| ObjectiveTo explore the completeness status of outcome data of those Cochrane systematic reviews and RCTs focused on TCM;to explore the related factors that may influence the completeness of outcome data;to provide references and suggestions for TCM researchers to report trials' outcome data completely,and to improve the reporting quality further.Methods1.Evaluation and reporting of incomplete outcome data:a study of Cochrane systematic reviews on TCMCochrane systematic reviews(CSRs)on TCM were searched according to search strategy from Cochrane library(DVD version April 2015).All CSRs focused on TCM that included at least one RCT and assessed incomplete outcome data of each included RCTs by applying a risk of bias tool were included in our study.We aim to classify each criteria that reviewers have been used to judge the risk of bias of incomplete outcome data to identify whether reviewers accurately and sufficiently assessed,and we also aim to find out how reviewers analyzed and handled the missing data when conducting a meta-analysis.2.Reporting status of completeness of outcome data:A survey in registered and subsequently published RCTs on TCMA systematic search of 16 registries was conducted from their set till July 2015 to obtain registered records of RCTs on TCM,and another search was subsequently conducted to find corresponding publications according to registered ID,title and basic information of sponsors in six databases including PubMed,Cochrane library,CNKI,VIP,Wanfang and SinoMed.After identifying RCTs that meet our inclusion criteria,we assessed the completeness status of outcome data reporting of each RCT,including the consistency of sample size in registered records and subsequently published RCTs,dropout status of participants,analysis strategy used in trial reports,and method of handling missing data in trial reports.Then we try to explore the potential factors that may induce incomplete outcome data.Results1.Evaluation and reporting of incomplete outcome data:a study of Cochrane systematic reviews on TCMA total of 63 CSRs on TCM were included and were published from 2009 to 2015,with most published in 2013(20/63,31.75%),and most included number of RCTs were less than 50(60/63,95.24%),and most were published in Chinese(30/63,47.62%).Most of the CSRs were done by 3 to 7 co-authors(55/63,87.30%),and the first/corresponding authors are coming from China(39/63,61.09%),Australia(11/63,17.46%)and British(6/63,9.52%)?The mean sample size of CSRs was 1774.The most frequent intervention was Chinese medicine(35/63,55.56%)and acupuncture(25/63,39.86%),and the most frequent conditions were diseases of the genitourinary system(15/63,23.81%).6.35%(4/63)of included CSRs did not report using risk of bias tool to assess methodological quality in method section.7.94%(5/63)did not report judgement criteria of risk of bias tool.Risk of bias of incomplete outcome data of RCTs reported in included CSRs judged as "unclear","high","low" risk were 44.44%(28/63),52.38%(33/63),85.71%(54/63)respectively.Only 63.49%(40/63)of the CSRs reported analysis strategy or method handling missing data in method section,of which only 44.44%(28/63)reported to use intention to treat(ITT)analysis,while only 23.81%(15/63)actually used this strategy.Only 22.22%(14/63)reported to use sensitivity analysis with consideration of missing data.2.Reporting status of completeness of outcome data:A survey in registered and subsequently published RCTs on TCMA total of 195 registered records and 207 subsequently published RCTs on TCM were included in our study.67.18%of the trials were registered retrospectively(131/195).The most were registered in clinicaltrials.gov(102/195,52.31%),ISRCTN(38/195,19.49%)and ChiCTR(36/195,18.46%).The number of publications showed a tendency of sustained growth.Primary sponsors were coming from 18 countries with most from China(74/195,37.95%).The majority were published in English(186/207,89.86%),designed as parallel-group(200/207,96.62%),single-center(99/207,47.83%)?two-armed(131/207,63.29%),and funded(197/207,95.17%).9.66%(20/207)did not specify primary and secondary outcomes.Numerical primary outcomes were most frequently reported(162/207,78.26%).The majority studies reported random number sequence generation(147/207,71.01%),allocation concealment(106/207,51.21%)and blinding(178/207,85.99%).Within 19 reported conditions musculoskeletal system and connective tissue(42/207,21.21%)diseases were most studied.The most studied interventions were acupuncture(122/207,58.94%),followed by Chinese medicine(86/207,18.36%).As to type of comparison,the majority were TCM versus standard treatment(n=86,41.55%)and TCM versus placebo(78/207,37.68%).Mean sample size of included published RCTs was 186.3,with only 43.96%(91/207)consistent with those specified in registered records.60.39%(125/207)reported sample size estimation and 35.27%(73/207)considered dropout rate.73.91%(153/207)reported participants flow by using flow chart,while 75.85%(157/207)reported in text.74.88%(155/207)assessed participants for eligibility.As for 76.33%(158/207)reporting the number of dropout,only 55.56%(115/207)specified reasons,and 83.09%(172/207)reported the number of participants included in main analysis.31.88%(66/207)and 14.98%(31/207)reported number and reason for exclusion respectively.Mean dropout rate was 11.48%with the highest rate reported in mental,behavioral and neurodevelopmental disorders(12.13%).70.53%(146/207)of the trials reported to use ITT analysis in method section,however,41.67%(50/120)of them were actually modified ITT analysis.Full ITT analysis was reported in 35.75%(74/207),and complete case analysis were 31.88%(66/207).Trials using complete case analysis reported a higher rate of positive results(42/66,63.64%).Within 6.28%(13/207)reported sensitivity analysis only one trial reported discrepancy compared with the result of primary analysis.54.59%(113/207)reported method of handling missing data in method section with the most frequently used method LOCF(last observation carried forward)(34/207,16.43%).3.Potential factors that may induce incomplete outcome reporting between registered and subsequently published TCM RCTsRCTs from German(n=11),Korea(n=5),Denmark(n=3),Norway(n=3),Canada(n=2),Spain(n=2),France(n=1),Japan(n=1),Netherland(n=1),Switzerland(n=1)and Thailand(n=1)all reported 0.00%risk of "incomplete outcome data",followed by USA(4/41,9.76%),while RCTs sponsored by Chinese(28/77,39.36%)reported the highest risk of "incomplete outcome data".Univariate analysis showed that,nationality of primary sponsor,publication language,reporting method of random number sequence generation,allocation concealment,blinding,sample size estimation,consider of dropout rate,reporting of participants flow,ITT analysis in method section and positive results,all may affect the risk of "incomplete outcome data"(P<0.05).Logistic regression analysis also showed that publication language,and blinding or not were associated with the risk of "incomplete outcome data"(P<0.05).ConclusionA lack of reporting and use of risk of bias tool in recent published TCM CSRs was found.Reviewers do misunderstand and misused judgement criteria for incomplete outcome data.The analysis strategy and method of handling missing data also have a lack of report in these CSRs.TCM RCTs have a lack of prospective registration but the reporting methodological information has been improved.However,researchers still insufficiently reported flow of participants,analysis strategy(including ITT analysis and sensitivity analysis)and method of handling missing data,moreover,ITT analysis till misused by some researchers.Sponsors from Chinese and RCTs published in Chinese likely to report incomplete outcome data.This was also associated with whether applying blinding method or not.In order to reduce the chance of misunderstanding the concept of incomplete outcome data and misusing of the principles of ITT analysis,to improve the credibility and reliability of TCM RCTs results,and provide reference for the decision making of evidence-based Chinse medicine,we suggest strengthening the scientific ability of TCM researchers by supporting more and more high-quality training and education opportunities. |
PDF Full Text Request