A Systematic Review of Reporting of Validation Efforts for Cancer Health Economic Decision Models.

OBJECTIVE: This study reviewed recently published cancer-related studies to identify the reporting profiles, influencing factors, and good practices of model validation efforts in health economic evaluations (HEEs). METHODS: A systematic review was conducted in PubMed, Ovid MEDLINE, Ovid Embase, and ScienceDirect to retrieve recent cancer-related HEEs published between 2021 and 2026. Reporting on model validation was extracted and evaluated using the International Society for Pharmacoeconomics and Outcomes Research and the Society for Medical Decision Making (ISPOR-SMDM) and Assessment of the Validation Status of Health-Economic decision models (AdViSHE) frameworks. Temporal trends, influencing factors of systematic reporting, and validation domain choices were analyzed using Fisher's exact test. We compared the observed counts of the five validation domains under two approaches: the broad-reporting ISPOR-SMDM claims versus the specific AdViSHE-verified details. Additionally, we performed a targeted analysis to identify examples of "good practice" validation in cancer HEE models. RESULTS: Of the 1178 identified studies, 356 (30.2%) reported model validation. Among the 356 studies, external and face validation were the most commonly reported domains (186/356 [52.2%] and 175/356 [49.2%], respectively). Only reporting proportions of internal validation and model fit showed significant increases over the past 5 years. In contrast, reporting of validation-related model adjustments significantly declined (5.0% to 1.5%), and remained the least reported transparency item (11/356, 3.1%). For the remaining domains and aspects, absolute counts rose with publication volume, but proportional reporting fluctuated, with no significant differences across years in the global tests. 43.3% (154/356) of studies included a separate validation section or reported using validation tools/guidelines. Pharmaceutical interventions and partitioned survival models were associated with lower proportions of systematic reporting. Independent and cross-validation were more frequently reported in digital health interventions (41.7% each) than in pharmaceutical ones (23.7% and 11.1%). Internal validation showed the largest under-reporting of specific details (133 studies under ISPOR-SMDM versus 56 under AdViSHE). Among selected good practice studies, the survival endpoint emerged as the most commonly used external validation measure (60.0%, 6/10). CONCLUSIONS: The reporting proportion of validation efforts in model-based HEEs has not improved significantly over the past 5 years, although the findings reflect reporting practices rather than actual execution.