Skip to content
Bending the Curve Trial by Trial
?

Stomach and gastro-oesophageal junction (upper gastrointestinal tract)

Gastric and gastro-oesophageal junction cancer

How H. pylori research, perioperative therapy, targeted treatment and immunotherapy shaped care for stomach and junctional cancers.

5-year relative survival · United States, SEER 8 registries

15.1%→36.42%

1975 (observed value; first year in the table) to 2018 (observed value; last year with an observed value in the table). Read these with the caveats under “Then and now”.Source

17 milestones, by decade

  • In 1984 Marshall and Warren reported spiral or curved bacilli in antral biopsy specimens from 58 of 100 consecutive consenting patients presenting for gastroscopy, found in almost all of those with active chronic gastritis, duodenal ulcer or gastric ulcer, and suggested the bacteria might be an important factor in the aetiology of those diseases; the abstract does not mention cancer. In 1991 a nested case-control study of 109 patients with gastric adenocarcinoma (excluding tumours of the gastroesophageal junction) reported earlier Helicobacter pylori infection in 84 percent of cases and 61 percent of matched controls (odds ratio 3.6, 95% CI 1.8 to 7.3); tumours of the junction and of the gastric cardia were not linked to infection. The 1991 study is observational and shows association.
  • On 22 September 2017 the FDA approved pembrolizumab as a single agent under its accelerated approval regulations for recurrent locally advanced or metastatic gastric or gastroesophageal junction adenocarcinoma with PD-L1 expression (combined positive score 1 or more), progression after two or more prior lines of therapy, based on tumour response rate and durability of response; the letter states that continued approval might depend on verification of clinical benefit in confirmatory trials. On 4 February 2022 a labelling supplement withdrew that third-line indication. The withdrawal letter releases the related postmarketing requirement because the indication was withdrawn from the label and states no reason for the withdrawal; the FDA list of withdrawn cancer accelerated approvals gives the same two dates.

Evidence current to October 2026 · Checked against primary sources to the extent possible

A guided tour of 8 milestones. You can pause or step through it yourself.

Filter milestonesshowing all 17
Treatment type
Setting

Molecular origins

  1. 1983 · Origin · UnderstandingH. pylori identified (1983-84), then linked to gastric cancerIn 1983 Warren and Marshall reported curved bacteria on the stomach lining in gastritis, and their 1984 series linked them with gastritis and ulcers. Later studies tied infection with this organism, Helicobacter pylori, to stomach cancer, and in 1994 an IARC evaluation classed the infection as carcinogenic to humans.Warren and Marshall's 1984 series found spiral bacilli in 58 of 100 consecutive gastroscopy patients and proposed a role in gastritis and ulcer. Parsonnet 1991 (nested case-control): prior infection in 84% of 109 gastric adenocarcinoma cases versus 61% of controls (OR 3.6, 95% CI 1.8-7.3), no association for GEJ or cardia tumours. Uemura 2001: cancer in 36 of 1246 infected, none of 280 uninfected. IARC 1994: Group 1.

Targeted therapy

  1. 2000 · De-escalation · SurgeryEndoscopic resection for early gastric cancer: Gotoda 2000 to JCOG0607A report from two large cancer centres in 2000 found no lymph node spread in a group of very early stomach cancers, which supported removing such tumours through the endoscope rather than by gastrectomy. A later single-arm trial of wider criteria reported high five-year survival, without a surgery comparison arm.Gotoda 2000 (5265 gastrectomy patients, two centres): none of 1230 well-differentiated intramucosal cancers under 30 mm had nodal metastasis (95% CI 0-0.3%). JCOG0607, single-arm confirmatory trial of ESD for expanded cT1a intestinal-type criteria (470 patients): 5-year OS 97.0% (95% CI 95.0-98.2) against a threshold of 86.1%; en bloc resection 99.1%; additional gastrectomy in 131 (28%).
  2. 2001 · Treatment · ChemotherapyINT-0116 adjuvant chemoradiotherapy, then ARTIST and CRITICSAn intergroup trial reported in 2001 that adding radiotherapy with chemotherapy after stomach surgery lengthened survival. Later trials asked whether radiotherapy adds benefit to chemotherapy in other settings, after more extensive lymph node surgery or alongside preoperative chemotherapy; in their main analyses they did not show a significant difference.INT-0116 randomised 556 patients with resected stomach or GEJ adenocarcinoma to surgery alone or adjuvant fluorouracil-leucovorin chemoradiotherapy: median OS 27 vs 36 months, HR for death 1.35 (95% CI 1.09-1.66; P=0.005); at over 10 years OS HR 1.32 (1.10-1.60). ARTIST (D2 surgery, 458): DFS not significantly prolonged (P = .0862). CRITICS (788): median OS 43 vs 37 months, HR 1.01 (0.84-1.22).
  3. 2004 · Understanding · Detection & preventionH. pylori eradication to prevent gastric cancer: trials 2004-2024Treating Helicobacter pylori infection has been tested as a way to prevent stomach cancer. A large placebo-controlled trial in China found no overall reduction. Later trials in people who had early cancers removed, in relatives of patients and in whole communities reported fewer stomach cancers with treatment, though the community effect was modest.Wong 2004 (1630 carriers, placebo-controlled, 7.5 years): 7 vs 11 cancers (P=.33); none vs 6 (P=.02) in the subgroup without precancerous lesions. After endoscopic resection, metachronous cancer: Choi 2018 HR 0.50 (95% CI 0.26-0.94), Fukase 2008 mITT HR 0.339 (0.157-0.729). Relatives of patients: Choi 2020 HR 0.45 (0.21-0.94). Linqu community trial: HR 0.86 (0.74-0.99).
  4. 2006 · Treatment · ChemotherapyMAGIC: perioperative ECF for resectable gastro-oesophageal cancerIn a randomised trial reported in 2006, people with operable cancer of the stomach, the junction or the lower gullet who had chemotherapy before and after surgery lived longer than people who had surgery alone. About 36 in 100 were alive at five years with chemotherapy, compared with 23 in 100.Randomised phase 3 trial of perioperative ECF (three cycles before and three after surgery; 250 patients) versus surgery alone (253) in resectable adenocarcinoma of the stomach, oesophagogastric junction or lower oesophagus. Median follow-up 4 years: OS HR 0.75 (95% CI 0.60-0.93; P=0.009), 5-year survival 36% vs 23%; progression HR 0.66 (0.53-0.81; P<0.001). Postoperative complications 46% vs 45%.
  5. 2007 · Treatment · ChemotherapyAdjuvant S-1 (ACTS-GC) and capecitabine-oxaliplatin (CLASSIC) after D2Two randomised trials in East Asia tested chemotherapy after stomach surgery with extended lymph node removal. One used the oral drug S-1 and the other used capecitabine with oxaliplatin; in each, people given chemotherapy had better outcomes than people who had surgery alone.ACTS-GC (Japan, stage II-III, D2; 529 S-1 vs 530 surgery only): 3-year OS 80.1% vs 70.1%, HR 0.68 (95% CI 0.52-0.87); 5-year OS 71.7% vs 61.1%, HR 0.669. CLASSIC (South Korea, China, Taiwan; stage II-IIIB, D2; 520 vs 515): 3-year DFS 74% vs 59%, HR 0.56 (0.44-0.72); 5-year OS 78% vs 69%, HR 0.66 (0.51-0.85).
  6. 2008 · Treatment · ChemotherapyREAL-2: oxaliplatin and capecitabine, advanced oesophagogastricIn 2008 a trial led by a UK research group, in about a thousand people with untreated advanced oesophagogastric cancer, swapped in a three-drug regimen oral capecitabine for infused fluorouracil and oxaliplatin for cisplatin. Each swap met the trial's noninferiority test; the version with both swaps had the longest median survival, 11.2 months.Two-by-two randomised phase 3 REAL-2: 1002 patients, untreated advanced oesophagogastric cancer; epirubicin plus cisplatin or oxaliplatin, with fluorouracil or capecitabine (ECF, ECX, EOF, EOX). Noninferiority OS: capecitabine vs fluorouracil HR 0.86 (95% CI 0.80 to 0.99); oxaliplatin vs cisplatin HR 0.92 (0.80 to 1.10); margin 1.23. Median OS 9.9, 9.9, 9.3, 11.2 months. EOX vs ECF HR 0.80 (0.66 to 0.97), P=0.02 (secondary).
  7. 2010 · Treatment · TargetedToGA: trastuzumab plus chemotherapy in HER2-positive gastric cancerIn 2010 a large international trial found that adding the antibody trastuzumab to chemotherapy lengthened median survival, from about 11 months to about 14 months, in advanced stomach or gastro-oesophageal junction cancers whose tumours carry the HER2 protein. European and US regulators approved the use in 2010.ToGA, open-label phase 3: 594 patients with HER2-positive (IHC or FISH) advanced gastric or GEJ cancer randomised 1:1 to capecitabine or fluorouracil plus cisplatin, with or without trastuzumab. Primary analysis (584 treated): median OS 13.8 vs 11.1 months, HR 0.74 (95% CI 0.60 to 0.91), p=0.0046. Grade 3 or 4 adverse events 68% vs 68%; cardiac adverse events 6% vs 6%.

