PREDICTION ACCURACY OF ADULT HEIGHT FROM BONE AGE IN BOYS. 566

Accuracy of the Bayley-Pinneau (BP), Tanner-Whitehouse (TW2), and Roche-Wainer-Thissen (RWT) adult height prediction models were compared for 23 normally growing boys every 8 mo from 8-16 yr. The RWT model was tested using the Greulich-Pyle (RWT-GP) and Fels (RWT-Fels) bone ages. To assure prediction accuracy, bone age assessments and height predictions were made by an originator/expert of each method. The influence of maturation on the predictions was assessed by grouping by bone age. Comparison of the average quadratic loss (cm²): Σi (y_i - f)²/n (where f = final height, y_i = height prediction [Illegible Text] calendar age c_i and n = # of observations) are presented in theTable. The smaller the absolute value the more accurate the prediction. The RWT and BP models, on average, overpredicted adult height in normally growing boys. The TW2 model generally underpredicted unless the bone age was corrected to American standards of maturation, when it slightly overpredicted though less than the RWT and BP models. Conclusion: our data are probably as accurate and precise as these methods permit; yet, biologically significant prediction error remains. Supported in part by NIH GCRC Grant RR 00847

Table 1