Abstract
The aim of this study was to evaluate the diagnostic value of three tumour markers, squamous cell carcinoma (SCC) antigen, carcinoembryonic antigen (CEA) and CYFRA 21.1, in lung cancer using a Bayesian analysis to obtain the predictive values for different pretest probabilities or prevalences. A cross-sectional study included 94 patients with lung cancer, 40 with benign lung disease, and 40 healthy controls. SCC antigen and CEA were measured in blood samples by microparticle enzyme immunoassay (MEIA), and CYFRA by enzyme-linked immunosorbent assay (ELISA). The results of tumour marker determinations were expressed as percentiles, and showed significantly higher levels in the cancer group than in the two control groups. Taking the 95th percentile of benign lung diseases as the cut-off point (specificity 95%), the following sensitivities were found: SCC 41%, CEA 31% and CYFRA 79%. After a Bayesian analysis, the best results for the three tumour markers were found in prevalences of 30-40%. The highest incremental gain was obtained by CYFRA (at prevalence of 36%, positive and negative predictive value approximately 90%). The three tumour markers were included in a stepwise regression analysis to predict lung cancer, and CYFRA was the only selected variable. We conclude that CYFRA 21.1 may be a useful marker in lung cancer when there is an intermediate pretest probability of disease.
