Risk of bias in randomized trials of pharmacological interventions in children and adults

Objective: To determine whether randomized controlled trials of pharmacologic interventions in children are more likely to be biased than similar trials in adults.

Study design: Trials involving only children and published in MEDLINE between January 2008 and October 2009 (n=100) were randomly selected and matched, by drug class and therapeutic area, with a similar trial completed in adults. The Cochrane risk of bias tool was used to compare the pediatric and adult trials.

Results: The characteristics of adult and pediatric trials included were similar, except that adult studies were more likely to be conducted in Europe and published in specialty journals. Two-thirds of all trials were single center, and 62% had 100 or fewer participants. Many trials had an unclear risk of bias for allocation concealment (65% adult, 52% pediatric). More pediatric trials had a low risk of bias for random sequence generation (59% pediatric, 41% adult, P=.002) and blinding of outcome assessment (63% pediatric, 48% adult, P=.04) than adult trials; however, a sensitivity analysis of trials published since 2008 (and so matched by year of publication) did not confirm this finding, suggesting year of publication was an important confounder.

Conclusions: When randomized controlled trials are matched for drug class and therapeutic area, trials involving children display a similar risk of bias. Differences in the risk of bias between pediatric and adult trials are not caused by differences in the capacity of researchers to conduct and report trials of high quality.