not formally accepted it as a standalone endpoint for accelerated approval.
ctDNA clearance as a new surrogate: in KN-671, pCR rate was 47% in ctDNA-negative patients vs 8% in ctDNA-positive.
Does everyone need adjuvant IO? De-escalation trials are in design (continue adjuvant based on ctDNA or pCR status).

NCCN 2026: resectable stage II-IIIB (EGFR/ALK-WT) = perioperative pembro+chemo (KN-671) / nivo+chemo (CM-77T) / durva+chemo (AEGEAN) Cat 1 preferred.

3.4 PD-L1 TPS <1% Subgroup

Mainstream: IO-chemo combo (KEYNOTE-189 non-sq TPS<1% 5y HR 0.55; KEYNOTE-407 sq TPS<1% HR ~0.79) remains standard. CheckMate-227 (nivo+ipi in PD-L1<1%, 5y OS 19% vs 7%) is a chemo-sparing alternative. POSEIDON shows tremelimumab adds ~3 months of OS in PD-L1<1% but is regionally constrained. Pure chemo is now inferior in every subgroup.

Controversies:

Is the KEYNOTE-189 PD-L1 TPS<1% OS benefit real IO contribution or immortal-time bias? The 5-year follow-up strengthens the IO signal, but subgroup HR confidence intervals are wide.
Do chemo-sparing IO-IO regimens (CM-227 / CM-9LA) truly beat full-dose IO-chemo in PD-L1<1%? No direct H2H; 9LA has higher immune toxicity but less marrow toxicity.
Is tremelimumab’s ~3-month OS gain in POSEIDON worth the toxicity? Real-world uptake is low.
PD-L1 <1% is a heterogeneous population (true biology vs assay false-negative) — biomarker refinement is a gap.

3.5 EGFR / ALK Post-Resistance

2026 mainstream: post-osi splits into resistance-mechanism-matched paths —

MET-amplified (12–30%) → savolitinib+osi (SACHI [PMID 41544643], mPFS 8.3 months), NMPA-approved 2026; global use is off-label.
MET overexpressing (no amplification) → telisotuzumab vedotin (LUMINOSITY [PMID 38843488], ORR 35%), FDA accelerated (Emrelis).
Unselected broad-spectrum → MARIPOSA-2 ami+chemo±laz (mPFS 6.3–8.3 months), Cat 1.
Any EGFRm post-chemo → datopotamab deruxtecan (TROPION-Lung01 [PMID 39250535], EGFRm subgroup PFS HR 0.38).
CNS-only progression → local therapy (SRS) + continue osi.

Post-lorlatinib ALK+: no approved targeted therapy. Chemo+IO is the fallback. 4th-gen ALK TKIs (NVL-655 etc.) remain in early-phase trials.

Important update: patritumab deruxtecan (HER3-DXd) HERTHENA-Lung01 [PMID 37689979] confirmed activity (ORR 29.8%), but Daiichi discontinued phase 3 NSCLC development in 2024 — not commercially available in 2026.

3.6 Rare Drivers: 1L Entry Status for 8 Branches

Driver	1L entry?	Key evidence
ALK	In	CROWN 5y PFS 60% [PMID 38819031], lorlatinib best
ROS1	In	TRIDENT-1 repotrectinib ORR 79% [PMID 38197815], best CNS activity
EGFR common	In	osi / osi+chemo / ami+laz, three-way choice
EGFR ex20	In	PAPILLON ami+chemo mPFS HR 0.40 [PMID 37870976]
RET fusion	In	LIBRETTO-431 selpercatinib vs chemo+pembro HR 0.46 [PMID 37870973]
BRAF V600E	In	PHAROS enco+bini ORR 75% [PMID 37270692]
MET ex14	Partial	capmatinib / tepotinib on phase II accelerated approval; no phase III
HER2 ex20	Not yet	2L T-DXd (DESTINY-Lung02) / zongertinib (Beamion LUNG-1); 1L still chemo-IO
KRAS G12C	Not yet	sotorasib (CB-200) / adagrasib (KRYSTAL-12) 2L PFS benefit but OS not significant; divarasib (phase 3 ongoing) may break through
NTRK	Basket	larotrectinib / entrectinib tumor-agnostic accelerated

