The Roche-Wainer-Thissen (RWT) prediction model, developed in 1975, predicts the adult stature of a child based on age, current stature, current weight, current skeletal age, and the average stature of the parents. Generally, the method has worked well; however, there are certain steps in the procedure that can be improved. Seven variations of the current version of the RWT prediction model are investigated and compared in terms of the accuracy and reliability of prediction, culminating in a recommendation for the prediction of adult stature in Caucasian Americans. The recommended method, called multivariate cubic spline smoothing [MCS2(1)], uses cubic splines in the smoothing part of the RWT prediction model, resulting in a simpler (i.e., fewer steps) method with smaller maximum deviations between predicted and actual adult statures than the current multivariate semi-metric smoothing (MS2) method.