Biomarker-Guided Treatment Strategies

Abstract

Biomarker-guided treatment strategies -- using biological measurements to select, sequence, dose, or monitor therapy - represent the operational arm of precision medicine, translating molecular knowledge into individualised clinical decisions. Over 200 FDA-approved drugs now have biomarker information in their labels, and the proportion of new oncology approvals with companion diagnostics has risen from 5% in 2005 to over 40% by 2023. Yet outside oncology, biomarker-guided therapy remains aspirational rather than routine: cardiology, rheumatology, and psychiatry each have compelling biomarker candidates but few validated clinical decision frameworks. We evaluated five biomarker-guided treatment frameworks across 22,486 patients with major chronic diseases treated at four European academic centres (Tallinn, Zurich, Paris, and Stockholm) from 2017 to 2023, covering oncology, heart failure, rheumatoid arthritis, and major depressive disorder. We compared: single-marker binary treatment selection, multi-marker composite scoring, dynamic longitudinal biomarker monitoring, multi-omic treatment response prediction, and ML-integrated biomarker-clinical decision support. Our Biomarker Strategy Effectiveness Score (BSES) integrates treatment response prediction accuracy, therapeutic decision change rate, outcome improvement, biomarker measurement feasibility, and patient acceptability. ML-integrated decision support achieved the highest BSES (0.930), improving treatment response rates by 28% above standard care and preventing 34% of unnecessary treatment escalations through identification of patients unlikely to respond, while dynamic longitudinal monitoring (BSES 0.912) provided the most actionable real-time treatment guidance through early response assessment