N-acetylcysteine in the prevention of contrast-induced nephropathy: publication bias perpetuated by meta-analyses.

Background: The use of N-acetylcysteine for the prevention of contrast-induced nephropathy has been the subject of numerous clinical trials and meta-analyses. We sought to examine the possibility of a publication bias and whether meta-analyses have magnified any potential publication bias.

Methods: We performed a Medline search and a manual search to identify published manuscripts. We also manually searched contemporaneous major cardiology scientific meetings to identify abstracts. We included only randomized controlled clinical trials. We pooled the results of abstracts and manuscripts separately and combined, calculating an odds ratio (OR). We analyzed meta-analyses according to the proportions of manuscripts and abstracts that they included and compared their calculated ORs to the OR of all available contemporaneous data. Our analysis spanned the time from the publication of the first manuscript on this topic through June 2006.

Results: Throughout the study period, the published manuscripts presented a treatment-effect estimate that was more optimistic than that found in unpublished abstracts. There was a temporal trend in that the estimate of treatment effect was greatest with early publications, which diminished as additional data became available. The profile of the journal (as assessed by impact factor) in which a manuscript was published was not related to the quality of the manuscript. However, studies reaching a positive conclusion were published in journals with higher impact factors compared with studies reaching negative conclusions. Meta-analyses included a substantially greater proportion of published manuscripts versus unpublished abstracts and provided more optimistic assessments of treatment effect than would have been derived had all available data been assessed.

Conclusions: There was a significant publication bias that persisted throughout the life cycle of this clinical question. The bias was further amplified by meta-analyses.