By Jonas Desloovere, Honours student Physical Education (Ghent University), and Dr. Pieter Vansteenkiste, postdoctoral researcher in sports science (Ghent University), expert in motor control and talent development.

Introduction

Predicting the adult height of a youth athlete can help coaches and sports organizations to make more informed decisions about developmental pathways, talent identification, and players’ positions. Based on biological principles such as genetics, anthropometry, and skeletal maturation, several methods have been developed to estimate adult height of youth athletes. Each method varies in the required input variables, practicality, accuracy, and invasiveness. Below we provide a short overview of the most common methods to estimate adult height for youth athletes.

Skeletal age

HOW IT WORKS By taking an X-ray of the hand and wrist, the degree of ossification of the growth plates can be compared to reference images to determine skeletal age. As skeletal age is strongly associated with biological maturity, reference tables or prediction equations can subsequently be used to predict the adult height. This method is still considered the gold standard for height prediction.

The skeletal age method is commonly used for children between 5 and 18 years old and is often applied during puberty, when interindividual differences in growth rates are large. Some of the most common methods to determine skeletal age and subsequently predict adult height are Bayley-Pinneau ⁽¹⁾, Greulich-Pyle ⁽²⁾, Tanner-Whitehouse ⁽³⁾, Fels ⁽⁴⁾, Roche-Wainer-Thissen ⁽⁵⁾, and BoneXpert ⁽⁶⁾. Some more recent methods use ultrasound instead of X-rays ⁽⁶⁾, and apply AI or deep learning systems to evaluate the radiographs ⁽⁷⁾.

ACCURACY The mean error of predicted adult height is approximately ±2.6 cm ⁽⁸⁾, which makes it the most accurate method for estimating adult height.

ADVANTAGES Still considered the gold standard with highest accuracy. Useful in children with atypical growth patterns.

DISADVANTAGES Requires medical intervention (X-ray). Impractical for frequent screening. No information about growth velocity. Interpretation has to be performed by a trained professional.

Maturity offset methods

HOW IT WORKS Maturity offset methods such as Mirwald ⁽⁹⁾, Moore ⁽¹⁰⁾, or Fransen ⁽¹¹⁾ estimate the biological maturity by predicting how many years a child is away from the Age at Peak Height Velocity (APHV). Based on the APHV, a reference table can then be used to estimate the remaining growth ⁽¹²⁾. Alternatively, knowing that APHV occurs at approximately 90% of our adult height, adult height can be predicted using this percentage and the current height ⁽¹³⁾.

The maturity offset method uses anthropometric variables: chronological age, body height, sitting height, leg length (derived from the previous measures), and body weight. This method is typically applied around the age of peak height velocity (APHV). It is most appropriate for boys aged 12-16 years, and girls aged 10-14 years.

ACCURACY The validity of this method decreases in children who are further away from the APHV. There is often an underestimation of predicted APHV in younger children and an overestimation at older ages. However, these biases are slightly lower when using the Moore and Fransen equations compared to the Mirwald equation. The estimation of APHV has a Mean Absolute Error that ranges between 0.5 and 1.81 years, with a larger error for early/late maturers compared to average maturers (boys). The maturity offset method can predict adult height within +/-5.3 cm in boys and +/-6.8 cm in girls ⁽¹²⁾.

ADVANTAGES Useful during PHV. Provides insight into biological maturation timing.

DISADVANTAGES Indirect prediction. Very accurate input of sitting height needed. Less accurate for early/late maturers.

Khamis-Roche method

HOW IT WORKS The Khamis-Roche method is a somatic method that was developed to predict adult height based on genetic and anthropometric data ⁽¹⁴⁾. The following variables are required: current height, body mass, chronological age, and average height of biological parents (mid-parent height), adjusted for sex.

It is a non-invasive method that is often used in sports sciences and paediatrics. This method is applicable to children between 4.0 and 17.5 years old and is considered one of the most accurate approaches for predicting adult height without measuring skeletal age.

ACCURACY For the average child, the prediction error is approximately ±2 cm. However, this error can increase in 11-15 year olds due to differences in maturation rate. In this age group, the median error is ±2.4-2.8 cm for the typical child, and may increase to ±5-7 cm for children in extreme growth percentiles, or who show an atypical growth pattern (e.g. early or late puberty). This corresponds to a prediction error of 1-3% of adult height ⁽¹⁵⁾.

ADVANTAGES Relatively high accuracy without skeletal age. Usable for a wide age range. Easy to apply (no specialist equipment).

DISADVANTAGES Height of both parents needed. Large error for children with atypical growth patterns. Less accurate during PHV.

