Advances in our knowledge of the molecular pathogenesis of cancer have led to increased interest in molecularly targeted agents (MTAs), which target specific oncogenic drivers and are now a major focus of cancer drug development. MTAs differ from traditional cytotoxic agents in various aspects, including their toxicity profiles and the potential availability of predictive biomarkers of response. The landscape of phase I oncology trials is evolving to adapt to these novel therapies and to improve the efficiency of drug development. In this Review, we discuss new strategies used in phase I trial design, such as novel dose-escalation schemes to circumvent limitations of the classic 3 + 3 design and enable faster dose escalation and/or more-precise dose determinations using statistical modelling; improved selection of patients based on genetic or molecular biomarkers; pharmacokinetic and pharmacodynamic analyses; and the early evaluation of efficacy — in addition to safety. Indeed, new expedited approval pathways that can accelerate drug development require demonstration of efficacy in early phase trials. The application of molecular tumour profiling for matched therapy and the testing of drug combinations based on a strong biological rationale are also increasingly seen in phase I studies. Finally, the shift towards multi-institutional trials and centralized study management results in consequent implications for institutions and investigators. These issues are also highlighted herein.