Cancer Care Northwest, Spokane Valley, WA, USA.
Department of Biostatistics and Bioinformatics, Duke University School of Medicine, Durham, NC, USA.
Oral Oncol. 2022 Sep;132:105955. doi: 10.1016/j.oraloncology.2022.105955. Epub 2022 Jun 22.
To develop and validate a prediction model to estimate overall survival (OS) with and without postoperative radiotherapy (PORT) for resected major salivary gland (SG) cancers.
Adults in the National Cancer Database diagnosed with invasive non-metastatic major SG cancer between 2004 and 2015 were identified. Exclusion criteria included prior malignancy, pT1N0 or unknown stage, no or unknown surgery, and neoadjuvant therapy. Cox proportional hazards models evaluated the effect of covariates on OS. A multivariate regression model was utilized to predict 2-, 5-, and 10-year OS. Internal cross-validation was performed using 50-50 hold-out and Harrell's concordance index.
18,400 subjects met inclusion criteria, including 9,721 (53%) who received PORT. Distribution of SG involvement was 86% parotid, 13% submandibular, and 1% sublingual. Median follow-up for living subjects was 4.9 years. PORT was significantly associated with improved OS for the following subgroups by log-rank test: pT3 (p < 0.001), pT4 (p < 0.001), high grade (p < 0.001), node-positive (p < 0.001), and positive margin (p < 0.001). The following variables were incorporated into a multivariate model: age, sex, Charlson-Deyo comorbidity score, involved SG, pathologic T-stage, grade, margin status, ratio of nodal positivity, and PORT. The resulting model based on data from 6,138 subjects demonstrated good accuracy in predicting OS, with Harrell's concordance index of 0.73 (log-rank p < 0.001).
This cross-validated prediction model estimates 2-, 5-, and 10-year differences in OS based on receipt of PORT for resected major SG cancers using readily available clinicopathologic features. Clinicians can utilize this tool to aid personalized adjuvant therapy decisions.
建立并验证一个预测模型，用于预测接受和不接受术后放疗（PORT）的切除性大涎腺癌患者的总生存（OS）。
从 2004 年至 2015 年期间，国家癌症数据库中诊断为侵袭性非转移性大涎腺癌的成年人被确定为研究对象。排除标准包括既往恶性肿瘤、pT1N0 或未知分期、无手术或未知手术、新辅助治疗。Cox 比例风险模型评估了协变量对 OS 的影响。利用多元回归模型预测 2 年、5 年和 10 年的 OS。采用 50-50 分割和 Harrell 一致性指数进行内部交叉验证。
符合纳入标准的 18400 例患者，其中 9721 例（53%）接受了 PORT。SG 受累的分布为 86%腮腺、13%颌下腺和 1%舌下腺。对存活患者的中位随访时间为 4.9 年。通过对数秩检验，PORT 与以下亚组的 OS 改善显著相关：pT3（p<0.001）、pT4（p<0.001）、高级别（p<0.001）、淋巴结阳性（p<0.001）和阳性切缘（p<0.001）。将以下变量纳入多变量模型：年龄、性别、Charlson-Deyo 合并症评分、受累 SG、病理 T 分期、分级、切缘状态、淋巴结阳性率和 PORT。该模型基于 6138 例患者的数据建立，在预测 OS 方面具有较好的准确性，Harrell 一致性指数为 0.73（对数秩 p<0.001）。
该经交叉验证的预测模型利用易于获得的临床病理特征，根据切除性大涎腺癌患者是否接受 PORT，预测 OS 的 2 年、5 年和 10 年差异。临床医生可以使用该工具来辅助制定个体化辅助治疗决策。
