The normal metabolism of drugs can generate metabolites that have intrinsic chemical reactivity towards cellular molecules, and therefore have the potential to alter biological function and initiate serious adverse drug reactions. Here, we present an assessment of the current approaches used for the evaluation of chemically reactive metabolites. We also describe how these approaches are being used within the pharmaceutical industry to assess and minimize the potential of drug candidates to cause toxicity. At early stages of drug discovery, iteration between medicinal chemistry and drug metabolism can eliminate perceived reactive metabolite-mediated chemical liabilities without compromising pharmacological activity or the need for extensive safety evaluation beyond standard practices. In the future, reactive metabolite evaluation may also be useful during clinical development for improving clinical risk assessment and risk management. Currently, there remains a huge gap in our understanding of the basic mechanisms that underlie chemical stress-mediated adverse reactions in humans. This review summarizes our views on this complex topic, and includes insights into practices considered by the pharmaceutical industry.

Acyl glucuronidation is the major metabolic conjugation reaction of most carboxylic acid drugs in mammals. The physiological consequences of this biotransformation have been investigated incompletely but include effects on drug metabolism, protein binding, distribution and clearance that impact upon pharmacological and toxicological outcomes. In marked contrast, the exceptional but widely disparate chemical reactivity of acyl glucuronides has attracted far greater attention. Specifically, the complex transacylation and glycation reactions with proteins have provoked much inconclusive debate over the safety of drugs metabolised to acyl glucuronides. It has been hypothesised that these covalent modifications could initiate idiosyncratic adverse drug reactions. However, despite a large body of in vitro data on the reactions of acyl glucuronides with protein, evidence for adduct formation from acyl glucuronides in vivo is limited and potentially ambiguous. The causal connection of protein adduction to adverse drug reactions remains uncertain. This review has assessed the intrinsic reactivity, metabolic stability and pharmacokinetic properties of acyl glucuronides in the context of physiological, pharmacological and toxicological perspectives. Although numerous experiments have characterised the reactions of acyl glucuronides with proteins, these might be attenuated substantially in vivo by rapid clearance of the conjugates. Consequently, to delineate a relationship between acyl glucuronide formation and toxicological phenomena, detailed pharmacokinetic analysis of systemic exposure to the acyl glucuronide should be undertaken adjacent to determining protein adduct concentrations in vivo. Further investigation is required to ascertain whether acyl glucuronide clearance is sufficient to prevent covalent modification of endogenous proteins and consequentially a potential immunological response.

Clinical decision making needs to be supported by evidence that treatments are beneficial to individual patients. Although randomized control trials (RCTs) are the gold standard for testing and introducing new drugs, due to the focus on specific questions with respect to establishing efficacy and safety vs. standard treatment, they do not provide a full characterization of the heterogeneity in the final intended treatment population. Conversely, real‐world observational data, such as electronic health records (EHRs), contain large amounts of clinical information about heterogeneous patients and their response to treatments. In this paper, we introduce the main opportunities and challenges in using observational data for training machine learning methods to estimate individualized treatment effects and make treatment recommendations. We describe the modeling choices of the state‐of‐the‐art machine learning methods for causal inference, developed for estimating treatment effects both in the cross‐section and longitudinal settings. Additionally, we highlight future research directions that could lead to achieving the full potential of leveraging EHRs and machine learning for making individualized treatment recommendations. We also discuss how experimental data from RCTs and Pharmacometric and Quantitative Systems Pharmacology approaches can be used to not only improve machine learning methods, but also provide ways for validating them. These future research directions will require us to collaborate across the scientific disciplines to incorporate models based on RCTs and known disease processes, physiology, and pharmacology into these machine learning models based on EHRs to fully optimize the opportunity these data present.

Induction of cytochrome P450 (P450) can impact the efficacy and safety of drug molecules upon multiple dosing with coadministered drugs. This strategy is focused on CYP3A since the majority of clinically relevant cases of P450 induction are related to these enzymes. However, the in vitro evaluation of induction is applicable to other P450 enzymes; however, the in vivo relevance cannot be assessed because the scarcity of relevant clinical data. In the preclinical phase, compounds are screened using pregnane X receptor reporter gene assay, and if necessary structure-activity relationships (SAR) are developed. When projects progress toward the clinical phase, induction studies in a hepatocyte-derived model using HepaRG cells will generate enough robust data to assess the compound's induction liability in vivo. The sensitive CYP3A biomarker 4β-hydroxycholesterol is built into the early clinical phase I studies for all candidates since rare cases of in vivo induction have been found without any induction alerts from the currently used in vitro methods. Using this model, the AstraZeneca induction strategy integrates in vitro assays and in vivo studies to make a comprehensive assessment of the induction potential of new chemical entities. Convincing data that support the validity of both the in vitro models and the use of the biomarker can be found in the scientific literature. However, regulatory authorities recommend the use of primary human hepatocytes and do not advise the use of sensitive biomarkers. Therefore, primary human hepatocytes and midazolam studies will be conducted during the clinical program as required for regulatory submission.