Radiographic Progression-Free Survival (rPFS) as a Surrogate Endpoint for Overall Survival (OS) in First-Line (1L) Metastatic Castration-Resistant Prostate Cancer (mCRPC): A Meta-Analysis and Surrogate Threshold Effect
Author(s)
Imtiaz Samjoo, BSc, MSc, PhD1, Elena Castro, MD, MSc, PhD2, Stefanie Paganelli, BSc1, Di Wang, MSc1, Anja Haltner, MSc3, Alexander Niyazov, MPH, PharmD4, Jane Chang, MPH4, Pedro Barata, MD, MSc, FACP5.
1EVERSANA, Burlington, ON, Canada, 2Hospital Universitario 12 de Octubre, Madrid, Spain, 3EVERSANA, New York, NY, USA, 4Pfizer, Inc., New York, NY, USA, 5University Hospitals Seidman Cancer Center, Cleveland, OH, USA.
1EVERSANA, Burlington, ON, Canada, 2Hospital Universitario 12 de Octubre, Madrid, Spain, 3EVERSANA, New York, NY, USA, 4Pfizer, Inc., New York, NY, USA, 5University Hospitals Seidman Cancer Center, Cleveland, OH, USA.
OBJECTIVES: To evaluate the validity of rPFS as a surrogate endpoint for OS in 1L asymptomatic or mildly symptomatic mCRPC, using methodology proposed by Germany’s Institute for Quality and Efficiency in Health Care (IQWiG).
METHODS: A systematic literature search in Ovid® including key grey literature sources identified randomized controlled trials (RCTs) evaluating 1L mCRPC treatments reporting both rPFS and OS. Trial-level correlations between rPFS and OS hazard ratios (HRs) were calculated using bivariate random-effects meta-analysis (BRMA) and weighted linear regression (WLR). Validity of the BRMA model was assessed using leave-one-out cross-validation (LOOCV), and correlation strength was interpreted per IQWiG criteria. The surrogate threshold effect (STE) was estimated to assess surrogacy. The primary analysis included trials meeting the proportional hazards (PH) assumption. A sensitivity analysis excluding trials identified as outliers (trials falling outside of the 95% confidence interval [CI] in the correlation plot) was conducted.
RESULTS: Eleven RCTs (n = 9,927 patients) were identified that met the eligibility criteria. One trial (PEACE-3) violated the PH assumption and was excluded from the primary analysis (n = 10 trials). A medium correlation between rPFS and OS was observed in the primary analysis (BRMA: R² [95% CI]: 0.78 [0.53, 0.90]; WLR: R² [95% CI]: 0.65 [0.40, 0.90]), with a corresponding STE of 0.83. Two additional trials (NCT02294461 and ERA 223) were identified as outliers. In the sensitivity analysis further excluding NCT02294461 and ERA 223 (n = 8 trials), a strong correlation was observed (BRMA: 0.92 [0.74, 0.97]; WLR: 0.91 [0.84, 0.99]). LOOCV demonstrated good predictive accuracy for OS in both the primary (80%) and sensitivity analysis (75%).
CONCLUSIONS: Results suggest rPFS is correlated to and a valid surrogate for OS in 1L mCRPC. A statistically significant effect in OS is therefore possible for a hypothetical trial demonstrating an upper confidence limit of HR < 0.83 in rPFS.
METHODS: A systematic literature search in Ovid® including key grey literature sources identified randomized controlled trials (RCTs) evaluating 1L mCRPC treatments reporting both rPFS and OS. Trial-level correlations between rPFS and OS hazard ratios (HRs) were calculated using bivariate random-effects meta-analysis (BRMA) and weighted linear regression (WLR). Validity of the BRMA model was assessed using leave-one-out cross-validation (LOOCV), and correlation strength was interpreted per IQWiG criteria. The surrogate threshold effect (STE) was estimated to assess surrogacy. The primary analysis included trials meeting the proportional hazards (PH) assumption. A sensitivity analysis excluding trials identified as outliers (trials falling outside of the 95% confidence interval [CI] in the correlation plot) was conducted.
RESULTS: Eleven RCTs (n = 9,927 patients) were identified that met the eligibility criteria. One trial (PEACE-3) violated the PH assumption and was excluded from the primary analysis (n = 10 trials). A medium correlation between rPFS and OS was observed in the primary analysis (BRMA: R² [95% CI]: 0.78 [0.53, 0.90]; WLR: R² [95% CI]: 0.65 [0.40, 0.90]), with a corresponding STE of 0.83. Two additional trials (NCT02294461 and ERA 223) were identified as outliers. In the sensitivity analysis further excluding NCT02294461 and ERA 223 (n = 8 trials), a strong correlation was observed (BRMA: 0.92 [0.74, 0.97]; WLR: 0.91 [0.84, 0.99]). LOOCV demonstrated good predictive accuracy for OS in both the primary (80%) and sensitivity analysis (75%).
CONCLUSIONS: Results suggest rPFS is correlated to and a valid surrogate for OS in 1L mCRPC. A statistically significant effect in OS is therefore possible for a hypothetical trial demonstrating an upper confidence limit of HR < 0.83 in rPFS.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
PT3
Topic
Clinical Outcomes
Topic Subcategory
Relating Intermediate to Long-term Outcomes
Disease
Oncology