肾肿瘤分期胸部CT:风险预测模型的外部验证与简化
本研究在1082例单侧强化肾肿块患者中外部验证Larcher列线图并开发简化模型。Larcher模型AUC为83.8%,但校准中度欠佳;新模型仅用肿瘤大小、淋巴结分期和全身症状,AUC相似(82.7%)且校准和临床净获益更优。低危组(肿瘤≤4 cm且无全身症状)胸部CT阳性风险<1%,或可省略胸部CT。
要点
- Larcher列线图预测胸部CT阳性准确但高估风险,校准欠佳。
- 简化模型仅用肿瘤大小、淋巴结分期和全身症状,校准更优。
- 低危组(≤4 cm且无全身症状)胸部CT阳性风险<1%,或可省略。
核心结论
- Larcher列线图准确但校准中度欠佳,简化模型精度相似且校准和临床净获益更优。
- 低危组(肿瘤≤4 cm且无全身症状)胸部CT阳性风险<1%,提示可省略胸部CT。
研究设计与样本
- 回顾性分析2011至2017年两中心连续1082例单侧强化肾肿块患者,均行分期胸部CT。
- 采用多变量逻辑回归建模,以AUC、校准和决策曲线分析评估模型性能。
主要发现
- Larcher列线图AUC为83.8%(95%CI 77.1-90.6),校准-in-the-large为-0.61,斜率0.82,高估风险且随预测风险增大而加重。
- 新模型AUC为82.7%(95%CI 75.5-90.0),校准和临床净获益优于Larcher模型。
- 新模型低危组预测胸部CT阳性风险<1%,高危组为1.9-79.9%。
关键数值与推荐
- 低危组定义:肿瘤大小≤4 cm且无全身症状,胸部CT阳性风险<1%。
- 新模型纳入变量:肿瘤大小、淋巴结分期和全身症状。
临床怎么用
- 简化模型使用转诊时可获得的客观变量,更易融入临床实践。
- 低危组(≤4 cm且无全身症状)患者或可省略胸部CT分期。
局限性
- 回顾性研究设计。
- 仅纳入两个中心的数据。
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
OBJECTIVES: To externally validate a nomogram recently proposed by Larcher et al. (BJU Int. 2017; 120: 490) and to develop a simplified model with comparable accuracy to guide on the need for staging chest computed tomography (CT) for patients with new renal masses. PATIENTS AND METHODS: We analysed the data of 1082 consecutive patients with unilateral enhancing renal masses referred to urology multidisciplinary team meetings at two centres between 2011 and 2017. All patients underwent a staging chest CT at diagnosis. We fitted multivariable logistic regression models and tested the Larcher model performance using area under the receiver-operating curve (AUC), calibration and decision curve analysis. RESULTS: Forty-two patients (3.9%) had a positive chest CT. The Larcher nomogram had an AUC of 83.8% (95% confidence interval [CI] 77.1-90.6), but was only moderately well calibrated (calibration-in-the-large = -0.61, slope = 0.82). Specifically, the nomogram overestimated the risk of positive chest CT, and the magnitude of miscalibration increased with increasing predicted risks. Using a stepwise backward approach, a new model was developed including tumour size, nodal stage and systemic symptoms. Compared with the Larcher model, the new model had a similar AUC (82.7% [95% CI 75.5-90.0]), but improved calibration and clinical net benefit. The predicted risk of positive chest CT was <1% in the low-risk group and 1.9-79.9% in the high-risk group. CONCLUSION: The Larcher nomogram is an accurate prediction tool that was moderately well calibrated with our dataset. However, our simplified model has similar accuracy and uses more objective variables available from referral, so may be easier to incorporate into clinical practice. The low-risk group from our model (tumour size ≤4 cm and no systemic symptoms) had a risk of positive chest CT <1%, suggesting these patients may forego chest CT.
引用
- Voss J, Drake T, Matthews H, et al. Chest computed tomography for staging renal tumours: validation and simplification of a risk prediction model from a large contemporary retrospective cohort. BJU Int. 2020. PMID: 31955483. doi:10.1111/bju.15001
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。