Longitudinal method

HOW IT WORKS The growth curve comparison (GCC) method ⁽¹⁶⁾ predicts adult height by analysing an individual’s longitudinal growth trajectory rather than relying on a single measurement. This method requires repeated height measurements across multiple years, which are then compared to a large reference dataset of growth curves from other children.

Using similarity metrics, the individual growth curve is matched with those of children showing the most similar growth patterns. The model then identifies a subset with comparable growth patterns and uses their observed adult heights and growth increments to estimate the future height of the child This method is currently not (yet) available on the Hylyght platform, but can be used on the SloFit website.

ACCURACY The GCC method’s prediction error is relatively high in prepubertal ages, due to uncertainty in pubertal timing. Accuracy improves significantly once children enter puberty and growth patterns become more informative.

ADVANTAGES Captures timing and tempo of growth, including variability in the onset PHV, without measuring biological maturity.

DISADVANTAGES Accuracy is lower in pre-pubertal children, requires longitudinal data, accuracy strongly depends on reference dataset.

AI-based methods

HOW IT WORKS Recent advances in artificial intelligence (AI) have enabled the development of highly sophisticated models for predicting adult height. These approaches typically use machine learning on large-scale longitudinal datasets containing anthropometric, demographic, and sometimes body composition data.

These models can learn complex, non-linear relationships between growth variables and final height, and can generate personalized growth trajectories rather than a single-point estimate ⁽¹⁷⁾. This method is currently not available on the Hylyght platform.

ACCURACY Evidence is still limited and model performance depends strongly on dataset quality.

ADVANTAGES Dynamic growth predictions. Allows personalized predictions.

DISADVANTAGES Requires specialized software. Currently too limited validation. Ethical and privacy considerations when using large-scale personal data.

Summary table

FAQ

What is the most accurate method to predict adult height?

Bone age assessment using an X-ray of the left wrist is still considered the most accurate method, with a typical prediction error of ±2.6 cm. However, its use is limited by the need for medical imaging and expert interpretation.

Which method is most practical in a sports setting?

Non-invasive methods such as the Mirwald and Khamis-Roche methods are most practical in field settings. They require only basic anthropometric measurements and can be applied quickly without medical equipment. Khamis-Roche is the easiest to apply as it requires only basic anthropometric measurements (height and weight) and parental data, avoiding the need for cumbersome sitting-height measurements. This also reduces both the measuring error and the time needed to test the athletes.

Can adult height be predicted accurately before puberty?

Prediction before puberty is less accurate for all methods because the timing and intensity of the pubertal growth spurt are still unknown. Accuracy improves significantly once children approach or enter Peak Height Velocity (PHV).

Why is biological maturity important in height prediction?

Children of the same chronological age can differ substantially in biological maturity. Methods that account for maturation (e.g. bone age or Mirwald) provide better insight into remaining growth potential and timing of growth spurts. Learn more about the importance of tracking growth and maturity here.

Which method should be used for early or late maturing children?

Bone age assessment is preferred in early or late maturers because it directly evaluates skeletal maturity. Methods like Mirwald and Khamis-Roche are less reliable in these populations.

How to use the Khamis-Roche method when the height of one or both of the parents is unknown?

If parental height is unknown, alternative methods can be used or the average height of the country of origin of the parent can be used. However, this will reduce prediction accuracy.

Why are longitudinal methods like GCC more accurate over time?

Growth curve comparison methods use repeated measurements, allowing them to capture individual growth patterns and pubertal timing. This improves prediction accuracy, especially during and after puberty.

Are AI-based methods better than traditional methods?

AI-based methods can achieve high accuracy (≈1.7–2.5 cm) by modelling complex growth patterns. However, their performance depends on data quality, and they are less transparent and less validated in practical sports settings.

What should be done if different methods give different predicted adult heights?

It is common for different methods to produce slightly different predictions, as each approach is based on different assumptions (e.g. skeletal maturity, genetics, or growth patterns). In such cases, results should be interpreted within their context. If you have a measure of bone age, this result should be most reliable. If bone age is not available, Khamis-Roche should be more reliable than Mirwald for very tall or short athletes, while Mirwald is should be more reliable when athletes are very early or late mature and close to their age of peak height velocity. In any case it is good practice to give a range estimation rather than a single number (e.g. “predicted between 178 and 182” rather than “predicted 180”). If the estimations are far apart, this might indicate an atypical profile such as tall/short for their age combined with being early/late mature. Follow-up on this athlete to see how the estimations of adult height evolve.

References

(1) Bayley, N., & Pinneau, S. R. (1952). Tables for predicting adult height from skeletal age. Revised for use with the Greulich-Pyle hand standards. The Journal of pediatrics.

