Control of selection bias in parallel-group controlled clinical trials in dogs and cats: 97 trials (2000-2005).

OBJECTIVE

To determine how selection bias (allocation bias) was controlled in published clinical trials.

DESIGN

Retrospective study.

SAMPLE POPULATION

97 parallel-group controlled clinical trials published from January 2000 through December 2005 in 12 peer-reviewed journals.

PROCEDURES

Journals were hand searched to identify eligible reports. Details concerning allocation of animals to study groups, baseline characteristics of the groups, and the number of animals allocated to each group were recorded.

RESULTS

Randomization was the stated method of allocating animals to groups in 87% of the reports, yet in only 11% of reports were both randomization of the group allocation process and concealment of the allocation sequence described. Studies reported as randomized were more likely to report baseline characteristics of the study groups for comparison than studies that did not report randomization (88% vs 54%). Studies in which baseline group characteristics were reported had more subjects allocated to each study group (median, 16) than those that did not (median, 11).

CONCLUSIONS AND CLINICAL RELEVANCE

Randomization was reported as the method of allocating study animals to groups in most publications, indicating that the potential power of randomization in controlling selection bias is appreciated by clinical investigators seeking to determine the efficacy of an intervention. However, in most reports, little corroborating information was included to support the claim. The absence of this information makes it difficult for practitioners to critically review the impact of bias on study results and make informed decisions regarding patients.