Contending paradigms for the interpretation of data on patient compliance with therapeutic drug regimens.

Electronic and chemical marker methods provide the first reliable measurements of drug exposure in ambulatory trials. These data contradict the usual claim in published drug trials of > 90% of patients having been satisfactorily compliant with the protocol-specified dosing regimen. Such exaggerated claims are based, usually, on count of returned dosing forms, which afford patients easy ability to manipulate by discarding or hoarding untaken doses. Electronic monitoring provides, for the first time, data on intervals between doses, revealing the 'drug holiday'--3 or more consecutive days without dosing--as a basis not only for lapsed therapeutic action, but as a pharmacodynamic trigger for hazardous rebound effects on recurrent first dose effects. Another new findings is the evident non-specificity of poor or partial compliance, the range and distributions of which appear to be hardly affected by drug, disease, prognosis, or symptoms. This finding contradicts often repeated but unsupported claims that noncompliance is a specific response to drug action, disease, prognosis or other treatment-related factors. New statistical methods are needed for trial design and analysis, to use drug exposure data as covariate information, to incorporate into drug labelling estimates of dose-related efficacy, holiday-related hazard, the limits of safe variation in dose-timing, and what one should best do when those limits are exceeded. Oral contraceptive labelling in the U.K. and U.S. is exemplar for this next step toward full-disclosure labelling.