OBJECTIVE: There is evidence that patients with cancer have worse survival if they have comorbid major depression, but uncertainty whether a reduction in depression severity improves survival. We aimed to address this question. METHODS: We did a secondary analysis of data from participants in the SMaRT Oncology-2 and 3 trials of depression treatment in patients with cancer and comorbid major depression (total n = 642). Participants' data were analysed as cohorts, defined by treatment (usual care or Depression Care for People with Cancer, an intensive treatment programme, in both trials) and cancer prognosis (good or poor, in SMaRT Oncology-2 and 3 respectively). We measured change in depression severity from randomisation to 12 weeks using Symptom Checklist Depression Scale (SCL-20) scores and assessed survival by linked mortality data. We used Cox regression to estimate the effect of a one-unit decrease in SCL-20 score on survival, controlling for measured confounders. RESULTS: We found no evidence for an association between improvement in depression and survival in any of the four cohorts, after adjusting for age, sex, primary cancer, baseline cancer severity and baseline depression severity. Pooling the cohorts in a fixed-effects meta-analysis yielded an estimated 7% reduction in the hazard of death per one-unit decrease in SCL-20 score. This finding was not statistically significant; the 95% confidence interval extended from a 26% decrease to an 18% increase in hazard of death. CONCLUSION: We found no evidence that reduction in severity of comorbid major depression is associated with longer survival in patients with cancer.