Gastric and gastroesophageal junction (GEJ) adenocarcinomas remain challenging to treat, even when diagnosed at a stage where curative surgery is possible. Perioperative chemotherapy with fluorouracil, leucovorin, oxaliplatin, and docetaxel (FLOT) has become an established treatment approach for patients with locally advanced, resectable disease. However, recurrence remains a concern, and integrating immunotherapy into perioperative treatment has become an important area of clinical investigation.
The phase III KEYNOTE-585 trial evaluated pembrolizumab in combination with perioperative chemotherapy in patients with locally advanced gastric or GEJ adenocarcinoma. Although the main cohort demonstrated an improvement in pathological complete response (pCR), the difference in event-free survival (EFS) did not reach statistical significance.
A complementary cohort of the trial investigated pembrolizumab specifically in combination with FLOT chemotherapy, providing additional data on the safety and potential efficacy of this approach.
In November 2026, ESMO Open published the final analysis of this cohort, titled “Pembrolizumab plus FLOT as neoadjuvant and adjuvant therapy in locally advanced gastric and gastroesophageal junction cancer: analysis of the FLOT cohort from the phase 3 KEYNOTE-585 study.”
The study was authored by S.-E. Al-Batran, K. Shitara, G. Folprecht, M. Moehler, E. Goekkurt, I. Ben-Aharon, S. Lonardi, S. Stein, A. Hubert, I. Chau, M. Mishaeli, L. Villanueva, P. Kavan, A. Wang, C.-S. Shih, K. Hatogai, and L. Wyrwicz.
Study Design and Patient Population
KEYNOTE-585 was a randomized, double-blind, placebo-controlled phase III study. The FLOT cohort enrolled patients across 45 medical centers in 11 countries. Eligible patients were adults with previously untreated, locally advanced, resectable gastric or GEJ adenocarcinoma, including Siewert type II and III tumors. Patients were required to have T3 or more advanced primary tumors and/or lymph node involvement, without evidence of distant metastatic disease.
A total of 203 patients were randomized 1:1 to receive pembrolizumab plus FLOT (n = 100) or placebo plus FLOT (n = 103). Patients received four cycles of FLOT during both the neoadjuvant and adjuvant phases, alongside pembrolizumab 200 mg or placebo administered every three weeks. The treatment strategy also included continued pembrolizumab or placebo during the adjuvant period.
The primary endpoint of the FLOT cohort was safety, while exploratory efficacy endpoints included pCR assessed by blinded independent central review, investigator-assessed EFS, and overall survival (OS). The FLOT cohort was not powered to independently establish efficacy, and the efficacy analyses were exploratory and descriptive, without formal hypothesis testing.
Efficacy Outcomes
At the February 16, 2024, data cutoff, the median follow-up was 43.8 months. Of the 202 patients who received treatment, 191 (94.6%) completed all planned neoadjuvant cycles, while 176 (87.1%) underwent surgery with curative intent. Overall, 153 patients (75.7%) proceeded to adjuvant treatment, and 101 (50.0%) completed the adjuvant phase.
Pathological Complete Response
The pCR rate was higher in the pembrolizumab plus FLOT group than in the placebo plus FLOT group, at 17.0% (95% CI, 10.2–25.8) versus 6.8% (95% CI, 2.8–13.5), respectively. The absolute difference was 10.2 percentage points (95% CI, 1.3–19.7).
Subgroup analyses generally favored pembrolizumab, with a larger numerical difference observed among patients with PD-L1 combined positive score (CPS) ≥10. In this subgroup, the difference in pCR rates was 31.3 percentage points (95% CI, 6.2–54.6). However, these findings require cautious interpretation because of the limited sample size.
R0 resection rates were similar between treatment groups, at 79.0% with pembrolizumab plus FLOT and 78.6% with placebo plus FLOT.
