Full publication of studies presented at biomedical meetings—updated systematic review of follow-up studies

Article type
Authors
Scherer R1, Meerpohl JJ2, Schmucker C2, Schwarzer G2, von Elm E3
1Bloomberg School of Public Health, Johns Hopkins University, Baltimore, USA
2German Cochrane Centre, Institute of Medical Biometry and Medical Informatics, University Medical Center, Freiburg, Germany
3Cochrane Switzerland, Institute of Social and Preventive Medicine, Lausanne University Hospital, Lausanne, Switzerland
Abstract
Background: Studies presented at scientific meetings are hard to identify in electronic literature searches. As part of the ‘grey literature’ they are often not used in systematic reviews. If full publication is dependent on factors such as magnitude or direction of results, reporting bias may occur. This potentially invalidates the results of meta-analyses aiming at an unbiased evidence base.

Objectives: To determine the full publication rate of studies presented at biomedical meetings and the time elapsed until full publication. To identify factors associated with full publication.

Methods: We searched major electronic databases through 11/2012 without language restriction for follow-up studies using cohorts of abstracts of biomedical meetings. We extracted data if publication rates were reported for a minimum follow-up of 24 months. Using random effects models, we calculated weighted estimates of publication rates and pooled odds ratios (OR; Mantel-Haenszel method) for factors associated with publication.

Results: A total of 264 studies following 147.529 abstracts from 667 individual meetings were included. The median publication rate after presentation was 42.3% (range 2.6–90.7). The weighted publication rate was 39.5% (95%CI: 37.4–41.6); individual estimates were heterogeneous (I2 = 98.3%, p < 0.0001). Publication activity starts to diminish after 60 months (Figure). In 20 studies including only RCT abstracts the weighted publication rate was 64.2% (95%CI: 56.7–71.0). Study characteristics associated with subsequent full publication were: significant vs. non-significant results (OR 1.9; 95%CI: 1.6–2.2), clinically relevant vs. not relevant results (1.6; 1.3–2.0); basic vs. human research (1.5; 1.3–1.7); RCT vs. observational study (1.4; 1.2–1.7). Abstract characteristics associated with full publication were: accepted vs. rejected for presentation (2.6; 2.2–3.2) and oral vs. poster presentation (1.7; 1.5–1.8).

Conclusions: A sizable proportion of research presented at biomedical meetings remains unpublished. Factors associated with full publication suggest that reporting bias is present even after study results have been presented to peers.