Immunotherapy

  1. 2013 · Treatment · TargetedREGARD and RAINBOW: ramucirumab after first-line therapyIn 2013 and 2014 two placebo-controlled trials tested ramucirumab, an antibody that blocks a blood-vessel growth receptor, in advanced stomach or junction cancer that had progressed on first-line chemotherapy. Alone it lengthened median survival modestly, from 3.8 to 5.2 months; with paclitaxel it added about two months, from 7.4 to 9.6 months.REGARD (355 patients, 2:1, best supportive care with ramucirumab 8 mg/kg or placebo): median OS 5.2 vs 3.8 months, HR 0.776 (95% CI 0.603 to 0.998), p=0.047. RAINBOW (665 patients, 1:1, paclitaxel with ramucirumab or placebo): median OS 9.6 vs 7.4 months, HR 0.807 (0.678 to 0.962), p=0.017. Hypertension was more frequent with ramucirumab in both trials.
  2. 2019 · Treatment · ChemotherapyFLOT4: perioperative FLOT versus ECF/ECX in gastric and GEJ cancerA German trial compared two chemotherapy combinations given before and after surgery for stomach or junction cancer. The docetaxel-containing combination, FLOT, was associated with longer survival than the older ECF or ECX combination, and serious side effects occurred at a similar rate in both groups.FLOT4 (phase 2/3; 716 patients with cT2 or higher and/or node-positive resectable gastric or GEJ adenocarcinoma) compared perioperative FLOT with ECF/ECX. Phase 3, primary endpoint OS (ITT): median 50 vs 35 months, HR 0.77 (95% CI 0.63-0.94). Phase 2 pathological complete regression (TRG1a): 16% vs 6% (p=0.02). Related serious adverse events 27% in both groups.

