OBJECTIVES: Joint clinical assessment (JCA) in Europe is quickly approaching with the first medicinal products set for assessment in 2025. In preparation for this, manufacturers continue to examine what PICOs will be needed as they look to create a roadmap for reimbursement. However, given the short timeframe from PICO confirmation to submission date (100 days), this creates challenges even with the most comprehensive planning. Given that SLRs are a vital and time-consuming part of this process, we anticipate that artificial intelligence (AI) will play a key role in successful implementation of the guidance. We propose the use of AI classifiers to semi-automate the abstract screening process to help streamline this process.

METHODS: To test this hypothesis, we utilised a previously conducted SLR and trained an AI classifier based on human decisions as per the agreed screening criteria. We then conducted an SLR update, with a human (first reviewer) and AI classifier (second reviewer) to ensure all articles were dual-screened and compared the responses.

RESULTS: Overall, >90% of decisions were matched between human and AI screeners. The remaining unmatched decisions suggested an over-inclusive approach by the AI classifier, further ensuring all relevant articles would be included. These results highlight that the use of AI classifiers in the role of the second reviewer on an SLR would result in a substantial reduction in time screening and streamline the SLR process, and we propose a framework by which we can implement this as part of the SLR methodology.

CONCLUSIONS: AI classifiers present a unique opportunity to streamline the abstract screening stage of an SLR update, and can drastically reduce the timeframe needed which is critical given the 100-day window for JCA dossier submission. Appropriately leveraging AI in the literature review component will be key to ensuring an efficient JCA submission process.