The Methods Of Age Estimation For Chinese Children And Adolescents

Objectives:As the most rapid growth period,the applications of age estimation technique during the child and adolescent phases are getting more and more attention in the filed of sports,forensic science and clinical medicine.By adopting the biological indicators on different dimensions,we aim to establish an age estimation system in Chinese children and adolescents with improved accuracy.Methods:1)A total of 14,970 children aged from 7 to 12 years old were included in the first study,which involved a large-scale school survey with a city-district-school-class sampling scheme conducted in Shanghai,China,in 2014.A gender-specific equation for age estimation was built upon a generalized linear model,in which Chronological age(CA)was the dependent variable,and the main effects and interaction of Tanner stages and percentage of the predicted mature height at a given age(PPMH)as predictors.The k-fold cross validation(k=5)procedure was applied to verify the estimation equation within the sample.The cross-sample validity of the equation was tested in another age-matched sample.The validation of models was measured by mean absolute error(MAE)and root of mean standard error(RMSE).2)A total of 138 children aged from 6 to 15 years were included in the second study.The skeletal age(SA)was evaluated by three methods:the Greulich and Pyle atlas method(GP),the Tanner and Whitehouse method version 3 on radius-ulna-short bones(TW3-Rus),and that on carpal bones(Carp)according to the characteristics on the X-ray of left hand.The dental age(DA)was evaluated by two methods:the Demirjian method and the Willems method by the orthopantomogram.Estimation methods were compared based on MAE,RMSE and R~2.Within each dimension(i.e.,SA and DA),a best indicator was selected to build a combined model.Gender-specific linear regression models were built,in which the chronological age(CA)was the dependent variable,and the multiple indicators as predictors.Further comparison of the age estimation accuracy was conducted between predictions by single measure and by combined measures.3)In the third study,130 healthy children with blood samples from the second study were included.The identification and validation of the age-related DNA Cp G loci were conducted through genome-wide DNA methylation profiling and digital PCR technique.The stepwise linear regression models were built and validated based on the combination of epigenetic biomarkers and biological indicators.Results:1)The MAE of the equation built based on anthropometric indicators was 0.60 years for boys and 0.59 years for girls,respectively.The results from 5-fold cross-validation showed that the MAE was 0.59-0.61 years for boys and 0.57-0.61 years for girls.The equations were further validated in another testing sample.The MAE in the cross-sample validation was 0.76 years for boys and girls.The paired t-tests indicated no significant differences between EA and CA,which supported the reliability of the equations.2)The TW3-Rus method was the best indicator for boys and the GP method for girls in all three SA measures.The Willems method was the better indicator in both genders in terms of the dental age evaluation.According to the results of analysis about boys,the MAE between EA and CA was 0.52 years,which decreased by 0.14 years from the best single method;for girls,the MAE between EA and CA was 0.43 years,which decreased by 0.09years from the best single method.3)After the identification and validation of the age-related DNA Cp G loci through genome-wide DNA methylation profiling and digital PCR technique,four Cp G sites(cg7164639,cg14517133,cg18647570 and cg22398226)were finally obtained.The final gender-specific linear regression model was established with the combination of skeletal and dental indictors selected in the previous analysis and the validated DNA Cp G sites.The estimation accuracies of final model as reflected by MAE were further improved by 0.03years from the combined method of SA and DA among boys,and were further improved by0.10 years among girls.Conclusion:1)In the first study,we successfully established an age estimation equation based on anthropometric indicators among Chinese children.Within-sample as well as cross-sample validations suggested this method provides accurate estimates comparable to skeletal age estimation considered as the golden standard.2)The combined measure of SA and DA has proved its significance in further improving accuracy,which provide us an appropriate option for the various applications.3)The findings indicated that the final model established on the epigenetic biomarkers and multiple biological indicators can achieve a higher accuracy in accordance with our goal.