Radiographic Progression-Free Survival (RPFS) as a Surrogate Endpoint for Overall Survival (OS) in Chemotherapy-Naïve Metastatic Castration-Resistant Prostate Cancer (MCRPC): A Correlation Meta-Analysis


Leung L1, Patel M2, Teitsson S3, Hofer K1, Qian A1, Kurt M4
1Evidinno Outcomes Research Inc., Vancouver, BC, Canada, 2Bristol Myers Squibb, Lawrenceville, NJ, USA, 3Bristol Myers Squibb, London, LON, UK, 4Bristol Myers Squibb, Princeton, NJ, USA

Presentation Documents

OBJECTIVES: To assess the correlation between treatment effects on rPFS and OS in chemotherapy-naïve mCRPC within published literature.

METHODS: A targeted literature review (TLR) was conducted from database inception to 12-14-2021 to identify aggregate-level data on rPFS and OS hazard ratios (HRs) from randomized controlled trials (RCTs) evaluating second-generation hormonal therapies, chemotherapies, or immunotherapies in chemotherapy-naïve adults with mCRPC. The correlation between rPFS HR and OS HR was assessed using bivariate random-effects meta-analysis (BRMA) and weighted linear regression (WLR) where each trial was weighted by its sample size. Validity and predictive accuracy of the estimated surrogacy equations were assessed via leave-one-out cross-validation (LOOCV). The primary analysis was conducted using all identified trials (n=23) while sensitivity analyses restricted the evidence base to trials that (1) met the proportional hazards assumption (n=13), (2) included death in the definition of rPFS (n=14), (3) had matching index dates between rPFS and OS (n=22), (4) did not allow treatment crossover (n=21), and (5) reported HRs for both endpoints (n=18).

RESULTS: Of 23 trials identified from the TLR, 21 were phase III, and two were phase II. In the primary analysis, the estimated correlation coefficient was 0.67 from BRMA (95% CI: 0.41-0.82) and 0.66 from WLR (95% CI: 0.46-0.90). The resulting surrogacy equation from WLR was log(HROS) = 0.05 + 0.33 × log(HRrPFS). In sensitivity analyses, correlation estimates ranged between 0.63 and 0.88 in BRMA, and 0.65 and 0.82 in WLR, with sensitivity analysis (1) reporting the strongest point estimates from both models. During LOOCV, alignment between observed and predicted OS HRs was 91.3% in the primary analysis and ≥85% across all analyses.

CONCLUSIONS: Results suggest a moderate-level positive correlation between rPFS and OS across both models and all analyses. Restricting the evidence base to trials meeting the proportional hazards assumption led to stronger correlation estimates.

Conference/Value in Health Info

2022-11, ISPOR Europe 2022, Vienna, Austria

Value in Health, Volume 25, Issue 12S (December 2022)




Clinical Outcomes

Topic Subcategory

Relating Intermediate to Long-term Outcomes


No Additional Disease & Conditions/Specialized Treatment Areas

Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on Update my browser now