Tests of Trend Between Disease Outcomes and Ordinal Covariates

• Activity Number: 43
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Naomi Brownstein*+
• Companies: The University of North Carolina
• Keywords: epidemiology methods ; regression ; ordinal predictors ; power

Abstract

Many epidemiological studies test trends when investigating the effect of a risk factor on a disease outcome. To reduce measurement error, a continuous exposure is commonly discretized. The effect of the discretized variable on the disease is modeled through logistic regression or survival analysis. A linear regression is then conducted by regressing the odds ratio or relative risk on the midpoint of the exposure interval. The trend test is then based on the slope of the regression line. In order to investigate the validity of this approach, we conducted simulation studies. We considered four approaches for the linear regression. The simulation studies show that the weighted estimator using the inverse of the estimated standard error as the weight is valid and more efficient than the others. The proposed method is applied to an epidemiological cohort study for illustration.