China approval lag (2026): repotrectinib approved in China 2025 (2 years after US), selpercatinib 2023, amivantamab 2025. Glecirasib (China-domestic KRAS G12C TKI) approved ahead of FDA-divarasib — a rare reversal.

NMPA-only drugs (China-exclusive): aumolertinib (3rd-gen EGFR TKI), furmonertinib, ensartinib, toripalimab, sintilimab, tislelizumab, camrelizumab, sugemalimab, penpulimab, cadonilimab (AK104, PD-1+CTLA-4 bispecific).

4. Research Gaps: 10 Unsolved Clinical Questions

The gaps below are specifically defined questions (not “more research is needed” clichés):

EGFRm 1L: FLAURA2 (osi+chemo) vs MARIPOSA (ami+laz) has no prospective H2H; clinicians are forced into cross-trial comparison while toxicity profiles are entirely different.
PD-L1 TPS<1% NSCLC: is the KEYNOTE-189 PD-L1-negative subgroup OS benefit real IO contribution, or is it confounded by control-arm crossover + immortal-time bias? A dedicated PD-L1-negative RCT is needed.
Perioperative IO de-escalation: is the adjuvant IO phase necessary for all patients, or only for ctDNA-positive / non-pCR patients? De-escalation trials are missing.
Post-lorlatinib ALK+ NSCLC: no approved targeted therapy; ~40% of relapses are driven by compound ALK mutations; 4th-gen ALK TKIs (NVL-655 etc.) in early trials.
KRAS G12C 1L: none of sotorasib / adagrasib / divarasib / glecirasib has phase 3 1L data beating IO-chemo — all remain 2L.
Biomarker-driven ADC sequencing: HER3 / TROP2 / c-MET expression thresholds across Dato-DXd / patritumab-DXd / teliso-V are not standardized; usage is empirical.
Chinese PD-1 inhibitors (sintilimab / tislelizumab / camrelizumab / toripalimab) vs pembrolizumab: each has independent phase 3 data but no H2H; price-vs-outcome decisions have no direct evidence.
PD-L1 IHC assay variability (22C3 vs SP263 vs SP142 vs Dako 28-8): particularly at the TPS 1-49% branching point — unresolved.
IO response and co-mutation effect (STK11 / KEAP1 / SMARCA4 loss): known to reduce IO benefit but not yet routinely incorporated into clinical decision algorithms.
CNS-penetrant next-gen TKIs (lorlatinib / repotrectinib) and SRS / WBRT sequencing: no prospective definition.

5. 2024–2026 Updates

5.1 FDA / NMPA New Approvals (10 key entries of 20)

Drug	Agency	Date	Indication / Supporting trial
datopotamab deruxtecan (Datroway)	FDA	2025-06-23	2L EGFRm NSCLC post-TKI and chemo / TROPION-Lung01
lazertinib + amivantamab (Lazcluze+Rybrevant)	FDA	2024-08-19	1L EGFRm NSCLC / MARIPOSA
repotrectinib (Augtyro)	FDA	2023-11-15 (NSCLC indication expanded 2024)	ROS1+ 1L / TRIDENT-1
tarlatamab (Imdelltra)	FDA	2024-05-16	SCLC (not within NSCLC scope)
zongertinib (Hernexeos)	FDA	2025-07-10	2L HER2-mutant NSCLC / Beamion LUNG-1
savolitinib+osimertinib	NMPA	2026-01	post-osi MET-amplified / SACHI
telisotuzumab vedotin (Emrelis)	FDA	2024-05-14	c-MET high-expressor 2L EGFR-WT NSCLC / LUMINOSITY
aumolertinib	NMPA	2020 (later-line updates 2024)	EGFRm NSCLC (China-exclusive 3rd-gen TKI)
furmonertinib	NMPA	2021	EGFRm NSCLC (China-exclusive 3rd-gen TKI)
neoadjuvant pembrolizumab (KN-671)	FDA	2023-10-16	Perioperative stage II-IIIB NSCLC

(This section shows 10 key approvals; full list of 20 follows chronological order.)