Event-Free Survival
Median EFS was not reached in the pembrolizumab group, compared with 31.6 months in the placebo group, with a hazard ratio (HR) of 0.75 (95% CI, 0.50–1.13). At 36 months, EFS rates were 59.0% (95% CI, 48.3–68.3) with pembrolizumab plus FLOT and 46.5% (95% CI, 36.2–56.1) with placebo plus FLOT.
Although EFS numerically favored pembrolizumab, the confidence interval crossed 1, and the exploratory design did not allow a definitive conclusion regarding treatment benefit.
Overall Survival
At the data cutoff, 84 of 203 patients (41.6%) had died. Median OS had not been reached in either treatment group, with an HR of 0.81 (95% CI, 0.53–1.25). The 36-month OS rate was 66.0% (95% CI, 55.8–74.4) with pembrolizumab plus FLOT compared with 59.2% (95% CI, 49.1–68.0) with placebo plus FLOT. These findings showed a numerical trend favoring pembrolizumab, but a survival benefit could not be established from this analysis.
Safety and Tolerability
Safety was the primary focus of the FLOT cohort, and the addition of pembrolizumab was associated with higher rates of several adverse events. Grade 3–4 treatment-related adverse events (AEs) occurred in 72.7% of patients receiving pembrolizumab plus FLOT and 62.1% receiving placebo plus FLOT. Among treated patients, grade 3–5 treatment-related AEs were reported in 75.8% and 63.1%, respectively.
The most frequently reported grade 3–4 treatment-related events included decreased neutrophil counts, neutropenia, diarrhea, vomiting, nausea, and fatigue. Serious treatment-related AEs were reported in 42.4% of the pembrolizumab group compared with 20.4% of the placebo group. Treatment-related AEs leading to discontinuation were also more frequent with pembrolizumab (32.3% versus 17.5%). Treatment-related deaths occurred in three patients (3.0%) receiving pembrolizumab and one patient (1.0%) receiving placebo.
Immune-mediated AEs or infusion reactions were reported in 44.4% and 28.2% of patients, respectively, with grade 3–5 events occurring in 20.2% versus 8.7%. Among patients who underwent surgery, surgery-related AEs were observed in 21.8% of the pembrolizumab group and 12.4% of the placebo group.
Overall, the investigators considered the treatment feasible, with an acceptable safety profile and no new safety signals. Nevertheless, the higher rates of serious adverse events and treatment discontinuation highlight the additional toxicity associated with pembrolizumab.
What Do These Findings Mean?
The final analysis of the KEYNOTE-585 FLOT cohort provides additional evidence regarding the combination of pembrolizumab with perioperative FLOT chemotherapy in locally advanced gastric and GEJ adenocarcinoma.
The higher pCR rate and numerical differences in EFS and OS were generally consistent with the trends observed in the main KEYNOTE-585 cohort. However, these results need to be considered alongside the study’s limitations. The FLOT cohort was designed primarily to assess safety rather than establish independent efficacy. In addition, the main KEYNOTE-585 cohort did not meet its primary EFS endpoint, limiting the conclusions that can be drawn from the exploratory FLOT analysis.
The authors also discussed these findings in the context of other perioperative immunotherapy studies, including MATTERHORN and DANTE. While these trials reported higher pathological response rates with the addition of immunotherapy to FLOT, differences in trial design, patient populations, and statistical methods make direct comparisons difficult. Importantly, a higher pCR rate cannot automatically be interpreted as evidence of improved long-term survival.
Takeaway
The final analysis of the KEYNOTE-585 FLOT cohort showed a numerically higher pCR rate and favorable EFS and OS estimates with perioperative pembrolizumab plus FLOT compared with FLOT alone in patients with locally advanced, resectable gastric or GEJ adenocarcinoma.
However, these exploratory findings did not establish a definitive survival benefit and were accompanied by increased treatment-related toxicity. Given the limited statistical power of the cohort and the results of the main KEYNOTE-585 study, the findings do not support a change in clinical practice but provide additional evidence for evaluating perioperative immunotherapy in this setting.

