The central problem PRIMA addresses is confounding. Two diseases can appear linked because one drives the other, or because shared genetic and environmental risks raise the likelihood of both. Traditional propensity score matching, which pairs exposed and unexposed individuals on measured clinical characteristics, cannot fully account for inherited liability that predates diagnosis. By adding polygenic risk scores computed for the specific exposure under study, PRIMA aims to create matched groups that are balanced not only clinically but also genetically.

The researchers evaluated the framework in two ways. First, they used liability-threshold simulations, a modelling approach that reflects how genetic and environmental influences combine to produce binary disease outcomes. In these simulations, PRIMA reduced bias arising from shared genetic liability and recovered effect estimates closer to the expected causal effects than clinical matching alone. The authors are careful to note that the method remained sensitive to unmeasured confounding, meaning that hidden factors not captured by the model could still distort results.

Second, the team applied PRIMA to disease-pair analyses in 333,370 unrelated UK Biobank participants of European ancestry. Integrating polygenic risk improved both genetic and clinical balance between matched groups, and it materially altered association estimates for a subset of disease pairs. The researchers compared PRIMA’s estimates against three established approaches: clinical matching, Mendelian randomization and survival analyses. PRIMA identified additional associations that Mendelian randomization did not detect and showed greater concordance with survival analyses.

The authors position PRIMA as a scalable complement to existing causal inference tools rather than a replacement for them. Their conclusions carry explicit caveats: causal interpretation remains conditional on appropriate model specification, sufficient overlap between matched groups and adequate control of confounding. Because the UK Biobank analyses were restricted to unrelated participants of European ancestry, the framework’s performance in other populations is not established by this study. As a methods paper, the work also reports simulations and observational comparisons rather than clinical trial evidence, so its findings concern statistical methodology and association estimates, not treatment outcomes. Still, by making genetic liability an explicit part of the matching process, PRIMA offers observational researchers a new lever for probing whether one disease plausibly leads to another.

Subject of Research: A polygenic risk-integrated propensity score matching framework for observational disease-pair analyses

Article Title: Resolving disease causality by integrating polygenic risk into propensity score matching

Article References: Kaufman, H., Rappoport, N., & Linial, M. (2026). Resolving disease causality by integrating polygenic risk into propensity score matching. Genome Medicine. https://doi.org/10.1186/s13073-026-01804-w

Image Credits: AI Generated

DOI: 10.1186/s13073-026-01804-w

Keywords: polygenic risk scores, propensity score matching, causal inference, UK Biobank, disease pairs, Mendelian randomization, confounding, observational studies, genetic liability, PRIMA, Resolving, disease