The evolution of clinical practice and time trends in drug effects

Evidence from contemporary studies suggests that the quantification of adverse drug effects through state-of-the-art non-experimental methods may be confounded by some aspects of clinical practice. Support for this hypothesis is manifest in the finding that drug effects sometimes show within-study time trends when clinical practice with regard to drug use is evolving. Three studies which show simultaneous trends in prescription frequency and drug effects are discussed as examples. These include a study of benzodiazepines in the etiology of hip fracture, postmenopausal estrogens as a preventive for major coronary disease and human insulin as a risk factor for severe hypoglycemia. For instance, during a period when clinical perceptions regarding the safety of benzodiazepine use in the elderly underwent gradual change, the relative risk for hip fracture among long half-life benzodiazepine users declined progressively from 2.0 (95% CI, 1.6-2.5) in 1977-79 to 1.3 (95% CI, 0.9-1.8) in 1984-85. The implications of these observations are set against the scientific objectives which guide etiologic research.