Tumor markers in lung cancer: Does the method of obtaining the cut-off point and reference population influence diagnostic yield?

Objectives: The aim of this study was to evaluate the clinical usefulness o f the tumor markers CA125, CEA, NSE, SCC, and TPS in a group of patients wi th lung cancer. We estimated the influence of the method for choosing the c ut-off point and of considering as a reference population either healthy co ntrols or patients with some form of non-neoplastic pulmonary disease (NNPD ). Design and methods: The tumor markers were determined using enzyme immunoas say techniques, and their diagnostic yield was evaluated using ROC curves a nd their correlation with the percentages between false and true positives. The diagnostic parameters of the tumor markers are presented in 116 patien ts with lung cancer and compared with a group of 25 healthy controls and an other group of 80 patients with some form of NNPD. We determined on the one hand the cut-off points resulting from the best sensitivity-specificity ba lance in the ROC curves and on the other those resulting from considering a specificity of 95%. With the two cut-offs we studied the different diagnos tic parameters: sensitivity, specificity and accuracy or area below the ROC curve. Results: Optimum diagnostic yield is obtained when we choose the cut-off po int determined by the best balance of sensitivity and specificity in the RO C curves and take a healthy population as a reference group. The cut-off va lues for CA125, CEA, NSE, SCC, and TPS were 24 U/mL, 2.8 ng/mL, 9.8 ng/mL, 1.6 ng/mL, and 67.8 U/L, respectively. Conclusions: Our results suggest that in future studies on tumor markers, a group of healthy subjects should be used as a reference population and ROC curves should be used to obtain the optimum cut-offs. Copyright (C) 1999 T he Canadian Society of Clinical Chemists.