Meta-analysis and use of tests of heterogeneity in neurosurgery

Abstract

The recent increase in implementation of evidence-based medicine in neurosurgery has led to an increase in awareness of the importance of meta-analysis. An integral component of meta-analysis is the test of heterogeneity. This test examines whether the apparent differences between the studies are significant enough to bias the outcome and conclusion of the meta-analysis. The author has examined four different tests of heterogeneity available in the scientific literature for binary data. In the context of neurosurgical data, the author found that Pearson's test was the most accurate in terms of Type I and Type II errors, as well as "goodness-of-fit" between the empirical distribution and approximate chi-squared distribution. Moreover, its ease of computation made this test a highly favorable test to be used in neurosurgical data analysis.