Prognostic model for long-term survival of locally advanced non-small-cell lung cancer patients after neoadjuvant radiochemotherapy and resection integrating clinical and histopathologic factors

Outcome of consecutive patients with locally advanced non-small cell lung cancer and histopathologically proven mediastional lymph node metastases treated with induction chemotherapy, neoadjuvant radiochemotherapy and thoracotomy at the West German Cancer Center between 08/2000 and 06/2012 was analysed. A clinico-pathological prognostic model for survival was built including partial or complete response according to computed tomography imaging (CT) as clinical parameters as well as pathologic complete remission (pCR) and mediastinal nodal clearance (MNC) as histopathologic factors. Proportional hazard analysis (PHA) and recursive partitioning analysis (RPA) were used to identify prognostic factors for survival. Long-term survival was defined as survival ≥ 36 months. A total of 157 patients were treated, median follow-up was 97 months. Among these patients, pCR and MNC were observed in 41 and 85 patients, respectively. Overall survival was 56 ± 4% and 36 ± 4% at 24 and 60 months, respectively. Sensitivities of pCR and MNC to detect long-term survivors were 38% and 61%, specificities were 84% and 52%, respectively. Multivariable survival analysis revealed pCR, cN3 category, and gender, as prognostic factors at a level of α < 0.05. Considering only preoperative available parameters, CT response became significant. Classifying patients with a predicted hazard above the median as high risk group and the remaining as low risk patients yielded better separation of the survival curves by the inclusion of histopathologic factors than by preoperative factors alone (p < 0.0001, log rank test). Using RPA, pCR was identified as the top prognostic factor above clinical factors (p = 0.0006). No long term survivors were observed in patients with cT3-4 cN3 tumors without pCR. pCR is the dominant histopathologic response parameter and improves prognostic classifiers, based on clinical parameters. The validated prognostic model can be used to estimate individual prognosis and forms a basis for patient selection for treatment intensification