前列腺影像报告和数据系统联合前列腺特异性抗原的Logistic回归模型评价外周带前列腺癌的诊断效能

目的 建立前列腺影像报告和数据系统第2版(PI-RADS v2)与前列腺特异性抗原(PSA)结合的Logistic回归预测模型,评价其对外周带前列腺癌的诊断能力.方法 回顾性分析经病理证实的外周带前列腺69例、非前列腺癌109例的患者术前MRI及前列腺特异性抗原资料.应用PI-RADS v2对外周带前列腺癌行发病风险评分.分析外周带前列腺癌组与非前列腺癌组PI-RADS v2评分、总PSA、游离与总PSA比值、前列腺特异性抗原密度(PSAD)及外周带前列腺特异性抗原密度(PZ-PSAD)指标的差异,选择具有统计学意义的指标作为自变量,以病理结果是否为前列腺癌作为因变量,拟建立四项Logistic回归模型:A、PI-RADS v2+总PSA;B、PI-RADS v2+游离与总PSA比值;C、PI-RADS v2+PSAD;D、PI-RADS v2+PZ-PSAD,建立Logistic回归模型产生的P和PI-RADS v2评分的受试者工作曲线,评估其诊断效能.结果 前列腺癌组与非前列腺癌组PI-RADS v2评分、总PSA、游离与总PSA比值、PSAD、PZ-PSAD差异有统计学意义(P<0.01),纳入因变量.外周带前列腺癌Logistic回归预测模型建立如下:A:Logit P=-6.825+1.024PI-RADS v2+0.223总PSA、B:Logit P=-4.354+1.586PI-RADS v2-12.7841游离与总PSA比值、C:Logit P=-8.993+1.630PI-RADS v2+17.091PSAD、D:Logit P=-9.434+1.596PI-RADS v2+10.494PZ-PSAD.A、B、C、D模型产生的Logit P预测概率,其受试者工作曲线下面积高于PI-RADS v2,差异具有统计学意义(Z=2.44、2.68、3.11、3.41,P<0.05).结论 联合PI-RADS v2评分与前列腺特异性抗原指标的Logistic回归预测模型对外周带前列腺癌的诊断效能优于单独使用PI-RADS v2评分,为可疑外周带前列腺癌患者行穿刺提供了更可靠的依据.

Value of PI-RADS v2 scores combined with prostate specific antigen in diagnosis of peripheral zone prostate cancer:a logistic regression analysis

Objective To assess the value of Prostate Imaging and Reporting and Data System:Version 2 (PI-RADS v2) combined with prostate specific antigen (PSA) in the diagnosis of peripheral zone (PZ) prostate cancer (PCa). Methods The preoperative magnetic resonance imaging and PSA data were ananlyzed for 69 patients with pathologically confirmed PCa and 109 non-PCa patients. PI-RADS v2 scores (1-5) was used to evaluate the risk of PZ PCa. The total PSA (tPSA) level, free to total PSA ratio (f/t PSA), PSA density (PSAD), PZ-PSAD and PI-RADS v2 scores were compared between the PCa and non-PCa patients. Logistic regression models were established with parameters that differed significantly the two groups. The receiver opearting characteristics (ROC) curve was constructed based on the P values derived from the logical regression models and PI-RADS scores to assess the diagnostic efficiency. Results PI-RADS v2 score, tPSA, f/t PSA, PSAD and PZ-PSAD differed significantly between the two groups (P<0.01). Four predictive multivariate models were established: Logit P=-6.825+1.024PI-RADS v2+0.223tPSA (A), Logit P=-4.354+1.586PI-RADS v2-12.7841f/tPSA (B), Logit P=-8.993+1.630PI-RADS v2+17.091PSAD (C), and Logit P=-9.434+1.596PI-RADS v2+10.494PZ-PSAD (D), whose area under the ROC curves was 0.908, 0.891, 0.944, and 0.961, respectively, all significantly greater than that of PI-RADS v2 score (P<0.05). Conclusion Compared with PI-RADS v2 score alone, the combination of PI-RADS v2 score and PSA in the logistic regression model can improve the diagnostic efficiency of PZ PCa and offers better confidence in the decision of biopsy in suspected cases.