用ROC曲线评价雌激素、卵泡刺激素、黄体生成素判断乳腺癌患者绝经状态

Evaluation of the role of estradiol, follicle stimulating hormone and luteinizing hormone in discriminating menopausal status of invasive breast cancer patients by receiver operating characteristics curve

目的应用ROC曲线评价雌二醇(estradiol,E2)、卵泡刺激素(follicle stimulating hormone,FSH)和黄体生成素(luteinizing hormone,LH)3项性激素对浸润性乳腺癌患者绝经状态的判断能力。方法分析本院128例绝经前及204例已绝经浸润性乳腺癌患者的性激素检测结果。应用ROC曲线评价单项性激素及它们联合的判断能力和判断点(Cut-offPoint)。结果单项E2、FSH和LH的ROC曲线下面积分别为0.9395(95%CI:0.9110～0.9679)、0.9714(95%CI:0.9514～0.9914)和0.9053(95%CI:0.8681～0.9424);Youden指数最大时的判断点分别为49.52pg/ml、21.96mU/ml和13.25mU/ml。FSH的ROC曲线下面积分别和E2(P=0.036)、LH相比(P<0.001),差别均具有统计学意义。经计算E2和FSH进入联合判断模型,联合判断的曲线下面积为0.9774(95%CI:0.9597～0.9952),Youden指数最大时的判断点的绝经概率(P绝经)=0.46。结论单项E2、FSH和LH均可判断乳腺癌患者的绝经状态,其中FSH的能力好于E2和LH。而E2和FSH的联合判断好于单项指标。由于LH的ROC曲线下面积最小,经计算没能进入联合判断模型,故不推荐用其判断乳腺癌患者的绝经状态。可根据临床需要,调整ROC曲线的绝经概率,选择判断点,并了解在该判断点下的误诊率或漏诊率的大小。

Objective To evaluate the roles of estradiol （E2）, follicle stimulating hormone （FSH） and luteinizing hormone （LH） in discriminating menopausal status of invasive breast cancer patients using receiver operating characteristics （ROC） curve. Methods Analysis was undertaken among 128 premenopausal and 204 postmenopausal invasive breast cancer patients. The efficient and useful cut-off points of the three sex hormones were achieved based on ROC curve. Results The areas under the ROC curve of E2, FSH and LH were 0.939 5 （95%CI： 0.911 0-0.967 9）, 0.9714 （95%CI： 0.951 4-0.991 4） and 0.905 3 （95%CI：0.868 1-0.942 4） respectively, and the cut-off point was 49.52 pg/ml, 21.96 mU/ml and 13.25 mU/ml respectively, with a maximum of the Youden index. The area under the ROC curve of FSH was significantly better than E2 （P=0.036） and LH （P〈0.001）. E2 and FSH were selected into the logistic regression model, the area under the ROC curve was 0.977 4 （95%CI： 0.959 7-0.995 2）. The probability of menopause （Pm） was 0.46 with a maximum of the Youden index. Conclusions E2, FSH and LH can effectively discriminate menopausal status of invasive breast cancer patients. The discrimination capabicity of FSH is significantly better than that of E2 or LH. The discrimination capability of combining E2 and FSH is better than using any of the three hormones independently. LH is not recommended to identify menopausal status, because it fails to be selected into the logistic regression model, and the area under the ROC curve is the least among the three hormones. The cut-off point may be adjusted according to the demand of clinical practice.