5.2 Key Conference Readouts (2024–2025, Downweighted Labeling)

Entries below are candidate-pool-only until peer-reviewed publication. Those with PMIDs have been promoted to the main library.

LAURA (ASCO 2024 LBA3) [PMID 38828946]: osi consolidation post-CRT, mPFS 39.1 vs 5.6 months, HR 0.16. Fills the PACIFIC gap for stage III EGFRm.
MARIPOSA OS update (ASCO 2025) [PMID 40923797]: ami+laz vs osi OS HR 0.75, 3y OS 60% vs 51%.
FLAURA2 final OS (ESMO 2025) [PMID 41104938]: osi+chemo mOS 47.5 vs 37.0 months.
CheckMate 816 OS (WCLC 2024) [PMID 40454642]: neoadj nivo+chemo OS HR 0.72.
KRYSTAL-12 (ESMO 2024) [PMID 40783289]: adagrasib vs docetaxel 2L KRAS G12C.
SACHI (WCLC 2025) [PMID 41544643]: savo+osi post-osi MET-amplified.

(This section extracts 6 key readouts; the full conference pool has 21 non-peer-reviewed PMIDs.)

5.3 Ongoing Phase III (2025-2026 Expected Readouts)

Selected 3 representatives from 90 ongoing phase III whose primary readout is expected in 2025-2026 and is likely to change practice:

NCT04853342 FORWARD — furmonertinib vs placebo adjuvant EGFRm (Allist, China)
NCT05840016 — AK112 (ivonescimab, PD-1+VEGF bispecific) + chemo vs tislelizumab+chemo in squamous NSCLC 1L (Akeso)
NCT06617416 — AK104 (cadonilimab) vs sugemalimab in unresectable stage III NSCLC consolidation
(This section shows 3 representative readouts expected in 2025-2026; the full pool has 90 ongoing phase III.)

AK112 (ivonescimab) is the biggest 2024-2025 dark horse in Chinese lung cancer — the PD-1+VEGF bispecific is challenging pembro across multiple domestic phase III trials. HARMONi-2 (2024 WCLC) H2H: ivonescimab+chemo vs pembro+chemo mPFS 11.14 vs 5.82 months. If overseas replication succeeds, the IO-chemo backbone may be rewritten.

6. Crosspoint Insights

6.1 Longitudinal × Cross-Sectional: Three Resonances Shaping 2026

Overlaying longitudinal paradigm evolution on the cross-sectional decision landscape reveals the 2026 NSCLC picture as three layered resonances:

Chemo ceiling (mOS ~8 months) → IO-chemo backbone (mOS ~22 months) pushed driver-negative + full-PD-L1-spectrum NSCLC to 2-3× OS.
Driver stratification from 2–3 (EGFR/ALK) → 8–10 (+ ROS1/MET/KRAS/HER2/RET/BRAF/NTRK/ex20) decomposed “unselected” NSCLC into 10+ subdiseases.
Early-stage NSCLC now covered by adjuvant osi (ADAURA) + perioperative IO-chemo + adjuvant alec (ALINA) — resectable NSCLC moved from “cut and pray” to “treat by subtype.”

Together these three resonances explain a clinical phenomenon: the 1L decision tree for a newly diagnosed stage IV NSCLC patient in 2026 has 3 more decision layers than in 2016 (driver panel → PD-L1 stratification → combo intensity → China accessibility branch).

