External validation of the International Metastatic Renal Cell Carcinoma (mRCC) Database Consortium prognostic model and comparison to four other models in the era of targeted therapy.
External validation of the International Metastatic Renal Cell Carcinoma (mRCC) Database Consortium prognostic model and comparison to four other models in the era of targeted therapy.
复制标题
DOI：
10.1200/jco.2011.29.15_suppl.4560
复制
发表时间：
2011-05
期刊：
影响因子：
--
通讯作者：
D. Heng;W. Xie;L. Harshman;G. Bjarnason;U. Vaishampayan;J. Lebert;L. Wood;F. Donskov;M. Tan;S. Rha;C. Wells;Yulei N. Wang;C. Kollmannsberger;B. Rini;T. Choueiri
中科院分区：
文献类型：
--
作者：
D. Heng;W. Xie;L. Harshman;G. Bjarnason;U. Vaishampayan;J. Lebert;L. Wood;F. Donskov;M. Tan;S. Rha;C. Wells;Yulei N. Wang;C. Kollmannsberger;B. Rini;T. Choueiri
4560背景：Heng et al JCO 2009预后模型是在VEGF靶向治疗时代开发的，本研究在将其性能与其他模型进行比较时作为外部验证。方法：1,028例以前未分析的患者用于外部验证总生存期（OS）的预后标准。使用C指数和净重新分类改进（NRI）将该模型的判别值与其他四个预后模型进行比较（Pencina et al 2008）。NRI指出了Heng等人的模型如何通过基于2年时观察到的OS将一定百分比的患者（pts）重新分类为更正确的风险组来改善其他模型。结果所有患者的中位OS为18.8个月。在多变量分析中，预先确定的Heng危险因素（贫血、血小板增多、嗜中性粒细胞增多、高钙血症、Karnofsky体力状态<80%和从诊断到治疗的时间< 1年）仍然是OS差的独立预测因素（p<0.05）。当患者分为3个风险类别时，有利（17%的患者）、中等（52%）和不良（31%）风险组的中位OS分别为44、21和8个月（p<0.0001，C指数=0.664）。NRI证明，与French和MSKCC模型相比，Heng模型能够更准确地重新分类患者，净23%和9.7%的患者，与CCF模型相比，重新分类13%的患者的准确性较低（但c指数较低），基于观察到的2年OS（见表）。缺失数据的5个插补数据集的敏感性分析产生了相似的结果。结论：Heng等人的模型是在靶向治疗时代得出的，目前已得到外部验证。c指数与主要在免疫治疗或早期VEGF靶向治疗时代衍生的其他模型相似。[表：见正文]。
4560 Background: The Heng et al JCO 2009 prognostic model was developed in the age of VEGF-targeted therapy and this study serves as an external validation while comparing its performance to other models. METHODS 1,028 previously unanalyzed patients were used to externally validate the prognostic criteria for overall survival (OS). The model's discriminatory value was compared to four other prognostic models using C-indices and Net Reclassification Improvement (NRI) (Pencina et al 2008). NRI indicates how the Heng et al model improves upon other models by reclassifying a percentage of patients (pts) into the more correct risk group based on the observed OS at 2 years. RESULTS The median OS of all pts was 18.8 mons. On multivariable analysis, the pre-identified Heng's risk factors (anemia, thrombocytosis, neutrophilia, hypercalcemia, Karnofsky performance status <80% and time from diagnosis to treatment < 1 year) continue to be independent predictors of poor OS (p<0.05). When pts were segregated into 3 risk categories, the median OS was 44, 21 and 8 mons in the favorable (17% of pts), intermediate (52%) and poor (31%) risk groups, respectively (p<0.0001, C-index=0.664). The NRI demonstrated that the Heng model was able to more accurately reclassify pts by a net of 23% and 9.7% of pts compared to the French and MSKCC models and was less accurate in reclassifying 13% of pts when compared to the CCF model (but lower c-index) based on observed 2-year OS (see table). Sensitivity analyses with 5 imputation datasets for missing data produced similar results. CONCLUSIONS The Heng et al model was derived in the targeted therapy era and is now externally validated. The c-index is similar to other models primarily derived in the era of immunotherapy or early VEGF-targeted therapy. [Table: see text].