Propensity Score Weighting Using Overlap Weights: A New Method Applied to Regorafenib Clinical Data and a...

Area of Science:

• Epidemiology
• Biostatistics
• Health Economics

Background:

• Inverse propensity score (PS) weights can be problematic with limited population overlap, leading to extreme values.
• Propensity score weighting using overlap weights (PSOW) addresses this by focusing on the point of highest mutual overlap.
• PSOW offers a potential advantage over traditional methods like trimming, target, or inverse weighting in specific scenarios.

Purpose of the Study:

• To evaluate the performance of the PSOW method using regorafenib effectiveness data.
• To compare PSOW against existing balancing methods in a real-world metastatic colorectal cancer (mCRC) dataset and a randomized clinical trial (RCT).
• To assess the cost-effectiveness of regorafenib versus placebo in mCRC patients.

Main Methods:

• Propensity scores were balanced using PSOW for key patient characteristics (age, sex, performance status, treatment lines, cancer location, KRAS mutation, time from metastases).
• Weighted Kaplan-Meier curves were generated and utilized in a 3-state partitioned survival model.
• The R code for the analysis is available.

Main Results:

• PSOW demonstrated superior performance compared to target or inverse PS weights, evidenced by effective sample size and improved weight distribution.
• Regorafenib showed enhanced survival in both the registry and RCT cohorts when compared to the RCT alone.
• The PSOW hazard ratio for overall survival (OS) was 0.53, offering a more conservative estimate than inverse (0.44) or target (0.27) weights.

Conclusions:

• This study marks the first application of PSOW in clinical data and cost-effectiveness analysis.
• The PSOW method is promising for analyses involving weak or small population overlap.
• PSOW facilitates feasible pharmacoeconomic modeling in challenging data scenarios.