Publication bias in reproductive medicine: from the European Society of Human Reproduction and Embryology annual meeting to publication.

BACKGROUND

Treatment decisions should ideally be based on well-designed randomized controlled trials (RCTs). Here we determine the rate of full publication of RCTs presented at annual meetings of the European Society of Human Reproduction and Embryology (ESHRE), identify potential bias against publishing non-significant results and results not favoring the experimental arm, quantify this bias in case it exists, and identify factors associated with time to publication.

METHODS

RCTs presented at ESHRE meetings 2003 and 2004 were recorded. Subsequent search in Medline, Cochrane Library and EMBASE was performed through December 2010 to identify full-text publication in a peer-review journal.

RESULTS

Among 155 abstracts describing RCTs 89 (57%) were published in full-text in a peer-review journal. Median time from presentation to publication was 15 months (range: 0-75). In bivariate analysis, only type of presentation and presence of outcomes favoring the experimental arm were related to publication rate. Studies presented orally or reporting a positive outcome in favor of the experimental arm were more likely to be published (P = 0.018 and 0.014, respectively). Results were consistent in a multivariable logistic regression, with odds ratio (OR) 2.51 [95% confidence interval (CI), 1.25-5.03] for oral versus poster presentations and OR 2.46 (95% CI, 1.23-4.95) for trials favoring versus not favoring the experimental arm. Kaplan-Meier curves revealed time to publication was shorter for oral presentations (log-rank test = 0.013) and trials favoring the experimental arm, compared with all others (log rank = 0.007).

CONCLUSIONS

RCTs with significant results in favor of the experimental arm are more likely to be published and are published sooner. Publication bias in reproductive medicine is a fact.