The premise for stratified medicine is that drug efficacy, drug safety, or both, vary between groups of patients, and biomarkers can be used to facilitate more targeted prescribing, with the aim of improving the benefit:risk ratio of treatment. However, many factors can contribute to the variability in response to drug treatment. Inadequate characterisation of the nature and degree of variability can lead to the
identification of biomarkers that have limited utility in clinical settings. Here, we discuss the complexities associated with the investigation of variability in drug efficacy and drug safety, and how consideration of these issues a priori, together with standardisation of phenotypes, can increase both the efficiency of stratification procedures and identification of biomarkers with the potential for clinical impact.