Computed Tomography Histogram Approach to Predict Lymph Node Metastasis in Patients With Clinical Stage IA Lung Cancer

Background. Quantitative computed tomography (CT) histogram analysis of tumors is reported to help distinguish between invasive and less invasive lung cancers. This study aimed to clarify whether CT histogram analysis of tumors can be used to classify patients with clinical stage 0 to IA non-small cell lung cancer according to pathologic lymph node (pN) status. Methods. Predictive factors associated with pN metastasis were identified from the derivation dataset including 629 patients with clinical stage 0 to IA non-small cell lung cancer who underwent complete resection with lymph node dissection (surgeries between 2008 and 2013). The validation dataset including 238 patients (surgeries between 2014 and 2015) were subsequently reevaluated. Clinicosurgical factors, including CT histogram analysis of tumors (CT value percentiles 2.5, 25, 50, 75, and 97.5, skewness, and kurtosis) were assessed. Results. Seventy-three patients (12%) in the derivation cohort and 35 patients (15%) in the validation cohort had positive nodes. The pN status significantly affected survival in the entire population: 5-year overall survival of 93.1% vs 71.1% and 5-year disease-free survival of 85.9% vs 43.1% for negative vs positive (both P <.001). On multivariate analysis in the derivation cohort, the 75th percentile CT value (P < .001), age (P = .003), and comorbidities (P = .006) were significantly associated with pN metastasis. The area under the curve and the cutoff level of the 75th percentile CT value relevant to pN metastasis were 0.729 and 1.5 HU, respectively, and the threshold value provided accuracy of 71% for the validation cohort. Conclusions. Histogram analysis of CT imaging metrics of tumors contributes to noninvasive prediction of pN metastasis in patients with clinical stage 0 to IA non-small cell lung cancer. (C) 2019 by The Society of Thoracic Surgeons