A prediction model for pathologic N2 disease in lung cancer patients with a negative mediastinum by positron emission tomography

Abstract

Introduction: Guidance is limited for invasive staging in patients with lung cancer without mediastinal disease by positron emission tomography (PET). We developed and validated a prediction model for pathologic N2 disease (pN2), using six previously described risk factors: tumor location and size by computed tomography (CT), nodal disease by CT, maximum standardized uptake value of the primary tumor, N1 by PET, and histology.

Methods: A cohort study (2004-2009) was performed in patients with T1/T2 by CT and N0/N1 by PET. Logistic regression analysis was used to develop a prediction model for pN2 among a random development set (n = 625). The model was validated in both the development set, which comprised two thirds of the patients and the validation set (n = 313), which comprised the remaining one third. Model performance was assessed in terms of discrimination and calibration.

Results: Among 938 patients, 9.9% had pN2 (9 detected by invasive staging and 84 intraoperatively). In the development set, univariate analyses demonstrated a significant association between pN2 and increasing tumor size (p < 0.001), nodal status by CT (p = 0.007), maximum standardized uptake value of the primary tumor (p = 0.027), and N1 by PET (p < 0.001); however, only N1 by PET was associated with pN2 (p < 0.001) in the multivariate prediction model. The model performed reasonably well in the development (c-statistic, 0.70; 95% confidence interval, 0.63-0.77; goodness of fit p = 0.61) and validation (c-statistic, 0.65; 95% confidence interval, 0.56-0.74; goodness-of-fit p = 0.19) sets.

Conclusion: A prediction model for pN2 based on six previously described risk factors has reasonable performance characteristics. Observations from this study may guide prospective, multicenter development and validation of a prediction model for pN2.