Conference abstracts of a new oncology drug do not always lead to full publication: Proceed with caution

BackgroundConference abstracts, often the first public record of a study, serve as a catalyst to initiate clinical and policy change. On average, 45% of all conference abstracts subsequently appear as full publications; however, the generalizability of this finding to studies of one intervention, in one population, is unknown. Our objectives were to determine the full publication rate of a cohort of abstracts, median time to publication, and predictors of these relationships.MethodsWe included the first 5 years of clinical abstract reports of rituximab for non-Hodgkin's lymphoma (NHL) from American Society of Hematology (ASH) meetings (1997–2001), identified all unique studies, and used electronic databases to identify full publications. We determined the full publication rate, median time to publication, and predictors of these outcomes.ResultsOf 109 abstracts representing 86 unique studies, the publication rate was 52.3% (45, 95% confidence interval [CI]: 41.3, 63.2), and the median time to publication, 1.4 years with 6.8 years' follow-up. Author affiliation with industry (odds ratio [OR] [95% CI] = 4.60 [1.32, 16.08] and presentation type (oral OR = 5.94 [1.31, 26.88], poster OR = 3.39 [1.24, 9.25]; reference, publication in conference abstract book only) independently predicted subsequent full publication in the adjusted analysis. We identified no predictors of time to publication.InterpretationWe suggest cautious consideration of data from conference proceedings to inform new technology clinical or policy decisions. Future work needs to examine the generalizability of our results to other diseases and technologies.