Smoleń Agata, Stachowicz Norbert, Czekierowski Artur, Kotarski Jan
Zakład Matematyki i Biostatystyki Medycznej, Uniwersytet Medyczny w Lublinie.
Ginekol Pol. 2010 Apr;81(4):254-61.
The aim was to evaluate the risk of ovarian tumor malignancy based on logistic regression analysis and to construct a practical tool which might be used at the bedside.
379 women with adnexal masses were enrolled to the study The results of the two-dimensional (2D) gray-scale ultrasound examination, color and spectral Doppler blood flow measurement, three-dimensional (3D) sonoangiography examination and serum levels of CA-125 measurement were analyzed. 160 out of 379 women with abnormal findings, (42.2%), were diagnosed with a malignant ovarian tumor and 219 (57.8%) were found to have a non-malignant adnexal mass. To improve the sensitivity and specificity of ultrasound scan findings, postmenopausal status and preoperative serum CA-125 levels, the risk of malignancy index (RMI) was calculated and compared to the result of logistic regression analysis. Furthermore, the nomograms applicable at the bedside for estimation of the probability of malignancy for the examined adnexal tumor were derived.
The highest values of the area under the receiver operating characteristic (ROC) curves in univariate analysis were 0.87 for serum levels of CA-125 and 0.83 for the color score. For the RML, the value of the area under ROC curve was 0.91. The results of logistic regression revealed that papillary growth, color score, age, pulsatility index (PI) and level of serum CA 125 were significant factors in the multivariate model. The obtained classification accuracies for MLRA model in the training set and the test set were 88% and 92%, respectively. The value of the area under the ROC curve was 0.94 and significantly differed from the value of area for CA-125 (p<0.05).
The results of the study suggest that the combined use of ultrasound parameters and serum level of CA-125 in the prognostic model improves the diagnostic precision in the primary discrimination of benign and malignant ovarian tumors in women.
旨在基于逻辑回归分析评估卵巢肿瘤恶性风险，并构建一种可在床边使用的实用工具。
379例附件包块女性纳入本研究。分析二维（2D）灰阶超声检查、彩色和频谱多普勒血流测量、三维（3D）血管造影检查结果以及血清CA-125测量水平。379例检查结果异常的女性中，160例（42.2%）被诊断为卵巢恶性肿瘤，219例（57.8%）为非恶性附件包块。为提高超声扫描结果、绝经后状态和术前血清CA-125水平的敏感性和特异性，计算恶性风险指数（RMI）并与逻辑回归分析结果进行比较。此外，还得出了适用于床边估计所检查附件肿瘤恶性概率的列线图。
单因素分析中，受试者操作特征（ROC）曲线下面积的最高值，血清CA-125水平为0.87，彩色评分值为0.83。对于RMI，ROC曲线下面积值为0.91。逻辑回归结果显示，乳头状生长、彩色评分、年龄、搏动指数（PI）和血清CA 125水平是多变量模型中的显著因素。训练集和测试集中MLRA模型的分类准确率分别为88%和92%。ROC曲线下面积值为0.
