Measurement of patient compliance and the interpretation of randomized clinical trials

Abstract

The aim of this review is to demonstrate why the management of compliance, although not an explicit feature of the rules of Good Clinical Practice, is essential to the successful conduct of clinical trials and in correct interpretation of the results. Methods to measure compliance in randomized clinical trials are also described. The relevant literature was retrieved by a manual search of the Cumulated Index Medicus 1975–1989 and a MEDLINE computer search of publications in 1990 using the Medical Subject Headings “patient compliance” and “clinical trials”. All retrieved articles are discussed. Research into patient compliance has stagnated because of the lack of a “gold” standard of measurement. Nevertheless, management of compliance at the different stages of clinical trials is necessary; at trial design, compliance should be taken into account in sample size calculations; during the conduct of a trial, compliance should be monitored in order to safeguard the power of the study; and in interpretation of trial results, compliance data are helpful both in order to avoid erroneous conclusions and to enrich the value of the data. Compliance should be measured in all limbs of randomized trials, including the placebo limb, without breaking trial blinding. A classification of compliance behaviour into six types (complier, partial complier, overuser, erratic user, partial dropout and dropout) is proposed, based on the changes in the risk-benefit ratio produced by non-compliance. The use of deuterium oxide, low dose phenobarbitone (2 mg per day), digoxin (2,2 μg per day) and the new electronic monitoring devices are suitable methods for measurement in clinical trials.