Precision & engineered

  1. 2020 · Treatment · Antibody–drug conjugateTrastuzumab deruxtecan in HER2-positive gastric cancerIn 2020 a randomised phase 2 trial reported that trastuzumab deruxtecan, an antibody linked to a chemotherapy drug, shrank tumours in about half of people with HER2-positive stomach cancer previously treated with trastuzumab, versus about one in seven with chemotherapy. A 2025 phase 3 trial reported longer survival than ramucirumab plus paclitaxel.DESTINY-Gastric01 (phase 2, 2:1, 187 treated, prior trastuzumab): ORR 51% vs 14% (P<0.001); median OS 12.5 vs 8.4 months, HR 0.59 (95% CI 0.39 to 0.88), P=0.01. DESTINY-Gastric04 (phase 3, 494 randomised, second line): median OS 14.7 vs 11.4 months vs ramucirumab plus paclitaxel, HR 0.70 (0.55 to 0.90), P=0.004; ILD or pneumonitis 13.9% vs 1.3%.
  2. 2021 · Treatment · ImmunotherapyFirst-line PD-1 plus chemotherapy in HER2-negative gastric cancerFrom 2021, trials adding a PD-1 immune checkpoint inhibitor to first-line chemotherapy reported longer survival in advanced HER2-negative stomach and junction cancers, with larger estimated benefit at higher PD-L1 scores. An East Asian trial improved progression-free but not overall survival. In 2025 US labels narrowed to PD-L1-positive tumours.Phase 3 trials of a PD-1 inhibitor added to first-line chemotherapy, HER2-negative disease. CheckMate 649, PD-L1 CPS 5 or more: OS HR 0.71 (98.4% CI 0.59 to 0.86). KEYNOTE-859, ITT: median OS 12.9 vs 11.5 months, HR 0.78 (95% CI 0.70 to 0.87). RATIONALE-305, TAP score 5% or more: 17.2 vs 12.6 months, HR 0.74. ATTRACTION-4 (Asia): OS HR 0.90 (95% CI 0.75 to 1.08).
  3. 2021 · Treatment · ImmunotherapyKEYNOTE-811: pembrolizumab plus trastuzumab and chemotherapyIn 2021 the US regulator approved pembrolizumab with trastuzumab and chemotherapy for HER2-positive stomach cancer based on tumour response: about three in four patients responded, versus about one in two when placebo was added instead. Later survival data, with the benefit mainly in PD-L1-positive tumours, led to a full approval in 2025 limited to those tumours.KEYNOTE-811 (phase 3, double-blind). First interim analysis (264 patients): ORR 74.4% vs 51.9%, P=0.00006 (FDA accelerated approval 5 May 2021). Later interim analyses (698 patients): PFS HR 0.72 (95% CI 0.60 to 0.87); OS HR 0.87 then 0.84, not significant. Final OS (ITT, FDA authors' summary): median 20.0 vs 16.8 months, HR 0.80 (0.67 to 0.94); CPS 1 or more HR 0.79, CPS below 1 HR 1.10.
  4. 2022 · Reversal · ImmunotherapyLater-line pembrolizumab in gastric cancer: approved, then withdrawnIn 2017 the US regulator gave pembrolizumab an accelerated approval for stomach cancer that had progressed after two or more treatments, based on tumour response. A separate 2018 randomised trial against paclitaxel in second-line treatment did not show significantly longer survival. The indication was withdrawn from the US label in February 2022; the FDA letter gives no reason.KEYNOTE-059 (phase 2, single-arm, 259 patients): ORR 11.6% (95% CI 8.0 to 16.1); 15.5% PD-L1-positive, 6.4% PD-L1-negative. KEYNOTE-061 (phase 3, second line, CPS 1 or more, 395 patients): median OS 9.1 vs 8.3 months for pembrolizumab vs paclitaxel, HR 0.82 (95% CI 0.66 to 1.03), one-sided p=0.0421; PFS HR 1.27. FDA accelerated approval 22 Sep 2017; withdrawn 4 Feb 2022.
  5. 2023 · Treatment · TargetedZolbetuximab for CLDN18.2-positive disease: SPOTLIGHT and GLOWIn 2023 two trials tested zolbetuximab, an antibody against the claudin 18.2 protein, with first-line chemotherapy in HER2-negative stomach or junction cancers that carry that protein. Both reported that the risk of progression or death was lower, by about a quarter in one and about a third in the other, and that the risk of death was also lower.SPOTLIGHT (565 patients, zolbetuximab plus mFOLFOX6 vs placebo plus mFOLFOX6, CLDN18.2-positive, HER2-negative): median PFS 10.61 vs 8.67 months, HR 0.75 (95% CI 0.60 to 0.94), p=0.0066; OS HR 0.75 (0.60 to 0.94), p=0.0053. GLOW (507 patients, with CAPOX): median PFS 8.21 vs 6.80 months, HR 0.687 (0.544 to 0.866); median OS 14.39 vs 12.16 months, HR 0.771 (0.615 to 0.965).
  6. 2025 · Treatment · Immunotherapy⚑ needs reviewMATTERHORN: perioperative durvalumab plus FLOT, and KEYNOTE-585Adding the immunotherapy drug durvalumab to FLOT chemotherapy before and after surgery improved event-free survival in a randomised trial of operable stomach or junction cancer, and the US regulator approved it in 2025. A trial of another drug, pembrolizumab, with chemotherapy reported less clear results.MATTERHORN (948 randomised, durvalumab or placebo with FLOT): primary endpoint EFS, 2-year EFS 67.4% vs 58.5%, HR 0.71 (95% CI 0.58-0.86; P<0.001); pCR 19.2% vs 7.2%; final OS HR 0.78 (0.63-0.96; p=0.021). KEYNOTE-585 final analysis (main cohort, 804): median OS 71.8 vs 55.7 months, HR 0.86 (0.71-1.06); EFS HR 0.81 (0.67-0.98; cohort not stated).
  7. 2026 · Treatment · ImmunotherapyHERIZON-GEA-01: zanidatamab, with or without tislelizumabIn 2026 a phase 3 trial reported that zanidatamab, an antibody that binds HER2 in two places, added to chemotherapy delayed progression compared with trastuzumab plus chemotherapy in HER2-positive cancers of the stomach, junction and oesophagus. Adding tislelizumab also lengthened survival at this early analysis. US approvals followed in August 2026.HERIZON-GEA-01 (open-label phase 3, 1:1:1; median follow-up 25.9 months) vs trastuzumab plus chemotherapy. PFS: median 12.4 months with zanidatamab plus tislelizumab and chemotherapy (HR 0.63, 95% CI 0.51 to 0.78) and 12.4 with zanidatamab plus chemotherapy (HR 0.65, 0.52 to 0.81) vs 8.1 months. OS (interim): 26.4 vs 19.2 months, HR 0.72 (0.57 to 0.90), P=0.004; zanidatamab plus chemotherapy HR 0.80 (0.64 to 1.01).

Then and now

  • 5-year relative survival · United States, SEER 8 registries

    • 15.1%

      1975 (observed value; first year in the table)

    All races, both sexes; SEER Cancer Stat Facts category 'stomach cancer' (the page does not list the ICD-O site codes or histologies included and does not say whether gastro-oesophageal junction tumours are counted as stomach). SEER 8 five-year relative survival series, 1975-2018. Age range is not stated for this series.

    Modeled trend value for 1975: 15.03%. The observed series moves from year to year (lowest 14.86% in 1977, highest 38.52% in 2017). Relative survival excludes deaths from other causes. 'Observed' is the table's label for the actual yearly value, as opposed to the Joinpoint modeled-trend value in the adjacent column. The column is headed 'Year' and the chart note says only 'SEER 8 5-Year Relative Survival Percent from 1975-2018'; the page does not say the year is the year of diagnosis. SEER 8 covers a limited set of registries. The page does not describe changes over time in subsite mix, histology, staging, imaging or endoscopy practice, does not apportion any change in this series to a cause, and gives no figure separately for gastro-oesophageal junction tumours. Read beside the mortality values. Compare only with other values from this same SEER 8 series, not with the SEER 21 value.

