Dataset Information

Sensitivity Analysis for Unmeasured Confounding in Meta-Analyses.

ABSTRACT: Random-effects meta-analyses of observational studies can produce biased estimates if the synthesized studies are subject to unmeasured confounding. We propose sensitivity analyses quantifying the extent to which unmeasured confounding of specified magnitude could reduce to below a certain threshold the proportion of true effect sizes that are scientifically meaningful. We also develop converse methods to estimate the strength of confounding capable of reducing the proportion of scientifically meaningful true effects to below a chosen threshold. These methods apply when a "bias factor" is assumed to be normally distributed across studies or is assessed across a range of fixed values. Our estimators are derived using recently proposed sharp bounds on confounding bias within a single study that do not make assumptions regarding the unmeasured confounders themselves or the functional form of their relationships with the exposure and outcome of interest. We provide an R package, EValue, and a free website that compute point estimates and inference and produce plots for conducting such sensitivity analyses. These methods facilitate principled use of random-effects meta-analyses of observational studies to assess the strength of causal evidence for a hypothesis.

SUBMITTER: Mathur MB

PROVIDER: S-EPMC7518377 | biostudies-literature | 2020

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Sensitivity Analysis for Unmeasured Confounding in Meta-Analyses.

Mathur Maya B MB VanderWeele Tyler J TJ

Journal of the American Statistical Association 20190430 529

Random-effects meta-analyses of observational studies can produce biased estimates if the synthesized studies are subject to unmeasured confounding. We propose sensitivity analyses quantifying the extent to which unmeasured confounding of specified magnitude could reduce to below a certain threshold the proportion of true effect sizes that are scientifically meaningful. We also develop converse methods to estimate the strength of confounding capable of reducing the proportion of scientifically m ...[more]

PMID: 32981992

Similar Datasets

Project description:BackgroundAlthough it has been established that low socioeconomic status is linked to increased risk of death after stroke, the mechanisms behind this link are still unclear. In this study we aim to shed light on the relationship between income level and survival after stroke by investigating the extent to which differences in stroke severity account for differences in survival.MethodsThe study was based on patients registered in Riksstroke (the Swedish stroke register) with first time ischemic stroke (n = 51,159) or intracerebral hemorrhage (n = 6777) in 2009-2012. We used causal mediation analysis to decompose the effect of low income on 3-month case fatality into a direct effect and an indirect effect due to stroke severity. Since causal mediation analysis relies on strong assumptions regarding residual confounding of the relationships involved, recently developed methods for sensitivity analysis were used to assess the robustness of the results to unobserved confounding.ResultsAfter adjustment for observed confounders, patients in the lowest income tertile had a 3.2% (95% CI: 0.9-5.4%) increased absolute risk of 3-month case fatality after intracerebral hemorrhage compared to patients in the two highest tertiles. The corresponding increase for case fatality after ischemic stroke was 1% (0.4-1.5%). The indirect effect of low income, mediated by stroke severity, was 1.8% (0.7-2.9%) for intracerebral hemorrhage and 0.4% (0.2-0.6%) for ischemic stroke. Unobserved confounders affecting the risk of low income, more severe stroke and case fatality in the same directions could explain the indirect effect, but additional adjustment to observed confounders did not alter the conclusions.ConclusionsThis study provides evidence that as much as half of income-related inequalities in stroke case fatality is mediated through differences in stroke severity. Targeting stroke severity could therefore lead to a substantial reduction in inequalities and should be prioritized. Sensitivity analysis suggests that additional adjustment for a confounder of greater impact than age would be required to considerably alter our conclusions.

Dataset Information

Sensitivity Analysis for Unmeasured Confounding in Meta-Analyses.

Publications

Sensitivity Analysis for Unmeasured Confounding in Meta-Analyses.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets