Article type
Year
Abstract
Objectives: To study differences between abstracts of randomized controlled studies (RCTs) in neonates submitted to the American Pediatric Society/Society for Pediatric Research (APS/SPR) annual conference and subsequent publication in peer-reviewed journal.
Methods: All RCTs involving neonates and published in the program issues of 1993/1994 were identified by hand searching. Medline, Embase and the Cochrane Library were searched to identify subsequent publication of these studies by names of authors and subject titles by March 1999; i.e. 5-6 years after the publication of the abstract.
Results: For the 1993/94 APS-SPR annual conferences, 141 abstracts of RCTs performed in neonates were submitted and 107 (76 %) were accepted for presentation (poster session 74, poster symposium 7, subspecialty platform 21, joint subspecialty platform 3, joint poster session. Authorship included a median of 4 (range 1 to 13) authors per study. Half of the RCTs enrolled <30 neonates. Only 73 (52%) of the studies were published by March of 1999 (38% within 1 year, 64% within 2 years, and 90% within 4 years of initial submission). Sixty four per cent of the published studies were published under a different title and 46% with an increased number of authors. Of the full publications 57% had < 5 authors. Thirty two percent of the abstracts not presented at the conference were subsequently published. One full publication was presented in two poster sessions at the meeting. In 39% of the published studies the numbers of enrolled patients differed from those reported in the abstract.
Discussion: The results of RCTs and high-quality meta-analyses of such trials are generally accepted as the best available evidence on which to base clinical decisions. The identified differences in numbers of enrolled patients in abstracts and mil publications in this study are of great concern. It has obvious implications for the accuracy of a meta-analysis that includes abstracts. Differences in the number of patients enrolled in RCTs as reported in abstracts and final reports should also be considered when quality assessments of RCTs are made. Differences may indicate elements of bias/poor data quality control with regards to; multiple looks at the data; changes in the definitions of outcomes; no preset sample size; closure of patient recruitment when statistical significance has been reached for the outcome under study and other sources of bias.
Methods: All RCTs involving neonates and published in the program issues of 1993/1994 were identified by hand searching. Medline, Embase and the Cochrane Library were searched to identify subsequent publication of these studies by names of authors and subject titles by March 1999; i.e. 5-6 years after the publication of the abstract.
Results: For the 1993/94 APS-SPR annual conferences, 141 abstracts of RCTs performed in neonates were submitted and 107 (76 %) were accepted for presentation (poster session 74, poster symposium 7, subspecialty platform 21, joint subspecialty platform 3, joint poster session. Authorship included a median of 4 (range 1 to 13) authors per study. Half of the RCTs enrolled <30 neonates. Only 73 (52%) of the studies were published by March of 1999 (38% within 1 year, 64% within 2 years, and 90% within 4 years of initial submission). Sixty four per cent of the published studies were published under a different title and 46% with an increased number of authors. Of the full publications 57% had < 5 authors. Thirty two percent of the abstracts not presented at the conference were subsequently published. One full publication was presented in two poster sessions at the meeting. In 39% of the published studies the numbers of enrolled patients differed from those reported in the abstract.
Discussion: The results of RCTs and high-quality meta-analyses of such trials are generally accepted as the best available evidence on which to base clinical decisions. The identified differences in numbers of enrolled patients in abstracts and mil publications in this study are of great concern. It has obvious implications for the accuracy of a meta-analysis that includes abstracts. Differences in the number of patients enrolled in RCTs as reported in abstracts and final reports should also be considered when quality assessments of RCTs are made. Differences may indicate elements of bias/poor data quality control with regards to; multiple looks at the data; changes in the definitions of outcomes; no preset sample size; closure of patient recruitment when statistical significance has been reached for the outcome under study and other sources of bias.