ABSTRACT
Aim: In systematic reviews and meta-analyses, variation (heterogeneity) in the primary studies is often a concern resulting from factors such as study design, data analysis methods, study quality, settings and interventions and/or patient characteristics. After determining the extent of heterogeneity, authors examine the causes of heterogeneity via sensitivity analysis, subgroup analysis and/or meta-regression analysis. There is no assessment of heterogeneity practices in gastroenterological literature; thus, we present this assessment.
Methods: We performed a PubMed search to identify systematic reviews published from 2011 to 2016 in the top 10 gastroenterology journals, as well as gastrointestinal topics in general medical journals and the PROSPERO trial registry. The first and second authors independently abstracted data elements, such as levels of inconsistency, sensitivity analysis, subgroup analysis and meta-regression analysis.
Results: The search identified 3154 studies; of these, 337 were eligible for inclusion. Included studies consisted of 306 from the gastroenterology literature, 19 studies from PROSPERO and 12 studies from the general internal medicine journals. A significant number of reviews in gastroenterology journals (31.05%, 95/306), internal medicine journals (16.67%, 2/12) and PROSPERO (52.63%, 10/19) conducted a meta-analysis despite having fewer than 10 primary studies. Most of the reviews in gastroenterology journals (81.05%, 248/306), general internal medicine journals (75%, 9/12) and PROSPERO results (73.68%, 14/19) used a combination of methods to assess heterogeneity. There were 20 various definitions of levels of inconsistency (I2) throughout all the results. Random effects was the most common model choice in gastroenterology journals reviews (57.84%; 177/306), internal medicine journals (75%, 9/12) and for PROSPERO results (41.11%, 8/19). The majority of reviews did not discuss the impact of heterogeneity on results in the gastroenterology journals (62.09%, 190/306), only one study in the general internal medicine journals and only one study in the PROSPERO results.
Conclusion: In gastroenterology journals and other journals printing gastrointestinal topic systematic reviews, most conducted statistical tests for heterogeneity; however, the statistical methods could be more robust and the impact of heterogeneity discussed more often in the article.