Evaluation And Correlation Analysis On Information Completeness Of Meta-analyses Abstracts And Full Texts

Objective:Meta-analyses are the important source of evidence in evidence-based clinical practice.Meta-analyses with poor report quality cannot reflect its value.Therefore,it is necessary to use existing standards to assess the reporting quality of abstracts and full texts of meta-analyses.In this study,intervention meta-analyses of tumor immunotherapy were selected as research objects,using PRISMA-A(PRISMA for Abstracts)and PRISMA(Preferred Reporting Items for Systematic Reviews and Meta-Analyses)to assess the reporting quality of abstracts and full texts from the perspective of information integrity.Furthermore,we explored the factors affecting the reporting quality of abstracts and full texts,and found the relationship between abstract reporting quality and full text reporting quality.Thus,we could understand the status of reporting quality of the meta-analyses of tumor immunotherapy interventions,and make suggestions for improving the reporting quality of the meta-analyses abstracts and full texts.Methods:English meta-analyses journal articles on immunotherapy intervention of tumors were retrieved in Pub Med,Embase and Cochrane Database of Systematic Review.Then,factors that might affect the quality of full texts and abstracts reporting were collected,including 21 characteristics involving basic article characteristics,author characteristics and journal characteristics.The adjusted PRISMA-A and PRISMA were used to assess the quality of abstracts completeness reporting and full text completeness reporting of included meta-analyses respectively,and the results of the reporting quality were measured by the total quality score and item reporting rate.The 75th percentile of the total quality score was used as the critical threshold,so that the included meta-analyses were divided into two groups of high and low reporting quality to analyze the significance of each characteristic.In univariate analysis,t-test was used for the comparison of the means of the measurement data between groups,and?~2 test or Fisher exact probability test was used for the comparison of the count data between the groups.Then the characteristics of P<0.05 were incorporated into the multivariate logistic regression to analyze the key influencing factors of reporting quality of meta-analyses.The correlation analysis between abstract reporting quality and full text reporting quality included correlation analysis between quality evaluation total scores and correlation analysis between reporting quality groups,and the scores of each abstract PRISMA-A item and the scores of the corresponding full text PRISMA item were compared using t-test.Results:6666 articles were found in the above databases and finally 216 meta-analyses were included after screening.The total reporting scores of abstracts were between 2.0 and10.5 points,with an average score of 6.68±1.11 points,and average report rate was 55.43%.Except for the 6 items with the higher reporting rate,which were item 10(result interpretation,98.62%),item 2(objectives,97.93%),item 6(included studies,93.78%)and so on,the reporting rate of other items were not satisfactory.The items with lower reporting rate were item 12(registration,1.38%),item 11(funding,5.99%),item 9(evidence strength and limitations,8.29%),and item 5(bias risk assessment method,9.22%).The five factors,i.e.,whether a meta-analyse has a structured abstract,whether it is funded,whether it is from the Cochrane database,the number of words in the abstract,and the number of pages of a meta-analyse,were significantly different between the high and low reporting quality groups(p<0.05).Multivariate analysis showed that the number of abstract words was closely related to the quality of the abstract report,which is a key influencing factor of the abstract reporting quality(OR=1.008,p=0.004)?The more words in the abstract of a meta-analyse,the higher probability of high-quality abstract reporting.The total scores of full text reporting in the included meta-analyses ranged from 7.5 to 27points,with an average score of 22.00±2.95 points.The average reporting rate of 27PRISMA's items was 81.46%.The items with the higher reporting rate were item 3(rationale,100.00%),item 21(synthesis of results,100.00%)and item 24(summary of evidence,100.00%),and the lowest is item 5(protocol and registration,15.28%).The three characteristics,i.e.,the number of abstract words,article page count,and number of citations were significantly different between the full text reporting quality groups(p<0.05),but multi-factor analysis did not find any influencing factors that could affect the full text report quality.The quality of the full text reporting of included meta-analyses were not related to the quality of the abstracts reporting at total quality assessment scores(r=-0.073,p=0.288)and reporting quality grouping(r=-0.028,p=0.684).The distribution of reporting rates for abstract items was more scattered than that of full text items,and the assessment scores of most abstract items were significantly lower than those of the corresponding full text items.Conclusions:The study pointed out the trend and problems of content items in the abstract reporting and the full text reporting of tumor immunotherapy meta-analyses.It was found that the reporting quality of the abstracts was poor and the reporting quality of the full texts was relatively satisfactory.It is necessary to strengthen the reporting of some items such as registration.The number of abstract words was a key influencing factor of the reporting quality of the abstracts.It is suggested for editors to recommend PRISMA and PRISMA-A for meta-analyses authors as reference,and appropriately relax the limit on the number of abstract words for meta-analyses to about 400 words to improve reporting quality.