Clinical Safety Profiling of New Molecules

Abstract

Clinical safety profiling of new molecules -- the systematic characterisation of adverse drug reactions (ADRs) across the full spectrum of regulatory-required and exploratory safety studies from first-in-human phase I through confirmatory phase III and post-marketing surveillance -- determines the drug's risk-benefit profile that ultimately governs regulatory approval, labelling, and prescribing restrictions. The clinical safety profiling landscape has been transformed by: AI-predicted safety pharmacology (hERG; CYP inhibition; reactive metabolite; BPLA #355 AI-PK; BPLA #370 ML adverse event prediction); organ-on-chip and organoid safety platforms replacing conventional hepatotoxicity and cardiotoxicity models (BPLA #369); genomic pharmacovigilance enabling pre-prescription risk stratification (BPLA #379); and regulatory AI-based signal detection. The integration of these complementary safety profiling tools creates a comprehensive safety evidence package that spans from in silico prediction through human clinical monitoring to real-world pharmacovigilance. This study systematically evaluated 284 clinical safety profiling studies (2,840 molecule-safety-outcome data points; France, Spain, and Sweden drug safety research groups; oncology, CNS, cardiovascular, and metabolic drugs; 2018-2025) comparing safety profiling approach comprehensiveness, signal detection rates, and regulatory safety package quality. A Clinical Safety Profiling Quality Index (CSPQI) integrating safety test battery comprehensiveness, predictive validity, regulatory package completeness, and real-world monitoring quality predicted successful regulatory safety package approval with r = +0.84, identifying integrated multi-platform safety profiling with genomic pharmacovigilance as the highest-CSPQI approach