Duplicate Patterns, Bias, and Inconsistencies among Chinese-Sponsored Randomized Controlled Trials

Article type
Authors
Jia Y1, Huang D1, Wen J1, Qureshi R1, Wang Y1, Rosman L2, Chen Q3, Robinson K4, Gagnier J5, Ehrhardt S1, Celentano D1
1Bloomberg School of Public Health, the Johns Hopkins University
2Welch Medical Library, the Johns Hopkins University
3Institute of Medical Information & Library, Chinese Academy of Medical Sciences & Peking Union Medical College
4School of Medicine, the Johns Hopkins University
5Michigan Medicine, School of Public Health, University of Michigan
Abstract
Background: Previous studies implied the existence of duplicates among research publications from China.

Objectives: To explore the patterns of duplicates, evaluate the existence of duplicate publication bias, and assesse the inconsistencies between duplicates and the original publications produced from Chinese-Sponsored Randomized Controlled Trials (CS-RCTs).

Methods: We searched trial registries for eligible CS-RCTs which evaluated drugs and were conducted between January 1, 2008 and December 31, 2014. Journal articles produced from CS-RCTs were identified from both English and Chinese bibliographic databases.
We identified the main article and duplicates of each CS-RCT. Duplicates were classified as four patterns: IMALAS (interim analysis without self-identification or cross-reference), re-publication (re-publication of the main article without self-identification or cross-reference), SALAMI (subgroup analysis without self-identification or cross-reference), and partial duplicates which shared a subset of participants with other articles and had no cross-reference. Duplicate publication bias was evaluated following a retrospective cohort study design. We hypothesized that a CS-RCT was more likely to have duplicate(s) if its initial journal article was positive. We also assessed the inconsistencies in reporting between the main articles and the duplicates in terms of treatment, efficacy outcomes, and adverse events.

Results: Among 470 CS-RCTs published as journal article(s), 55 (11.7%) had 75 duplicates. Fifteen (20.0%), 33 (44.0%), 25 (33.3%), and 2 (2.7%) out of 75 duplicates were IMALASes, re-publications, SALAMIs, and partial duplicates, respectively. After adjusting for covariates, among CS-RCTs that were initially published in Chinese, CS-RCTs were 2.35 (95% CI: 1.04~5.30) times more likely to have duplicate(s) if their first article was positive. Among CS-RCTs that were initially published in English, CS-RCTs were 0.99 (95% CI: 0.31~3.15) times more likely to have duplicate(s) if their first article was positive.
Among 51 eligible duplicates, 14 (27.5%) and 3 (5.9%) reported inconsistent doses and schedules comparing with main articles, respectively. Among 25 eligible duplicates, 9 (36%) reported inconsistent outcomes comparing with main articles. Among 15 eligible duplicates, 11 (73.3%) reported inconsistent adverse events comparing with main articles, of which (26.7%) reported completely different adverse events comparing with main articles.

Conclusions: At least 11.7% of CS-RCTs registered in trial registries have at least one duplicate. The most prevailing duplicate pattern was re-publication. There was evidence supporting duplicate publication bias among CS-RCTs initially published in Chinese. The inconsistencies between the main articles and duplicates implied the inaccurate reporting of CS-RCTs.

Patient or healthcare consumer involvement: None