Selecting External Controls for Internal Cases Using Stratification Score Matching Methods

• Type of Material: Article
• Authors: Stefanie A Busgang, Icahn School of Medicine at Mount Sinai; Lance Waller, Emory University; Elena Colicino, Icahn School of Medicine at Mount Sinai; Ralph D'Agostino Jr, Wake Forest School of Medicine; Irva Hertz-Picciotto, University of California Davis; Chris Gennings, Icahn School of Medicine at Mount Sinai
• Language: English
• Date: 2022-03-01
• Publisher: MDPI
• Publication Version: Preprint (Prior to Peer Review)
• Copyright Statement: © 2022 by the authors.
• License: Creative Commons Attribution 4.0 International
• Final Published Version (URL): http://dx.doi.org/10.3390/ijerph19052549
• Title of Journal or Parent Work: INTERNATIONAL JOURNAL OF ENVIRONMENTAL RESEARCH AND PUBLIC HEALTH
• Volume: 19
• Issue: 5
• Grant/Funding Information: This work was supported by grants U2CES026555-01, U2CES026560-01, and P30ES023515 from the National Institute of Environmental Health Sciences; grants R01ES015359, UG3OD023365, UH3OD023365, P30ES023513, P01ES011269, and U54HD079125 from the National Institutes of Health; R829388 and R833292, RD83543201 from the United States Environmental Protection Agency Science to Achieve Results Grants; and grant P30CA012197 from the National Cancer Institute, and R01 ES032242 from the National Institute of Environmental Health Sciences.
• Keywords: Environmental Sciences & Ecology; BIAS; Public, Environmental & Occupational Health; registry-based; Life Sciences & Biomedicine; stratification score matching; external controls; HEART-FAILURE; PROPENSITY-SCORE; Environmental Sciences; Science & Technology
• Research Categories: Health Sciences, Public Health

Abstract

Rare-disease registries can be useful for studying the associations between environmental exposures and disease severity, but often require the addition of a healthy comparison control group. Defining a surrogate control group, matched and balanced on potentially confounding variables, would allow for the comparison of exposure distributions with cases from a registry. In the present study, we assess whether controls selected externally, using stratification score (SS) matching, can serve as effective proxies for internal controls. In addition, we use methyl paraben (MEPB) to compare the estimated associations between an externally matched sample and case–control frequencies in a cohort with internally matched controls. We started with 225 eligible cases of autism spectrum disorder (ASD) from Childhood Autism Risks from Genetics and the Environment (CHARGE), 241 internal controls from CHARGE, and 265 external controls from the National Health and Nutrition Examination Survey (NHANES) cycles 2005–2016. We calculated the SSs using demographic covariates and matched 1:1 using a caliper method without a replacement. The distribution of the covariates and the mean squared error of the paired differences (MSEpaired) in the SSs between the internal and external group were similar (MSEpaired = 0.007 and 0.011, respectively). The association between MEPB and ASD compared to the controls was similar between the externally matched SS pairs and the original frequency matched cohort. Controls selected externally, via SS matching, can provide a comparable bias reduction to that provided by the internal controls, and therefore may be a useful strategy in situations when the internal controls are not available.