Meta-analysis: Principles and procedures

In the citing paper

“Only one (1%) of the 93 included articles addressed all key items [29]”

In the citing paper

“Sixteen articles (17%) provided recommendations on the key item “protocol development” (Table 3), [29–44] with publication dates between 1994 and 2015 (median year of publication 2009).”

In the citing paper

“The articles recommended that the protocol should state the objectives, hypotheses to be tested and the rationale of the review, [29] and that it should describe eligibility criteria [29, 33, 35, 36, 38, 39, 41], define the type of studies to be included [35, 37, 42, 44], and give the reason when including observational studies in the review [35, 37, 40].”

In the citing paper

“Twenty articles (22%) gave recommendations on the key item “research question” [29, 30, 35, 36, 38–43, 45–54], with publication dates between 1997 and 2015 (median year of publication 2009).”

In the citing paper

“Nine articles (10%) made recommendations on how to perform data extraction [29, 33–35, 38, 43, 47, 48, 74], with publication dates between 1995 and 2016 (median year of publication 2009).”

In the citing paper

“The majority of articles discussed aspects general to systematic reviews including advantages and limitations of different literature sources and databases, search tools, the importance of identifying unpublished studies, literature searching techniques including how to build a search string and reporting [12, 29, 33–36, 46, 48, 50, 51, 53, 54, 57, 59–62, 67, 68].”

In the citing paper

“Twenty-two articles (24%) provided recommendations on the key item “study eligibility”, [12, 29, 30, 33, 34, 37–39, 41–43, 46–48, 54, 55, 61, 65, 70–73] with publication dates between 1995 and 2016 (median year of publication 2009).”

In the citing paper

“Thirty-three articles (35%) made recommendations about the key item “search strategy” [12, 29, 30, 33–36, 38, 39, 43, 46–51, 53–69], with publication dates between 1995 and 2016 (median year of publication 2007).”

In the citing paper

“Five articles gave recommendations about how to assess eligibility: it should be blind [29, 30, 46], independent [29, 33, 46, 48], and performed by two reviewers [29, 30, 33, 48].”

In the citing paper

“Most articles considered the combination of RCTs and non-randomised designs [29, 35, 38, 39, 49, 51, 56, 61, 75, 81].”

In the citing paper

“Several articles recommended that the data should be extracted by two independent reviewers to avoid errors [29, 33], or that blinding those assessing methodological quality to the names of the authors, their institutions and the names of journals would lead to more consistent extraction [29, 34].”

In the citing paper

“Twenty-five (27%) articles gave recommendations on item “considering different study designs” [8, 29, 34, 35, 37–39, 42, 44, 49, 51, 56, 61, 65, 70, 73, 75, 76, 77–83], with publication dates between 1995 and 2016 (median year of publication 2007).”

In the citing paper

“Recommendations were made mainly on 2 topics: whether different study designs should be included in a systematic review; [29, 34, 37–39, 49, 56, 65, 73, 75] and whether different study designs should be analysed together in a single meta-analysis [8, 34, 35, 37, 38, 42, 49, 51, 56, 61, 70, 75–78].”

In the citing paper

“It was generally accepted that data extraction should be performed using a standardised form [29] to avoid omissions [74].”

In the citing paper

“One article recommended neither including different study designs in systematic reviews nor combining their results in meta-analyses [29].”

In the citing paper

“Twenty (22%) articles reported on item “publication bias” [16, 29, 33, 34, 36, 46, 48–50, 55, 56, 61, 70, 78, 85, 95–99], with publication dates between 1994 and 2012 (median year of publication 2001).”

In the citing paper

“Thirty-nine articles (42%) made recommendations on the assessment of quality and risk of bias of individual studies, [12, 16, 29, 30, 32–35, 37, 38, 43, 47–50, 52–54, 56, 64–66, 70, 71, 73, 75, 77, 82, 84–94] with publication dates between 1994 and 2016 (median year of publication 2009).”

In the citing paper

“The majority of articles advised against the use of scales but 12 articles recommended their use [29, 30, 48–50, 52, 53, 56, 66, 77, 86, 91].”

In the citing paper

“One common recommendation was to explore heterogeneity in sensitivity analyses, comparing studies stratified by design or clinical features [12, 16, 29, 34, 39, 51, 55, 56, 66, 77, 85, 100, 102].”

In the citing paper

“Thirty-nine articles (42%) provided recommendations on heterogeneity and its assessment, [8, 12, 16, 29, 30, 33, 34, 36, 38, 39, 41–44, 47, 49–52, 54–56, 61, 66, 70, 72, 77, 82, 85, 87, 98, 100–107] with publication dates between 1994 and 2016 (median year of publication 2006).”

In the citing paper

“Fifty-two articles (56%), provided recommendations on the statistical analysis in the context of meta-analysis of observational studies [8, 12, 16, 29–31, 33–36, 38, 41–44, 47, 48, 52, 54–56, 61, 66, 70, 72, 73, 77, 79–83, 85, 87, 96, 100, 101, 103–105, 107–118], with publication dates between 1994 and 2016 (median year of publication 2006).”

In the citing paper

“Only one article, published in 1997, addressed each of the 10 methodological aspects we examined [29].”

In the citing paper

“Second, the use of a random-effects model for meta-analysis of observational studies by default was frequently recommended [8, 16, 29, 34, 70, 77, 100, 101].”

In the citing paper

“Although many authoritative articles providing sound advice exist, such as those included in this review by Greenland [85], Egger et al. [29, 97], and a series in Research Synthesis Methods [37, 40, 65, 71, 75], the inexperienced reviewer may find them difficult to identify among the many other papers with contradictory advice.”

In the citing paper

“Visual inspection of funnel plots and Egger's tests were used to estimate publication bias [34].”

In the citing paper

“The presence of publication bias, when there were 3 or more independent estimates available, was assessed by visual inspection of the funnel plot of the meta-analyses and by using Egger's test (p b 0.05) (Egger, Smith, & Phillips, 1997).”