In two of the situations that we presented, this apparent effectiveness was not confirmed in subsequent large-scale randomized controlled IPI 145 trials conducted to evaluate these findings. Indeed, the numerous observational studies of hormone replacement therapy

(HRT), indicated for menopausal symptoms, and suggesting cardiovascular benefits, were clearly flawed; the WHI randomized trials did not confirm such benefit. Similarly, the observational studies of inhaled corticosteroid treatment, indicated for asthma Inhibitors,research,lifescience,medical but used in COPD without evidence, suggested spectacular benefits of these drugs on reducing all-cause mortality, benefits which were subsequently not Inhibitors,research,lifescience,medical corroborated by the large TORCH randomized trial. Currently, history may be repeating itself with the anti-diabetic medication metformin which has been the subject of several observational studies that reported impressive reductions in the incidence of and mortality from cancer. These spectacular “beneficial” Inhibitors,research,lifescience,medical anti-cancer effects are clearly again the result of time-related biases which tend to exaggerate the benefits observed with a drug. Yet, these observational

studies form the basis for the conduct of large-scale randomized trials currently Inhibitors,research,lifescience,medical underway. Interestingly, with such promising findings from observational studies, many animal studies are conducted to understand and describe

possible mechanisms by which, for instance, metformin could prevent or Inhibitors,research,lifescience,medical slow cancer progression, or physiological explanations of the possible effects of inhaled corticosteroids on systemic inflammation in COPD and the potential benefit on mortality. Such research brings greater momentum to the new indication, also eventually leading to large trials. However, it is imperative first to carry out critical assessments of the observational study methods, for which possible methodological explanations for these “spectacular” results have received little attention (see Box 1). While these biases are well-known in pharmacoepidemiology and have been described extensively in different therapeutic areas,31,34,35,63,64 they do not seem to have yet sufficiently penetrated different subspecialty fields such as diabetes, cancer, pulmonary medicine, etc. Box 1. How to Detect Immortal Time Bias During Peer Review If a cohort study reports extreme beneficial effects (relative risks < 0.