A systematic review and time-specific meta-analysis of hepatocellular carcinoma risk prediction models

Background: To date, various scoring models for medium- and long-term hepatocellular carcinoma (HCC) risk prediction have been developed, validated in different populations, and subjected to rigorous evaluation to systematically evaluate the performance and methodological quality of liver cancer prediction models and to provide a reference for clinicians when selecting appropriate risk assessment tools. Methods: PubMed, the Cochrane Library, and Web of Science were used to search for articles published up to December 31, 2025. Two researchers independently screened the literature and extracted any relevant data. The Prediction Model Risk of Bias Assessment Tool (PROBAST) was used to evaluate the risk of bias and the applicability of the included studies. The area under the receiver operating characteristic curve (AUC) for the validation group was analyzed using a meta-analysis based on a random-effects model. Additionally, publication bias assessments and sensitivity analyses were also conducted. Results: A total of 58 studies were included in the analysis. Among these, 10 had high applicability, whereas the remaining studies had low applicability. The PROBAST assessment indicated that most of the included studies had a high risk of bias, primarily owing to their retrospective design, inadequate sample size, and improper management of missing data. Additionally, nearly one-third of the studies lacked internal validation. After a meta-analysis of the AUC and after excluding studies with a heterogeneity exceeding 50%, LSM-HCC had the highest 3-year AUC, aMAP had the highest 5-year AUC, and CU-HCC had the highest 10-year AUC. Conclusions: The development and validation of current HCC risk prediction models are currently inadequate. Future research should closely adhere to established reporting standards, and the procedures used by prediction models to create high-quality risk prediction tools should be applicable to clinical practice. There are large variations among existing HCC risk prediction models, and their predictive performances are time dependent. Typically, short-term predictions are superior to long-term forecasts, which strongly suggests that model selection should align with clinical objectives.