Design: multicentre, randomised, open-label phase 3 trial
Population: 724 patients with stage II–III gastric or gastro-oesophageal junction adenocarcinoma after D2 gastrectomy
Primary outcome: overall survival
Effect: 5-year overall survival 70.9% vs 62.9%, hazard ratio 0.74 (95% CI 0.58–0.95), p=0.018

Adjuvant chemotherapy after D2 gastrectomy is standard, but whether S-1 alone is enough for stage II–III disease has divided practice between regions. CAPITAL randomised 724 patients with pathological stage II–III gastric or gastro-oesophageal junction adenocarcinoma, one to one, to S-1 with oxaliplatin or S-1 alone, with overall survival as the primary endpoint and follow-up now at a median of 74 months.

Five-year overall survival was 70.9% (95% CI 66.0–76.1) with the doublet against 62.9% (57.8–68.5) with S-1 alone — hazard ratio 0.74 (0.58–0.95, p=0.018). Five-year disease-free survival was 66.2% against 55.6%, hazard ratio 0.76 (0.61–0.96). The cost is real: grade 3–4 treatment-related adverse events occurred in 25% of the doublet group against 13%, with grade 3–4 neutropenia in 13% against 7%.

An eight percentage point gain in five-year survival is a large effect in this setting, and the toxicity is familiar and manageable rather than novel. The population is the thing to hold in mind — all had undergone D2 lymphadenectomy, which is the standard in East Asian centres and in high-volume Indian units but not universal. Where the surgery was less than a D2, this trial does not tell you what the adjuvant regimen should be, and the fitness needed to complete oxaliplatin after a gastrectomy is not a given.

Confirm the lymphadenectomy was a D2 before applying this result
Assess fitness for oxaliplatin explicitly after gastrectomy — weight loss and poor intake are the limiting factors
Warn about neutropenia specifically; it was the dominant grade 3–4 event
Baseline and monitor for peripheral neuropathy, which will determine whether the course is completed
Quote the absolute figures in consent: about eight more patients alive at five years per hundred treated

Why it matters

It settles a regional divergence in adjuvant practice with an overall survival endpoint and six years of follow-up.

The statistics, in plain English

The hazard ratio of 0.74 has an upper bound of 0.95, so the benefit is real but the interval sits close enough to 1.0 that the true effect could be modest. The absolute difference — 70.9% against 62.9% — is the number for a consent conversation, and it corresponds to treating about 13 patients to have one more alive at five years. Set that against grade 3–4 toxicity in one in four rather than one in eight, which is roughly one extra serious toxicity for every eight treated.