    Source

  • 5-year relative survival · United States, SEER 8 registries

    • 36.42%

      2018 (observed value; last year with an observed value in the table)

    All races, both sexes; SEER Cancer Stat Facts category 'stomach cancer' (the page does not list the ICD-O site codes or histologies included and does not say whether gastro-oesophageal junction tumours are counted as stomach). SEER 8 five-year relative survival series, 1975-2018. Age range is not stated for this series.

    Modeled trend value for 2018: 39.54%. The observed value is lower than the 2017 value (38.52%); the table shows no observed value after 2018 (the modeled-trend column continues beyond 2018). Relative survival excludes deaths from other causes. 'Observed' is the table's label for the actual yearly value, as opposed to the Joinpoint modeled-trend value in the adjacent column. The column is headed 'Year' and the chart note says only 'SEER 8 5-Year Relative Survival Percent from 1975-2018'; the page does not say the year is the year of diagnosis. SEER 8 covers a limited set of registries. The page does not describe changes over time in subsite mix, histology, staging, imaging or endoscopy practice, does not apportion any change in this series to a cause, and gives no figure separately for gastro-oesophageal junction tumours. Read beside the mortality values. Compare only with other values from this same SEER 8 series, not with the SEER 21 value.

    Source

  • 5-year relative survival · United States, SEER 21 registries (excluding IL)

    • 39.8%

      2016-2022 (headline value on the page)

    All races, both sexes; stomach cancer, all stages together. By stage at diagnosis in the same table (SEER Combined Summary Stage): localized 78.1% (32% of cases), regional 39.0% (23%), distant 8.1% (35%), unknown stage 33.6% (9%). The page states no age range, histology or subsite for these figures.

    A different registry set (SEER 21 excluding Illinois) and period from the SEER 8 series above: keep it separate and do not subtract it from, or chain it with, the 1975 or 2018 values. Relative survival excludes deaths from other causes. The page states that 2016-2022 is the data period for this figure and does not say whether these are diagnosis years. The page text gives 32.4% of cases diagnosed at the local stage; the table rounds to 32%.

    Source

  • Age-standardised mortality · United States (U.S. Mortality data shown on the SEER page; not a SEER registry series)

    • 8.5per 100,000

      1975 (observed value; first year in the table)

    All races, both sexes; stomach cancer deaths; age-adjusted rate per 100,000 (the page text read does not name the standard population).

    1975 is the highest observed value in the table (8.50); modeled trend value 8.42. The page gives 2.7 per 100,000 for 2020-2024 deaths and reports age-adjusted death rates falling on average 2.0% each year over 2015-2024. For comparison with the survival values, the page's SEER 8 incidence series falls from 11.54 (1975) to 7.29 (2023) per 100,000, and the page text reports new cases rising on average 1.1% each year over 2014-2023. Death-certificate coding of stomach versus junction tumours and the changing mix of gastric subsites are not described on the page. Mortality reflects incidence as well as survival.

    Source

  • Age-standardised mortality · United States (U.S. Mortality data shown on the SEER page; not a SEER registry series)

    • 2.6per 100,000

      2024 (observed value; last year in the table)

    All races, both sexes; stomach cancer deaths; age-adjusted rate per 100,000 (the page text read does not name the standard population).

    The table prints 2.60 for 2024; modeled trend value 2.56; the lowest observed value in the table is 2.59 (2023). Same series and caveats as the 1975 value above. The page gives 2.7 per 100,000 as the rate for 2020-2024 deaths combined.

    Source

Survival figures are affected by earlier detection (lead time) and overdiagnosis, and by changes in who is diagnosed. For screened cancers, read them alongside mortality.

Evidence chains

A lane follows one defined population and setting, where each new trial tests against what was then standard. Compare only within a lane, and read every comparison as illustrative.

HER2-positive gastric or GEJ cancer, first-line

Adults with previously untreated, advanced HER2-positive gastric or gastro-oesophageal junction adenocarcinoma (ToGA, KEYNOTE-811; HERIZON-GEA-01 also enrolled gastro-oesophageal adenocarcinoma) · Metastatic · endpoint: OS

Results are not comparable across trials.

The control arms differ: ToGA compared chemotherapy (capecitabine or fluorouracil with cisplatin) with and without trastuzumab, whereas KEYNOTE-811 and HERIZON-GEA-01 both used trastuzumab plus chemotherapy as the control. KEYNOTE-811 was double-blind and placebo-controlled; ToGA and HERIZON-GEA-01 were open-label. HER2 testing rules, eras and partner chemotherapy differ. KEYNOTE-811's plotted medians are its final overall survival analysis as reported in an FDA authors' summary, and its exploratory PD-L1 subgroup results are not drawn. HERIZON-GEA-01's figures are an interim analysis; the plotted pair is zanidatamab plus tislelizumab plus chemotherapy versus trastuzumab plus chemotherapy, and its zanidatamab-plus-chemotherapy arm, which did not differ significantly in overall survival at this analysis, is not drawn. Compare only each trial's own two arms; the chart is not a pooled or network estimate.

Each pair shows one trial: its control arm and its experimental arm. Yesterday's experimental arm is often today's control. Populations, definitions and follow-up differ.

05101520253035OS, median (months)201020212026ToGAKEYNOTE-811HERIZON-GEA-01
  • experimental arm
  • control arm
  • different endpoint from the lane's
  • step line through experimental values
OS, median, across 3 trials in this lane. ToGA (2010): experimental 13.8 months vs control 11.1 months; KEYNOTE-811 (2021): experimental 20 months vs control 16.8 months; HERIZON-GEA-01 (2026): experimental 26.4 months vs control 19.2 months.
Data table
TrialYearEndpointControl (months)Experimental (months)
ToGA2010OS11.1 months13.8 months
KEYNOTE-8112021OS16.8 months20 months
HERIZON-GEA-012026OS19.2 months26.4 months

HER2-negative gastric or GEJ cancer, first-line

Adults with previously untreated, advanced gastric or gastro-oesophageal junction adenocarcinoma that was HER2-negative (CheckMate 649 enrolled non-HER2-positive disease and also included oesophageal adenocarcinoma, about 70% gastric by primary site; GLOW enrolled only CLDN18.2-positive tumours) · Metastatic · endpoint: OS

Results are not comparable across trials.

