Evaluating Metformin Efficacy in ALS Using Real-World Data: A Causal Inference Approach

Background and ObjectivesObservational data can provide evidence of treatment effects in a low-burden, cost-effective manner. Using a causal inference framework to address selection bias, we illustrate the approach in ALS by examining previously researched associations, baseline function and sex, and evaluating metformins survival benefits. MethodsData are from the ALS Natural History Study collected in the United States, Israel, and Italy from 2015 to 2025. We employ propensity score matching, two approaches to address loss to follow-up (i.e. a naive exclusion and principal stratification), and randomization-based inference to evaluate difference in 18-month restricted mean survival time. ResultsFor the function investigation, estimated survival is 16.39, 16.14, 16.09, and 16.24 months for the high ALSFRS-R group and 14.04, 14.42, 14.72, and 14.60 months for the low ALSFRS-R group using the non imputed and imputed datasets 1-3, respectively (p<0.0001, N=450; p<0.00001, N=1,485; p<0.00001, N=1,492; p<0.00001, N=1,600). Using principal stratification, estimated survival is 16.38, 16.14, 16.09, and 16.23 months for the high ALSFRS-R group and 14.04, 14.42, 14.72, and 14.59 months for the low ALSFRS-R group with the non imputed and imputed datasets 1-3, respectively (p<0.0001, N=449; p<0.00001, N=1,483; p< 0.00001, N=1,491; p<0.00001, N=1,596). For the sex investigation, estimated survival is 15.20, 15.16, 15.20, and 15.17 months for females and 15.41, 14.84, 14.91, and 14.97 months for males using the non imputed and imputed datasets 1-3, respectively (p{approx}0.594, N=612; p{approx}0.155, N=2,081; p{approx}0.200, N=2,104; p{approx}0.359, N=2,111). Using principal stratification, estimated survival is 15.20, 15.16, 15.20, and 15.17 months for females and 15.41, 14.83, 14.91, and 14.97 months for males with the non imputed and imputed datasets 1-3, respectively (p{approx}0.598, N=612; p{approx}0.146, N=2,079; p{approx}0.195, N=2,102; p{approx}0.364, N=2,110). For the metformin investigation, estimated survival is 14.77, 14.87, and 14.57 months for metformin users and 14.59, 14.43, and 13.97 months for non metformin users using imputed datasets 1-3, respectively (p{approx}0.391, N=289; p{approx}0.257, N=279; p{approx}0.191, N=301). Using principal stratification, estimated survival is 14.75, 14.85, and 14.55 months for metformin users and 14.68, 14.58, and 13.97 months for non metformin users with imputed datasets 1-3, respectively (p{approx}0.458, N=286; p{approx}0.351, N=275; p{approx}0.201, N=300). DiscussionWe reject the sharp null hypothesis for the function investigation and fail to reject it for the sex and metformin investigations. Trial Registration InformationClinicalTrials.gov Identifier: NCT05966038.