MATTERHORN：EFS与最终OS

Figure 1. Kaplan–Meier Estimates of Event-free Survival (Full Analysis Population). The threshold of significance for this analysis was 0.0239. The analysis was based on blinded independent central review assessments, on local pathological testing (if it was clinically required), or both. The hazard ratio and corresponding confidence interval were estimated from a Cox proportional-hazards model, with adjustments for geographic region, clinical lymph-node status, and programmed death ligand 1 (PD-L1) expression. The confidence interval for the hazard ratio was calculated with the use of a profile likelihood approach. A hazard ratio of less than 1 favored durvalumab. A two-sided P value was calculated with a stratified log-rank test with adjustments for geographic region, clinical lymph-node status, and PD-L1 expression. At the data-cutoff date, in the event-free survival analysis, data from three participants (0.6%) in the durvalumab group and seven participants (1.5%) in the placebo group were censored at the ran­ domization date because of a missing disease assessment before or after the start of neoadjuvant treatment (surgery was not attempt­ ed, and they did not die or have a non-RECIST progression); data from three participants (0.6%) and six participants (1.3%), respective­ ly, were censored at the surgery date because of a missing adjuvant postsurgery baseline assessment; and data from eight participants (1.7%) and six participants (1.3%) who withdrew consent were censored at the last disease assessment available before withdrawal from the trial (Table S6). Tick marks indicate censored observations. CI denotes confidence interval, FLOT fluorouracil, leucovorin, oxaliplatin, and docetaxel, and NR not reached.

Figure 2. Event-free Survival in Participant Subgroups (Full Analysis Population). Subgroups defined according to sex, age at randomization, geographic region, clinical lymph-node status, PD-L1 expression, tumor lo­ cation, and Eastern Cooperative Oncology Group (ECOG) performance-status score were prespecified in the protocol. ECOG performance- status scores range from 0 to 5, with higher scores indicating greater disability. The analysis was performed with a Cox proportional- hazards model with treatment as the only covariate. A hazard ratio of less than 1 favored durvalumab. The confidence interval was cal­ culated with the use of a profile likelihood approach. Because event-free survival, overall survival, and pathological complete response were the only end points that were part of the multiple testing procedure, the widths of the confidence intervals for other end points or subgroups have not been adjusted for multiplicity and may not be used in place of hypothesis testing. The gray band represents the 95% confidence interval for the hazard ratio calculated for all participants. The analysis was based on blinded independent central review assessment, on local pathological testing (if clinically required), or both. The proportional-hazards assumption that included the interac­ tion of time by treatment in general was tested and was satisfied for most event-free survival subgroups (see Fig. S5 in the Supplemen­ tary Appendix). PD-L1 expression was assessed according to the Tumor Area Positivity (TAP) score,19 which is determined by visual aggregation and estimation of the area covered by PD-L1 positive tumor cells and tumor-associated immune cells relative to the total tumor area on the immunohistochemical slide. NC denotes not calculable.

Figure 3. Kaplan–Meier Estimates of Overall Survival and Pathological Complete Response (Full Analysis Population). Panel A shows Kaplan–Meier estimates of overall survival calculated with an unstratified Cox proportional-hazards model. A hazard ratio of less than 1 favors durvalumab plus FLOT. The threshold of significance for this analysis was P<0.0001. Tick marks indicate censored observations. The two-sided P value was calculated with a stratified log-rank test with adjustments for geographic region, clinical lymph-node status, and PD-L1 expression. Data from 9 participants (1.9%) in the durvalumab group and 17 participants (3.6%) in the placebo group were censored because the participants withdrew from the trial before death; the data were censored at the last date at which the participants were known to be alive (Table S6). Some of these participants were able to be followed for survival status after having withdrawn from the trial; if not, where applicable, survival status was checked from publicly available death registries. Panel B shows the percentages of participants in each group who had a pathological complete response. The threshold of significance for this analysis was P = 0.001 . The analysis for relative risk was performed with a log-binomial regression model with adjustment for stratification factors, which included geographic region, clinical lymph-node status, and PD-L1 expression at randomization. For pathological complete response, 43 participants (9.1%) in the durvalumab group and 46 participants (9.7%) in the placebo group did not have surgery attempted and did not provide a surgical specimen for central assessment of pathological response and were classified as not having had a response; 19 participants (4.0%) and 28 participants (5.9%) in the durvalumab and placebo groups, respectively, had surgery attempted but had a missing sample; and 24 participants (5.1%) and 24 participants (5.1%) had surgery attempted but did not have evaluable samples and were classified as not having had a response (Table S7).

