Fluorescence in Situ Hybridization Testing Algorithm Improves Lung Cancer Detection in Bronchial Brushing Specimens

Rationale: Bronchoscopically collected cytology specimens are commonly used to obtain a diagnosis of cancer in patients with pulmonary lesions. However, the sensitivity of cytology is suboptimal, especially for peripheral lesions less than 2 cm in diameter. Objectives: We assessed the performance of a testing algorithm using cytology and fluorescence in situ hybridization (FISH) as part of clinical practice. Methods: Bronchial brushing specimens (n = 343) were obtained from patients undergoing bronchoscopy for indeterminate pulmonary lesions. Routine cytology was performed and specimens without a positive diagnosis (n = 294) were analyzed by FISH, using residual brushing material. Pathology-confirmed lung cancer or clinical/radiographic evidence of disease was considered diagnostic of malignancy. Measurements and Main Results: Routine cytology had a sensitivity and specificity of 41% (23 of 56) and 100% (45 of 45) for central lesions and 20% (26 of 133) and 100% (109 of 109) for peripheral nodules, respectively. FISH detected an additional 32% of lung cancers (18 central and 43 peripheral) not detectable by cytology alone, while producing false positive diagnoses in 22% (110 of 45) and 6% (6 of 109) benign central and peripheral lesions, respectively. In peripheral nodules, FISH detected (relative to routine cytology) an additional 44% (15 of 34) and 28% (25 of 91) of lung cancers less than 2 cm and 2 cm or more n size, respectively. A positive FISH result had a likelihood ratio of 1.45 and 5.87 for central and peripheral lesions and 3.44 and 15.38 for peripheral nodules less than 2 cm and 2 cm or more in size, respectively. Conclusions: FISH testing significantly increases the detection of lung cancer over routine cytology alone. It is especially useful for peripheral nodules.