METHODS: Across a statewide consortium, data was prospectively collected on stage II to III NSCLC patients who received CRT between 2012 and 2019. Early poor outcomes included hospice enrollment or death within 3 months of completing CRT. Logistic regression models were used to assess predictors in prognostic models. LASSO regression with multiple imputation were used to build a final multivariate model, accounting for missing covariates.
RESULTS: Of the 2267 included patients, 128 experienced early poor outcomes. Mean age was 71 years and 59% received concurrent chemotherapy. The best predictive model, created parsimoniously from statistically significant univariate predictors, included age, ECOG, planning target volume (PTV), mean heart dose, pretreatment lack of energy, and cough. The estimated area under the ROC curve for this multivariable model was 0.71, with a negative predictive value of 95%, specificity of 97%, positive predictive value of 23%, and sensitivity of 16% at a predicted risk threshold of 20%.
CONCLUSIONS: This multivariate model identified a combination of clinical variables and patient reported factors that may identify individuals with inoperable NSCLC undergoing curative intent chemo-radiotherapy who are at higher risk for early poor outcomes.
方法:在全州范围的联盟中,前瞻性收集了2012年至2019年接受CRT治疗的II至III期NSCLC患者的数据.早期不良结果包括临终关怀登记或完成CRT后3个月内死亡。Logistic回归模型用于评估预后模型中的预测因子。采用多元插补的LASSO回归建立最终的多变量模型,考虑缺失的协变量。
结果:在2267名患者中,128经历了早期的不良结果。平均年龄71岁,59%接受同步化疗。最好的预测模型,从统计上显著的单变量预测因子简约地创建,包括年龄,ECOG,规划目标量(PTV),平均心脏剂量,预处理缺乏能量,还有咳嗽.该多变量模型的估计ROC曲线下面积为0.71,阴性预测值为95%,97%的特异性,23%的阳性预测值,在20%的预测风险阈值下,敏感性为16%。
结论:该多变量模型确定了临床变量和患者报告因素的组合,这些因素可能会确定无法手术的非小细胞肺癌患者接受根治性化疗放疗,其早期不良预后风险较高。