Drug-disease interactions (DDSIs) occur when drug effects, such as risks for rare but severe adverse effects, are altered by a preexisting disease. DDSIs affect up to 50% of older adults and have been associated with increased mortality and use of health services. DDSIs are of particular concern in older adults because both polypharmacy and chronic illness become progressively more prevalent with advanced age. Although drug labels, treatment guidelines, and drug information compendia include warnings and contraindications for many thousands of drug-disease combinations, extremely little evidence and research exists on their clinical relevance. DDSI guidance generally relies on case reports, pharmacological mechanism, or structural similarity to related drugs, and thus commonly represents untested hypotheses rather than evidence from well-designed population-based studies. As a result, physicians and their patients are often unable to distinguish between warnings for clinically relevant DDSIs that should be followed to avoid increased risk for adverse drug effects, and warnings for purely theoretical DDSIs that should be ignored in order to allow initiation of treatment with the otherwise indicated drug of choice. Better evidence on DDSIs is thus urgently needed to allow physicians to recommend evidence-based personalized therapy for their patients. Using existing data resources on millions of patients from Medicare (US) and the Clinical Practice Research Datalink (UK), the proposed study will use four carefully selected examples of highly prevalent drugs to demonstrate a new methodological framework for the systematic assessment of DDSIs from large observational datasets: Metformin and renal impairment increasing risk of lactic acidosis (Aim 1), Z-drugs and osteoporosis increasing risk of hip fracture (Aim 2), systemic corticosteroids and peptic ulcer disease increasing risk of gastrointestinal bleeding (Aim 3), and allopurinol and renal impairment reducing risk of dialysis or kidney transplant (Aim 4). These examples were selected considering a number of explicit criteria including severity of the adverse outcome, disagreement about relevance in the literature and clinical practice, ability to measure disease and adverse clinical outcome in the databases, and availability of therapeutic alternatives that do not share the hypothesized DDSI. We included interactions across a spectrum of prevalence and expected effect sizes to evaluate the performance of the proposed approach in different situations and sought some effects very likely to be absent (Aim 1) or present (Aim 2) to show we can reproduce expected findings, and uncertain effects (Aims 3 and 4). The proposed study puts forward a novel framework that comprehensively classifies DDSIs according to their underlying biological mechanisms and represents the first systematic attempt to apply modern epidemiological and statistical methods to the examination of DDSIs. Its results will begin a line of work that will ultimately enable physicians to practice evidence-based personalized medicine by providing reliable data on the effects of patient-specific comorbidities on the safety and effectiveness of their therapeutic regimens.
Drug-disease interactions affect a large and growing number of older adults and have been associated with severe adverse outcomes including increased mortality. Warnings and contraindications are in place for many thousands of drug-disease combinations but, despite their substantial public health impact, virtually none of these warnings are based on evidence from well-designed studies. The proposed study will use four carefully selected examples of hypothesized drug-disease interactions of widely used drugs to begin a line of work that leverages extremely large data resources together with innovative methods from multiple fields to generate definitive quantitative data on the effects of patient-specific comorbidities on the safety and effectiveness of their therapeutic regimens.