(2) Greulich WW, Pyle SI. Radiographic atlas of skeletal development of the hand and wrist. Redwood City, CA: Stanford University Press; 1959

(3) Tanner JM, Whitehouse RH, Marshall WA, Carter BS. Prediction of adult height from height, bone age, and occurrence of menarche, at ages 4 to 16 with allowance for midparent height. Archives Of Disease in Childhood. 1975;50(1):14–26. doi:10.1136/adc.50.1.14

(4) Chumela, W. C., Roche, A. F., & Thissen, D. (1989). The FELS method of assessing the skeletal maturity of the hand‐wrist. American Journal of Human Biology, 1(2), 175-183.

(5) Roche, A. F., Wainer, H., & Thissen, D. (1975). The RWT method for the prediction of adult stature. Pediatrics, 56(6), 1026-1033.

(6) Cumming, S., Pi-Rusiñol, R., Rodas, G., Drobnic, F., & Rogol, A. D. (2024). The validity of automatic methods for estimating skeletal age in young athletes: a comparison of the BAUSport ultrasound system and BoneXpert with the radiographic method of Fels. Biology of Sport, 41(1), 61-67.

(7) Suh, J., Heo, J., Kim, S. J., Park, S., Jung, M. K., Choi, H. S., ... & Kim, H. S. (2023). Bone age estimation and prediction of final adult height using deep learning. Yonsei Medical Journal, 64(11), 679.

(8) Lolli, L., Johnson, A., Monaco, M., Cardinale, M., Di Salvo, V., & Gregson, W. (2021). Tanner–Whitehouse and Modified Bayley–Pinneau adult height predictions in elite youth soccer players from the Middle East. Medicine and science in sports and exercise, 53(12), 2683-2690.

(9) Mirwald RL, Baxter-Jones AD, Bailey DA, Beunen GP. An assessment of maturity from anthropometric measurements. Med Sci Sports Exerc 2002;34:689–94.

(10) Moore, S. A., McKay, H. A., Macdonald, H., Nettlefold, L., Baxter-Jones, A. D., Cameron, N., & Brasher, P. (2015). Enhancing a somatic maturity prediction model. Medicine & science in sports & exercise, 47(8), 1755-1764.

(11) Fransen, J., Bush, S., Woodcock, S., Novak, A., Deprez, D., Baxter-Jones, A. D., ... & Lenoir, M. (2018). Improving the prediction of maturity from anthropometric variables using a maturity ratio. Pediatric Exercise Science, 30(2), 296-307.

(12) Sherar, L. B., Mirwald, R. L., Baxter-Jones, A. D., & Thomis, M. (2005). Prediction of adult height using maturity-based cumulative height velocity curves. The Journal of pediatrics, 147(4), 508-514.

(13) Sanders, J. O., Qiu, X., Lu, X., Duren, D. L., Liu, R. W., Dang, D., ... & Cooperman, D. R. (2017). The uniform pattern of growth and skeletal maturation during the human adolescent growth spurt. Scientific reports, 7(1), 16705.

(14) Khamis HJ, Roche AF. Predicting adult stature without using skeletal age: The Khamis-Roche method. Pediatrics 1994;94:504–7.

(15) Towlson, C., Salter, J., Ade, J. D., Enright, K., Harper, L. D., Page, R. M., & Malone, J. J. (2021). Maturity-associated considerations for training load, injury risk, and physical performance in youth soccer: One size does not fit all. Journal of sport and health science, 10(4), 403-412.

(16) Mlakar M, Gradišek A, Luštrek M, Jurak G, Sorić M, Leskošek B, e.a. Adult height prediction using the growth curve comparison method. PLoS ONE. 2023;18(2):e0281960. doi:10.1371/journal.pone.0281960

(17) Chun D, Jung HW, Kang J, Jang WY, Kim J. Artificial intelligence for pediatric height prediction using large-scale longitudinal body composition data. Digital Health. 2025;11:20552076251395975. doi:10.1177/20552076251395975

Malina RM, Kozieł SM, Králik M, Chrzanowska M, Suder A. Prediction of maturity offset and age at peak height velocity in a longitudinal series of boys and girls. American Journal Of Human Biology. 2020;33(6):e23551. doi:10.1002/ajhb.23551

Cumming SP, Lloyd RS, Oliver JL, Eisenmann JC, Malina RM. Bio-banding in sport: Applications to competition, talent identification, and strength and conditioning of youth athletes. Strength Cond J 2017;39:34–47.

Carvalho, H. M., Galvão, L. G., Karasiak, F. C., Lima, A. B., & Gonçalves, C. E. (2025). Is the Maturity Offset Equation Valid and Useful? A Simulation-Based and Longitudinal Evaluation with Implications for Youth Sport.

⤶ Terug to the Growth Knowledge Base