Publication bias in blood and marrow transplantation

Abstract

Only a small proportion of abstracts lead to full publication. Abstracts with "positive" results are more likely to be published than other abstracts, leading to publication bias. To date, this issue has not been examined in the blood and marrow transplantation (BMT) literature. We hypothesized that because BMT centers are often based at academic centers, the proportion of abstracts leading to publication will be high. All abstracts presented at the Canadian Blood and Marrow Transplant Group biannual meetings in 2002, 2004, and 2006 were reviewed and categorized by study type, funding source, single-center or multicenter study, form of presentation, and positive or negative results, using the authors' definitions. To determine publication, each reference was searched on multiple databases (MEDLINE, EMBASE, Web of Science, and CINAHL) by first, second, and final author names. Two authors performed abstract categorization and searching, and disagreements were resolved by consensus. Of the 141 abstracts reviewed, only 43 were published (30.4%). Twenty-one studies were published from 2002 (36.8%), compared with 12 from 2004 (24.0%) and 10 from 2006 (29.4%) (P = .35). Neither positive results nor the number of involved centers were associated with the likelihood of publication. Clinical studies (retrospective or prospective) were more likely to be published than nonclinical studies (P = .014). Funded studies and oral presentations were more likely to be published (P = .009 and .004, respectively). A low rate of publication is seen in the field of BMT. Studies with clinical outcomes, externally funded studies, and studies presented orally were more likely to be published. However, there was no publication bias in favor of studies with positive results. Publication bias should be evaluated further at larger BMT meetings, and efforts should be made to encourage full publication of scientific abstracts.