Article type
Year
Abstract
Background: One of the main aspects of publication bias is underreporting of randomised clinical trials with negative results (RCT-NR). Due to an increased awareness of the importance of publication bias, researchers and entities are now more prone to promote registration of protocols of clinical trials and publication of their results.
Objectives: Primary: TTo assess changes across time in the prevalence of published RCT-NRs. Secondary: To describe the main characteristics of published RCT-NR (e.g. sample size, medical specialty, journal of publication). To describe the concordance between results and conclusions for the main outcome in abstracts of published RCT-NRs.
Methods: We planned a retrospective, cross-sectional, comparative study of a random sample of RCTs from PubMed database, published in two different periods of time the years: 1998 (first period) and 2009 (second period). We excluded references with no abstract available and studies with designs other than RCT. The main outcome was the difference in percentage of RCT-NRs in the two periods. A total sample size of 964 RCTs, 482 from each period would have a minimum of 80% power to detect a difference greater than or equal to 10% in the main outcome, assuming a 20% of RCT-NRs in the first period, a 5% type I error and allowing for 40% attrition. Two authors will independently extract data. The chi-squared test will be used to compare the data of the main outcome. We will contrast the trial result associated with the main outcome with the authors’ conclusions.
Results: In April 2013, we obtained 49 363 references of possible RCTs after the application of a validated RCT filter. We identified 623 references from the first period and 2000 from the second period. We randomly selected 482 references from each period.
Conclusions: We plan to present the results and conclusions of this study during the 2013 Cochrane Colloquium.
Objectives: Primary: TTo assess changes across time in the prevalence of published RCT-NRs. Secondary: To describe the main characteristics of published RCT-NR (e.g. sample size, medical specialty, journal of publication). To describe the concordance between results and conclusions for the main outcome in abstracts of published RCT-NRs.
Methods: We planned a retrospective, cross-sectional, comparative study of a random sample of RCTs from PubMed database, published in two different periods of time the years: 1998 (first period) and 2009 (second period). We excluded references with no abstract available and studies with designs other than RCT. The main outcome was the difference in percentage of RCT-NRs in the two periods. A total sample size of 964 RCTs, 482 from each period would have a minimum of 80% power to detect a difference greater than or equal to 10% in the main outcome, assuming a 20% of RCT-NRs in the first period, a 5% type I error and allowing for 40% attrition. Two authors will independently extract data. The chi-squared test will be used to compare the data of the main outcome. We will contrast the trial result associated with the main outcome with the authors’ conclusions.
Results: In April 2013, we obtained 49 363 references of possible RCTs after the application of a validated RCT filter. We identified 623 references from the first period and 2000 from the second period. We randomly selected 482 references from each period.
Conclusions: We plan to present the results and conclusions of this study during the 2013 Cochrane Colloquium.