Here we summarise the CONSORT-AI and SPIRIT-AI extensions – guidelines for clinical trial protocols for interventions involving artificial intelligence (AI).
The original CONSORT (Consolidated Standards of Reporting Trials) and SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) statements are guidelines for protocols of randomised controlled trials. They were developed according to the EQUATOR (Enhancing Quality and Transparency of Health Research) Network methodological framework for guideline development.
These statements have now been extended to incorporate interventions involving AI. This process began with a literature search and expert consultation. This generated a list of candidate items – 29 for CONSORT-AI and 26 for SPIRIT-AI. These items were then presented to stakeholders in two online Delphi surveys. In these surveys, stakeholders were asked to rate the importance of each item. The final consensus meeting was held at the start of the year to finalise both guidelines.
CONSORT-AI
CONSORT-AI provides evidence-based recommendations to improve the completeness of the reporting of randomised control trials. The CONSORT statement was first introduced in 1996 and has since undergone several updates. CONSORT-AI is an international initiative to support the existing CONSORT 2010 statement and to also extend this guidance to support the reporting of clinical trials for AI interventions.
The CONSORT-AI extension specifically recommends that 14 new checklist items should be added to the existing CONSORT 2010 statement (11 extensions and 3 elaborations). The report was published in BMJ.
SPIRIT-AI
The SPIRIT statement published in 2013 provides minimum guidance applicable for all clinical trial interventions. SPIRIT-AI is an international initiative to extend upon the existing SPIRIT 2013 statement in order to develop consensus-based AI-specific protocol guidance.
The SPIRIT-AI extension specifically includes 15 new items that should be addressed for clinical trial protocols of AI interventions and added to the current SPIRIT 2013 guidance (12 extensions and 3 elaborations). The report was published in BMJ.
Conclusion
In summary, these guidelines recommend that investigators should provide clear descriptions of the AI intervention, the setting in which the AI intervention is integrated, the handling of inputs and outputs, the human-AI interaction and the provision of an analysis of error cases.
These guidelines are widely endorsed and also have been instrumental in ensuring transparency for evaluating new interventions. Their extensions will continue to help promote transparency and completeness in reporting clinical trials for AI interventions. Additionally, they will assist editors and peer reviewers to understand, interpret and critically review the quality of clinical trial design.
Image credit: By freepik – www.freepik.com
