Ospina 2006.

Methods	Identification of subsequent full‐length publications Searched electronic database CINAHL, MEDLINE, Embase to May 2004 Search completed by two investigators Searched by all authors and keywords Matched abstract to full‐length publication by One author Contents Methodology
Data	Included 383 abstracts presented at the 1995 to 2003 Society for Academic Emergency Medicine meetings Included all abstracts of controlled clinical trials and randomized clinical trials
Comparisons	Proportion of abstracts published, overall and by meeting Median time to publication Survival analysis of publication rate 'Positive' versus not 'positive' Authors endorsing treatment versus not Parallel design versus crossover or factorial design Multi‐centered versus single center US versus non‐US origin University affiliation versus no university affiliation
Outcomes	194 of 383 abstracts published 23/33 abstracts presented at the 1995 meeting, 24/40 at the 1996 meeting, 28/50 at the 1997 meeting, 27/42 at the 1998 meeting, 28/43 at the 1999 meeting, 27/44 at the 2000 meeting, 17/36 at the 2001 meeting, 17/52 at the 2002 meeting, and 3/43 at the 2003 meeting published Proportion of abstracts published by time Median time to publication = 32 months (95% CI = 23 to 41 months) Survival analysis of proportion published at 108 months = 62.1% Factors related to proportion of abstracts published included 84/161 'positive' (defined as significant results) versus 110/222 not 'positive' abstract results published 80/154 'positive' (defined as experimental better than control) versus 4/7 not 'positive' abstract results published 108/210 abstracts with authors endorsing the treatment versus 86/173 abstracts with authors not endorsing the treatment published 164/327 abstracts with parallel design versus 30/56 abstracts with crossover or factorial design published 29/61 abstracts with multiple centers versus 165/322 abstracts with a single center published 169/339 abstracts originating in the US versus 25/44 abstracts not originating in the US published 138/253 abstracts with university affiliation versus 31/86 abstracts without a university affiliation published
Notes	Emergency medicine Funding not reported
Risk of bias
Item	Authors' judgement	Description
Sampling method?	Yes	Included all abstracts that described specific study designs, so low risk of bias.
Search for publications?	Yes	Searched 3 databases.
Follow‐up time?	Yes	All meetings before 2001 had at least 48 months follow‐up. The other meetings had less than 48 months follow‐up.
Matching?	Yes	Matched by 3 different criteria.
Adjustment for confounding?	Yes	Examined association of meeting year, positive results, sample size, author conclusions, study design, multi‐center status, number of authors, university affiliation, and US origin with publication using multivariable logistic regression analyses.