Analysis of Dose–Response Relationship Based on Categorical Outcomes

Abstract

In an eloquent article prepared in defense of the dichotomy, Lewis (2004) wrote that one of the most important ways in which we learned to understand the world was to describe complicated phenomena using simple categories. Thus, it is hardly surprising that medical researchers often seek to categorize data in their attempt to make sense of unfamiliar measurement scales and treatment effects of uncertain implication. For this reason, threshold values based on continuous measurements are frequently used to help guide the decision to initiate medical interventions. Examples include a diastolic blood pressure greater than 90, a fasting cholesterol level higher than 200, and a CD4 count lower than 200. Normal ranges were constructed to screen subjects for possible lab abnormalities. Even though this black-and-white dichotomy appears to be crude in many situations, its simplicity helps human minds make decisions, decisions that are often binary in nature.