The relative efficacies of cytotoxic chemotherapy regimens in the treatment of advanced breast cancer are generally assessed by comparing response rates in randomised trials. Treatment attempts to prolong survival but trials rarely demonstrate a statistically significant survival advantage: it has been argued that chemotherapy does not prolong survival. The correlation between response rates and survival has been examined by reviewing 79 comparisons between arms with unequal response rates in 50 published trials of chemotherapy in advanced breast cancer. In 73% of comparisons the group with the higher response rate also demonstrated the longer median survival (P less than 0.001). Weighted linear regression showed a statistically significant relationship between relative response rates and survival (P less than 0.001). The number of patients in a comparison did not influence this relationship.