Establishment Of Diagnostic Model For Girls With Central Precocious Puberty

Objective To establish a logistic regression model for preliminary screening diagnosis of girls with precocious puberty using bone age,gynecological ultrasound,hormone levels and other auxiliary examinations.After screening,Gn RH stimulation test was been done to improve the accuracy rate of central precocious puberty(CPP)diagnosis,and reach the goal of an economic,effective,early diagnosis and treatment of CPP.Finally improve patient compliance.Methods From June 2015 to December 2018,767 precocious puberty girls who visited the outpatient department of child precocious puberty of Subei hospital,clinical medical college of Yangzhou university were studied.They were classified into CPP group and PPP group.The Wilcoxon rank sum test was used to compare the clinical characteristics between the two groups.Logistic regression analysis was used to screen for risk factors of CPP,and finally a logistic regression model was obtained.The receiver operating characteristic curve(ROC)was used to evaluate the diagnostic value of the model for CPP.The best diagnostic boundary was determined based on the cut value.The paired comparison method was used to compare the diagnostic value of the diagnostic model.Results 1.Among 767 precocious puberty girls,146 patients were CPP,and621 were PPP.Breast emerging age(years),chronological age(years),height(cm),weight(kg),body mass index(BMI,kg/m~2),menarche,advanced bone age(month),endometrial appearance,uterine length(cm),uterine volume(ml),ovarian volume(ml),large follicle(diameter>4 mm)proportion,estradiol(pmol/L),basal LH values(IU/L),basal FSH values(IU/L),and basal LH values/basal FSH values were statistically significant between CPP group and PPP group.logistic regression found chronological age(OR=1.953,95%CI:1.289-2.961),advanced bone age(OR=1.156,95%CI:1.107-1.207),uterine volume(OR=2.480,95%CI:1.436-4.282),BMI(OR=0.812,95%CI:0.684-0.965),estradiol(OR=1.012,95%CI:1.005-1.020)and basal LH value/basal FSH value×100(OR=1.069,95%CI:1.050-1.088)were the risk factors of diagnosing CPP,and established a diagnostic model Y=-9.184+0.670×chronological age(years)+0.145×advanced bone age(month)+0.908×uterine volume(ml)+0.012×estradiol(pmol/L)-0.208×BMI(kg/m~2)+0.067×basal LH value/basal FSH value×100.The ROC curve suggested that the model had a predicted AUC of 0.941(P<0.001)for diagnosing CPP,and the corresponding yoden index(0.783)with a cut value of-0.69.The sensitivity was 89.6%and specificity was88.7%.2.According to the age of visit,there were 535 girls of precocious puberty whose age were younger than 8 years old.Among these cases,logistic regression found that advanced bone age(OR=1.130,95%CI:1.084-1.179),uterine volume(OR=2.998,95%CI:1.586-5.668),basal LH value/basal FSH value×100(OR=1.054,95%CI:1.035-1.074)were the risk factors of diagnosing CPP,and established the diagnostic model Y1=-6.984+0.123×advanced bone age(month)+1.098×uterine volume(ml)+5.300×basal LH value/basal FSH value×100;the ROC curve indicated that the area under the curve(AUC)of model Y1 was 0.888(P<0.001);the cut-off value was-0.51,the corresponding yoden index was the largest(0.670),and the sensitivity and specificity were 75.5%and 91.5%.232 girls with precocious puberty were older than or equal to 8 years old,logistic regression found that BMI(OR=0.571,95%CI:0.371-0.881),advanced bone age(OR=1.246,95%CI:1.114-1.394),uterine volume(OR=4.639,95%CI:1.515-14.207),estradiol(OR=1.016,95%CI:1.004-1.028),basal LH value/basal FSH value×100(OR=1.108,95%CI:1.060-1.157)were the risk factors of diagnosing CPP,and another diagnostic model were established.Y2=-0.722-0.560×BMI(kg/m~2)+0.220×advanced bone age(month)+1.534×uterine volume(ml)+10.200×basal LH value/basal FSH value×100+0.016×estradiol(pmol/L).The ROC curve suggested that AUC of model Y1 was 0.976(P<0.001),the cut-off value was-0.39,the corresponding yoden index was the largest(0.838),and the sensitivity and specificity were 95.3%and 88.5%.3.In 49 CPP girls younger than 8 years old,the diagnostic accuracy of model Y and Y1 was 97.9%and75.5%,respectively.In 86 CPP girls aged 8 years or older,the diagnostic accuracy of model Y and Y2 was 96.5%and 95.3%,respectively.Conclusion This study established the diagnostic models for CPP girls:Y=-9.184+0.670×chronological age(years)+0.145×advanced bone age(month)+0.908×uterine volume(ml)+0.012×estradiol(pmol/L)-0.208×BMI(kg/m~2)+0.067×basal LH value/basal FSH value×100.The diagnostic cutoff values were-0.69.The relevant indicators of precocious puberty girls who were suspected of CPP were brought into the model.If the result was greater than the diagnostic threshold,Gn RH stimulation test should be performed to determine whether it is CPP.