OBJECTIVES: SLRs are the cornerstone of HTA decision-making but are labor intensive. AI/ML could expedite SLR conduct, however, HTA acceptability of AI/ML is unclear. Our 2022 review was updated to examine recent changes in HTA guidance on the use of AI/ML in SLRs.

METHODS: HTA guidance documents from EUnetHTA, JCA (European Union), NICE (England), HAS (France), IQWiG (Germany), NCPE (Ireland), SMC (Scotland), TLV (Sweden), CADTH (Canada) and PBAC (Australia) were reviewed for guidance on AI/ML in HTA-compliant SLRs. Cochrane guidance was sought as it is often referenced by HTA bodies.

RESULTS: Only IQWiG (2023) explicitly refers to AI/ML for HTA SLRs, stating that validated randomized clinical trial (RCT) classifiers can be used for screening. AI/ML prioritization of relevant records in screening should be tested on an individual-case basis. Although the NICE HTA SLR guidance (2023) provides no information on AI/ML, the NICE Guideline development manual (2024) supports the use of ML for prioritizing references for screening and for automated exclusion of references, provided classifiers' performance characteristics are known. Caution is advised if classifiers are used on data of a different type to the development dataset. If used to prioritize relevant records and define a stopping criterion after which studies are automatically excluded, the methods and stopping threshold should be documented. No other HTA bodies provide AI/ML guidance. The Cochrane Handbook (2023) advises using its RCT Classifier to identify RCTs from titles/abstracts. Automated study prioritization is allowed. Automated exclusion of records based on a stopping criterion, automated data extraction, and use of Large Language Models for screening, are not recommended. Cochrane SLR authors are allowed to use generative AI in reporting.

CONCLUSIONS: Only IQWiG provided updated AI/ML guidance. Future NICE HTA SLR guidance could follow the direction of the NICE Guidelines development manual, recommending some AI/ML, however, guidance from other HTA bodies is still unclear.