Kerkmeijer Linda G W, Monninkhof Evelyn M, van Oort Inge M, van der Poel Henk G, de Meerleer Gert, van Vulpen Marco
Department of Radiation Oncology, University Medical Center Utrecht, Heidelberglaan 100, 3584 CX, Utrecht, The Netherlands.
Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, The Netherlands.
World J Urol. 2016 Jun;34(6):789-95. doi: 10.1007/s00345-015-1691-4. Epub 2015 Sep 29.
Current models for prediction of prostate cancer-specific survival do not incorporate all present-day interventions. In the present study, a pre-treatment prediction model for patients with localized prostate cancer was developed.
From 1989 to 2008, 3383 patients were treated with I-125 brachytherapy (n = 1694), external beam radiotherapy (≥74 Gy, n = 336) or radical prostatectomy (n = 1353). Pre-treatment parameters (clinical T-stage, biopsy grade, PSA and age) were related to the hazard of mortality by multivariate Cox proportional hazard analysis. The PRetreatment Estimation of the risk of Death In Cancer of the prosTate (PREDICT) model was developed. The predictive accuracy of the model was assessed by calibration and discrimination and compared to the Ash risk classification system.
Of the 3383 patients analyzed, 2755 patients (81 %) were alive at the end of follow-up, 149 patients (4 %) died of prostate cancer and 365 patients (11 %) died of other causes, and for 114 patients (3 %) cause of death was unknown. Median follow-up time was 7.6 years. After correction for overoptimism, the c-statistic of the prediction model for prostate cancer-specific mortality was 0.78 (95 % CI 0.74-0.82), compared to 0.78 (95 % CI 0.75-0.81) for the risk classification system by Ash et al. The PREDICT model showed better calibration than the Ash risk classification system.
The PREDICT model showed a good predictive accuracy and reliability. The PREDICT model might be a promising tool for physicians to predict disease-specific survival prior to any generally accepted intervention in patients with localized prostate cancer.
目前用于预测前列腺癌特异性生存的模型并未纳入所有当今的干预措施。在本研究中，我们开发了一种针对局限性前列腺癌患者的治疗前预测模型。
1989年至2008年期间，3383例患者接受了碘-125近距离放疗（n = 1694）、外照射放疗（≥74 Gy，n = 336）或根治性前列腺切除术（n = 1353）。通过多变量Cox比例风险分析，将治疗前参数（临床T分期、活检分级、前列腺特异性抗原和年龄）与死亡风险相关联。由此开发了前列腺癌死亡风险的治疗前估计（PREDICT）模型。通过校准和区分评估该模型的预测准确性，并与Ash风险分类系统进行比较。
在分析的3383例患者中，2755例患者（81%）在随访结束时仍存活，149例患者（4%）死于前列腺癌，365例患者（11%）死于其他原因，114例患者（3%）的死亡原因不明。中位随访时间为7.6年。校正过度乐观偏差后，前列腺癌特异性死亡率预测模型的c统计量为0.78（95%CI 0.74 - 0.82），而Ash等人的风险分类系统的c统计量为0.78（95%CI 0.75 - 0.81）。PREDICT模型显示出比Ash风险分类系统更好的校准。
PREDICT模型显示出良好的预测准确性和可靠性。对于局限性前列腺癌患者，在进行任何普遍接受的干预之前，PREDICT模型可能是医生预测疾病特异性生存的一个有前景的工具。
