[Many sources of bias in medical research: experience from systematic reviews].

A well-conducted systematic review requires a scrupulous assessment of the design of included studies. This may unveil major issues in how studies were planned, conducted and reported. This section presents a few examples. 1) A Cochrane review on pain and sedation management in the newborn identified a study described as a randomized trial, which later, following communication with authors and the editor-in-chief, turned out to be observational. 2) Poor evaluation of heterogeneity and active placebo when pooling studies on inhalation of saline solution for bronchiolitis led to clinical implementation of treatments later shown not to be effective. 3) A Cochrane review on methylphenidate for attention deficit hyperactivity disorder in adults did not identify problems with blinding and a »wash-out« period, resulting in erroneous conclusions. The review was therefore retracted. Although as important as benefits, harms of interventions are often given less attention in trials and systematic reviews.