The Quasi-cohort Approach in Pharmacoepidemiology:... : Epidemiology (original) (raw)
Pharmacoepidemiology
Upgrading the Nested Case–Control
From the aDepartment of Epidemiology and Biostatistics, McGill University, Montreal, Quebec, Canada; and bMcGill Pharmacoepidemiology Research Unit, Centre for Clinical Epidemiology, Jewish General Hospital, Montreal, Quebec, Canada.
Submitted 05 May 2014; accepted 19 July 2014; posted 17 December 2014.
Funding: This research was funded in part by a grant from the Canadian Institutes of Health Research (CIHR) and the Canadian Foundation for Innovation (CFI). The author is the recipient of the James McGill Chair award.
Correspondence: Samy Suissa, Centre for Clinical Epidemiology, Jewish General Hospital, 3755 Cote Ste-Catherine, H4.61, Montreal, Quebec, Canada H3T 1E2. E-mail: [email protected].
Abstract
Observational studies of drug effects conducted using health care mega-databases often involve large cohorts with multiple time-varying exposures and covariates. These present formidable technical challenges in data analysis, necessitating sampling approaches such as nested case–control designs. The nested case–control approach is, however, baffling to medical journal readers, particularly the comparisons involving “cases” versus “controls” and the convoluted way in which forward-looking relations from exposure to outcome are extracted from backward-looking data. I propose a “quasi-cohort” approach involving alternative ways of data presentation and analysis that are more in line with the underlying cohort design, including the computation of quasi-rates, rate ratios, and quasi-rate differences. I illustrate the quasi-cohort approach using data from a study of pneumonia risk associated with inhaled corticosteroid use in a cohort of 163,514 patients with chronic obstructive pulmonary disease, including 20,344 who had the outcome event of pneumonia hospitalization during more than 304 million person-days of follow-up.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.