Analysis Of The Reporting Quality And Statistical Reporting Quality Of Observational Studies And Its Impact Factors

BackgroundIn recent years,some published biomedical research results cannot be reproduced by other researchers,a phenomenon that has attracted widespread attention from researchers,journal editors and the public.The poor reporting quality of literature,especially statistical reporting,is one of the causes of this problem.Studies with poor reporting quality can not only affect the secondary use of study data in systematic review and meta-analysis,but also limit the application of research results in clinical practice.Studies on assessing the reporting quality and statistical reporting quality of observational studies published in the journal of nutrition have not been published.Objective(1)To assess the reporting quality of observational studies published in the journals of high-impact factor nutrition and to explore the factors affecting the reporting quality;(2)To analyze the types,numbers and frequencies of the statistical methods used in observational studies published in the journals of high-impact factor nutrition;(3)To access the statistical reporting quality of observational studies published in the journals of high-impact factor nutrition and to explore the factors affecting the statistical reporting quality;(4)We hope that the study will not only draw attention from researchers,journal editors and peer reviewers in nutrition to the reporting quality and statistical reporting quality of observational studies published in this field,but also provide the theoretical basis for health care workers and decision-makers to evaluate correctly and apply rationally the findings of observational studies.Methods(1)According to the impact factors of Web of Science in 2018,the top five nutrition journals were selected: American Journal of Clinical Nutrition(impact factors = 6.568);Clinical Nutrition(impact factor = 6.402);International Journal of Behavioral Nutrition and Physical Activity(impact factor = 6.037);International Journal of Obesity(impact factor = 4.514);European Journal of Nutrition(impact factor = 4.449).Observational studies related to diet and nutrition from January 1,2014 to December 16,2019 published in five journals in PubMed database were retrieved.The language limited to English.(2)According to the inclusion and exclusion criteria,the screening of the literature and the extraction of data were carried out independently by two researchers.Information extracted from articles included: journal type,year of publication,type of study design,number of authors,the affiliation of the first author,the origin of the corresponding author,number of author affiliation,whether statisticians or epidemiologists were involved,international collaboration,and funding support.(3)The reporting quality and statistical reporting quality of the included articles were evaluated using the STROBE-nut statement and self-designed statistical reporting quality evaluation checklist.(4)For the categorical variables,rate,constituent ratio,odds ratio(OR)and 95% confidence interval(CI)were calculated and for continuous variables,the mean and standard deviation were calculated.The chi-square test or Fisher's exact test was used to compare the composition ratio and rate between different groups of categorical variables.Student t-test or analysis of variance was used to compare the mean between different groups of continuous variables.Multivariate linear regression analysis was used to identify the factors affecting the reporting quality score and statistical reporting quality score,and the mean difference(MD)and 95%CI were calculated.After the STROBE-nut score and statistical score were divided into the high-quality group and the low-quality group,the impact factors related to reporting quality were explored by using multivariate Logistic analysis,and the OR and 95%CI were calculated.Results 1.Basic information of included articlesA total of 1417 articles were identified in initial retrieval.After excluding 951 articles,466 qualified articles were included in the study.The American Journal of Clinical Nutrition accounted for 34.5%(161 articles),Clinical Nutrition accounted for 26.4%(123 articles),International Journal of Behavioral Nutrition and Physical Activity accounted for 1.9%(9 articles),International Journal of Obesity accounted for 2.4%(11 articles)and European Journal of Nutrition accounted for 34.8%(162 articles).Before and after the publication of the STROBE-nut statement,194 and 272 articles were published,respectively.2.The STROBE-nut reporting quality of included articles(1)Of the 58 items in the STROBE-nut statement,the reporting rate of 26 items was above 90%.The following items had a reporting rate of 100%,including item 1.b(title/abstract),item 2(background/rationale),item 3(study objectives),item 4(study design),item 5(study settings),item nut-5(study settings),item 6.a(study participant),item nut-6(study participant),item 15(outcome data),item nut-17(sensitivity analysis and data imputation),item 18(key results),item 20(interpretation of results),item nut-20(the nutritional relevance of the findings),item nut-22.1(the procedure for consent and study approval from ethics committee).Items with a reporting rate of less than 20% were item 10(the calculation of sample size),item 14.b(the number of participants with missing data),item 16.c(translating of absolute risk),item nut-12.2(the method for energy adjustments and intake modeling),item nut-8.2(food composition data and the conversion procedure);(2)Since the publication of STROBE-nut statement,the reporting rate for the following items has increased: description of matching criteria and number(item 6.b,95.4% vs.98.9%,OR=4.08,95%CI:1.28-13.04),the assessment of non-dietary data and timing of assessment(item nut-8.5,75.8% vs.84.2%,OR=1.71,95%CI: 1.08-2.73),the use of flow diagram(item 13.c,19.1% vs.38.2%,OR=2.47,95%CI: 1.66-3.69),conversion of absolute risk(item 16.c,9.3% vs.15.8%,OR=1.77,95%CI: 1.03-3.06),and online supplementary information(item nut-22.2,51.0% vs.67.6%,OR=2.00,95%CI: 1.38-2.92);(3)The mean of the STROBE-nut score in included articles of this study was(44.78±3.40),accounting for 77.21% of the full score.There were 314 articles in the high-quality group(67.4%)and 152 articles in the low-quality group(32.6%).Univariate analysis showed the STROBE-nut score of American Journal of Clinical Nutrition adopting the STROBE-nut statement was higher than the other four journals that not adopted the statement(P<0.05),the score after the publication of the STROBE-nut statement was higher than before(P=0.025),the score of cohort studies was higher than case-control studies and cross-sectional studies(P=0.003),articles with the participation of statistician or epidemiologist scored higher(P=0.001)and articles with fund supporting had a higher score(P=0.026).