Meta-Analyses in Plastic Surgery: Can We Trust Their Results?

Purpose: The objectives of this manuscript is to assess the overall quality of meta-analyses in plastic surgery from 2007-2017, assess whether there has been an improvement in quality over time, and evaluate variables that may be associated with scientific quality.

Methods: A systematic review of meta-analyses was undertaken using a computerized search of Medline, Embase, Cochrane Database for Systematic Reviews. Articles from seven plastic surgery journals published between the years 2007 to 2017 were included. Publication descriptors (author, year, country of publication), methodological and statistical methods were extracted. Each article was then assessed using the A Measurement Tool to Assess Systematic Reviews (AMSTAR) instrument.

Results: A total of 67 studies were included. The number of meta-analyses increased consistently between 2007 and 2017 with the majority of studies coming from the United States. Most studies were outcome based, assessing a single intervention, from the journal Plastic & Reconstructive Surgery, pooled a mean of 21 primary studies (range: 2-134), and utilized a mean of 2465 patients (range: 44-14884). Most meta-analyses analyzed primary studies in the middle tiers of evidence levels (II to IV), with a small percentage analyzing randomized controlled trials (16.4%). Random effect modeling was most commonly used (47.8%) and meta-analyses generally had positive (82.1%) and significant results (74.6%). Meta-analyses evaluated clinical (80.6%), methodological (65.6%), and statistical heterogeneity (50.7%) variably in terms of appropriateness and a substantial portion did not acknowledge or report methodological (7.5%) and statistical heterogeneity (25.4%). AMSTAR scores ranged between two and ten, with a mean of 6.7 out of 11. AMSTAR scores were correlated with year of publication (p=0.04, R=0.25). Multivariable linear analysis indicated that more recent studies, studies that included a rationale for statistical pooling, and studies that properly managed methodological heterogeneity were correlated with higher AMSTAR scores (r=0.66, p<0.01).

Conclusions: The quality and number of meta-analyses have increased; however, despite an improvement in quality, the overall quality of most meta-analyses remains low. Meta-analyses should utilize proper data pooling methods and account for clinical heterogeneity appropriately. Readers, authors, reviewers, and journal editors should utilize validated instruments to evaluate meta-analysis to uphold methodological integrity.