ABSTRACT
Background: Combining multiple tumor markers increases sensitivity for lung cancer diagnosis in the cost of false positive. However, some would like to check as many as tumor markers in the fear of missing cancer. We though to propose a panel of fewer tumor markers for lung cancer diagnosis. Methods: Patients with suspected lung cancer who simultaneously underwent all six tests [carcinoembryonic antigen (CEA), cytokeratin-19 fragment (CYFRA), squamous cell carcinoma-associated antigen (SCC), neuron-specific enolase (NSE), pro-gastrin-releasing peptide (ProGRP), and sialyl Lewis-X antigen (SLX)] were included. Tumor markers with significant impact on the lung cancer in a logistic regression model were included in our panel. Area under the curve (AUC) was compared between our panel and the panel of all six. Results: We included 1,733 [median 72 years, 1,128 men, 605 women, 779 (45%) confirmed lung cancer]. Logistic regression analysis suggested CEA, CYFRA, and NSE were independently associated with the lung cancer diagnosis. The panel of these three tumor markers [AUC =0.656, 95% confidence interval (CI): 0.630-0.682, sensitivity 0.650, specificity 0.662] had better (P<0.001) diagnostic performance than six tumor markers (AUC =0.575, 95% CI: 0.548-0.602, sensitivity 0.829, specificity 0.321). Conclusions: Compared to applying all six markers (at least one marker above the upper limit of normal), the panel with three markers (at least one marker above the upper limit of normal) led to a better predictive value by lowering the risk of false positives.
ABSTRACT
BACKGROUND: For simultaneous prediction of phenotypic drug susceptibility test (pDST) for multiple anti-tuberculosis drugs, the whole genome sequencing (WGS) data can be analyzed using either catalogue-based approach, wherein one causative mutation suggests resistance, (e.g., WHO catalog) or non-catalogue-based approach using complicated algorithm (e.g., TB-profiler, machine learning). The aim was to estimate the predictive ability of WGS-based tests with pDST as the reference, and to compare the two approaches. METHODS: Following the systematic literature search, the diagnostic test accuracies for 14 drugs were pooled using a random-effect bivariate model. RESULTS: Out of 779 articles, 44 articles with 16,821 specimens for meta-analysis and 13 articles not for meta-analysis were adopted. The areas under summary receiver operating characteristic curve suggested "excellent" (0.97-1.00) for 2 drugs (isoniazid 0.975, rifampicin 0.975), "very good" (0.93-0.97) for 8 drugs (pyrazinamide 0.946, streptomycin 0.952, amikacin 0.968, kanamycin 0.963, capreomycin 0.965, para-aminosalicylic acid 0.959, levofloxacin 0.960, ofloxacin 0.958), and "good" (0.75-0.93) for 4 drugs (ethambutol 0.926, moxifloxacin 0.896, ethionamide 0.878, prothionamide 0.908). The non-catalogue-based and catalogue-based approaches had similar ability for all drugs. CONCLUSION: WGS accurately identifies isoniazid and rifampicin resistance. For most drugs, positive WGS results reliably predict pDST positive. The two approaches had similar ability.