The PD-L1 populations differ and are reported with different scoring systems: CheckMate 649's plotted result is in patients with a combined positive score (CPS) of 5 or more, KEYNOTE-062's is in CPS 1 or more, while KEYNOTE-859, RATIONALE-305 and ATTRACTION-4 are plotted for all randomised patients regardless of PD-L1 (RATIONALE-305 also reported a tumour area positivity score of 5% or more, not drawn). ATTRACTION-4 enrolled people in Japan, South Korea and Taiwan; GLOW enrolled only CLDN18.2-positive tumours, a biomarker-selected group, and used capecitabine plus oxaliplatin. Chemotherapy backbones, eras and follow-up differ, and the plotted analyses are interim for KEYNOTE-859 and GLOW. KEYNOTE-062's plotted pair is pembrolizumab plus chemotherapy versus chemotherapy; its pembrolizumab-alone arm is not drawn. SPOTLIGHT reports overall survival as a hazard ratio only and is not drawn. REAL-2 is on the lane as a milestone but is not drawn: it did not restrict or test HER2 status, compared chemotherapy regimens rather than a new drug class, and included other oesophagogastric cancers. Compare only each trial's own two arms; the chart is not a pooled or network estimate.

Each pair shows one trial: its control arm and its experimental arm. Yesterday's experimental arm is often today's control. Populations, definitions and follow-up differ.

0510152025OS, median (months)202120212021202120212023CheckMate 649KEYNOTE-859RATIONALE-305ATTRACTION-4KEYNOTE-062GLOW
  • experimental arm
  • control arm
  • different endpoint from the lane's
  • step line through experimental values
OS, median, across 6 trials in this lane. CheckMate 649 (2021): experimental 14.4 months vs control 11.1 months; KEYNOTE-859 (2021): experimental 12.9 months vs control 11.5 months; RATIONALE-305 (2021): experimental 15 months vs control 12.9 months; ATTRACTION-4 (2021): experimental 17.45 months vs control 17.15 months; KEYNOTE-062 (2021): experimental 12.5 months vs control 11.1 months; GLOW (2023): experimental 14.39 months vs control 12.16 months.
Data table
TrialYearEndpointControl (months)Experimental (months)
CheckMate 6492021OS11.1 months14.4 months
KEYNOTE-8592021OS11.5 months12.9 months
RATIONALE-3052021OS12.9 months15 months
ATTRACTION-42021OS17.15 months17.45 months
KEYNOTE-0622021OS11.1 months12.5 months
GLOW2023OS12.16 months14.39 months

Advanced gastric or GEJ cancer, after first-line chemotherapy

Adults with advanced gastric or gastro-oesophageal junction adenocarcinoma whose disease progressed after first-line chemotherapy (REGARD, RAINBOW, KEYNOTE-061; DESTINY-Gastric04 enrolled HER2-positive disease after trastuzumab-based therapy) · Metastatic · endpoint: OS

Results are not comparable across trials.

The controls differ: placebo with best supportive care (REGARD), placebo plus paclitaxel (RAINBOW), paclitaxel alone (KEYNOTE-061) and ramucirumab plus paclitaxel (DESTINY-Gastric04). DESTINY-Gastric04 enrolled only HER2-positive disease after trastuzumab-based therapy, and KEYNOTE-061's plotted result is in the PD-L1 CPS 1 or more population, for which the prespecified significance threshold for overall survival was not reached. KEYNOTE-061 is drawn at the year of its milestone (2022, the withdrawal of the later-line indication), not its own publication year. DESTINY-Gastric01 (randomised phase 2, patients with at least two earlier therapies) is not drawn because the line of therapy differs; KEYNOTE-059 (single-arm) and JAVELIN Gastric 100 (maintenance after induction chemotherapy) are not drawn. Eras and regions differ. Compare only each trial's own two arms; the chart is not a pooled or network estimate.

Each pair shows one trial: its control arm and its experimental arm. Yesterday's experimental arm is often today's control. Populations, definitions and follow-up differ.

05101520OS, median (months)2013201320202022REGARDRAINBOWDESTINY-Gastric04KEYNOTE-061
  • experimental arm
  • control arm
  • different endpoint from the lane's
  • step line through experimental values
OS, median, across 4 trials in this lane. REGARD (2013): experimental 5.2 months vs control 3.8 months; RAINBOW (2013): experimental 9.6 months vs control 7.4 months; DESTINY-Gastric04 (2020): experimental 14.7 months vs control 11.4 months; KEYNOTE-061 (2022): experimental 9.1 months vs control 8.3 months.
Data table
TrialYearEndpointControl (months)Experimental (months)
REGARD2013OS3.8 months5.2 months
RAINBOW2013OS7.4 months9.6 months
DESTINY-Gastric042020OS11.4 months14.7 months
KEYNOTE-0612022OS8.3 months9.1 months

Resectable gastric or GEJ cancer, perioperative therapy

Adults with previously untreated, resectable gastric or gastro-oesophageal junction adenocarcinoma (MAGIC also included lower oesophageal tumours) · Perioperative · endpoint: OS

Results are not comparable across trials.

The trials report overall survival in different forms. FLOT4 and KEYNOTE-585 give medians and are the plotted pair; MATTERHORN gives 2-year survival rates and is listed beside the chart, not on the same axis. MAGIC (surgery alone versus perioperative ECF) reports 5-year survival rates only (36% versus 23%, no medians) and is left off the chart because it cannot share an axis with the medians; it remains a milestone of the lane. Controls differ: ECF or ECX (FLOT4), placebo plus chemotherapy (KEYNOTE-585, main cohort of 804 only) and placebo plus FLOT (MATTERHORN). KEYNOTE-585's overall survival result did not show a significant difference (the hazard ratio interval includes 1), MATTERHORN's overall survival was a secondary endpoint at an interim analysis, and the KEYNOTE-585 record is marked needs-review. MATTERHORN and KEYNOTE-585 are drawn at their shared milestone year (2025). Compare only each trial's own two arms; the chart is not a pooled or network estimate.

Each pair shows one trial: its control arm and its experimental arm. Yesterday's experimental arm is often today's control. Populations, definitions and follow-up differ.

