A new measure of between‐studies heterogeneity in meta‐analysis

Abstract

Assessing the magnitude of heterogeneity in a meta-analysis is important for determining the appropriateness of combining results. The most popular measure of heterogeneity, I(2) , was derived under an assumption of homogeneity of the within-study variances, which is almost never true, and the alternative estimator, R^I, uses the harmonic mean to estimate the average of the within-study variances, which may also lead to bias. This paper thus presents a new measure for quantifying the extent to which the variance of the pooled random-effects estimator is due to between-studies variation, R^b, that overcomes the limitations of the previous approach. We show that this measure estimates the expected value of the proportion of total variance due to between-studies variation and we present its point and interval estimators. The performance of all three heterogeneity measures is evaluated in an extensive simulation study. A negative bias for R^b was observed when the number of studies was very small and became negligible as the number of studies increased, while R^I and I(2) showed a tendency to overestimate the impact of heterogeneity. The coverage of confidence intervals based upon R^b was good across different simulation scenarios but was substantially lower for R^I and I(2) , especially for high values of heterogeneity and when a large number of studies were included in the meta-analysis. The proposed measure is implemented in a user-friendly function available for routine use in r and sas. R^b will be useful in quantifying the magnitude of heterogeneity in meta-analysis and should supplement the p-value for the test of heterogeneity obtained from the Q test. Copyright © 2016 John Wiley & Sons, Ltd.

RETRACTED ARTICLE: Could Intensive Blood Pressure Control Really Reduce Diabetic Retinopathy Outcomes? Evidence from Meta-Analysis and Trial Sequential Analysis from Randomized Controlled Trials

“…We assessed the within- and between-study variation or heterogeneity by testing Cochran’s Q statistic [ 21 , 22 ]. Heterogeneity was quantified with the I 2 metric, which was independent of the number of studies in the systematic review [ 23 ]. The pooled OR was estimated using fixed effects (FE, Mantel and Haenszel) and random effects (RE, DerSimonian and Laird) models.…”

Living or deceased organ donors in liver transplantation for hepatocellular carcinoma: a systematic review and meta-analysis

“…As demonstrated by Cochrane handbook, an I 2 <40% likely indicates mild heterogeneity and I 2 40% indicates moderate or substantial heterogeneity. 31 We employed fixed effect model when there was mild heterogeneity, otherwise, the random effect model was employed.…”

Dietary interventions to prevent and manage diabetes in worksite settings: a meta-analysis

“…Study-specific effect sizes were pooled using random-effects models 42) (%metaanal macro available at ). Cochran's Q-statistics, R B , a variant of I 2 with better statistical properties 43 , 44) , and the coefficient of variation between studies were used to assess the between-study heterogeneity. Meta-regression 45) (%metareg macro available at ) was used to explore the potential sources of heterogeneity, including the study design (pre-post/randomized controlled trial), location (US/Europe/other), frequency of intervention (per month), duration of intervention (in months), eligibility criteria (mixed/high risk/diabetic), size of worksite (in number of employees), mean age (in years), female participants (%), intervention level (environment/individual), and intervention unit (group/individual).…”