SUMMARY: Evaluation of continuous representations of exposure is often of interest in epidemiologic research to delineate possible causal relationships. These representations are usually highly correlated, so a method to compare their independent contributions to development of disease must be able to control for these correlations. A method that combines categorization of levels of exposure to facilitate interpretation and polynomial functions to achieve maximum levels of control is described. Its utility is demonstrated by examining results from a simulation and by analyzing an epidemiological data set.

KEY WORDS: case-control study; etiologic fraction; lung cancer; polynomial; simulation; smoking