020406080OS, median (months)20192025FLOT4-AIOKEYNOTE-585
  • experimental arm
  • control arm
  • different endpoint from the lane's
  • step line through experimental values
OS, median, across 2 trials in this lane. FLOT4-AIO (2019): experimental 50 months vs control 35 months; KEYNOTE-585 (2025): experimental 71.8 months vs control 55.7 months.
Data table
TrialYearEndpointControl (months)Experimental (months)
FLOT4-AIO2019OS35 months50 months
KEYNOTE-5852025OS55.7 months71.8 months
MATTERHORN2025OS (not plotted: different measure or time point)70.4 %75.7 %

Gastric cancer, adjuvant chemotherapy after D2 gastrectomy

Adults with stage II or III gastric cancer after curative gastrectomy with extended (D2) lymph-node dissection, in East Asia (ACTS-GC in Japan, stage II-III; CLASSIC in South Korea, China and Taiwan, stage II-IIIB) · Adjuvant · endpoint: OS

Results are not comparable across trials.

Both trials compared adjuvant chemotherapy with no further treatment after D2 surgery and are shown at 5 years (survival rates, not medians). ACTS-GC tested one year of S-1 and was stopped at its first interim analysis, with a 5-year update; CLASSIC tested capecitabine plus oxaliplatin and reported its 5-year result as final. Countries, stage limits and chemotherapy differ. These results are specific to this surgical setting: the intergroup INT-0116 trial (adjuvant chemoradiotherapy, with more limited lymph-node surgery in most patients) and its later trials ARTIST and CRITICS are not drawn here because they are different comparisons and report medians or subgroup hazard ratios. Compare only each trial's own two arms.

Each pair shows one trial: its control arm and its experimental arm. Yesterday's experimental arm is often today's control. Populations, definitions and follow-up differ.

020406080100OS at 60 months (%)20072007ACTS-GCCLASSIC
  • experimental arm
  • control arm
  • different endpoint from the lane's
  • step line through experimental values
OS at 60 months, rate, across 2 trials in this lane. ACTS-GC (2007): experimental 71.7 % vs control 61.1 %; CLASSIC (2007): experimental 78 % vs control 69 %.
Data table
TrialYearEndpointControl (%)Experimental (%)
ACTS-GC2007OS61.1 %71.7 %
CLASSIC2007OS69 %78 %

Standard of care, then and now

How the usual sequence of treatment changed for one setting. Choose a point in time; each step links to the milestone behind it.

HER2-positive gastric or GEJ cancer: first-line and later treatment

Adults with advanced HER2-positive gastric or gastro-oesophageal junction adenocarcinoma · Metastatic

  1. Immunotherapy11st line: pembrolizumab + trastuzumabKEYNOTE-811: pembrolizumab plus trastuzumab and chemotherapy (2021)
  2. Immunotherapy21st line: zanidatamab + chemo ± tislelizumabHERIZON-GEA-01: zanidatamab, with or without tislelizumab (2026)
  3. Antibody–drug conjugate3After trastuzumab: T-DXdTrastuzumab deruxtecan in HER2-positive gastric cancer (2020)
Today. Based only on the trials cited in each step. HERIZON-GEA-01 compared zanidatamab-based regimens with trastuzumab plus chemotherapy; the trials were not compared with each other.
Text version
  • 2010: 1st line: trastuzumab + chemotherapy
  • 2021: 1st line: pembrolizumab + trastuzumab → After trastuzumab: T-DXd
  • Today: 1st line: pembrolizumab + trastuzumab → 1st line: zanidatamab + chemo ± tislelizumab → After trastuzumab: T-DXd

Resectable gastric or GEJ cancer: perioperative treatment

Adults with resectable gastric or gastro-oesophageal junction adenocarcinoma · Perioperative

  1. Immunotherapy1Perioperative FLOT + durvalumabMATTERHORN: perioperative durvalumab plus FLOT, and KEYNOTE-585 (2025)
Today. Based only on the trials cited in each step. MATTERHORN added durvalumab to FLOT; KEYNOTE-585, which added pembrolizumab to chemotherapy, reported less clear results.
Text version
  • 2006: Perioperative ECF chemotherapy
  • 2019: Perioperative FLOT chemotherapy
  • Today: Perioperative FLOT + durvalumab

Reversals and dead ends

Approaches that were abandoned or withdrawn, and trials whose results were later questioned, are part of the story.

  • 2022

    Later-line pembrolizumab in gastric cancer: approved, then withdrawn

    In 2017 the US regulator gave pembrolizumab an accelerated approval for stomach cancer that had progressed after two or more treatments, based on tumour response. A separate 2018 randomised trial against paclitaxel in second-line treatment did not show significantly longer survival. The indication was withdrawn from the US label in February 2022; the FDA letter gives no reason.

    KEYNOTE-059 (phase 2, single-arm, 259 patients): ORR 11.6% (95% CI 8.0 to 16.1); 15.5% PD-L1-positive, 6.4% PD-L1-negative. KEYNOTE-061 (phase 3, second line, CPS 1 or more, 395 patients): median OS 9.1 vs 8.3 months for pembrolizumab vs paclitaxel, HR 0.82 (95% CI 0.66 to 1.03), one-sided p=0.0421; PFS HR 1.27. FDA accelerated approval 22 Sep 2017; withdrawn 4 Feb 2022.

Today and tomorrow

Standard of care today

Perioperative
Resectable gastric or gastro-oesophageal junction adenocarcinoma, regulatory documents read. FDA, 25 November 2025: durvalumab (BLA 761069/S-052) in combination with FLOT (fluorouracil, leucovorin, oxaliplatin and docetaxel) as neoadjuvant and adjuvant treatment, followed by single-agent durvalumab, for adults with resectable gastric or gastroesophageal junction adenocarcinoma; the indication wording in the letter carries no PD-L1 or other biomarker limit. EMA: the Imfinzi procedural-steps document lists an accepted extension of indication to the same combination for adults with resectable gastric or gastro-oesophageal junction adenocarcinoma, based on interim results from the MATTERHORN study, with opinion on 29 January 2026 and Commission decision on 12 March 2026 (column order confirmed on the rendered page). The perioperative chemotherapy trials (MAGIC, FLOT4), MATTERHORN and KEYNOTE-585 are covered in the milestone timeline. No guideline text was read for this setting.
US Food and Drug Administration 2025EMA 2026
Metastatic

