A prognostic index for second-line chemotherapy of NSCLC was previously developed, based on individual patient data (IPD) of nine randomized trials. In order to validate the prognostic score in an external dataset, we analysed IPD of a non-inferiority phase III trial comparing vinflunine vs. docetaxel in second-line treatment of advanced NSCLC. Primary endpoint of this analysis was overall survival (OS). The following variables were considered for survival analysis and score calculation: gender, performance status, stage of disease, tumour histology, type of first-line treatment, response to first-line treatment. Cox model, stratified by treatment arm, was used for multivariate analysis. Individual prognostic scores were derived, and patients were divided into 3 categories: <5 (best), 5-9 (intermediate), >9 (worst). All 551 patients enrolled in the trial had complete information for the calculation of prognostic score. Median OS in the whole group was 6.9 months, with similar efficacy in the two treatment arms. Median OS was 12.9, 6.9 and 3.8 months in the best, intermediate and worst category, respectively. Cox model showed a significant effect comparing intermediate vs. best category (Hazard Ratio 1.79, 95%CI 1.31-2.47, p=0.0003) and comparing worst vs. best category (Hazard Ratio 3.25, 95%CI 2.18-4.83, p<0.0001). The C-index of the model was high (0.926), indicating a good discrimination according to the proposed three risk categories. Prognostic ability of our score for candidates to second-line treatment in advanced NSCLC was successfully validated, allowing the identification of subgroups of patients with more vs. less favourable outcome. Prognostic score could be useful in daily decision-making in clinical practise, because a better understanding of factors conditioning life expectancy of patients could greatly help a careful evaluation of risks and benefits associated with therapeutic decisions.
Copyright © 2012 Elsevier Ireland Ltd. All rights reserved.