Department of Maternal-Child and Urological Sciences, Sapienza University Rome, Policlinico Umberto I Hospital, Rome, Italy.
Cancer Prognostics and Health Outcomes Unit, Division of Urology, University of Montréal Health Center, Montréal, Québec, Canada.
Prostate. 2022 Jun;82(10):1051-1059. doi: 10.1002/pros.24350. Epub 2022 Apr 11.
Contemporary seminal vesicle invasion (SVI) rates in National Cancer Comprehensive Network (NCCN) high-risk prostate cancer (PCa) patients are not well known but essential for treatment planning. We examined SVI rates according to individual patient characteristics for purpose of treatment planning.
Within Surveillance, Epidemiology, and End Results (SEER) database (2010-2015), 4975 NCCN high-risk patients were identified. In the development cohort (SEER geographic region of residence: South, North-East, Mid-West, n = 2456), we fitted a multivariable logistic regression model predicting SVI. Its accuracy, calibration, and decision curve analyses (DCAs) were then tested versus previous models within the external validation cohort (SEER geographic region of residence: West, n = 2519).
Out of 4975 patients, 28% had SVI. SVI rate ranged from 8% to 89% according to clinical T stage, prostate-specific antigen (PSA), biopsy Gleason Grade Group and percentage of positive biopsy cores. In the development cohort, these variables were independent predictors of SVI. In the external validation cohort, the current model achieved 77.6% accuracy vs 73.7% for Memorial Sloan Kettering Cancer Centre (MSKCC) vs 68.6% for Gallina et al. Calibration was better than for the two alternatives: departures from ideal predictions were 6.0% for the current model vs 9.8% for MSKCC vs 38.5% for Gallina et al. In DCAs, the current model outperformed both alternatives. Finally, different nomogram cutoffs allowed to discriminate between low versus high SVI risk patients.
More than a quarter of NCCN high-risk PCa patients harbored SVI. Since SVI positivity rate varies from 8% to 89%, the currently developed model offers a valuable approach to distinguish between low and high SVI risk patients.
目前尚不清楚国家综合癌症网络（NCCN）高危前列腺癌（PCa）患者的精囊侵犯（SVI）率，但这对于治疗计划至关重要。我们根据患者的个体特征检查了 SVI 率，以进行治疗计划。
在监测、流行病学和最终结果（SEER）数据库（2010-2015 年）中，确定了 4975 名 NCCN 高危患者。在开发队列（SEER 居住地地理区域：南部、东北部、中西部，n=2456）中，我们拟合了一个多变量逻辑回归模型，用于预测 SVI。然后，在外部验证队列（SEER 居住地地理区域：西部，n=2519）中，对该模型的准确性、校准和决策曲线分析（DCAs）进行了测试，以评估其与之前模型的对比。
在 4975 名患者中，有 28%的患者存在 SVI。根据临床 T 分期、前列腺特异性抗原（PSA）、活检 Gleason 分级组和阳性活检核心百分比，SVI 率从 8%到 89%不等。在开发队列中，这些变量是 SVI 的独立预测因子。在外部验证队列中，当前模型的准确率为 77.6%，而 Memorial Sloan Kettering Cancer Centre（MSKCC）为 73.7%，Gallina 等人的模型为 68.6%。与两种替代方案相比，校准更好：当前模型的预测偏差为 6.0%，MSKCC 为 9.8%，Gallina 等人的模型为 38.5%。在 DCA 中，当前模型优于两种替代方案。最后，不同的诺模图截止值可以区分低 SVI 风险和高 SVI 风险患者。
超过四分之一的 NCCN 高危 PCa 患者存在 SVI。由于 SVI 阳性率从 8%到 89%不等，因此当前开发的模型提供了一种有价值的方法来区分低 SVI 风险和高 SVI 风险患者。
