PURPOSE: In most published series of cases, the accuracy of EBUS in the diagnosis of thoracic malignancies is reported taking into consideration only those patients in whom a valid sample was obtained during the exploration. In this report we compare the accuracy of EBUS in an intention-to-diagnose basis (ITD cases including all patients intended to be staged) and in cases with a valid sample of cells obtained during the procedure (per-protocol cases, PP).
METHODS: Cross sectional analysis on a prospectively recorded database of patients with known or suspected thoracic malignancies. Cases were discussed and selected for EBUS during a weekly multidisciplinary meeting. EBUS was indicated for diagnosis of accessible mediastinal masses or for mediastinal staging in all patients with positive mediastinal FDG-PET or mediastinal lymph nodes over 14mm. The reference standard was the histological study of lymph nodes resected at thoracotomy or the presence of malignant cells in EBUS samples or in biopsies taken at mediastinoscopy. Cases without a valid sample at EBUS (ITD series) were considered as false negatives. Sensitivity (S), negative predictive value (NPV) and negative likelihood ratio (LR-) and their 95%CI were calculated and cusum graphics were constructed for ITD and PP series of cases, considering 1 true positive and true negative results and 0 false negative and no valid sample.
RESULTS: 83 cases were included in the analysis. In the ITD series, results were as follows S: 0.72 (0.58-0.83); NPV: 0.69 (0.54-0.81); LR-: 0.27 (0.17-0.43), in the PP series, S: 0.88 (0.73-0.95); NPV: 0.86 (0.70-0.95); LR-: 0.12 (0.05-0.27) Cusum graphics for both series clearly showed a learning curve in the ITD series.
CONCLUSIONS: Accuracy of EBUS is lower in intention-to-diagnose series of cases.
CLINICAL IMPLICATIONS: These values should be reported for valid cost-effectiveness analysis of EBUS in the diagnosis of thoracic malignancies.
DISCLOSURE: The following authors have nothing to disclose: Rosa Cordovilla, Gonzalo Varela, Carmen García-Macías, Jose María González-Ruiz, Manuel Lanchas, Aldo Mateo Torracchi, Dolores Ludeña, Angel Rodríguez-Encinas
No Product/Research Disclosure Information