Purpose: When mediastinal lymph nodes are clinically uninvolved in the setting of inoperable non-small cell lung cancer, whether conventional radiation techniques or three-dimensional dose-escalation techniques are used, the benefit of elective nodal irradiation is unclear. Inclusion of the clinically negative mediastinum in the radiation portals increases the risk of lung toxicity and limits the ability to escalate dose. This analysis represents an attempt to use clinical characteristics to estimate the risk of subclinical nodal involvement, which may help determine which patients are most likely to benefit from elective nodal irradiation. Methods: From 1987 to 1990, 346 patients undergoing complete resection of non-small cell lung cancer underwent a preoperative computed tomographic scan revealing no clinical evidence of N2/N3 involvement. Multivariate regression and regression tree analyses attempted to define which patients were at highest risk for subclinical mediastinal involvement (N2) and which patients were at highest risk for subclinical N1 and/or N2 involvement (N1/N2). Immunohistochemical data suggest that the conventional histopathologic techniques used during this study somewhat underestimate the true degree of lymph node involvement; therefore, a third end point was also evaluated: N1 involvement and/or N2 involvement and/or local-regional recurrence (N1/N2/LRR). Results: Regression analyses revealed that the following factors were independently associated with a high risk of more advanced disease: positive preoperative bronchoscopy (N2, p = 0.02; N1/N2, p < 0.0001; N1/N2/LRR, p < 0.001) and tumor grade 3/4 (N1/N2/LRR, p < 0.01). A regression tree analysis was then used to separate patients into risk groups with respect to N1/N2/LRR. Conclusion: In inoperable non-small cell lung cancer, the patients for whom mediastinal radiation therapy may most likely be indicated are those with a positive preoperative bronchoscopy, especially with large (> 3 cm) primary tumors.