Research Ethics Committee and Community Engagement Readiness for Artificial Intelligence in African Clinical Trials: A Systematic Review (Preprint)

BACKGROUND Sub-Saharan Africa bears over 20% of the global burden of disease but hosts only 2% to 3% of global clinical trials. While artificial intelligence (AI) is increasingly integrated into trial design, recruitment, safety monitoring, and data synthesis, the operational readiness of African Research Ethics Committees (RECs) and the alignment of community engagement (CE) models with AI paradigms remain insufficiently synthesized. OBJECTIVE To systematically appraise and synthesize empirical evidence and formal policy guidance on: (1) African REC capacity to evaluate AI-integrated clinical trials; (2) documented socio-technical oversight challenges; and (3) AI-adapted community engagement models across Africa. METHODS Conducted per PRISMA 2020 guidelines and registered on PROSPERO (PROSPERO ID: CRD420261530035). Eight databases (PubMed/MEDLINE, Scopus, Web of Science, CINAHL, Cochrane Library, Global Health, AJOL, and WHO Afro) and six grey repositories were searched from January 1, 2000, to September 30, 2026. The SPIDER framework guided selection. Editorials, commentaries, opinion pieces, news features, non-African studies, and non-clinical AI papers were strictly excluded. Methodological quality was appraised using MMAT (2018) for primary empirical studies and AACODS for policy frameworks. Synthesis followed Braun and Clarke’s Reflexive Thematic Analysis, with confidence assessed via GRADE-CERQual. RESULTS Of 1,912 records identified, 18 met strict inclusion criteria (12 primary empirical African studies and 6 formal policy frameworks). Although binding national guidance has emerged (e.g., South African NDoH 2024 Ethics Guidelines), operational REC readiness remains low. Key findings demonstrate severe technical literacy deficits among REC members, absence of specialized AI review SOPs in most institutions, algorithmic opacity, data extractivism risks, and liability voids. Community engagement remains largely conceptual, with traditional static consent models failing to address dynamic algorithms. GRADE-CERQual confidence in overall evidence was moderate. CONCLUSIONS African RECs face significant operational bottlenecks in overseeing AI trials. Priorities include institutionalizing competency-based AI training, establishing regional computational advisory panels under AVAREF, enforcing data sovereignty, and co-designing relational, Ubuntu-based community engagement frameworks. CLINICALTRIAL This was a systematic review and registered on the PROSPERO site (PROSPERO ID: CRD420261530035)