Meta-analysis Using Flexible Random-effects Distribution Models.

Hisashi Noma,Kengo Nagashima,Shogo Kato,Satoshi Teramukai,Toshi A Furukawa

doi:10.2188/jea.je20200376

Abstract

BackgroundIn meta-analysis, the normal distribution assumption has been adopted in most systematic reviews of random-effects distribution models due to its computational and conceptual simplicity. However, this restrictive model assumption is possibly unsuitable and might have serious influences in practices.MethodsWe provide two examples of real-world evidence that clearly show that the normal distribution assumption is explicitly unsuitable. We propose new random-effects meta-analysis methods using five flexible random-effects distribution models that can flexibly regulate skewness, kurtosis and tailweight: skew normal distribution, skew t-distribution, asymmetric Subbotin distribution, Jones–Faddy distribution, and sinh–arcsinh distribution. We also developed a statistical package, flexmeta, that can easily perform these methods.ResultsUsing the flexible random-effects distribution models, the results of the two meta-analyses were markedly altered, potentially influencing the overall conclusions of these systematic reviews.ConclusionThe restrictive normal distribution assumption in the random-effects model can yield misleading conclusions. The proposed flexible methods can provide more precise conclusions in systematic reviews.

Highlights

In meta-analysis in medical studies, random-effects models have been the primary statistical tools for quantitative evaluation of treatment effects that account for between-studies heterogeneity.[1,2] Conventionally, the normal distribution assumption has been adopted in most systematic reviews due to its computational and conceptual simplicity.[2,3] the shape of the randomeffects distribution reflects how the treatment effects parameters are distributed in the target population, and are directly associated with the fundamental heterogeneity of treatment effects
We propose random-effects meta-analysis methods with flexible distribution models that can flexibly express skewness, kurtosis, and tailweight: (1) skew normal distribution,[14,15] (2) skew t-distribution,[14,16] (3) asymmetric Subbotin distribution,[14,17] (4) Jones–Faddy distribution,[18] and Address for correspondence
The flexible random-effects distribution models To address the restriction problem of the normal distribution, we propose random-effects meta-analysis methods using five flexible random-effects distribution models based on Bayesian methodology

Summary

Introduction

In meta-analysis in medical studies, random-effects models have been the primary statistical tools for quantitative evaluation of treatment effects that account for between-studies heterogeneity.[1,2] Conventionally, the normal distribution assumption has been adopted in most systematic reviews due to its computational and conceptual simplicity.[2,3] the shape of the randomeffects distribution reflects how the treatment effects parameters (eg, mean difference, log relative risk) are distributed in the target population, and are directly associated with the fundamental heterogeneity of treatment effects. In meta-analysis, the normal distribution assumption has been adopted in most systematic reviews of randomeffects distribution models due to its computational and conceptual simplicity. This restrictive model assumption is possibly unsuitable and might have serious influences in practices

Methods

Results

Conclusion