Introducing propensity score matching with progression rate matching (PSM-with-PRM): a novel method to better align uncontrolled extensions of randomized controlled trials data with external comparators

Introduction : Randomized studies that are followed by an open-label extension (OLE) period provide valuable data for new therapies. External natural history (NatHx) data sources are useful for interpreting study outcomes, however challenges may exist due to potential differences in patient populations. Propensity score matching (PSM) is a statistical method used to balance characteristics between cohorts, traditionally leveraging baseline covariates only. This study proposes a novel methodology to incorporate Year-1 progression rate matching (PRM) into a PSM comparison using 3 years data. Introduction of a progression-anchored covariate allows the NatHx comparator to more effectively reflect observed placebo group progression during the randomized clinical trial (RCT) period, offering improved face-validity of assessment at later timepoints. Methods : This method is illustrated with an example in which the 3-year effectiveness of troriluzole among patients with spinocerebellar ataxia (SCA) was evaluated as compared with a NatHx cohort. The PSM cohort was selected such that the distribution of Year 1 progression rates in the external comparator group were aligned with those observed in the placebo group during the 1 year RCT. Estimated treatment effects as measured by the modified functional Scale for the Assessment and Rating of Ataxia (f-SARA) outcome measure at Years 2 and 3 were calculated with both the standard PSM and “PSM-with-PRM” in the population adjustment. Results : There were 101 troriluzole-treated subjects from Study BHV4157-206 and 303 subjects from a global NatHx cohort included in the base-case PSM analysis. Least squares (LS) mean change differences (troriluzole – NatHx) in change from baseline f-SARA were − 0.61 at 1-year, -1.02 at 2-year, and − 1.33 at 3-years, favoring troriluzole (each p < 0.01). Implementation of PSM-with-PRM (n = 273 matched subject) resulted in treatment differences of -0.75 and − 0.96 at years 2 and 3 years, favoring troriluzole (each p < 0.01); Year 1 progression rates were comparable as designed (treatment difference of -0.02, p = 0.91). Conclusions : Addressing residual heterogeneity following indirect treatment comparisons with a PRM approach can enhance the face validity of analyses utilizing an external control group, as illustrated with a use case for PSM. This improves the interpretability of OLE findings in the context of real-world evidence, further supporting regulatory, payer and clinical decision-making.