Letters

Time to publication of studies was not affected by whether results were positive

EDITORThe time to publication measured by Stern and Simes began with approval of the project by the ethics committee; thus the interval embraced all phases of research and analysis.¹ This interval differs from that used by several other ethics committees cited ^{2 3} and in most previous reports of time to publication, which have begun with an analysis of completed data, not approval by the ethics committee.⁴

View larger version (23K): [in this window] [in a new window]   Relation of effect size ratio of studied treatment to speed of publication

We examined the fate of all 493 completed research studies submitted by members of staff from 103 American medical schools for consideration to the 1991 meeting of the Society of Academic Emergency Medicine.⁵ We searched Index Medicus on line in 1996 to determine which studies had been published, and we sent questions to the authors of all unpublished studies. We evaluated the methodology and quality of all submitted abstracts using a blinded delphi panel, and we calculated the effect size ratio reported in each (percentage efficacy of the intervention divided by that of the control).

Altogether 179 studies of the 493 submitted were accepted for presentation; 223 of the submitted studies were published in 44 peer reviewed journals (39 in specialties other than emergency medicine), with impact factors of 0.23 to 24.5. The mean time to publication from the time of submission to the time of the meeting did not differ between controlled studies with positive outcomes and those with negative outcomes (19.1 (SD 12.5) months v 15.6 (11.7) months, P=0.20). The impact factors of the publishing journals did not differ between studies with positive and negative results. Among published trials with control groups and a calculable effect size ratio, months to publication was poorly correlated with the effect size ratio (r=0.13; r=0.05 for non-emergency medicine journals only) (figure). Stern and Simes obtained outcomes for 70% of eligible studies; we obtained them for all. The studies we examined had all completed the final analysis, whereas only 43% of the authors' eligible studies had. The largest difference, however, is probably that the publication interval in the authors' study incorporated all phases of the research, and its length may be due to researchers' enthusiasm for implementing logistics. Investigators finding early positive results may enrol patients faster, conclude studies and analyse results early, and submit the studies more promptly than other investigators. This represents a different phenomenon than that analysed previously and does not focus on bias by the journal or peer reviewer.