A general framework for the use of logistic regression models in meta-analysis

Where individual participant data are available for every randomised trial in a meta-analysis of dichotomous event outcomes, “one-stage” random-effects logistic regression models have been proposed as a way to analyse these data. Such models can also be used even when individual participant data are not available and we have only summary contingency table data. One benefit of this one-stage regression model over conventional meta-analysis methods is that it maximises the correct binomial likelihood for the data and so does not require the common assumption that effect estimates are normally distributed. A second benefit of using this model is that it may be applied, with only minor modification, in a range of meta-analytic scenarios, including meta-regression, network meta-analyses and meta-analyses of diagnostic test accuracy. This single model can potentially replace the variety of often complex methods used in these areas. This paper considers, with a range of meta-analysis examples, how random-effects logistic regression models may be used in a number of different types of meta-analyses. This one-stage approach is compared with widely used meta-analysis methods including Bayesian network meta-analysis and the bivariate and hierarchical summary receiver operating characteristic (ROC) models for meta-analyses of diagnostic test accuracy.

Download Full-text

Retrieval of Individual Participant Data for Exercise Meta-Analyses May Not Be Worth the Time and Effort

BioMed Research International ◽

10.1155/2016/5059041 ◽

2016 ◽

Vol 2016 ◽

pp. 1-5 ◽

Cited By ~ 3

Author(s):

George A. Kelley ◽

Kristi S. Kelley

Keyword(s):

Logistic Regression ◽

Rheumatic Diseases ◽

Gold Standard ◽

Odds Ratio ◽

Meta Analysis ◽

Individual Participant Data ◽

Individual Participant ◽

Using Data ◽

Meta Analyses

Purpose. While individual participant data (IPD) meta-analyses are considered the gold standard for meta-analysis, the feasibility of obtaining IPD may be problematic.Methods. Using data from a previous meta-analysis of 29 studies on exercise in adults with arthritis and other rheumatic diseases, the percentage of studies in which useable IPD was provided was calculated.Results. Eight of 29 authors (28%, 95% CI = 11% to 44%) provided IPD. Using logistic regression, neither year of publication (odds ratio = 1.05, 95% CI = 0.90 to 1.27,p=0.58) nor country (odds ratio = 1.36, 95% CI = 0.20 to 10.9,p=1.00) was significantly associated with the obtainment of IPD.Conclusions. The retrieval of IPD for exercise meta-analyses may not be worth the time and effort. However, further research is needed before any final recommendations can be made.

Download Full-text

Estimating interactions in individual participant data meta-analysis. A comparison of methods in practice.

10.21203/rs.3.rs-317690/v1 ◽

2021 ◽

Author(s):

Ruth Walker ◽

Lesley Stewart ◽

Mark Simmonds

Keyword(s):

Random Effects ◽

Meta Analysis ◽

Individual Preference ◽

Detailed Examination ◽

International Studies ◽

Individual Participant Data ◽

Stage Models ◽

One Stage ◽

Interaction Terms ◽

Individual Participant

Abstract Medical interventions may be more effective in some types of individuals than others and identifying characteristics that modify the effectiveness of an intervention is a cornerstone of precision or stratified medicine. The opportunity for detailed examination of treatment-covariate interactions can be an important driver for undertaking an individual participant data (IPD) meta-analysis, rather than a meta-analysis using aggregate data. A number of recent modelling approaches are available. We apply these methods to the Perinatal Antiplatelet Review of International Studies (PARIS) Collaboration IPD dataset and compare estimates between them. We discuss the practical implications of applying these methods, which may be of interest to aid meta-analysists in the use of these, often complex models. Models compared included the two-stage meta-analysis of interaction terms and one-stage models which fit multiple random effects and separate within and between trial information. Models were fitted for nine covariates and five binary outcomes and results compared. Interaction terms produced by the methods were generally consistent. We show that where data are sparse and there is low heterogeneity in the covariate distributions across trials, the meta-analysis of interactions may produce unstable estimates and have issues with convergence. In this IPD dataset, varying assumptions by using multiple random effects in one-stage models or using only within trial information made little difference to the estimates of treatment-covariate interaction. Method choice will depend on datasets characteristics and individual preference.