6.2 Clinical Takeaways for Junior-Mid Oncologists

“Panel first, then decide” is SoC — opening IO-chemo in 2026 without comprehensive molecular profiling is wrong; missing EGFR / ALK gives the patient the wrong regimen.
Do not give perioperative IO to EGFRm / ALK+ — all five perioperative trials excluded them. These two driver-positive subgroups go adjuvant osi / alec instead.
PD-L1 TPS 50% is not “IO-mono exclusive” — IO-chemo data in fit patients are more consistent, especially for non-squamous.
Post-osi requires typing before treating — MET-amplified → SACHI, MET-overexpressing → teliso-v, broad-spectrum → MARIPOSA-2 or Dato-DXd.
KRAS G12C is still 2L only — do not use in 1L; 1L remains IO-chemo based on histology + PD-L1.
Perioperative therapy in 2026 is SoC that was unthinkable 5 years ago — 3-4 cycles of pre-op IO-chemo is the new starting point.
Chinese PD-1 accessibility × price: sintilimab / tislelizumab / camrelizumab / toripalimab are reasonable choices backed by their own phase III data but without H2H vs pembro. Insurance- and accessibility-driven decisions are the real-world norm.
AK112 (ivonescimab) may be the next breakthrough — but until global replication succeeds, treat it cautiously.

7. Sources

The 111-trial metadata in this report is independently verified via PubMed and ClinicalTrials.gov. Each [PMID xxxxxxxx] in the body can be opened directly on PubMed for verification.

Published trials: 111, covering 2002-2026
Ongoing phase III: 90 (primary completion 2025-2026)
FDA/NMPA 2024-2026 approvals: 20
2024-2025 key conference readouts: 21 (those with PMID available have been promoted to published trials)
Research gaps: 10

7.1 PMID Reference Table (ascending order)

The table below is the PMID list bracket-cited in the body; each can be clicked through to PubMed for verification.

