Development and External Validation of a Prognostic Classifier for Elderly Head-and-Neck Cancer Patients Undergoing (Chemo)radiation

Abstract

Elderly head-and-neck squamous cell carcinoma (HNSCC) patients form a heterogeneous cohort, and survival estimation is often challenging due to underlying comorbidities. Prognostic classifiers and nomograms may therefore be useful for overall survival (OS) estimation in order to provide shared decision-making in the clinical routine.A total of 284 elderly HNSCC patients aged ≥65 years who received a curative (chemo)radiation between 2010 and 2020 at a tertiary cancer center were used for the development of a survival classifier. On the basis of a multivariate Cox regression analysis, significant parameters were identified for which points were given according to the beta regression values. The derived classifier was then validated in a second, external cohort consisting of 217 elderly HNSCC patients undergoing (chemo)radiation. Based on the cumulative data of 501 patients, a nomogram for the 2-year and 4-year OS was created. We then examined in a third independent cohort whether the classifier could also stratify the prognosis of surgically treated elderly HNSCC patients without adjuvant (chemo)radiation (n = 169).In the multivariate backward stepwise Cox regression with likelihood ratio tests using P < 0.1 as inclusion criterion, the Karnofsky Performance Status (KPS, HR = 2.654, P < 0.001), the age-adjusted Charlson Comorbidity Index (CCI, HR = 2.598, P < 0.001) and the pre-radiotherapy CRP serum concentration (HR = 1.634, P = 0.064) were significant prognostic parameters for OS. Following the beta regression values, a KPS ≤70% and a CCI ≥6 points were given 1 point, while a CRP concentration ≥5 mg/L was given 0.5 points. The median OS was 107 (0 points), 34 (0.5 points), 28 (1 point), 11 (1.5 points), 9 (2 points) and 6 months (2.5 points), respectively. In order to obtain considerably distinct prognostic subgroups, 3 prognostic groups were created: A favorable (0 points), an intermediate (0.5-2 points) and a poor (2.5 points) subgroup. While the median OS for the favorable group amounted to 107 months, it was 28 and 6 months for the intermediate and poor cohorts, respectively (P < 0.001, log-rank test). In the external cohort, the median OS was found to range at 130, 29 and 9 months for the favorable, intermediate and poor group, respectively (P = 0.005). Using the aggregated data of both cohorts, a nomogram based on the KPS, CCI and CRP was created for the 2-year and 4-year OS that exhibited a concordance index of 0.65 (Harrell's C). For the surgically treated cohort of elderly HNSCC patients, only the KPS ≤70% (HR = 5.950, P = 0.002) but not the CCI ≥6 points (HR = 1.937, P = 0.232) or the baseline CRP serum value (HR = 1.743, P = 0.328) were significant prognosticators.We developed and externally validated a clinically feasible survival score for elderly HNSCC patients undergoing (chemo)radiation. Both the score and nomogram may be useful aiding shared decision-making of radiation oncologists and medical oncologists treating elderly HNSCC patients.