The More the Merrier: Integrating Multiple Models of Skeletal Maturity Improves the Accuracy of Growth Prediction. Issue 7 (9th August 2021)

• Record Type: Journal Article
• Authors: Munger, Alana M.; Yu, Kristin E.; Li, Don T.; Furdock, Ryan J.; Boeyer, Melanie E.; Duren, Dana L.; Weber, David R.; Cooperman, Daniel R.
• Is Part Of: Journal of pediatric orthopaedics. Volume 41:Issue 7(2021)
• Page Start: e580
• Page End: e584
• Publication Date: 2021-08-09
• Subjects: pediatric orthopaedics -- growth prediction -- limb length discrepancy -- scoliosis -- Greulich and Pyle Atlas -- the Fels Method -- and the Sanders Hand Classification; Pediatric orthopedics -- Periodicals; 618.927
• DOI: 10.1097/BPO.0000000000001847
• Languages: English
• ISSNs: 0271-6798

Abstract

Abstract : Background: Multiple systems using radiographic skeletal markers to measure development have been described, including the Greulich and Pyle Atlas (GP), the Fels Method (Fels), and the Sanders Hand Classification (Sanders). The purpose of this study was to quantitatively assess whether the integration of skeletal maturity assessment methods and demographic variables improves the accuracy of pediatric growth predictions over the use of skeletal markers or chronologic age alone. Methods: The Brush Inquiry contains prospectively collected longitudinal data on children who lived in Cleveland, Ohio between 1926 and 1942. A total of 16 boys and 29 girls were selected for study. All had age, height, and an anteroposterior radiograph of the hand at each of 3 visits. Those visits occurred at 85%, 90%, and 95% of final height. We determined the growth completed at each visit by dividing the height observed by the final height at skeletal maturity. Boys and girls were analyzed separately using chronologic age, height, GP, Fels, and Sanders. The residual difference between the height predicted and actual height, as well as the SD of the prediction error of the cohort at each time point was calculated. To account for multiple visits from each subject, all linear models were produced using the generalized estimating equations (GEEs) procedure. Results: For boys, age, GP, and Fels performed similarly in predicting growth remaining at all 3 time points. For girls, age, GP, and Fels performed similarly in predicting growth remaining at the 85% and 95% time points; however, the Fels Method demonstrated improved performance at the 90% time point compared with chronologic age ( P =0.0076) and GP alone ( P =0.0155). For both boys and girls, the most accurate multivariate GEE model with the lowest SD of prediction error integrated Fels, age, GP, Sanders, and height. Conclusions: The most accurate multivariate GEE model of growth prediction for both boys and girls integrated Fels, age, GP, Sanders, and height. When calculating the amount of growth remaining, it is prudent to integrate multiple systems for greater predictive accuracy. Level of Evidence: Level III.