Search

Among the observational studies of drug effects in chronic diseases, many of them have found effects that were exaggerated or wrong. Among bias responsible for these errors, the immortal time bias, concerning the definition of exposure and exposure periods, is relevantly important as it usually tends to wrongly attribute a significant benefit to the study drug (or exaggerate a real benefit). In this article, we define the mechanism of immortal time bias, we present possible solutions and illustrate its consequences through examples of pharmacoepidemiological studies of drug effects., (© 2014 Société Française de Pharmacologie et de Thérapeutique.)

Pharmacological treatment of chronic obstructive pulmonary disease (COPD) relies principally on long-acting bronchodilators. Inhaled corticosteroids (ICS) were introduced for COPD two decades ago, despite the fact that no randomized trial had yet assessed their efficacy for this indication. Since then, the numerous randomized trials and meta-analyses performed to justify their use in COPD have been contradictory and controversial. Moreover, observational studies have reported efficacy rates so exceptional that they are almost too good to be true. These studies contain important methodological flaws that produce the appearance of efficacy. The randomized trials infringe the fundamental principle of intention-to-treat analysis, an analysis necessary to prevent important biases. Two other complications are the interruption of treatment at the moment of randomization and the use of a run-in period; in both cases, the withdrawal of treatment can introduce bias. The observational studies reporting phenomenal reductions in mortality with ICS were distorted by "immortal time" bias. Finally, recent data suggest that the effect of ICS/bronchodilator combinations is due mainly to the effect of the long-acting bronchodilator. Given the absence of proof of the efficacy of inhaled corticosteroids in COPD and their associated risks, especially of ocular damage and pneumonia, and particularly among the elderly, as well as the high doses currently prescribed in COPD, it is difficult to recommend their use in this indication. They should be prescribed in COPD for at most a limited population of patients.