目的: 预测非转移性T_3a肾细胞癌患者术后3年肿瘤特异性生存期(cancer specific survival, CSS)。方法: 选择2013年3月至2021年2月在北京大学第三医院泌尿外科行手术治疗且术后组织病理学诊断证实为T_3aN0-1M0的肾细胞癌患者的病例资料进行回顾性分析，共收集到符合要求的患者336例，以4 ∶ 1的比例随机划分为训练队列(268例)和内部验证队列(68例)。采用双向Lasso回归筛选变量绘制预测T_3aN0-1M0肾细胞癌患者术后3年CSS的列线图。利用一致性指数(C指数)、时间依赖性ROC曲线下面积(area under the curve，AUC)、校准曲线、决策曲线(decision curve analysis, DCA)来评估列线图的判别和校准能力及临床效益。根据列线图得分划分危险分层，并用Kaplan-Meier生存分析和Log-rank检验比较不同危险分层患者的无进展生存期(progress free survival, PFS)和CSS。结果: 训练队列和验证队列的基线资料组间差异均无统计学意义(P＞0.05)。根据Lasso回归筛选结果，最终纳入肿瘤最大径、组织学分级、肉瘤样变、T_3a特征、淋巴结转移5个变量构建列线图。列线图的训练队列和内部验证队列的一致性指数分别为0.808(0.708~0.907)、0.903(0.838~0.969)；3年肿瘤特异性生存期的AUC分别为0.843(0.725~0.961)、0.923(0.844~1.002)；各队列的校准曲线均显示实际CSS与预测概率之间具有高度的一致性；各队列的DCA曲线均显示列线图在临床上具有良好的净效益；两个队列共计336例患者，其中肿瘤特异性死亡35例，术后复发69例。根据列线图将0~117分的患者划分为低危组，119-284分的患者划分为高危组，其中低危组(282例)肿瘤特异性死亡16例，术后复发36例，高危组(54例)肿瘤特异性死亡19例，术后复发33例。低危组和高危组的PFS和CSS差异有统计学意义(P＜0.000 1)。结论: 构建并验证了预测非转移性T_3a肾细胞癌患者术后3年CSS的列线图预测模型，可协助临床精准评估此类患者的远期预后。

Objective: To predict the 3-year cancer-specific survival (CSS) of patients with non-metastatic T_3a renal cell carcinoma after surgery. Methods: A total of 336 patients with pathologically confirmed T_3a N0-1M0 renal cell carcinoma (RCC) who underwent surgical treatment at the Department of Urology, Peking University Third Hospital from March 2013 to February 2021 were retrospectively collected. The patients were randomly divided into a training cohort of 268 cases and an internal validation cohort of 68 cases at an 4 ∶ 1 ratio. Using two-way Lasso regression, variables were selected to construct a nomogram for predicting the 3-year cancer-specific survival (CSS) of the patients with T_3aN0-1M0 RCC. Performance assessment of the nomogram included evaluation of discrimination and calibration ability, as well as clinical utility using measures such as the concordance index (C-index), time-dependent area under the receiver operating characteristic curve [time-dependent area under the curve (AUC)], calibration curve, and decision curve analysis (DCA). Risk stratification was determined based on the nomogram scores, and Kaplan-Meier survival analysis and Log-rank tests were employed to compare progression-free survival (PFS) and cancer-specific survival (CSS) among the patients in the different risk groups. Results: Based on the Lasso regression screening results, the nomogram was constructed with five variables: tumor maximum diameter, histological grading, sarcomatoid differentiation, T_3a feature, and lymph node metastasis. The baseline data of the training and validation sets showed no statistical differences (P>0.05). The consistency indices of the column diagram were found to be 0.808 (0.708- 0.907) and 0.903 (0.838-0.969) for the training and internal validation sets, respectively. The AUC values for 3-year cancer-specific survival were 0.843 (0.725-0.961) and 0.923 (0.844-1.002) for the two sets. Calibration curves of both sets demonstrated a high level of consistency between the actual CSS and predicted probability. The decision curve analysis (DCA) curves indicated that the column diagram had a favorable net benefit in clinical practice. A total of 336 patients were included in the study, with 35 cancer-specific deaths and 69 postoperative recurrences. According to the line chart, the patients were divided into low-risk group (scoring 0-117) and high-risk group (scoring 119-284). Within the low-risk group, there were 16 tumor-specific deaths out of 282 cases and 36 postoperative recurrences out of 282 cases. In the high-risk group, there were 19 tumor-specific deaths out of 54 cases and 33 post-operative recurrences out of 54 cases. There were significant differences in progression-free survival (PFS) and cancer-specific survival (CSS) between the low-risk and high-risk groups (P < 0.000 1). Conclusion: A nomogram model predicting the 3-year CSS of non-metastatic T_3a renal cell carcinoma patients was successfully constructed and validated in this study. This nomogram can assist clinicians in accurately assessing the long-term prognosis of such patients.