PMID	Trial / Paper	Year	Journal	Section
11784875	ECOG 1594	2002	NEJM	§2.1 Chemo plateau
12837811	TAX 326	2003	JCO	§2.1
18506025	JMDB / Scagliotti 2008	2008	JCO	§2.1
19692680	IPASS	2009	NEJM	§2.2 EGFR TKI
21783417	OPTIMAL	2011	Lancet Oncol	§2.2
22285168	EURTAC	2012	Lancet Oncol	§2.2
22341744	PARAMOUNT	2012	Lancet Oncol	§2.1
25470694	PROFILE 1014	2014	NEJM	§2.3 ALK
26028407	CheckMate-017	2015	NEJM	§2.4 IO
26412456	CheckMate-057	2015	NEJM	§2.4
26712084	KEYNOTE-010	2016	Lancet	§2.4
27718847	KEYNOTE-024	2016	NEJM	§2.4, §3.1
28586279	ALEX	2017	NEJM	§2.3 ALK
28885881	PACIFIC	2017	NEJM	§2.2
29151359	FLAURA primary	2018	NEJM	§2.2
29658856	KEYNOTE-189 primary	2018	NEJM	§2.4, §3.1
29863955	IMpower150	2018	NEJM	§2.4, §3.1
30280635	KEYNOTE-407 primary	2018	NEJM	§2.4, §3.1
30955977	KEYNOTE-042	2019	Lancet	§3.1
31562796	CheckMate-227 primary	2019	NEJM	§2.4, §3.4
31751012	FLAURA OS	2019	NEJM	§2.2
32469185	VISION tepotinib	2020	NEJM	§2.3, §3.6
32877583	GEOMETRY mono-1 capmatinib	2020	NEJM	§2.3, §3.6
32955177	ADAURA primary	2020	NEJM	§2.2, §3.3
32997907	IMpower110	2020	NEJM	§2.4
33207094	CROWN primary	2020	NEJM	§2.3 ALK
33476593	CheckMate 9LA primary	2021	Lancet Oncol	§2.4, §3.1
33872070	KEYNOTE-024 5y	2021	JCO	§2.4
34096690	CodeBreaK 100	2021	NEJM	§2.3 KRAS
34555333	IMpower010 primary	2021	Lancet	§2.5, §3.3
35403841	CheckMate 816 primary	2022	NEJM	§2.5, §3.3
35917647	ORIENT-11 final OS	2022	JTO	§2.4
36108662	KEYNOTE-091	2022	Lancet Oncol	§2.5, §3.3
36223558	CheckMate-227 5y	2023	JCO	§2.4, §3.4
36327426	POSEIDON primary	2023	JCO	§2.4, §3.1, §3.4
36735893	KEYNOTE-407 5y	2023	JCO	§2.4, §3.1
36764316	CodeBreaK 200	2023	Lancet	§2.3 KRAS, §3.6
36809080	KEYNOTE-189 5y	2023	JCO	§2.4, §3.1, §3.4
37270692	PHAROS enco+bini	2023	JCO	§2.3 BRAF, §3.6
37272513	KEYNOTE-671 primary	2023	NEJM	§2.5, §3.3
37272535	ADAURA OS	2023	NEJM	§2.2, §3.3
37611121	Divarasib phase 1/2	2023	NEJM	§2.3 KRAS, §3.6
37689979	HERTHENA-Lung01	2023	JCO	§2.2, §3.5
37694347	DESTINY-Lung02	2023	JCO	§2.3 HER2, §3.6
37870973	LIBRETTO-431	2023	NEJM	§2.3 RET, §3.6
37870974	AEGEAN primary	2023	NEJM	§2.5, §3.3
37870976	PAPILLON	2023	NEJM	§2.3 EGFR ex20, §3.6
37879444	MARIPOSA-2	2024	Ann Oncol	§2.2, §3.5
37937763	FLAURA2 primary	2023	NEJM	§2.2, §3.2
38197815	TRIDENT-1	2024	NEJM	§2.3 ROS1, §3.6
38227033	Neotorch	2024	JAMA	§2.5, §3.3
38598794	ALINA	2024	NEJM	§2.3 ALK, §3.3
38749033	CheckMate 77T	2024	NEJM	§2.5, §3.3
38819031	CROWN 5y	2024	JCO	§3.6
38828946	LAURA	2024	NEJM	§2.2
38843488	LUMINOSITY teliso-v	2024	JCO	§2.2, §3.5
38924756	MARIPOSA primary	2024	NEJM	§2.2, §3.2
39243945	POSEIDON 5y	2025	JTO	§3.4
39250535	TROPION-Lung01	2025	JCO	§2.2, §3.5
39270380	CheckMate 9LA 5y	2024	JCO	§2.4
39288781	KEYNOTE-671 OS	2024	Lancet	§2.5, §3.3
39362249	GEOMETRY mono-1 final	2024	Lancet Oncol	§3.6
39581197	RATIONALE-315 interim	2025	Lancet Respir Med	§2.5, §3.3
40293180	Beamion LUNG-1 zongertinib	2025	NEJM	§2.3 HER2, §3.6
40446184	IMpower010 5y	2025	JCO	§3.3
40454642	CheckMate 816 OS	2025	NEJM	§2.5
40480428	PHAROS updated	2025	JCO	§3.6
40783289	KRYSTAL-12	2025	Lancet	§2.3 KRAS, §3.6
40923797	MARIPOSA OS	2025	NEJM	§2.2, §3.2
41104938	FLAURA2 OS	2026	NEJM	§2.2, §3.2
41486797	2025 NMA pembro+chemo	2025	Tuberk Toraks	§3.1
41544643	SACHI	2026	Lancet	§2.2, §3.2, §3.5

7.2 Verification

Every PMID is directly accessible via https://pubmed.ncbi.nlm.nih.gov/{PMID}/
Every NCT id via https://clinicaltrials.gov/study/{NCT_id}/
If you find a discrepancy between a PMID in this report and its PubMed entry (trial name / year / conclusion), please flag it for correction.

