The Absolute Effect / Evidence
ATTRACTION-4
Clinical question: In first-line her2-negative advanced/recurrent gastric/gej cancer in east asia, does the investigational strategy improve clinically meaningful outcomes versus its control?
Registry: NCT02746796 · One underlying study; aliases and follow-up publications consolidated.
Phase III, double-blind, placebo-controlled RCT; 724 patients. Evidence cut-off: 13 November 2026.
Design, population and eligibility
Phase III, double-blind, placebo-controlled RCT; 724 patients.
Previously untreated HER2-negative unresectable advanced or recurrent gastric/GEJ cancer in Japan, South Korea and Taiwan.
Nivolumab + SOX or CAPOX versus placebo + SOX or CAPOX.
Nivolumab 360 mg IV every 3 weeks with oxaliplatin-based S-1 or capecitabine chemotherapy.
placebo + SOX or CAPOX.
Primary endpoint: Co-primary centrally assessed PFS and OS.
Key efficacy outcomes
Initial PFS 10.45 vs 8.34 months, HR 0.68; OS 17.45 vs 17.15 months, HR 0.90 (95% CI 0.75–1.08), p=0.26. At 3-year follow-up, PFS 10.94 vs 8.48 months (HR 0.67); OS remained similar (HR 0.89, 95% CI 0.75–1.05).
Comparator: 8.34. The bars are a simple display of the first reported pair only; consult the endpoint and time horizon above.
Safety and treatment delivery
Grade 3–4 neutrophil decrease 20% vs 16%; platelet decrease 9% vs 9%; treatment-related deaths 3 vs 3.
Clinical interpretation and limitations
Interpret the reported hazard ratios in the context of the prespecified endpoint, statistical hierarchy, follow-up maturity, crossover/subsequent therapy and enrolled population. Median survival differences are not risk differences and must not be converted into NNT.
Absolute effect
No NNT/NNH is displayed unless an absolute event-risk difference at a clearly stated fixed timepoint can be calculated from reported trial data. Median survival differences and hazard ratios alone do not support an NNT. The visual above uses reported outcome data only.
Primary publication and follow-up
Registry / follow-up / authoritative source
Source-level verification status: primary publication identified and key quoted outcomes transcribed; where full-text toxicity or procedural details were not available, the page explicitly flags the limitation rather than inferring values.
Current-practice context · 13 November 2026
This page separates historical trial results from contemporary practice. Verify the applicable jurisdictional guideline, regulatory approval, biomarker assay, liver function, comorbidity and treatment availability before applying the evidence. NCCN Guidelines Insights are not a substitute for the full NCCN guideline.
Guideline/regulatory reconciliation and browser-rendered QA remain scheduled for the subsequent quality-control pass.