| PurposeSystematic review is the highest level evidence source of evidence-based medicine,which is of great significance for guiding medical decision-making.Publication bias is one of the key factors affecting the quality and reliability of systematic reviews.In this study,1)literature analysis was conducted to understand the current situation of TCM systematic review published in Chinese and propose improvement methods;2)systematically sort out several common statistical methods for detecting and correcting publication bias from the perspective of the author and the reader,analyze their advantages,disadvantages and applicable conditions,and evaluate the measures to reduce publication bias and the risk formation of potential publication bias;3)to compare the differences in publication bias between Chinese and western medicine in four types of systematic reviews from 2001 to the first half of 2018,and to evaluate and compare the potential risks of publication bias in these four types of systematic reviews;MethodsWanfang Database was searched for systematic reviews published in the year of 2001-2004,2005-2007,2010,2014,and first half of 2018.Random samples from all therapeutic systematic reviews and Meta-analysis documents taking TCM therapies as interventional measures,in the above five time periods(30,30,50,50,15 papers respectively),were identified.Funding information,conflict of interest.searching of grey literature,methods dealing with publication bias were extracted and evaluated by reviewers in pairs independently.Agreements were achieved.The chi-square test or fisher’s exact probability test was used for the comparison between the two groups.The literature related to publication bias was searched,and the key words included"publication bias"."Funnel plot","Egger’s","fail-safe "Rank correlation","Begg method","Peters’s","Rucker’s","Harbord’s" and "Schwarzer’s".Methodological literature on publication bias and methodological reports related to systematic review were preliminarily screened out,and various statistical algorithm literature on publication bias and literature related to publication bias were included,and literature referring to publication bias and its algorithm in general system review were excluded.Various statistical algorithms for publication bias in the literature and factors related to the potential risk of publication bias were extracted.Various statistical algorithms for publication bias were systematically sorted out according to outcome type,and their advantages and disadvantages and applicable conditions were discussed.The articles on judging the size of publication bias were extracted according to the literature,and the foreign experts were discussed.The evaluation elements were sorted out according to the extracted articles and expert opinions.According to the statistical methods of publication bias in study 2.the risk assessment elements of potential publication bias were systematically sorted out and formed,and the systematic reviews retrieved from English database PubMed and Chinese database wanfang were evaluated.The random seeds were randomly included into the systematic review of Chinese traditional medicine,Chinese western medicine,English Chinese traditional medicine and English western medicine according to the inclusion and exclusion criteria.All included systematic reviews were understood and compared in terms of prevention,reduction and detection.The adequacy of measures to reduce publication bias and the potential risk of publication bias were compared.Sixteen participants of the project checked and reached a consensus after extracting the included literature in pairs and back to back,and a third party discussed or made a decision if they had different opinions.Frequency(composition ratio)was used to test the consistency of judgment results and kappa value was calculated.ResultsA total of 175 papers were included and extracted.Generally speaking,publication bias was not given enough attention in the included studies.In the overall five time periods,there have been improvements in the searching for grey literature.However more aspects were still problematic,including inadequate adoption of statistical methods for detecting publication bias(55%),neglecting the reporting of funding(41.7%)and conflict of interests(0%),and neglecting the reporting of language limitation(35.4%).In systematic reviews with positive results Meta-analysis of primary outcomes,the proportion of negative/unclear attitudes of conclusions in reviews that did not attempt to detect publication bias(6%)is 1/4 of that(25%)in reviews that attempted.Funnel plot is the most commonly used statistical method to detect publication bias,but the use conditions and causes of funnel plot asymmetry should be considered in its use.In addition,the continuity outcome indicator can be Egger test,Begg’s test,and the binary outcome indicator can be Harbord test,Peters test.Rucker test and other methods.However.the test efficiency of these statistical methods still needs to be further verified.The risk of publication bias can be assessed qualitatively from the perspectives of systematic literature retrieval,Meta analysis results,detection of publication bias results and author conclusions.English western medicine reported higher fund sources(75%)than Chinese medicine(71%),Chinese medicine(67%)and Chinese western medicine(57%).Since 2001,MEDLINE use has increased from 67%to 96%.The overall retrieval rate of the Cochrane library from 2001 to 2004 was only 8%.By the first half of 2018,the Cochrane database had been explicitly searched for systematic reviews of 73%Chinese traditional medicine,75%Chinese western medicine.91%English traditional medicine and 87%English western medicine.Compared with 2001 to the first half of 2018,the use of EMBASE has almost doubled(15%to 36%),but this contribution rate mainly comes from the western medicine systematic review published in Chinese and English.For the Chinese medicine systematic review published in Chinese or English,the retrieval rate of EMBASE is very low,and there is no room for improvement,less than 1%.Only 13 of the total included articles were explicitly indexed from the clinical trial registry.For the retrieval of gray literature,the retrieval rate of degree and conference papers in Chinese and traditional Chinese medicine is 100%by default due to the inclusion of Chinese database,which is higher than 26%in English western medicine.No publication bias was detected in 45%Chinese traditional medicine,22%Chinese western medicine,37%English Chinese traditional medicine systematic review and 21%English western medicine systematic review.The efforts of systematic review makers to reduce publication bias are insufficient:the proportion of western medicine in English is 53%,that of traditional Chinese medicine in English is 63%,that of traditional Chinese medicine in Chinese is 78%.and that of western medicine in Chinese is 75%.The estimated risk of publication bias was 68%for Chinese medicine,43%for English medicine.34%for English medicine,and 32%for Chinese medicine.ConclusionThe publication bias in the systematic review of TCM has been slightly improved in terms of the sources of report funds and the retrieval of grey literature,but there are still many problems in the use of improper methods to detect publication bias and the neglect of the discussion on publication bias.The review that detects publication bias will be more cautious to give a positive conclusion.About publication bias overall improved,but the Chinese system of traditional Chinese medicine were reviewed on the publication bias value is less Chinese medicine class,English class of traditional Chinese medicine and western medicine in English class system,we can consider from the Angle of the qualitative and quantitative risk evaluation for publication bias,from all aspects to reduce the publication bias efforts,improve the system reliability of evidence is summarized in this paper,at the same time,strengthen the methodology training related to publication bias... |
PDF Full Text Request