Download Full-text

Abstract P343: Comparison of Two-stage and One-stage Meta-analyses: An Example of eGFR-Cardiovascular Mortality Association (for CKD-PC collaborators)

Circulation ◽

10.1161/circ.125.suppl_10.ap343 ◽

2012 ◽

Vol 125 (suppl_10) ◽

Author(s):

Yingying Sang ◽

Kunihiro Matsushita ◽

Bakhtawar K Mahmoodi ◽

Brad C Astor ◽

Josef Coresh ◽

...

Keyword(s):

Random Effects ◽

Cardiovascular Mortality ◽

Cox Model ◽

Meta Analysis ◽

Fixed Effect ◽

Two Stage ◽

Statistical Estimates ◽

One Stage ◽

Individual Participant ◽

Meta Analyses

Purpose: Individual participant data (IPD) meta-analysis provides precise statistical estimates. Two approaches to meta-analyze IPD are currently used. The 1-stage approach fits a regression model to a pooled dataset including all studies. The 2-stage approach fits models in individual studies and meta-analyzes the estimates. However, their comparability has not been well described. We compare these methods for eGFR-cardiovascular mortality association in the CKD Prognosis Consortium (CKD-PC). Methods: For the 1-stage method, we fitted a Cox stratified model, allowing each study to have a unique baseline hazard but assuming a common hazard ratio (HR) for eGFR across studies. For the 2-stage method, we first fitted a Cox model in each study, and then meta-analyzed HRs using a fixed-effect (assuming one true HR for eGFR across studies) and a random-effects (allowing some variance of true HR) model. eGFR was fitted as linear splines in all models. Results: In a sample of 18 of 46 cohorts joining CKD-PC (191,276 participants and 8,732 cardiovascular deaths [CHD, stroke, or heart failure]), these methods gave nearly identical estimates except for the width of the confidence intervals ( Figure ). The 95% CIs were wider as methods made fewer assumptions - narrowest for 1-stage method, slightly wider for the fixed-effect 2-stage and wider for the random-effects 2-stage method. Conclusion: The two-stage and one-stage meta-analyses provided nearly identical estimates for the eGFR-cardiovascular mortality relationship. The random-effects 2-stage method will provide conservative estimates with wider 95% CIs but this is necessary in the presence of heterogeneity.

Download Full-text

Arm Based on LEg blood pressures (ABLE-BP): can systolic leg blood pressure measurements predict systolic brachial blood pressure? Protocol for an individual participant data meta-analysis from the INTERPRESS-IPD Collaboration

BMJ Open ◽

10.1136/bmjopen-2020-040481 ◽

2021 ◽

Vol 11 (3) ◽

pp. e040481

Author(s):

Sinead T J McDonagh ◽

James P Sheppard ◽

Fiona C Warren ◽

Kate Boddy ◽

Leon Farmer ◽

...

Keyword(s):

Blood Pressure ◽

Public Involvement ◽

Meta Analysis ◽

Patient And Public Involvement ◽

Prognostic Models ◽

Individual Participant Data ◽

Dissemination Strategy ◽

Cross Sectional ◽

Individual Participant ◽

Meta Analyses

IntroductionBlood pressure (BP) is normally measured on the upper arm, and guidelines for the diagnosis and treatment of high BP are based on such measurements. Leg BP measurement can be an alternative when brachial BP measurement is impractical, due to injury or disability. Limited data exist to guide interpretation of leg BP values for hypertension management; study-level systematic review findings suggest that systolic BP (SBP) is 17 mm Hg higher in the leg than the arm. However, uncertainty remains about the applicability of this figure in clinical practice due to substantial heterogeneity.AimsTo examine the relationship between arm and leg SBP, develop and validate a multivariable model predicting arm SBP from leg SBP and investigate the prognostic association between leg SBP and cardiovascular disease and mortality.Methods and analysisIndividual participant data (IPD) meta-analyses using arm and leg SBP measurements for 33 710 individuals from 14 studies within the Inter-arm blood pressure difference IPD (INTERPRESS-IPD) Collaboration. We will explore cross-sectional relationships between arm and leg SBP using hierarchical linear regression with participants nested by study, in multivariable models. Prognostic models will be derived for all-cause and cardiovascular mortality and cardiovascular events.Ethics and disseminationData originate from studies with prior ethical approval and consent, and data sharing agreements are in place—no further approvals are required to undertake the secondary analyses proposed in this protocol. Findings will be published in peer-reviewed journal articles and presented at conferences. A comprehensive dissemination strategy is in place, integrated with patient and public involvement.PROSPERO registration numberCRD42015031227.

