初诊前列腺癌骨扫描的必要性分析

Necessity analysis of bone scan in patients with newly diagnosed prostate cancer

目的建立用于预测中国初诊前列腺癌骨转移风险的回归树模型，以减少不必要的骨扫描。方法在2005至2011年复旦大学附属肿瘤医院住院的501例初诊前列腺癌中采用分类及回归树（CART）分析建立Fudan回归树模型，同时验证Briganti回归树模型，并比较两个模型在曲线下面积（AUC）及临床应用价值上的优劣性。结果本组骨转移的发生率为27．5％（138／501）。Fudan回归树模型、Briganti回归树模型及单以骨转移相关症状（SRE）为标准预测骨转移的准确度分别为0．813、0．691及0．645，三者间比较差异有统计学意义（P〈0．05）。当决策阈值概率取值范围为（24．2％，36．8％）时，Fudan回归树模型具有更低的漏诊率及骨扫描过度检查率。结论Fudan回归树模型具有较高的预测价值，同时能减少中国初诊前列腺癌不必要的骨扫描。

Objective To construct a classification and regression tree （CART） to predict the occurrences of bone metastases in patients with newly diagnosed prostate cancer so as to reduce unnecessary bone scans. Methods CART analyses were performed in 501 subjects from 2005 to 2011 of Fudan University Shanghai Cancer Center to establish Fudan CART model and externally validate Briganti＇s CART model. The both models were compared with regards to the area under the curve （AUC） and their clinical values. Results The rate of bone metastasis was 27.5% （ 138/501 ）. The predictive accuracy of Fudan CART model, Briganti＇s CART model and skeleton-related events （SRE） model was O. 813, 0. 691 and O. 645 respectively. There were statistically significant differences （ P 〈 0. 05 ）. Fudan CART model had a lower missed diagnosis and an over-examination rate of bone scan within the probability threshold （Pt） range of 24. 2% to 36. 8%. Conclusion With a higher predictive value, Fudan CART model may be employed to reduce the unnecessary bone scans for Chinese patients with newly diagnosed orostate cancer.