Exploring the Relationship Between Progression-Free Survival and Overall Survival in Breast Cancer Patients Treated with Fulvestrant or Anastrozole

Due to the limitations of measuring overall survival (OS), including the potential requirements for larger study populations and longer follow-up periods, progression-free survival (PFS) is often used as a surrogate for OS. The relationship between PFS and OS in advanced cancer, including breast cancer, has been supported by previous analyses and the use of PFS as a proxy for OS is supported by the US Food and Drug Administration. “However, increased PFS does not necessarily always translate into improved OS,” said Dr Mario Ouwens of AstraZeneca in Mölndal, Sweden. In this analysis, Dr Ouwens and colleagues further examined the relationship between PFS and OS, with the goal of using PFS as a predictor of OS using data from the phase 2 FIRST study of fulvestrant 500 mg versus anastrozole as first-line treatment in hormone receptor–positive (HR+) advanced breast cancer.

Although the FIRST study enrolled 205 postmenopausal women with HR+ advanced breast cancer, to optimize the homogeneity of the sample population, this analysis only evaluated endocrine-naïve patients in the study (n = 73 [72%] fulvestrant; n = 80 [78%] anastrozole). The relationship between PFS and OS was determined by substituting the linear expression into each other and using Weibull parametric fits and linear regression. PFS and OS data from a previous study of anastrozole in a similar population were applied to validate the Weibull and linear regression model.

For OS, there were 44 events for fulvestrant and 59 for anastrozole (hazard ratio [HR], 0.63; 95% confidence interval [CI], 0.42-0.93). For PFS, there were 44 events for fulvestrant and 65 for anastrozole (HR, 0.52; 95% CI 0.35-0.77). Using log cumulative hazard plots, relationships between OS and PFS could be derived from a Weibull model, such that the survival percentage (S) was given by fulvestrant ln time_OS(S) = 0.82 + 0.80 ln time_PFS(S); and anastrozole ln time_OS(S) = 0.97 + 0.79 ln time_PFS(S). Additionally, when a strictly linear section was investigated, the following relationships were derived for each survival percentage (S): fulvestrant ln time_OS(S) = 0.77 + 0.63 ln time_PFS(S); and anastrozole ln time_OS(S) = 0.95 + 0.67 ln time_PFS(S). For all equations, time is given in days/1000. Applying both models to other clinical data for anastrozole showed a good fit, potentially extending this relationship beyond the FIRST study.

This analysis demonstrates a novel, validated approach by which the relationship between PFS and OS in patients receiving first-line treatment with fulvestrant or anastrozole can be modeled. These results add to the acceptance of PFS as a predictor of OS in this setting, and although the exact regression coefficients would not apply to other studies, “the methodological approach can be applied to predict OS when trial data is relatively immature given more mature PFS data,” said Dr Ouwens.




  • Ouwens M, et al. ESMO 2016. Abstract 250P.