Advanced gastric or gastro-oesophageal junction adenocarcinoma, first line, HER2-negative or unselected, regulatory documents read. FDA letters:

  • nivolumab, 16 April 2021 (BLA 125554/S-091), with fluoropyrimidine- and platinum-containing chemotherapy for advanced or metastatic gastric cancer, gastroesophageal junction cancer and oesophageal adenocarcinoma
  • pembrolizumab, 16 November 2023 (BLA 125514/S-143), with fluoropyrimidine- and platinum-containing chemotherapy for locally advanced unresectable or metastatic HER2-negative gastric or gastroesophageal junction adenocarcinoma
  • tislelizumab-jsgr (BLA 761417), with platinum and fluoropyrimidine-based chemotherapy for unresectable or metastatic HER2-negative gastric or gastroesophageal junction adenocarcinoma whose tumours express PD-L1 (1 or more) (the FDA approval-package page shows 26 December 2024
  • the corrected letter that could be read was e-signed 2 January 2025)
  • zolbetuximab-clzb, 18 October 2024 (BLA 761365), with fluoropyrimidine- and platinum-containing chemotherapy for HER2-negative disease whose tumours are claudin 18.2 positive by an FDA-approved test. On 22 May 2025 (pembrolizumab, S-182) and 23 May 2025 (nivolumab, S-133) the FDA revised the labels so that these first-line indications apply to tumours expressing PD-L1 (pembrolizumab CPS 1 or more
  • nivolumab PD-L1 1 or more)
  • the letters present these as labelling revisions, not new indications, and describe the revised populations as those with a favourable risk-benefit assessment. EMA procedural-steps documents: nivolumab with chemotherapy for HER2-negative disease with PD-L1 CPS 5 or more (Commission decision 19 October 2021)
  • pembrolizumab with chemotherapy for HER2-negative disease with CPS 1 or more (Commission decision 23 November 2023)
  • tislelizumab with chemotherapy for HER2-negative disease with a tumour area positivity score of 5% or more (Commission decision 25 November 2024)
  • and the Vyloy product page: marketing authorisation issued 19 September 2024 for HER2-negative, claudin 18.2-positive disease. Later EU variations were not read, so the EU and US PD-L1 populations should not be assumed to match today
US Food and Drug Administration 2021US Food and Drug Administration 2023US Food and Drug Administration 2024US Food and Drug Administration 2025US Food and Drug Administration 2024US Food and Drug Administration 2025US Food and Drug Administration 2025European Medicines Agency 2021European Medicines Agency 2023European Medicines Agency 2024European Medicines Agency 2024
Metastatic

HER2-positive advanced gastric, gastro-oesophageal junction or oesophageal adenocarcinoma, first line, regulatory documents read. Trastuzumab (Herceptin):

  • FDA, 20 October 2010 (BLA 103792/5250), for HER2-overexpressing metastatic gastric or gastroesophageal junction adenocarcinoma in patients who have not received prior treatment for metastatic disease
  • EMA, Commission decision 19 January 2010 (variation II/0047), for use with capecitabine or 5-fluorouracil and cisplatin, with HER2 overexpression defined as IHC 2+ with a confirmatory FISH-positive result, or IHC 3+. Pembrolizumab with trastuzumab and fluoropyrimidine- and platinum-containing chemotherapy: FDA accelerated approval on 5 May 2021 (S-097, 21 CFR 601.41)
  • traditional approval on 19 March 2025 (S-170) for tumours expressing PD-L1 (CPS 1 or more), where the corrected letter, e-signed 3 April 2025, states that the effective action date remains 19 March 2025 (the letter's text also refers to the original letter as dated 18 March 2025) and contains a 'Subpart E Fulfilled' section (the letter does not use the word 'converted')
  • EMA, Commission decision 23 August 2023 (II/0133, based on interim KEYNOTE-811 results
  • the scope text states no PD-L1 threshold). Zanidatamab-hrii (BLA 761416/S-002) and tislelizumab-jsgr (BLA 761232/S-018), 25 August 2026: first-line HER2-positive unresectable locally advanced or metastatic gastric, gastroesophageal junction or oesophageal adenocarcinoma, with zanidatamab, fluoropyrimidine- and platinum-containing chemotherapy and tislelizumab (IHC 3+ or IHC 2+/ISH-positive), or zanidatamab with chemotherapy alone (IHC 3+), as detected by an FDA-authorised test. TOGA, KEYNOTE-811 and HERIZON-GEA-01 are in the milestone timeline. EU documents for zanidatamab were not read
US Food and Drug Administration 2010European Medicines Agency 2010US Food and Drug Administration 2021US Food and Drug Administration 2025European Medicines Agency 2023US Food and Drug Administration 2026US Food and Drug Administration 2026
Relapsed / refractory

Advanced gastric or gastro-oesophageal junction adenocarcinoma, later lines, regulatory documents read. Ramucirumab (Cyramza):

  • FDA, 21 April 2014 (BLA 125477), as a single agent after prior fluoropyrimidine- or platinum-containing therapy, and 5 November 2014 (S-2), with paclitaxel after progression on or after prior fluoropyrimidine- or platinum-containing chemotherapy
  • the EMA product page gives 19 December 2014 as the date the marketing authorisation was issued and lists advanced gastric or junctional adenocarcinoma after prior platinum and fluoropyrimidine chemotherapy among its indications. Trastuzumab deruxtecan (Enhertu): FDA, 15 January 2021 (S-011), for locally advanced or metastatic HER2-positive gastric or gastroesophageal junction adenocarcinoma after a prior trastuzumab-based regimen
  • EMA, Commission decision 12 December 2022 (II/0012), for monotherapy in advanced HER2-positive disease after a prior trastuzumab-based regimen. Pembrolizumab as a single agent for recurrent locally advanced or metastatic gastric or gastroesophageal junction adenocarcinoma with PD-L1 CPS 1 or more after two or more prior lines: accelerated approval on 22 September 2017 (S-24, based on tumour response rate and durability of response) and withdrawal by labelling supplement on 4 February 2022 (S-125), with no reason stated in the letter. RAINBOW, REGARD, DESTINY-Gastric01 and the pembrolizumab withdrawal are in the milestone timeline. Trifluridine/tipiracil and other later-line approvals were not read
US Food and Drug Administration 2014US Food and Drug Administration 2014European Medicines Agency 2014US Food and Drug Administration 2021European Medicines Agency 2022US Food and Drug Administration 2017US Food and Drug Administration 2022
Any setting
Guideline documents are cited by identifier only; their text was not read, so no recommendation is attributed to them, and whether they are the current editions was not checked beyond a PubMed search on 21 October 2026: the ESMO Clinical Practice Guideline on gastric and oesophageal cancer (Ann Oncol, published online 15 October 2026, PMID 42880714); the ASCO guideline update on immunotherapy and targeted therapy for advanced gastroesophageal cancer (J Clin Oncol 2026;44(12):1145-1165, PMID 41747202) and its living guideline version 2026.1.2 (J Clin Oncol, PMID 42607281). A corresponding NCCN summary for gastric cancer was not found in that search.
Guideline: Al-Batran SE, Bencivenga M, Crosby T, Dinis-Ribeiro M, Huguet F, Lordick F, et al 2026Guideline: Shah MA, Kennedy EB, Deighton D, Jhawer M, Kadakia KC, Mukherjee S, et al 2026Guideline: Shah MA, Marzalik JS, Kennedy EB, Deighton D, Hamedi Z, Jhawer M, et al 2026