(4)The results of multivariate linear regression analysis showed that the STROBE-nut score of articles published in American Journal of Clinical Nutrition increased by 2.203 scores compare with articles published in European Journal of Nutrition(P<0.001),the STROBE-nut score increased by 1.422 scores after the publication of the STROBE-nut statement(P<0.001),and the score of articles with the participation of statistician or epidemiologist increased by 0.630 scores(P=0.043).(5)The results of multivariate Logistic regression analysis revealed that the reporting quality of observational studies published in American Journal of Clinical Nutrition was higher than that of the articles published in European Journal of Nutrition(OR=4.298,95%CI: 2.376-7.774,P<0.001);the reporting quality after the publication of the STROBE-nut statement was higher than before(OR=2.571,95%CI: 1.624-4.071,P<0.001);the reporting quality of articles with the participation of statistician or epidemiologist was higher than that of the literature without the participation of statistician or epidemiologist(OR=1.671,95%CI: 1.067-2.617,P=0.025).3.The types and frequencies of statistical methods used in included articlesAmong the included articles,each article used 4 statistical methods on average.The five most commonly used statistical methods were trend test(52.4%),contingency tables analysis(51.3%),Logistic regression analysis(36.9%),linear regression analysis(36.3%),and Cox proportional hazards model(35.0%).4.The statistical reporting quality of included articles(1)In statistical reporting items of included articles,three items with the highest reporting rate were confidence interval(97.6%),name of statistical program or software(97.2%),and exact P-value(97.0%).The three most overlooked items were the sample size calculation/power analysis(11.6%),outliers(37.8%),and one/two tailed for test(48.1%).In the multivariate analysis,collinearity between independent variables,coding of variables,the validation of the statistical models,goodness of fit test and methods of independent variables entering the models were seldom reported in included articles;(2)The statistical reporting quality score of included articles ranged from 15.00% to 94.12%,with an average score of(46.19±12.46)%,accounting for 46.19% of the full score.There were ten articles in the high-quality group(2.1%)and 456 articles in the low-quality group(97.9%);(3)The results of multivariate linear regression analysis showed that compared with the score of articles published in European Journal of Nutrition,the score of articles published in Clinical Nutrition was 3.560 lower(P=0.018),the International Journal of Behavioral Nutrition and Physical Activity 8.627 lower(P=0.039),the International Journal of Obesity 11.743 lower(P=0.001);the score of cohort study was 0.748 higher than that of the cross-sectional study(P=0.037);compared with articles whose corresponding author was from Asia,the score of articles whose corresponding author was from Europe decreased by 2.786 scores(P=0.043)and the score of articles whose corresponding author was from Africa increased by 20.734 scores(P=0.004).5.The analyses about the relatioanship between the STROBE-nut reporting quality and statistical reporting quality of included articlesThe results of linear correlation showed that there was a weak positive correlation between the reporting quality and the statistical reporting quality of observational studies published in the journal of high-impact factor nutrition.ConclusionThe STROBE-nut reporting quality of observational studies published in the journal of high-impact factor nutrition was good,with reporting rates of over 90% in 26 items,but the reporting quality of methods,results and discussion section needed to be improved,especially in the following items including the calculation of sample size,the number of participants with missing data,translating of absolute risk,the method for energy adjustments and intake modeling,food composition data and the conversion procedure.The publication of the STROBE-nut statement has improved the reporting quality;working with statistician or epidemiologist can improve the reporting quality;journals that adopted the STROBE-nut statement have higher reporting quality.Observational studies published in the journal of high-impact factor nutrition used more statistical methods,with an average of 4.Sophisticated statistical methods such as Logistic regression analysis,multiple linear regression analysis and Cox proportional hazards model were widely used,but the quality of statistical reporting was not high.The report rate of collinearity between independent variables,coding of variables,the validation of the statistical models,goodness of fit test and methods of independent variables entering the models was low.Researchers in the field of nutrition should pay attention to the reporting quality and statistical reporting quality of observational studies,and it is suggested that more nutrition journals should actively adopt the STROBE-nut statement and statistical reporting quality checklist.Authors,journal editors and peer reviewers should strictly comply with the STROBE-nut statement when conducting,writing and reviewing observational studies papers to produce more high-quality observational studies that provide better evidence for public health and clinical decision-making and provide better services for guiding medical and health practice.