Receiver operating characteristic curve analysis in bone age estimation for girls with idiopathic precociouspuberty

Objective To investigate the diagnostic value of the receiver operating characteristic (ROC) curve analysis of bone age for female idiopathic precocious puberty. Methods Fifty-five girls aged3. 2~9. 9(7. 55 ± 1. 90) years with idiopathic precocious puberty were enrolled. Among them 36 cases under 8 (6. 07± 1. 72) years old had breast development ; 20 cases aged 5. 0~9. 75 (8. 75 ± 1. 26) years experienced menophania. Four control groups were designed : Control group A had 14 cases over 10(11. 45 ± 1. 05) years old with menophania; Control group B 25 cases aged 8~9 (9. 2 ± 0. 67) years with low-weight ; Control group C 32 cases aged 8~10 (9. 11 ± 0. 74) years with normal weight ; Control group D 12 cases aged 8~10 (9. 61 ±0. 66) years with over-weight. The first radiographs of the left hand and wrist were used for bone age estimation with a single-blind analysis using the Tanner and Whitehouse(TW2) methods. Systems of RUS, carpal, and 20 bones were used for bone ages. Five decision thresholds were set in each system. Results Four points were relatively high for both sensitivity and specificity: (1)RUS of >P90 (Sen = 0. 836, Spe = 0. 916), (2)carpale of >P90 (Sen = 0. 746, Spe =0. 916),(3) 20 bone of >P90 (Sen = 0. 746, Spe = 0. 964) and (4) >P75 (Sen = 0. 982,Spe = 0. 783); For the area under a ROC curve, RUS of A13 = 0. 939 ± 0. 019 (95%CI:0. 902~0. 977), carpale of A7 =0. 899 ± 0. 028 (95%CI:0. 845 ~ 0. 954), 20 bone of A20 = 0. 958 ± 0. 014 (95%CI: 0. 930 ~ 0. 986), F =2. 03, P=0. 13. Conclusion The bone age of RUS and 20 bones were high in accuracy for diagnosing idiopathic precocious puberty. The decision threshold of RUS of >P90 was the best diagnostic point when both sensitivity and specificity were considered. Key words: age determination by skeleton; idiopathic precocious puberty; receiver operating characteristic (ROC) curve