The role of artificial intelligence in clinical trials: opportunities

Abstract

Artificial intelligence (AI) is now embedded across the clinical trial lifecycle, from target discovery and protocol design to recruitment, conduct, and post-marketing surveillance. Adoption has outpaced validation, and the randomised controlled trial (RCT) evidence base remains geographically uneven. For India, the question is no longer whether to adopt AI in trials, but how to do so without importing algorithms never validated on Indian patients. Narrative review of peer-reviewed literature (2018–2026) indexed in PubMed and major journal families, supplemented by official Indian regulatory documents (CDSCO, ICMR, MeitY). The review maps AI applications across the trial lifecycle, examines RCT evidence, and situates findings in India's regulatory context. A 2024 scoping review identified 86 AI RCTs (2018-2023); 81 per cent reported a positive primary endpoint, although 63 per cent were single-centre and only 19 per cent of post-2021 trials cited CONSORT-AI.¹ Machine learning predicts early trial termination with area under the curve 0.80.² The first generative-AI-discovered small molecule, rentosertib for idiopathic pulmonary fibrosis, completed phase 2a in 2025.³ Counter-examples persist: the Epic Sepsis Model showed sensitivity of 33 per cent on external validation.⁴ India's sepsis mortality (36.3 per cent overall; 50.8 per cent in septic shock) and projected antimicrobial-resistance deaths (11.8 million in South Asia by 2050) make trial-grade AI validation a national imperative. AI will reshape clinical trials, but only prospective, externally validated, reporting-standard-compliant studies, increasingly in Indian populations, will turn potential into evidence that regulators and clinicians can act on.