Establishment of a prediction model for biochemical recurrence after radical prostatectomy for prostate cancer
摘要
目的 建立基于超声影像特征-临床病理多模态融合的前列腺癌根治术(Radical prostatectomy,RP)后生化复发(Biochemical recurrence,BCR)预测模型,为中国患者提供精准、经济的术后风险分层工具.方法 回顾性纳入2018 年1 月至2023 年12 月行腹腔镜/机器人辅助 RP 的 320 例局限性前列腺癌患者(BCR 组 108 例,非 BCR 组212 例).收集临床病理资料(如术前 PSA、术后 Gleason 评分)、多参数超声特征(灰阶、彩色多普勒、弹性成像、超声造影)及 MRI 参数(DWI、ADC 值),通过单因素及多因素 Logistic 回归筛选 BCR 独立预测因素,构建列线图模型并评估其诊断效能.结果 BCR 发生率为 33.8%(108/320),中位复发时间 13 个月.多因素分析显示,f/t PSA 比值(OR=20.797)、术后 Gleason 评分(OR=6.889)、DWI 高信号(OR=11.678)、Adler 血流分级Ⅲ级(OR=7.786)及手术切缘阳性(OR=3.495)是 BCR 的独立预测因素(均 P<0.001).基于上述因素构建的列线图模型 AUC 为0.918(95%CI:0.885~0.951),灵敏度88.9%,特异度78.3%,显著优于单一指标(P<0.001).结论 本研究建立的多模态融合模型可精准预测中国前列腺癌患者 RP 术后 BCR 风险,为优化术后随访策略及个体化辅助治疗提供重要依据.
Objective To develop a multimodal prediction model for biochemical recurrence(BCR)after radical prostatectomy(RP)based on ultrasound imaging features and clinicopathological factors,providing an accurate and cost-effective risk stratification tool for Chinese patients.Methods A total of 320 patients with localized prostate cancer who underwent laparoscopic/robot-assisted RP between January 2018 and December 2023 were retrospectively enrolled(108 in the BCR group and 212 in the non-BCR group).Clinicopathological variables,including preoperative prostate-specific antigen(PSA)level and postoperative Gleason score,multiparametric ultrasonographic features(grayscale ultrasonography,color Doppler ultrasonography,elastography,and contrast-enhanced ultrasonography),and magnetic resonance imaging(MRI)parameters,including diffusion-weighted imaging(DWI)and apparent diffusion coefficient(ADC)values,were collected.Univariate and multivariate Logistic regression analyses were performed to identify independent predictors of BCR,and a nomogram model was constructed to evaluate its diagnostic performance.Results:The BCR incidence was 33.8%(108/320)with a median recurrence time of 13 months.Multivariate analysis showed that f/t PSA ratio(OR=20.797),postoperative Gleason score(OR=6.889),DWI hyperintensity(OR=11.678),Adler blood flow grade Ⅲ(OR=7.786),and positive surgical margin(OR=3.495)were independent predictors of BCR(all P<0.001).The nomogram model based on these factors had an AUC of 0.918(95%CI:0.885~0.951),with a sensitivity of 88.9%and specificity of 78.3%,significantly outperforming single indicators(P<0.001).Conclusion The multimodal fusion model established in this study can accurately predict the risk of biochemical recurrence following radical prostatectomy in Chinese prostate cancer patients.This model may facilitate postoperative risk stratification,optimize follow-up strategies,and support individualized decisions regarding adjuvant therapy.