新发转移性去势敏感性前列腺癌多变量预后模型的开发与验证
基于LATITUDE III期随机对照试验数据,纳入1058例新发转移性去势敏感性前列腺癌患者,按70:30随机分为训练集743例和测试集315例,采用弹性网回归筛选变量并构建Cox回归模型预测总生存期。最终模型含11个预后变量,预测2年和3年总生存期的时依AUC分别为0.74和0.72,影像学无进展生存期分别为0.72和0.77。
要点
- 模型含11个预后变量:体力状态、骨转移数目、Gleason评分、肝转移、最痛评分、白蛋白、乳酸脱氢酶、前列腺特异性抗原、血红蛋白和治疗方案。
- 测试集预测2年和3年总生存期的时依AUC分别为0.74(95%CI 0.67-0.80)和0.72(95%CI 0.65-0.77)。
- 预测2年和3年影像学无进展生存期的时依AUC分别为0.72(95%CI 0.65-0.77)和0.77(95%CI 0.70-0.82)。
核心结论
- 在新发转移性去势敏感性前列腺癌患者中开发并验证了预后模型,该模型对总生存期和影像学无进展生存期具有较高预测准确性。
- 模型包含常用临床和实验室参数,可指导这些患者未来试验参与的风险分层。
研究设计与样本
- 使用LATITUDE III期随机对照试验数据,该试验将新发转移性去势敏感性前列腺癌男性随机分配至ADT联合阿比特龙或ADT联合安慰剂。
- 无缺失数据的患者共1058例,按70:30比例随机分为训练集743例和测试集315例。
- 采用弹性网回归进行变量选择,随后用所选变量拟合多变量Cox回归模型预测总生存期。
主要发现
- 最终模型包含11个预后变量:体力状态、骨转移数目、Gleason评分、肝转移 presence、最痛评分、白蛋白、乳酸脱氢酶、前列腺特异性抗原、血红蛋白和治疗方案。
- 测试集预测2年和3年总生存期的时依AUC分别为0.74(95%CI 0.67-0.80)和0.72(95%CI 0.65-0.77)。
- 测试集预测2年和3年影像学无进展生存期的时依AUC分别为0.72(95%CI 0.65-0.77)和0.77(95%CI 0.70-0.82)。
临床怎么用
- 作者认为,该模型包含常用临床和实验室参数,可指导新发转移性去势敏感性前列腺癌患者未来试验参与的风险分层。
局限性
- 原文未报告局限性
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Metastatic castrate sensitive prostate cancer (mCSPC) is a heterogeneous disease state with variable prognosis. Although several life-prolonging systemic agents are available, there is no robust multivariable model to predict prognosis and improve risk stratification in mCSPC. The objective of this study was to build and validate a multivariable prognostic model to predict overall survival (OS) in mCSPC. METHODS: We used data from LATITUDE, a phase III randomized controlled trial in which men with de novo mCSPC were randomly allocated to either ADT plus abiraterone or ADT with placebo. Patients with non-missing data (n = 1,058) were randomly split in a 70:30 ratio to training (n = 743) and testing (n = 315) sets. Elastic net regression was used for variable selection. A multivariable Cox regression model for OS was then fitted using the selected variables. The predictive accuracy of the model was assessed on the testing set using the time-dependent area under curve (tAUC) with bootstrapped confidence intervals [CI] primarily for OS and secondarily for radiographic progression-free survival (rPFS). RESULTS: The 11 prognostic variables in the final model were performance status, number of skeletal metastases, Gleason score, presence of liver metastasis, worst pain score, albumin, lactate dehydrogenase, prostate-specific antigen, hemoglobin, and treatment regimen. The tAUC for predicting OS at 2- and 3-years was 0.74 (95% CI, 0.67-0.80) and 0.72 (95% CI, 0.65-0.77), respectively. The tAUC for rPFS at 2- and 3-years was 0.72 (95% CI, 0.65-0.77) and 0.77 (95% CI, 0.70-0.82), respectively. CONCLUSIONS: A prognostic model for men with de novo mCSPC was developed and validated in an independent testing set. Our model had high accuracy for predicting OS and rPFS. The model includes commonly used clinical and laboratory parameters and can guide risk stratification of these patients for participation in future trials.
引用
- Roy S, Sun Y, Wallis CJD, et al. Development and validation of a multivariable prognostic model in de novo metastatic castrate sensitive prostate cancer. Prostate Cancer Prostatic Dis. 2022. PMID: 35790787. doi:10.1038/s41391-022-00560-3
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。