Factors Related To Risk Of Bias In Randomized Controlled Trials Of Chinese Herbal Medicine

Objectives: Assess the risk of bias(Ro B)in randomized controlled trial(RCT)of Chinese herbal medicine and explore the associations between trial characteristics and the size and direction of treatment effect estimates.Methods: A meta-epidemiological study searched five databases.Meta-analyses based on RCT with Chinese herbal medicine on intervention effect are eligible.Reviewers assessed the methodological quality of included RCTs and extracted treatment effect estimates independently.We explored the associations between trial characteristics and treatment effect estimates using two-step method.Differences in treatment effect estimates were measured with ratio of risk ratio(RRR)for binary outcome and differences in standardized mean difference(SMD)for continuous outcome.We performed subgroup analyses according to outcome category,publication language and funding support status.Associations between treatment effect estimates and trial characteristics were presented with forest plots.Results: This study identified 91 meta-analyses(1338 RCTs)with 17 disease areas.We explored the associations between 14 trial characteristics and treatment effect estimates.Around 98% RCTs performed high or unclear Ro B on allocation concealment and blinding.More than 80% RCTs did not report related information about intention to treat analysis and funding support.The meta-epidemiological study:(1)Binary outcome: we did not find the significant associations between trial characteristics and treatment effect estimates.However,subgroup analyses found that larger treatment effect estimates(RRR : 1.05,95%CI: 1.01,1.10)were significantly associated with RCTs of prescription in English publication.RCTs with low Ro B for blinding showed significant relations with larger treatment effect estimates [blinding of patients and personnel(RRR:1.42,95%CI:1.07,1.88);blinding of outcome assessors(RRR:1.28,95%CI:1.04,1.58)] in Chinese publication.(2)Continuous outcome: i)Specifical trial characteristics for RCT with Chinese herbal medicine: RCT with Chinese medicine diagnosis criteria showed larger treatment effect estimates(SMD:0.23,95%CI:0.06,0.41)and such association was only found in studies with subjective outcome,Chinese publication and no funding support.ii)Cochrane Ro B related trial characteristics: We found significant relation between RCTs with low Ro B for allocation concealment and lower treatment effect estimates(SMD:-0.70,95%CI:-0.99,-0.42),such relation only being found in objective outcome rather than subjective outcome.RCTs with low Ro B on incomplete outcome data showed much larger treatment effect estimates(SMD:0.30,95%CI:0.18,0.43)than that of high or unclear Ro B and such associations only being found in studies with subjective outcome,English publication and no funding support.In studies with objective outcome,Chinese publication and funding support,RCTs with low Ro B for selective outcome reporting showed larger treatment effect estimates.iii)Other general trial characteristics: Compared with single-center RCTs,multicenter RCTs showed lower treatment effect estimates(SMD:-0.33,95%CI:-0.61,-0.05)and such association only being found in studies with Chinese publication and funding support.RCTs with funding support were significantly associated with lower treatment effect estimates(SMD:-0.22,95%CI:-0.41,-0.02),which only was found in subjective outcome.Conclusions: Chinese medicine diagnosis criteria,allocation concealment,incomplete outcome data,number of centers and funding support were associated with treatment effect estimates.The methodological and reporting quality were dissatisfied.Specific guideline for critically appraising RCT of Chinese herbal medicine is needed.