Implications of M Bias in Epidemiologic Studies: A Simulation Study
Authors:
- M. Alan BrookhartMBM. Alan Brookhart
- Sebastian SchneeweissSSSebastian Schneeweiss
- Xiaojuan MiXMXiaojuan Mi
- Soko SetoguchiSSSoko Setoguchi
Open Access
- 25 October 2012
- journal article
- research article
- Published by Oxford University Press (OUP) in American Journal of Epidemiology
- Vol. 176 (10) , 938-948
Abstract
Collider-stratification bias arises from conditioning on a variable (collider) which opens a path from exposure to outcome. M bias occurs when the collider-stratification bias is transmitted through ancestors of exposure and outcome. Previous theoretical work, but not empirical data, has demonstrated that M bias is smaller than confounding bias. The authors simulated data for large cohort studies with binary exposure, an outcome, a collider, and 2 predictors of the collider. They created 178 scenarios by changing the frequencies of variables and/or the magnitudes of associations among the variables. They calculated the effect estimate, percentage bias, and mean squared error. M bias in these realistic scenarios ranged from −2% to −5%. When the authors increased one or both relative risks for the relation between the collider and unmeasured factors to ≥8, the negative bias was more substantial (>15%). The result was substantially biased (e.g., >20%) if an unmeasured confounder that was also a collider was not adjusted to avoid M bias. In scenarios resembling those the authors examined, M bias had a small impact unless associations between the collider and unmeasured confounders were very large (relative risk > 8). When a collider is itself an important confounder, controlling for confounding would take precedence over avoiding M bias.Keywords
This publication has 34 references indexed in Scilit:
- Confounding Control in Healthcare Database ResearchMedical Care, 2010
- Illustrating bias due to conditioning on a colliderInternational Journal of Epidemiology, 2009
- High-dimensional Propensity Score Adjustment in Studies of Treatment Effects Using Health Care Claims DataEpidemiology, 2009
- Long-term persistence with statin treatment in a not-for-profit health maintenance organization: A population-based retrospective cohort study in IsraelClinical Therapeutics, 2008
- Use of antidepressants and risk of lung cancerCancer Causes & Control, 2007
- The design of simulation studies in medical statisticsStatistics in Medicine, 2006
- Variable Selection for Propensity Score ModelsAmerican Journal of Epidemiology, 2006
- National Trends in Statin Use by Coronary Heart Disease Risk CategoryPLOS Medicine, 2005
- The Risk of Cancer in Users of StatinsJournal of Clinical Oncology, 2004
- Serum Total Cholesterol Concentrations and Awareness, Treatment, and Control of Hypercholesterolemia Among US AdultsCirculation, 2003