Closing

NSCLC over the past 25 years has been oncology’s most rapidly changing field — from four-way-tied chemo doublets in 2002 to 10 drivers × 3 IO strategies × the perioperative matrix in 2026, decision-making complexity has grown exponentially.

The flip side of complexity is precision. The “possible median OS” a newly diagnosed stage IV NSCLC patient can now expect has moved from ~8 months in 2002 to ~47 months in EGFRm (FLAURA2 final OS), ~12 months + ADC in HER2 ex20, ~34 months + repotrectinib in ROS1, and 30% more cure in resectable disease via perioperative therapy.

This report’s value is not in exhaustively listing every trial (PubMed does that) but in compressing 25 years of evolution + current decisions + unsolved gaps into the cognitive bandwidth of a single read. The next time you face a newly diagnosed NSCLC patient, every branch of the decision tree has this map to consult, trace, and question.

Clinician × AI = Research Superpower + Clinical Decision Amplifier

—— Dual Brain Lab · 2026-04-17

Immunotherapy on Dual Brain Lab

Biliary Tract Cancer Clinical Trial Timeline: A 16-Year Evolution Map

Biliary Tract Cancer Clinical Trial Timeline: In-depth Report

1. One-sentence definition

2. Vertical: timeline of five treatment paradigms

2.1 Adjuvant chemo evolution (2015-2024): from GEMOX/gemcitabine failure → capecitabine tacit ascent → S-1 / GemCis-CRT East-West divergence

2.2 Advanced 1L (2010-2025): GemCis dominates 12 years → three triplet challenges all fail → IO double-hit forms class effect

2.3 Advanced 2L (2019-2024): FOLFOX marginal + nal-IRI East-West divergence + IO/regorafenib monotherapy historical value

