Meta-analysis of binary outcomes via generalized linear mixed models: a simulation study

Bakbergenuly, Ilyas; Kulinskaya, Elena

doi:10.1186/s12874-018-0531-9

Cited by 56 publications

(52 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…11,51,52 While these have theoretical and inferential advantages, they present considerable computational problems, and simulations show improvements only in the presence of small and sparse data. 53,54 An additional requirement of the two-stage procedure, when applied in multi-parameter meta-analyses, is the knowledge of the within-study covariances. Methods for estimating them from published data were developed in multivariate meta-analysis, 5,31,55 and in dose-response meta-analysis, and can be applied in this general model.…”

Section: Discussionmentioning

confidence: 99%

An extended mixed‐effects framework for meta‐analysis

Sera

Armstrong

Blangiardo

et al. 2019

Statistics in Medicine

197

144

View full text Add to dashboard Cite

Standard methods for meta‐analysis are limited to pooling tasks in which a single effect size is estimated from a set of independent studies. However, this setting can be too restrictive for modern meta‐analytical applications. In this contribution, we illustrate a general framework for meta‐analysis based on linear mixed‐effects models, where potentially complex patterns of effect sizes are modeled through an extended and flexible structure of fixed and random terms. This definition includes, as special cases, a variety of meta‐analytical models that have been separately proposed in the literature, such as multivariate, network, multilevel, dose‐response, and longitudinal meta‐analysis and meta‐regression. The availability of a unified framework for meta‐analysis, complemented with the implementation in a freely available and fully documented software, will provide researchers with a flexible tool for addressing nonstandard pooling problems.

show abstract

Section: Discussionmentioning

confidence: 99%

An extended mixed‐effects framework for meta‐analysis

Sera

Armstrong

Blangiardo

et al. 2019

Statistics in Medicine

197

144

View full text Add to dashboard Cite

show abstract

“…While a forest plot was generated for each strata of interest, results were not pooled if there was substantial methodological or clinical heterogeneity in either the design or populations of the studies or if there is substantial statistical heterogeneity. As the more common I 2 statistic is not available for GLMMs,13 statistical heterogeneity was assessed using τ 2 , a measure of interstudy heterogeneity 14. The decision to pool was made based on an assessment of clinic heterogeneity and a τ 2 <4.…”

Section: Methodsmentioning

confidence: 99%

Tuberculosis case fatality in India: a systematic review and meta-analysis

et al. 2020

View full text Add to dashboard Cite

IntroductionThe WHO End TB Strategy calls for a global reduction in the case fatality ratio (CFR) below 5%. India accounts for a third of global tuberculosis (TB) deaths. This systematic review estimated CFRs among Indian patients with TB both during and after treatment.MethodsWe systematically searched Medline, Embase and Global Health for eligible studies published between 1 January 2006 and 8 January 2019, including both cohort studies and intervention study control arms that followed Indian patients with TB for fatality either during treatment or post-treatment. From relevant studies we extracted CFRs in addition to study demographics. Study quality was assessed using modified Scottish Intercollegiate Guidelines Network cohort criteria. Sufficiently homogenous studies were pooled using a random effect generalised linear mixed model. A meta-regression was performed to associate study characteristics with resulting CFRs.Results218 relevant studies were identified, of which 211 provided treatment phase CFRs. Most patients (92.4%) were treated in the public sector. Quality concerns were identified in 74% of papers. We estimated a pooled treatment phase CFR of 5.16% (95% CI 4.20% to 6.34%) which fell to 3.78% (2.77% to 5.16%) when restricted to 52 high-quality studies. Treatment phase CFRs were higher for paediatric (n=27, 6.50% (2.65% to 10.36%)), drug-resistant (n=43, 14.06% (10.15% to 19.49%)) and HIV-infected (n=35, 10.91% (7.68% to 15.50%)) patients. Nineteen post-treatment CFR studies were too heterogeneous to pool except when restricting to three high-quality studies (2.69% (–0.79% to 6.18%)). Poor study quality (OR=2.27 (2.01 to 2.57)) and tertiary centres patients (OR=1.15 (1.03 to 1.28)) were significantly associated with increased treatment phase case fatality.ConclusionsCase fatality is a critical measure of the quality of TB care. While India’s treatment CFRs are in line with WHO targets, several key patient groups remain understudied and most studies suffer from methodological issues. Increased high-quality reporting on patient outcomes will help improve the evidence base on this topic.

show abstract

“…This approximation is valid when the total number of events is small relative to the group sizes. In that case the non-central hypergeometric distribution can be approximated by a binomial distribution 37 . We will refer to this approach as NCH-NMA.…”

Section: Software For Fitting Mh-nmamentioning

confidence: 99%

Network meta‐analysis of rare events using the Mantel‐Haenszel method

Efthimiou

Rücker

Schwarzer

et al. 2019

Statistics in Medicine

View full text Add to dashboard Cite

The Mantel-Haenszel method has been used for decades to synthesize data obtained from studies that compare two interventions with respect to a binary outcome. It has been shown to perform better than the inverse-variance method or Peto's odds ratio when data is sparse. Network meta-analysis (NMA) is increasingly used to compare the safety of medical interventions, synthesising for example data on mortality or serious adverse events. In this setting, sparse data occur often and yet there is to-date no extension of the Mantel-Haenszel method for the case of NMA. In this paper we fill this gap by presenting a Mantel-Haenszel NMA method for odds ratios. Similarly to the pairwise Mantel-Haenszel method, we assume common treatment effects. We implement our approach in R, and we provide freely available, easy-to-use routines. We illustrate our approach using data from two previously published networks. We compare our results to those obtained from three other approaches to NMA: NMA with non-central hypergeometric likelihood, an inverse-variance NMA and a Bayesian NMA with a binomial likelihood. We also perform simulations to assess the performance of our method and compare it with alternative methods. We conclude that our Mantel-Haenszel NMA method offers a reliable approach to the network meta-analysis of binary outcomes, especially in the case or sparse data, and when the assumption of methodological and clinical homogeneity is justifiable.

show abstract

Meta-analysis of binary outcomes via generalized linear mixed models: a simulation study

Cited by 56 publications

References 39 publications

An extended mixed‐effects framework for meta‐analysis

An extended mixed‐effects framework for meta‐analysis

Tuberculosis case fatality in India: a systematic review and meta-analysis

Network meta‐analysis of rare events using the Mantel‐Haenszel method

Contact Info

Product

Resources

About