# Do trial benefits predict real-world gains in metastatic castration resistant prostate cancer

## TL;DR

This study examines how well clinical trial results for prostate cancer drugs predict real-world patient outcomes, finding that trial benefits can underestimate real-world gains when selection bias is considered.

## Contribution

The study introduces a method to adjust for treatment selection bias, revealing that real-world benefits of mCRPC drugs can exceed clinical trial estimates.

## Key findings

- Real-world OS gains were 15% lower than RCT benefits when treatment selection bias was unaccounted.
- After adjusting for selection bias, real-world OS gains were 28% greater than RCT OS and rPFS benefits.
- The association between treatment and OS gains increased with the duration a therapy was on the market.

## Abstract

It is important to understand the relationship between drug efficacy measured in randomized clinical trials (RCTs) and real-world drug effectiveness. We estimate how RCT overall survival (OS) and RCT radiographic progression-free survival (rPFS) benefits predict the association between treatments and real-world OS gains for metastatic castration-resistant prostate cancer (mCRPC) drugs.

Using the National Cancer Institute list of approved cancer drugs and the National Comprehensive Cancer Network Treatment Guidelines, we identified all pharmaceutical therapies for mCRPC approved between 2010 and 2019. We obtained RCT OS and rPFS hazard ratios from the pivotal trials used for Food and Drug Administration (FDA) approval, and we estimated real-world OS hazard ratios using the Optum Clinformatics Extended DataMart Databases. We modeled real-world OS hazard ratios as a function of both RCT OS and RCT rPFS hazard ratios using Cox proportional hazards regressions, adjusted for year of diagnosis, age, race, and Elixhauser Comorbidity Index.

When we did not account for nonrandom real-world selection of patients into receiving a newly approved therapy (ie, “treatment selection bias”), real-world OS gains were 15% lower than associated RCT OS and RCT rPFS benefits. However, after accounting for treatment selection bias in real-world settings, real-world OS gains were almost 28% greater than RCT OS and RCT rPFS benefits. Association between treatment and OS gains increased the longer a new therapy was on the market.

After adjusting for treatment selection bias, RCT OS and RCT rPFS estimates serve as useful, or even conservative, predictors of RW OS gains.