Article type
Year
Abstract
Background: previous studies have hinted at the challenge raised by duplicates among Chinese-sponsored drug-related randomized controlled trials (CS-RCTs).
Objectives: to evaluate the duplicate patterns and duplicate publication bias among CS-D-RCTs. We hypothesized that CS-D-RCTs with positive results are more likely to have duplicates than ones with negative results.
Methods: we conducted a retrospective cohort study among CS-D-RCTs in the primary registries of the International Clinical Trials Registry Platform of the World Health Organization.
We defined an eligible CS-D-RCT as an RCT where:
1) one of the experimental interventions was a drug or biological;
2) at least one sponsor was located in mainland China;
3) at least one recruitment center was located in mainland China; and
4) it was conducted between 1 January 2008 and 31 December 2014.
We mapped eligible records to journal articles indexed in three English bibliographic databases and four Chinese bibliographic databases. We tailored the search strategy for each bibliographic database.
Journal articles that corresponded to one CS-D-RCT formed a cluster. We defined the main article as the one with the largest sample size or earliest publication date. We defined a duplicate as an article that overlapped with a previous one without cross-reference. We compared participants, interventions, and outcomes between the main article and duplicates to determine the duplicate pattern.
We considered a CS-D-RCT to be positive if at least one of the primary outcomes was positive. We used logistic regression models to evaluate the duplicate publication bias, adjusting for covariates such as sample size, number of recruitment centers, and funding source.
Results: we identified 1003 eligible CS-D-RCTs registered in four primary registries: 621 from the Chinese Clinical Trial Registry, 324 from ClinicalTrials.gov, 52 from the ISRCTN registry, and six from the Australian New Zealand Clinical Trials Registry. Among the 548 (55%) CS-D-RCTs that were published, 109 (19.9%) had at least one duplicate. Among the 769 journal articles, 186 (24.2%) were duplicates. Among the 109 clusters with duplicates: 36 were translations without cross-reference; 17, 3, and 23 were 'salami' publications with a subset of outcomes, interventions, and participants from the main article, respectively; 24 and six were 'imalas' publications (i.e. 'salami' backwards, where new data are added repeatedly to existing data) with increasing sample size and interventions, respectively. Adjusting for covariates, the risk of having duplicates among CS-D-RCTs with positive results was 5.71 (95% CI: 4.34 to 6.67) times the risk among CS-D-RCTs with negative results.
Conclusions: the three prevailing duplicate patterns among CS-D-RCTs are translations, 'salami', and 'imalas' publications. CS-D-RCTs with positive results were more likely to have duplicates. Systematic reviewers should be alert of the possible duplicates regarding CS-D-RCTs.
No patient/healthcare consumer was involved in this study.
Objectives: to evaluate the duplicate patterns and duplicate publication bias among CS-D-RCTs. We hypothesized that CS-D-RCTs with positive results are more likely to have duplicates than ones with negative results.
Methods: we conducted a retrospective cohort study among CS-D-RCTs in the primary registries of the International Clinical Trials Registry Platform of the World Health Organization.
We defined an eligible CS-D-RCT as an RCT where:
1) one of the experimental interventions was a drug or biological;
2) at least one sponsor was located in mainland China;
3) at least one recruitment center was located in mainland China; and
4) it was conducted between 1 January 2008 and 31 December 2014.
We mapped eligible records to journal articles indexed in three English bibliographic databases and four Chinese bibliographic databases. We tailored the search strategy for each bibliographic database.
Journal articles that corresponded to one CS-D-RCT formed a cluster. We defined the main article as the one with the largest sample size or earliest publication date. We defined a duplicate as an article that overlapped with a previous one without cross-reference. We compared participants, interventions, and outcomes between the main article and duplicates to determine the duplicate pattern.
We considered a CS-D-RCT to be positive if at least one of the primary outcomes was positive. We used logistic regression models to evaluate the duplicate publication bias, adjusting for covariates such as sample size, number of recruitment centers, and funding source.
Results: we identified 1003 eligible CS-D-RCTs registered in four primary registries: 621 from the Chinese Clinical Trial Registry, 324 from ClinicalTrials.gov, 52 from the ISRCTN registry, and six from the Australian New Zealand Clinical Trials Registry. Among the 548 (55%) CS-D-RCTs that were published, 109 (19.9%) had at least one duplicate. Among the 769 journal articles, 186 (24.2%) were duplicates. Among the 109 clusters with duplicates: 36 were translations without cross-reference; 17, 3, and 23 were 'salami' publications with a subset of outcomes, interventions, and participants from the main article, respectively; 24 and six were 'imalas' publications (i.e. 'salami' backwards, where new data are added repeatedly to existing data) with increasing sample size and interventions, respectively. Adjusting for covariates, the risk of having duplicates among CS-D-RCTs with positive results was 5.71 (95% CI: 4.34 to 6.67) times the risk among CS-D-RCTs with negative results.
Conclusions: the three prevailing duplicate patterns among CS-D-RCTs are translations, 'salami', and 'imalas' publications. CS-D-RCTs with positive results were more likely to have duplicates. Systematic reviewers should be alert of the possible duplicates regarding CS-D-RCTs.
No patient/healthcare consumer was involved in this study.