RESUMEN
Objective To investigate the related factors on effects of uterine artery embolization (UAE) in the treatment of dysmenorrhea in patients with adenomyosis, and to construct and validate the efficacy prediction model. Methods A total of 127 cases of adenomyosis patients with symptoms of dysmenorrhea in Guangzhou No.1 People′s Hospital and Nanfang Hospital of Southern Medical University from June 1999 to December 2009 were reviewed. The evaluation standard was to improve the degree of dysmenorrhea, the related factors of efficacy were analysed. Combined with artificial neural network theory, the effect prediction model was constructed, and the effectiveness of the model was evaluated using receiver operating characteristic (ROC) curve, and the effectiveness of the cut-off point was calculated. The model was validated by 68 cases of patients with adenomyosis in the Nanfang Hospital from January 2010 to November 2014. Results (1) In 127 cases of dysmenorrhea patients, UAE treatment was effective in 98 cases, effective rate was 77.2% (98/127). (2) Age was an independent predictor of effective UAE treatment (HR=1.129, P=0.026);in the range of this study, the greater the age, the higher the UAE treatment efficiency. (3) The developing situation of ovary branches of uterine artery was an independent predictor of effective UAE treatment (HR=0.460, P=0.020), the efficiency of patients whose intraoperative bilateral uterine artery ovarian branch did not develop was 89.7%(35/39), the efficiency of patients whose unilateral uterine artery ovarian branch was developing was 84.1% (37/44) and the efficiency of patients whose bilateral uterine artery ovarian branch were developing was 59.1% (26/44). (4) Blood supply of adenomyosisis was an independent predictor of effective UAE treatment (HR=0.313, P=0.001). Type Ⅰ (bilateral predominated) patients, efficiency was 93.5%(43/46);typeⅡ(bilateral balanced) patients, efficiency was 78.0%(39/50);type Ⅲ (unilateral predominated) patients, efficiency was 51.6% (16/31). (5) UAE for the treatment of adenomyosis efficacy of artificial neural network prediction model was constructed, the model′s area under the ROC curve was 0.808, the optimal cut-off point was 0.669 13. Actual verification of the model, sensitivity was 96.5%, specificity was 81.8%, positive predictive value was 96.5%and negative predictive value was 81.8%, the total accuracy was 94.1%. Conclusions (1) Age, the developing situation of ovary branches and blood supply of adenomyosis are the independent predictors of effective UAE treatment. (2) The artificial neural network prediction model is satisfied with the accuracy and the accuracy of prediction.