Analysis of ROC Curves Applied to the Diagnosis of Lyme Disease

Abstract

I in the analysis of receiver operating I acteristic (ROC) curves is an ongoing activity in our education of residents in laboratory medicine; therefore, I was interested to see the data presented by Sivak et al 1 concerning the accuracy of IgM immunoblotting in confirming the clinical diagnosis of early Lyme disease. Areas under ROC curves can readily be calculated using spreadsheet methodologies and elementary principles of integral calculus 2 that date to the time of Newton and Leibniz. Many residents in pathology and laboratory medicine have difficulty with numerical images, as opposed to cellular images, and do not easily grasp the concepts of these area calculations. To assist them in appreciating the principle of area under the ROC curve, my colleagues and I developed an empirical technique for such measurements that is extremely accurate (error rate, ±