Characterizing exposure-outcome associations to identify biologically relevant cutoffs.

Determining cutoffs for dichotomization or multi-level categorization of continuous measures can be challenging and often results in arbitrarily defined values. In this article, we describe the application of using association curves determined from overlapping exposure bins to define relevant serological cutoffs. However, this method applies to most continuous measures in experimental and clinical contexts. The presented concepts are then illustrated using previously published data on viral serology (Epstein-Barr virus and cytomegalovirus) and multiple sclerosis.