Observational studies dominate liver transplantation (LT) research because randomized controlled trials are often infeasible. However, non-random treatment allocation introduces substantial confounding, limiting causal inference.

We systematically reviewed propensity score matching (PSM) methods in LT research. We searched six databases from inception through March 31, 2026, with an emphasis on studies published from 2011 onward. We predefined eligibility criteria, screening procedures, and data-extraction domains. Reporting followed PRISMA 2020. Because we evaluated methodological practices rather than treatment effects, we did not conduct a meta-analysis or a formal risk-of-bias assessment.

We included 475 PSM-based studies. Propensity scores were estimated using logistic regression in 334 studies (70.3%). Nearest-neighbor matching was the dominant approach, reported in 371 studies (78.1%), and 318 studies (66.9%) specified a caliper. Three hundred and four studies (64.0%) reported standardized mean differences before and after matching, and 311 studies (65.5%) described missing-data handling. In contrast, only 66 studies (13.9%) reported overlap or common-support assessment. Sensitivity analyses were performed in 116 studies (24.4%), and 210 studies (44.2%) used statistical methods that explicitly accounted for the matched or weighted design. Inverse probability of treatment weighting (IPTW) was reported in 34 studies (7.1%). No study used machine learning or other flexible methods for propensity score estimation. Limitations include reliance on published reports, heterogeneous reporting practices, and the absence of a formal study-level risk-of-bias assessment.

Although PSM reporting has improved substantially, diagnostic and inferential practices remain inconsistent. Key components of contemporary causal inference, including overlap assessment, sensitivity analyses, and appropriate post-matching inference, remain underused. Future LT research should move beyond covariate balancing alone and adopt more comprehensive causal-inference frameworks to strengthen the validity and interpretability of observational evidence (INPLASY registration number #202680025).