Receiver operating characteristic curve analysis in the prediction of carpal tunnel syndrome: a model for reporting...

Receiver operating characteristic (ROC) curves were used to predict the risk of carpal tunnel syndrome (CTS). Patients were classified clinically as: (1) normal exam and no symptoms (169 hands); (2) having a motor and/or sensory deficit typical of CTS (115 hands); (3) having a history characteristic of CTS (156 hands); and (4) nondiagnostic symptomatology (122 hands). Electrophysiological studies consisted of median and ulnar motor, sensory, and palmar measurements. Group mean values for group 1 differed significantly from groups 2 and 3 (not 4) for all measurements, but values overlapped considerably. Median distal motor latency (DMML) combined with median-ulnar palmar latency differences (MUPLD) had significantly superior discriminant power than other measurements and correlated highly for all groups (r values = 0.71-0.73). These variables were used to construct ROC curves and prediction tables. The approach used allows one to assign a percentage risk of having a CTS and can be used in outcome studies.