Table 1. Baseline Demographic and Clinical Characteristics in the Full Analysis Population.*

* FLOT denotes fluorouracil, leucovorin, oxaliplatin, and docetaxel. † Race or ethnic group was reported by the patient. The category “Other” included any patient-reported race or ethnic group that did not fit in any of the categories specified on the basis of definitions from the Food and Drug Administration. ‡ Eastern Cooperative Oncology Group (ECOG) performance-status scores range from 0 to 5, with higher scores indi­ cating greater disability. § Siewert type 1 indicates an adenocarcinoma of the lower esophagus with the center located within 1 to 5 cm above the anatomical gastroesophageal junction; Siewert type 2 indicates a true carcinoma of the cardia at the gastroesoph­ ageal junction, with the tumor center within 1 cm above and 2 cm below the gastroesophageal junction; and Siewert type 3 indicates a subcardial carcinoma with the tumor center between 2 cm and 5 cm below the gastroesophageal junction, which infiltrates the gastroesophageal junction and lower esophagus from below. ¶ Tumor staging was performed according to the eighth edition of the American Joint Committee on Cancer AJCC Cancer Staging Manual.17 ‖ The lymph-node status was recorded at randomization with the use of the interactive response technology system or the randomization and trial supply management system. ** Programmed death ligand 1 (PD-L1) expression (assessed according to the Tumor Area Positivity [TAP] score,19 which is determined by visual aggregation and estimation of the area covered by PD-L1 positive tumor cells and tumor-associated immune cells relative to the total tumor area on the immunohistochemical slide) was measured by the VENTANA PD-L1 (SP263) CDx assay and recorded at randomization with the use of the interactive response technology system, randomiza­ tion and trial supply management system, or an electronic case-report form or with data from an external vendor from samples collected on or before randomization. †† Microsatellite instability was measured by a clinical trial assay based on FoundationOne CDx in a research use only capacity.

* Included are adverse events reported during the overall treatment period that had an onset date on or after the first dose of investigational treatment, as well as adverse events that had an onset date before the first dose but that increased in severity on or after the first dose up to and including 90 days after the last dose or until initiation of the first subsequent anticancer therapy (excluding palliative radiotherapy), whichever occurred first. † One participant in the placebo group received a single dose of durvalumab and was therefore included in the durvalumab group for the safety analysis. ‡ This category excludes infusion or hypersensitivity reactions. § A surgical delay was defined as surgery occurring more than 8 weeks (56 days) after the last dose of neoadjuvant treatment.

Figure 1: Trial profile Data cutoff: Dec 20, 2024. From The New England Journal of Medicine, Janjigian YY, Al-Batran SE, Wainberg ZA, Muro K, Molena D, Van Cutsem E, Hyung WJ, Wyrwicz L, Oh DY, Omori T, Moehler M, Garrido M, Oliveira SCS, Liberman M, Oliden VC, Smyth EC, Stein A, Bilici M, Alvarenga ML, Kozlov V, Rivera F, Kawazoe A, Serrano O, Heilbron E, Negro A, Kurland JF, Tabernero J, for the MATTERHORN Investigators, Perioperative Durvalumab in Gastric and Gastroesophageal Junction Cancer, 393, 217–230, Copyright © 2025, Massachusetts Medical Society. Reprinted with permission from Massachusetts Medical Society. FLOT=fluorouracil, leucovorin, oxaliplatin, and docetaxel. ITT=intention-to-treat. *One participant was randomised twice in error and excluded at the level of the ITT analysis set. †Reasons are provided in the appendix (pp 8–9). ‡Reasons were not mutually exclusive for participants with multiple reasons, although these were only counted once per category. §Completed all cycles with all FLOT components.

Figure 1: Trial profile Data cutoff: Dec 20, 2024. From The New England Journal of Medicine, Janjigian YY, Al-Batran SE, Wainberg ZA, Muro K, Molena D, Van Cutsem E, Hyung WJ, Wyrwicz L, Oh DY, Omori T, Moehler M, Garrido M, Oliveira SCS, Liberman M, Oliden VC, Smyth EC, Stein A, Bilici M, Alvarenga ML, Kozlov V, Rivera F, Kawazoe A, Serrano O, Heilbron E, Negro A, Kurland JF, Tabernero J, for the MATTERHORN Investigators, Perioperative Durvalumab in Gastric and Gastroesophageal Junction Cancer, 393, 217–230, Copyright © 2025, Massachusetts Medical Society. Reprinted with permission from Massachusetts Medical Society. FLOT=fluorouracil, leucovorin, oxaliplatin, and docetaxel. ITT=intention-to-treat. *One participant was randomised twice in error and excluded at the level of the ITT analysis set. †Reasons are provided in the appendix (pp 8–9). ‡Reasons were not mutually exclusive for participants with multiple reasons, although these were only counted once per category. §Completed all cycles with all FLOT components.

