Mapping information-sharing methods for sparse evidence meta-analysis: a scoping review and implications for health technology assessment

Health technology assessment (HTA) and clinical decision-making depend heavily on robust evidence synthesis. However, traditional meta-analyses often struggle with data sparsity, particularly for rare indications or specific subgroups. While advanced statistical methods using information sharing can address this, their mathematical complexity limits practical application. This scoping review maps these techniques to practical evidence scarcity scenarios. We searched PubMed and Embase from inception to November 6, 2025, supplemented by backward and forward citation searching, for methodological studies focusing on information-sharing methods in evidence synthesis of therapeutic interventions. Eligibility was determined using a prespecified operational definition requiring a target evidence base, a related evidence source, a specified statistical sharing mechanism, and a conceptual no-sharing comparator. Included studies were mapped to specific synthesis challenges and categorized into four domains: Population and Indication, Intervention and Structure, Outcomes, and Evidence Integration. Challenges, domains, and information-sharing mechanisms were coded as non-mutually exclusive labels. A multi-stage Sankey diagram was constructed to visualize connections between clinical challenges, mathematical relationships, and statistical solutions. The review included 140 reports, comprising 131 original methodological articles and 9 methodology reviews. Among the 131 articles, Bayesian frameworks were used in 89 (67.9%) and frequentist frameworks in 38 (29.0%), frequently employing informative or power priors to incorporate external evidence. Among all 140 reports, oncology was the most frequent clinical application area (28, 20.0%). Methods were classified into four core analytical relationships, which were not mutually exclusive: functional links (46, 35.1%), exchangeability assumptions (38, 29.0%), prior distributions (21, 16.0%), and multivariate correlations (65, 49.6%). Forty-nine articles (37.4%) included a simulation study. The mapping demonstrates how distinct challenges, such as disconnected networks or a lack of long-term endpoints, have been addressed in the methodological literature by specific information-sharing models. By organizing the methodological literature into a structured framework, this review provides an evidence map of information-sharing methods for sparse evidence settings. The mapping may help HTA practitioners identify candidate analytical strategies for addressing sparse, indirect, or immature evidence. Not applicable.