Logistic regression is commonly used to estimate an adjusted relative risk (aRR) in cohort studies by calculating an adjusted odds ratio (aOR). This approach, however, usually overestimates the aRR, sometimes dramatically, when incidence of disease is greater than 10%. A method to correct the aOR from logistic regression in cohort studies of common outcomes (JAMA 1998;280(9):1690-1) was proposed recently and is becoming widely adopted by epidemiologists and medical researchers. We compared several methods for estimating an aRR in cohort studies of common outcomes, including the suggested correction of aOR from logistic regression, stratified analysis, and poisson regression. The advantages and disadvantages related to each potential approach will be illustrated. For example, the aOR correction method produces a correct estimate of the RR when no confounding occurs, however, it does not provide a correct point estimate and confidence interval (CI) of an aRR when confounding exists. Also, an effect modification may exist on the RR scale, but is not apparent on the odds ratio scale. Stratified analysis is simple, however, potential confounders must be categorical. Poisson regression provides a correct point estimate of aRR, however, in some cases provides a CI wider than the specified (1-alpha)x100% level. Our study proposes using the bootstrap method to obtain an estimate of the CI for an aRR estimated from poisson regression. Examples are used to illustrate this method.
Learning Objectives: At the conclusion of the session, the participant (learner) in this session will be able to: 1. list the advantages and disadvantages of methods used to estimate the adjusted relative risk in cohort studies of common outcomes; 2. choose appropriate methods to estimate the adjusted relative risk and its confidence interval in cohort studies of common outcomes.
Keywords: Epidemiology, Methodology
Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: None
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.