Publication bias in abstracts presented to the annual meeting of the American Academy of Orthopaedic Surgeons.

PURPOSE To examine possible causes of publication bias in the orthopaedic literature so as to avoid inappropriate clinical decisions based on reviews of the literature. METHODS Two reviewers independently reviewed abstracts presented to the 1999 American Academy of Orthopaedic Surgeons annual meeting. Data pertaining to sample size, statistical significance, study setting, country of origin, outcome, study type, and sponsorship were extracted from each abstract. The publication rate was measured after 5 years, by electronic searching and author contact. Predictors of publication were identified using logistic regression analysis. RESULTS Of the 318 abstracts listed in the proceedings, 175 (55%) were published within 5 years. Publication was associated with positive rather than neutral outcomes (odds ratio, 1.62; 95% confidence interval, 1.01-2.59; p<0.05) and with the reporting of statistical significance (odds ratio, 2.05; 95% confidence interval 1.24-3.39; p=0.005). Sponsorship, country of origin, sample size, study setting, and study type did not significantly influence the publication rate. CONCLUSION Evidence of publication bias exists in abstracts presented at the 1999 American Academy of Orthopaedic Surgeons annual meeting. Clinical decisions based on the literature may be biased due to an over-representation of studies with positive outcomes.