Prognostic score for survival with pulmonary carcinoids: the importance of associating clinical with pathological characteristics

OBJECTIVES: Lung carcinoids (LCs) are staged using the non-small-cell lung cancer tumour/node/metastasis staging system; the possibility of an LC-specific staging system is still being debated. The goal of our study was to construct a composite prognostic score for LC. METHODS: From January 2002 to December 2014, data from 293 patients who underwent surgical treatment for LC in 7 research institutes were retrospectively analysed. A panel of established prognostic factors in addition to lymph node metastasis patterns (single/multiple N1-N2 station, skip metastasis, lobe specific), numbers of lymph nodes resected and the ratio between the numbers of metastatic lymph nodes and the numbers of lymph nodes resected (node ratio) were correlated to overall survival (OS) and disease-free survival (DFS). The log-hazard ratio (HR), obtained from the Cox model, was used to derive weighting factors for a continuous prognostic index, designed to identify differential outcome risks. The score was dichotomized according to maximally selected log-rank statistics. RESULTS: Pathological analysis showed typical carcinoids in 223 (76.1%) and atypical carcinoids in 70 (23.9%) patients; the tumour/node/metastasis pattern was stage I in 72.4%, stage II in 18.1%, stage III in 9.5% and stage IV in 0.03% cases. The median numbers of lymph nodes resected was 12 (range 0-53); hilar and mediastinal node metastases were identified in 14% and 6.8% of cases, respectively. Overall, the 5-year OS and 5-year DFS rates were 90.6% and 76.7%, respectively. At multivariable analysis, sex, age, pathological T stage and node ratio were significantly related to a better OS; age, histological type, pathological T stage and node ratio were related to DFS. These factors were used to generate the prognostic score, which showed statistically significant differences between the high-risk and low-risk groups: 5-year OS = 96.6% if score <3.1 vs 63.5% if score >= 3.1 [P<0.0001; HR 17.56, 95% confidence interval (CI) 5.45-56.53]; 5-year DFS 92.3% if score <1.5 vs 52.5% if score >= 1.5 (P<0.0001; HR 7.95, 95% CI 3.48-18.16). CONCLUSIONS: The proposed prognostic scores seem to be effective in predicting outcomes for patients with LCs.