2.4 Biomarker-matched precision therapy (2018-2026): eight pathways + 2/3 ICC three-subtype reshaping

2.4.1 FGFR2 pathway (ICC-exclusive, 10-15% of ICC patients)

2.4.2 IDH1 pathway (ICC-enriched, 10-20% of ICC patients)

2.4.3 HER2 pathway (GBC-enriched, 15-20% of GBC patients; lower in ICC/ECC)

2.4.4 BRAF V600E pathway (ICC-enriched, 3-5% of ICC)

2.4.5 NTRK fusion pathway (cross-subtype, <1% of BTC)

2.4.6 MSI-H / dMMR / TMB-H pathway (cross-subtype, BTC MSI-H ~2-3%)

2.4.7 NRG1 fusion pathway (cross-subtype, <1% of BTC)

2.4.8 KRAS G12C pathway (cross-subtype, ~1-2% of BTC)

2.4.9 RET fusion pathway (cross-subtype, <1% of BTC)

2.5 Ongoing phase III and pending readouts (key 2026-2028 inflection points)

3. Horizontal: 2026 current decision landscape (six dimensions)

3.1 Newly diagnosed BTC: immediate comprehensive molecular profiling

3.2 Advanced 1L: the two PD-(L)1 choices in the IO + GemCis class effect for fit patients

3.3 Adjuvant regimen: capecitabine vs S-1 vs GemCis-CRT three paths

3.4 Precision therapy eight pathways: which biomarker → which drug

3.5 Advanced 2L+: FOLFOX + East-West nal-IRI divergence + IO/regorafenib fallback

3.6 Three-subtype differences (GBC / ICC / ECC): biomarker enrichment and decision triage

4. Research Gaps: the ten unresolved clinical questions

5. 2024-2026 latest developments

5.1 FDA / NMPA new approvals (ten key entries)

5.2 Key conference readouts (2024-2026, lower-weighted tagging)

5.3 Ongoing phase III (2025-2028 readout highlights)

6. Convergent insights and judgments

6.1 Vertical × Horizontal: the 2026 BTC landscape is shaped by three “resonances”

6.2 Clinical decision takeaways (for junior-mid oncologists)

7. Information sources

7.1 Report-body citation list (sorted by PMID)

7.2 Verification conventions

The trial timeline lives here

Closing

Esophageal Cancer Trial Timeline: A Dual-Track Map of 30 Years and 42 RCTs

Esophageal Cancer Clinical Trial Timeline — In-Depth Report

1. One-sentence definition

2. Longitudinal: five paradigm shifts along the timeline

2.1 Definitive CRT era established (1992–2007): making “no surgery” an option, and the dose paradox

2.2 Global neoadjuvant CRT (2012–2018): CROSS rewrote the world + East Asia split off

2.3 Intra–East Asia ESCC divergence (2024): JCOG1109 NExT broke the “RT always helps” assumption

2.4 Advanced IO big three explosion (2019–2024): 4 years and 8 positive phase III trials reset SoC from CF/TP to IO+chemo

2.5 IO enters the perioperative setting (2021–2024): CheckMate-577 adjuvant rewrote + ESCORT-NEO neo-adj established

2.6 Surgical technique and local therapy (2012–2020): MIE / RAMIE / proton vs IMRT

2.7 EGFR death in ESCC (2017–2020): POWER + SCOPE-1

3. Cross-sectional: 2026 decision landscape (six dimensions)

3.1 Newly diagnosed advanced ESCC 1L: 8 IO+chemo phase III trials in parallel, how to choose under fragmented PD-L1 metrics

3.2 Advanced EAC / GEJ AC 1L: KEYNOTE-590 all-comers + borrowing from gastric

3.3 Advanced 2L+: post-1L-IO is the largest ESCC clinical gap of the next 5 years

3.4 Perioperative LA-ESCC: East Asian three-country split + CheckMate-577 adjuvant global SoC

3.5 Unresectable LA ESCC def-CRT: 50 Gy + CF dose unchanged for 40 years + KN-975 pending

3.6 Surgical technique: MIE vs RAMIE vs open

4. Research Gaps: the ten unresolved clinical questions

5. 2024–2026 latest developments

5.1 Recent FDA / NMPA approvals (ESCC / EC-related excerpts)

5.2 Key conference readouts (2024–2026, flagged as lower-tier)

5.3 Ongoing phase III (selected 2026–2028 readouts)

6. Intersecting insights and judgments

6.1 Longitudinal × cross-sectional: the 2026 EC landscape is shaped by four “resonances”

6.2 Clinical decision takeaways (for junior-mid oncologists)

7. Sources

7.1 Citation index (sorted by PMID)

7.2 Verification conventions

The clinical trial timeline lives here

Closing

Gastric Cancer Clinical Trial Timeline: 58 RCTs Across 25 Years Mapping a Subtyping Revolution

Gastric Cancer Clinical Trial Timeline — In-depth Report

1. One-sentence definition

2. Longitudinal: timeline of five treatment-paradigm shifts

2.1 Chemotherapy backbone era (2001-2010): adjuvant chemoRT / perioperative chemo / HER2 — three starting points

2.2 East–West adjuvant / perioperative divergence takes shape (2010-2019): CLASSIC / FLOT4 / ARTIST / RESOLVE / PRODIGY

2.3 HER2 era’s 10-year leap (2010-2025): ToGA → T-DXd → KEYNOTE-811 / zanidatamab

2.4 IO + chemo 1L rewrite (2021-2024): four phase IIIs concordantly positive

2.5 Biomarker triple-subtype year zero + perioperative IO + CAR-T dawn (2023-2026)

3. Horizontal: the 2026 decision landscape (six dimensions)