Predictive model for mediastinal lymph node status at the time of mediastinoscopy

Abstract

18004 Background: Mediastinal lymph node (N2) positivity in non-small cell lung cancer (NSCLC) patients is suspected based on imaging such as CT or PET scan, with confirmation by mediastinoscopy. However, the most accurate clinical information in predicting N2 status is controversial. Methods: We reviewed 147 candidates for NSCLC resection (2000–2005) who had clinical database information available and had undergone mediastinoscopy. Using suspected clinical predictors of mediastinal metastasis available prior to mediastinoscopy, we constructed a predictive model of N2 status. Results: The largest N2 node short-axis diameter on CT was by far the most influential factor in the model. Three other predictors for N2 node positivity were significant (p<0.05) in univariate analysis: indistinct tumor borders and mediastinal invasion on CT, and mediastinal PET scan positivity. However, all were less influential than N2 size on CT. Using logistic regression, these factors can be used to predict probability of positive N2 biopsy in an individual patient. The resulting diagnostic test had a ROC (receiver operator characteristic) area of 0.80 and optimal sensitivity-specificity pairing of 75% and 73%. 35% (51/147) of patients analyzed had at least one N2 node positive at mediastinoscopy. 39% (57/147) of patients had PET scan data available, and 82% (120/147) had CT data available. Conclusions: Of available data in early-stage NSCLC patients, mediastinal lymph node size on CT scan was more important than PET scan or other CT scan findings in predicting probability of positive mediastinoscopy. A predictive model is useful in more accurately determining need for invasive staging by mediastinoscopy. No significant financial relationships to disclose.