Nonlinear modeling of alcohol consumption for analysis of beverage type effects and beverage preference effects.

In a previous report (Kimball and Friedman, Am J Epidemiol, 1992;135:1279-86), linear models for relating health outcomes to alcohol consumption were proposed for differentiating between beverage type effects and beverage preference effects. The models were applied to data relating serum high density lipoprotein cholesterol to alcohol consumption. In this report, those models are extended to the nonlinear case and are applied to data from the 1982 Maryland Hypertension Survey relating systolic blood pressure to alcohol consumption.