Relationship between OS and PFS in metastatic breast cancer (MBC): Review of FDA submission data.

1035 Background: An improvement in overall survival (OS) in a randomized controlled study is the gold standard endpoint for the regular approval of new treatments in MBC. Progression-free survival (PFS) has been used for approval of second and third-line indications or accelerated approval of first-line indications. We have further explored the relationship between OS and PFS in MBC. Methods: We reviewed 12 randomized clinical trials in 8,431 patients with MBC. The trials led to new drug approvals or were presented at an ODAC meeting. All of these trials had PFS as a primary or secondary endpoint. Due to important limitations (e.g., lack of endpoint definitions, inconsistent PFS definitions, or insufficient information on trial statistical design) with some published trials, we limited our evaluation and endpoints correlation to data reviewed by the FDA. The relationship between PFS and OS in MBC was investigated using a weighted linear regression model, where each trial was weighted by the sample size. The coefficient of determination (R2) was calculated to explain the variability of the treatment effect on PFS with respect to the treatment effect on OS, where treatment effects were derived as hazard ratios. Analyses were also performed within lines of therapy and for differences in median PFS and OS time. Results: A total of 16 PFS/OS comparison pairs, 10 in first line therapy and 6 in second/third line therapy. The medians (range) in the difference of median PFS and OS for first line are 2.3 (0.3, 5.6) months and 1.7 (−1.7, 5.4) months, respectively. For second/third line, the medians (range) in the difference of median PFS and OS are 1.9 (1.5, 2.8) months and 2.5 (−0.1, 3.3) months, respectively. Conclusions: Our evaluation is limited to 12 trials that were submitted to support approval in MBC. Although the number of comparison pairs was limited, the results (R=0.28) fail to demonstrate an association between PFS and OS. Less than 10% (R2=0.079) of the variation in OS is explained by the variation in PFS, suggesting that the relationship between PFS and OS is weak in MBC. The lack of association is especially apparent for second/third line therapy MBC.