Download Full-text

Deriving percentage study weights in multi-parameter meta-analysis models: with application to meta-regression, network meta-analysis and one-stage individual participant data models

Statistical Methods in Medical Research ◽

10.1177/0962280216688033 ◽

2017 ◽

Vol 27 (10) ◽

pp. 2885-2905 ◽

Cited By ~ 8

Author(s):

Richard D Riley ◽

Joie Ensor ◽

Dan Jackson ◽

Danielle L Burke

Keyword(s):

Meta Analysis ◽

Information Matrix ◽

Indirect Evidence ◽

Parameter Estimates ◽

Individual Participant Data ◽

One Stage ◽

Individual Participant ◽

Meta Regression ◽

Multiple Treatments ◽

Analysis Models

Many meta-analysis models contain multiple parameters, for example due to multiple outcomes, multiple treatments or multiple regression coefficients. In particular, meta-regression models may contain multiple study-level covariates, and one-stage individual participant data meta-analysis models may contain multiple patient-level covariates and interactions. Here, we propose how to derive percentage study weights for such situations, in order to reveal the (otherwise hidden) contribution of each study toward the parameter estimates of interest. We assume that studies are independent, and utilise a decomposition of Fisher’s information matrix to decompose the total variance matrix of parameter estimates into study-specific contributions, from which percentage weights are derived. This approach generalises how percentage weights are calculated in a traditional, single parameter meta-analysis model. Application is made to one- and two-stage individual participant data meta-analyses, meta-regression and network (multivariate) meta-analysis of multiple treatments. These reveal percentage study weights toward clinically important estimates, such as summary treatment effects and treatment-covariate interactions, and are especially useful when some studies are potential outliers or at high risk of bias. We also derive percentage study weights toward methodologically interesting measures, such as the magnitude of ecological bias (difference between within-study and across-study associations) and the amount of inconsistency (difference between direct and indirect evidence in a network meta-analysis).

Download Full-text

An individual participant data (IPD) meta-analysis on the attachment network to multiple caregivers

10.31234/osf.io/8w5a3 ◽

2020 ◽

Author(s):

Or Dagan ◽

Pasco Fearon ◽

Carlo Schuengel ◽

Marije Verhage ◽

Glenn I. Roisman ◽

...

Keyword(s):

Meta Analysis ◽

Developmental Outcomes ◽

Research Synthesis ◽

Individual Participant Data ◽

Attachment Relationship ◽

Attachment Research ◽

Individual Participant ◽

The Subject ◽

Meta Analyses

Since the seminal 1992 paper by van IJzendoorn, Sagi, and Lambermon, putting forward the “The multiple caretaker paradox”, relatively little attention has been given to the potential joint effects of the role early attachment network to mother and father play in development. Recently, Dagan and Sagi-Schwartz (2018) have published a paper that attempts to revive this unsettled issue, calling for research on the subject and offering a framework for posing attachment network hypotheses. This Collaboration for Attachment Research Synthesis project attempts to use an Individual Participant Data meta-analyses to test the hypotheses put forward in Dagan and Sagi-Schwartz (2018). Specifically, we test (a) whether the number of secure attachments (0,1, or 2) matter in predicting a range of developmental outcomes, and (b) whether the quality of attachment relationship with one parent contributes more than the other to these outcomes.

Download Full-text