Trials to watch

  • FORTITUDE-101 NCT05052801

    In previously untreated advanced gastric or gastro-oesophageal junction adenocarcinoma with FGFR2b overexpression, does bemarituzumab added to mFOLFOX6 differ from placebo with mFOLFOX6 in overall survival? Registry record: randomised, double-blind, placebo-controlled phase 3 (Amgen); status completed, with primary completion June 2025 and study completion April 2026 (both actual); enrolment 547 (actual); the primary outcome listed is overall survival; no results posted. A PubMed search on 21 October 2026 found FGFR2b prevalence analyses from the screening phase but no primary results paper.

  • STAR-221 NCT05568095

    In previously untreated locally advanced unresectable or metastatic gastric, gastro-oesophageal junction or oesophageal adenocarcinoma, does domvanalimab and zimberelimab with chemotherapy (FOLFOX or CAPOX, investigator's choice) differ from nivolumab with chemotherapy in overall survival? Registry record: randomised, open-label phase 3 (Arcus Biosciences); status completed, with primary completion August 2026 (actual); enrolment 1,040 (actual); the primary outcome listed is overall survival; no results posted. A PubMed search on 21 October 2026 found a separate phase 2 report of the combination but no phase 3 results paper.

  • DESTINY-Gastric05 NCT06731478

    In first-line HER2-positive unresectable locally advanced or metastatic gastric or gastro-oesophageal junction cancer, does trastuzumab deruxtecan with chemotherapy, with or without pembrolizumab, differ from chemotherapy with trastuzumab, with or without pembrolizumab, in progression-free survival? Registry record: randomised, open-label phase 3 (Daiichi Sankyo) with main and exploratory cohorts; status recruiting; enrolment 726 (estimated); the primary outcome listed is progression-free survival; estimated primary completion June 2028; no results posted.

Pioneers

  • Barry J. Marshall

    born 1951 · Co-author of the 1984 Lancet series on curved bacilli in gastritis and ulcer, Nobel laureate in Physiology or Medicine, 2005

  • J. Robin Warren

    1937-2024 · First author of the 1983 Lancet letter on curved bacilli in active chronic gastritis, Nobel laureate in Physiology or Medicine, 2005

References

Show all 93 references

Loading references… Open the list as a page

Every figure was checked against its primary source to the extent possible. Items that could not be fully confirmed are marked “needs review”. Educational history, not medical advice.

Report an error

Links open the primary source. A flag marked ⚑ needs review means a value or claim has not yet been independently confirmed.

2025TreatmentImmunotherapy

majorPerioperativeNeeds review

MATTERHORN: perioperative durvalumab plus FLOT, and KEYNOTE-585

Adding the immunotherapy drug durvalumab to FLOT chemotherapy before and after surgery improved event-free survival in a randomised trial of operable stomach or junction cancer, and the US regulator approved it in 2025. A trial of another drug, pembrolizumab, with chemotherapy reported less clear results.

MATTERHORN (948 randomised, durvalumab or placebo with FLOT): primary endpoint EFS, 2-year EFS 67.4% vs 58.5%, HR 0.71 (95% CI 0.58-0.86; P<0.001); pCR 19.2% vs 7.2%; final OS HR 0.78 (0.63-0.96; p=0.021). KEYNOTE-585 final analysis (main cohort, 804): median OS 71.8 vs 55.7 months, HR 0.86 (0.71-1.06); EFS HR 0.81 (0.67-0.98; cohort not stated).

Standard before

The MATTERHORN report describes perioperative FLOT as a standard therapy for resectable gastric and GEJ adenocarcinoma, with high recurrence rates.

What changed

On 25 November 2025 the FDA approved durvalumab with FLOT as neoadjuvant and adjuvant treatment, followed by single-agent durvalumab, for adults with resectable gastric or GEJ adenocarcinoma. The MATTERHORN authors call perioperative durvalumab plus FLOT a new standard treatment option.

Why it mattered

MATTERHORN added an immune checkpoint inhibitor to the FLOT backbone and reported better event-free survival, a higher pathological complete response rate and, in the final analysis, better overall survival. KEYNOTE-585, which tested pembrolizumab with perioperative chemotherapy, is shown beside it; the confidence interval of its final overall survival hazard ratio includes 1.

Caveats

  • The FDA letter does not name the supporting trial and its indication text carries no PD-L1 restriction in the passage read; it contains no accelerated-approval wording, so the approval type 'regular' is an inference.
  • In the 2025 report, overall survival was an interim secondary analysis: the P value of 0.03 did not cross the prespecified threshold of P<0.0001, and the piecewise hazard ratios were 0.99 for months 0 to 12 and 0.67 from month 12. The final analysis (2026) reports a hazard ratio of 0.78 with p=0.021 against a threshold of p<0.0499.
  • KEYNOTE-585 enrolled a main cohort (804) and a separate FLOT cohort (203); the overall survival figures quoted are for the main cohort. The two trials differ in design, chemotherapy backbone and endpoints and are not compared directly.
  • ClinicalTrials.gov lists 957 participants for MATTERHORN; the papers report 948 randomised of 1258 enrolled.
  • The KEYNOTE-585 abstract's event-free survival hazard ratio is described by its authors as a trend toward longer EFS.