Analysis of Continuous Covariates and Dose-Effect Analysis

When analyzing data from an epidemiological study, some features are rather specific for a particular study design. Those are dealt with among others in chapters “Descriptive Studies,” “Cohort Studies,” “Modern Epidemiological Study Designs,” and “Survival Analysis” of this handbook. Other features are generally relevant; see chapters “Rates, Risks, Measures of Association,” and “Impact and Confounding and Interaction.” This chapter focuses on the analysis of continuous covariates where it will be discussed how such variables can be modeled to capture their potential association with an outcome of interest and to best describe the shape of such an association. We present classical methods based on categorization and subsequent contingency table analysis. The major part of the chapter, however, deals with the analysis of continuous covariates using regression models commonly used in epidemiology (see also chapter “Regression Methods for Epidemiological Analysis” of this handbook). Each of the proposed techniques to model continuous covariates is illustrated by a real data example taken from a case-control study on laryngeal cancer and smoking as well as alcohol consumption that has been conducted in Germany during the 1990s. The chapter ends with practical recommendations and conclusions.