Comparison of the accuracy of clinical prediction of survival and palliative prognostic index for patients with head and neck squamous cell carcinoma in the end-of-life setting

Abstract

ObjectiveIn the end-of-life stage of head and neck squamous cell carcinoma (HNSCC), predicting survival is essential to determine treatment procedure and place of care. Several reports have compared actual survival (AS) and clinical prediction of survival (CPS), a subjective prognostic prediction by attending physicians. However, specific studies focusing on patients with HNSCC are limited. Likewise, a comparison of the accuracy of CPS and palliative prognostic index (PPI), a prognostic tool using subjective assessment, has not been sufficiently investigated. This study aimed to clarify the correlation between AS and CPS/PPI and compare the accuracy of CPS and PPI in end-stage HNSCC. MethodsThis retrospective study included patients with HNSCC in the end-of-life setting. Patients were recruited from the National Hospital Organization Shikoku Cancer Center between April 2011 and March 2019. Data on basic demography and clinical parameters when patients decided to start end-of-life care at the head and neck oncology division were collected. We examined the correlation between AS and CPS using Spearman's correlation coefficients. The area under the receiver operating characteristic curve of CPS and PPI for 30-day survival prediction were compared for predictive accuracy. ResultsAmong 98 eligible patients, 59 patients were enrolled in this study and analyzed. Of the 59 patients, CPS and PPI were calculated for 30 patients, whereas, only the PPI was calculated for 29 patients. The median AS and CPS were 35 (IQR: 9–73) days and 30 (IQR: 7–83) days, respectively. CPS and PPI (30 cases) were moderately correlated (r = 0.72, p<0.01). AS and CPS/PPI (30 cases) were significantly correlated (p<0.01) and showed a strong correlation (r = 0.86 and 0.80, respectively). In the 30-day survival prediction, the AUROCs of CPS and PPI (30 cases) were 0.967 (95%CI: 0.919–1) and 0.884 (95%CI: 0.767–1), respectively. Both CPS and PPI (30 cases) showed high accuracy in predicting the 30-day prognosis, with no significant difference (p = 0.077). The AUROC of PPI (59 cases) was 0.840 (95%CI: 0.711–0.969). ConclusionsAS and CPS/PPI showed significant correlations. The high accuracy of CPS may have been influenced by the fact that multiple head and neck cancer specialists at a comprehensive cancer center estimated CPS. Both CPS and PPI showed high prognostic accuracy in predicting 30-day survival. This suggests that PPI is useful in centers among physicians and healthcare workers unfamiliar with head and neck cancer.