Figure 2: Overall survival and overall survival among participant subgroups, including participants with PD-L1 TAP ≥1% or <1%, in the durvalumab plus FLOT group versus the placebo plus FLOT group in the ITT population Overall survival (A), subgroup analysis of overall survival (B), and overall survival among participants with PD-L1 TAP ≥1% (C) and PD-L1 TAP <1% (D) in the durvalumab plus FLOT group versus the placebo plus FLOT group. Data cutoff: Sept 1, 2025. An HR of <1 favours durvalumab plus FLOT. (A) Overall survival maturity was 37%. Median overall survival was NR (95% CI NR–NR) in either treatment group. (B) The grey band represents the 95% CI for the HR for the ITT population. (C) Median overall survival was NR (95% CI NR–NR) in either treatment group. (D) Median overall survival in months was NR (95% CI 43·66–NR) with durvalumab plus FLOT and NR (95% CI 21·72–NR) with placebo plus FLOT. ECOG=Eastern Cooperative Oncology Group. FLOT=fluorouracil, leucovorin, oxaliplatin, and docetaxel. HR=hazard ratio. ITT=intention-to-treat. NR=not reached. RMST=restricted mean survival time. TAP=tumour area positivity. *In censored participants.

Figure 2: Overall survival and overall survival among participant subgroups, including participants with PD-L1 TAP ≥1% or <1%, in the durvalumab plus FLOT group versus the placebo plus FLOT group in the ITT population Overall survival (A), subgroup analysis of overall survival (B), and overall survival among participants with PD-L1 TAP ≥1% (C) and PD-L1 TAP <1% (D) in the durvalumab plus FLOT group versus the placebo plus FLOT group. Data cutoff: Sept 1, 2025. An HR of <1 favours durvalumab plus FLOT. (A) Overall survival maturity was 37%. Median overall survival was NR (95% CI NR–NR) in either treatment group. (B) The grey band represents the 95% CI for the HR for the ITT population. (C) Median overall survival was NR (95% CI NR–NR) in either treatment group. (D) Median overall survival in months was NR (95% CI 43·66–NR) with durvalumab plus FLOT and NR (95% CI 21·72–NR) with placebo plus FLOT. ECOG=Eastern Cooperative Oncology Group. FLOT=fluorouracil, leucovorin, oxaliplatin, and docetaxel. HR=hazard ratio. ITT=intention-to-treat. NR=not reached. RMST=restricted mean survival time. TAP=tumour area positivity. *In censored participants.

Figure 3: Event-free survival by pathological outcomes and nodal staging in the durvalumab plus FLOT group versus the placebo plus FLOT group in the central pathology analysis set and nodal involvement evaluable analysis set Event-free survival among participants who reached pathological complete response (A), non-pathological complete response (B), major pathological response (C), any pathological response (D), with no nodal involvement (E), and with nodal involvement (F) in the durvalumab plus FLOT group versus the placebo plus FLOT group. Data cutoff: Dec 20, 2024. (A) Median event-free survival was NR (95% CI NR–NR) in either treatment group. (B) Median event- free survival in months was NR (95% CI 40·74–NR) with durvalumab plus FLOT and NR (95% CI 32·20–NR) with placebo plus FLOT. (C) Median event-free survival was NR (95% CI NR–NR) in either treatment group. (D) Median event- free survival in months was NR (95% CI NR–NR) with durvalumab plus FLOT and NR (95% CI 35·65–NR) with placebo plus FLOT. (E) Median event-free survival was NR (95% CI NR–NR) in either treatment group. (F) Median event-free survival in months was 33·84 (95% CI 23·33–NR) with durvalumab plus FLOT and 25·56 (95% CI 18·83–29·14) with placebo plus FLOT. FLOT=fluorouracil, leucovorin, oxaliplatin, and docetaxel. HR=hazard ratio. NR=not reached. *Analysis based on a total of 11 events.

Baseline demographics and clinical characteristics, ITT population. Full table and footnotes span PDF pages 6–7.

Baseline demographics and clinical characteristics, ITT population. Full table and footnotes span PDF pages 6–7.

Adverse event summary, safety analysis set. Full table and footnotes span PDF pages 12–13.

Adverse event summary, safety analysis set. Full table and footnotes span PDF pages 12–13.