Journal Logo


Prediction of adult stature and noninvasive assessment of biological maturation


Author Information
Medicine & Science in Sports & Exercise: February 1997 - Volume 29 - Issue 2 - p 225-230
  • Free


Elite performance, as well as the performance characteristics of growing children and adolescents, is associated with biological maturity status (5,13). An accurate assessment of the biological maturity status of growing children is thus important in the study of growth and performance and also for those involved in the guidance of young children in sports. Skeletal maturity is the best single estimator of the biological maturity status because it spans childhood through adolescence. Unfortunately, the method requires a small radiation exposure and specialized training in the use of one of the available systems. Secondary sex characteristics, excellent indicators of maturity status, are limited to the pubertal years, allow classification only in five stages from immature to adult status, and in many settings have cultural restrictions in their use.

Roche et al. (16) proposed the ratio of present stature/adult stature as a measure of biological maturity because it has the same value in all adults and increases monotonically with age. The long duration of several serial growth studies has allowed the development of systems for the prediction of adult stature, but in the most accurate systems presently in use skeletal age (SA) is one of the important predictors(1,17,20-22). Based on data of the Fels Longitudinal Study, Wainer et al. (23) demonstrated that in American white children reasonable accuracy can be obtained in predicting adult stature when SA is replaced by chronological age(CA), whereas Khamis and Roche (12) showed that when current stature, current weight, and midparent stature are used as predictors the errors of prediction are only slightly larger than those for the Roche-Wainer-Thissen (RTW) method (17) which requires SA.

The percentage of adult stature attained at a given age is thus an indicator of biological maturity status. Youths who are closer to their adult or mature stature compared with others of the same CA are advanced in maturity status, while those who have attained a smaller percentage of adult stature compared with others of the same CA are slowed in their maturity progress. Variation in the percentage of adult stature attained at a given CA reflects variation in the rate or tempo of growth, i.e., individuals who are growing at a faster rate are closer to their adult stature than those who are growing at a slower rate.

Although estimates of the percentage of adult stature attained at a given CA during childhood and youth require longitudinal data, the use of predicted adult stature has potential application in the sport sciences. Such an approach may be useful, for example, in distinguishing youngsters who are tall at a given CA because they are genetically tall or who are tall at a given CA because they are maturationally advanced compared with their peers, i.e., they have attained a greater percentage of their predicted adult stature at that CA.

The present study considers the prediction of adult stature in boys 13-16 yr of age, the ages when participation in a variety of youth sports among boys is generally high. These are also the ages when maturity-associated variation in size and performance, especially in strength and power tasks, favors boys who are advanced in biological maturation. The early maturing boys often dominate many sports, e.g., baseball, football, soccer, and ice hockey(13) at these ages. Since the use of SA requires minimal radiation exposure and the use of secondary sex characteristics often has cultural limitations in their assessment, the proposed method has potential use as an indicator of maturity status in field studies of performance in youth sports.

A method was developed to predict adult stature in the absence of SA in the Leuven Longitudinal Study of Belgian Boys. The accuracy of the noninvasive method, labeled the Beunen-Malina (BM) method, was then compared to the accuracy of the Tanner-Whitehouse (TW) method(20-22) in which SA, current stature, and CA are used as predictors.


Subjects. Subjects included 102 Flemish-speaking boys who were participants in the Leuven Growth Study of Belgian Boys(7,15). These subjects are also current participants in an extension of the study, which included follow-up at 30 and 35 yr of age (4). A detailed description of the design, sample, and measurement techniques are provided elsewhere(7,15). Informed parental consent was obtained each year prior to the children's participation in the study. Of the total sample of 588 boys who were followed at annual intervals between 13 and 18 yr, 278 Flemish-speaking subjects were also examined at 30 yr. Complete anthropometric data, SA assessments, static strength, and stature measured at the age of 30 yr were available for 102 subjects. These data were used for the present analyses. It can be questioned if this sample of 102 subjects used for further analysis is representative for the total sample followed longitudinally during adolescence and at age 30 yr. At each age level, mean differences in predictor variables used in the BM-method of this sample and of the total longitudinal series were tested for significance. For the BM-method predictor variables (stature, sitting height, subscapular skinfold, triceps skinfold, and chronological age) the age-specific means of this sample did not differ from the means of the total longitudinal series except for stature([horizontal bar over]X this sample = 152.14, [horizontal bar over]X total = 153.68) and chronological age ([horizontal bar over]X this sample = 12.78,[horizontal bar over]X total sample = 12.85) at ± 12 yr. Furthermore, adult stature (stature at 30 yr) of this sample and of the longitudinal series did not differ significantly. Consequently, the sample on which all further analysis is based is a fairly good representation of the total group followed longitudinally during adolescence and at 30 yr.

Measurements. Postereoanterior radiographs (dose equivalent < 10 mrem) of the left hand and wrist were taken annually according to the protocol of Tanner et al. (21). SA assessments were made by a single trained observer using the Tanner-Whitehouse II technique(22). The reproducibility of the assessor was well within those reported by other trained observers, and his estimations agreed closely (83-84% interobserver agreement) with those of the originators of the method (2). The anthropometric techniques are described by Ostyn et al. (15). Arm pull strength was used as an indicator of static strength. In this test the subject pulled maximally with one arm against a dynamometer fixed to the wall while the other arm was extended horizontally and used for support.

Adult stature (AS) was predicted according to the Tanner-Whitehouse technique (ASTW) (20,22) using current stature, current SA, and current CA as the predictors. As demonstrated by Wainer et al.(23) adult stature can be predicted using the RWT technique (17) by replacing SA with CA without severe loss in accuracy except in the period of the adolescent growth spurt and thereafter. Consequently, the same procedure was also used to predict adult stature according to the TW method by replacing SA with CA (predicted adult stature Tanner-Whitehouse method, SA substituted by CA (ASCA)) in the regression equation. In regression analysis, it is acceptable to replace individual values by an average value in the case of missing observations. By definition, CA corresponds, of course, to the average SA in a certain age group of the population.

Statistical analysis. Multiple regression techniques were used to predict adult stature from observations made at 13, 14, 15, and 16 yr. Since stature also increases in this sample after 18 yr (6), stature at 30 yr was taken as adult stature. Selection of predictor variables was based on the tracking of somatic dimensions (14), the age-specific associations between skeletal maturity and somatic dimensions(9), and on somatic structure(10,11) observed in this age range. A four-step procedure was adopted. First, age-specific multiple regressions (stepwise and maximum R method) were calculated (18,19), which served to select the most important predictors. Body mass, lengths (stature, sitting height, leg length), breadths (biacromial diameter, bicristal diameter, biepicondylar humerus, and bicondylar femur), circumferences (upper arm, chest, thigh, and calf), skinfolds (triceps, subscapular, suprailiac, and calf) and static strength (arm pull) were used as possible predictor variables. Based on the significant changes in R and the decrease in the SEE of the age-specific analyses, it was decided to include five predictors. Furthermore, at each age level, current stature, sitting height, the triceps and subscapular skinfolds, and CA were among the most important predictors. Second, final Beunen-Malina adult stature (ASBM) multiple regressions were then calculated with those five characteristics entering the regressions. Third, the P25 to P75 and P5 to P95 boundaries of error prediction were subsequently calculated for ASTW, ASBM, and ASCA. Fourth, correlations between ASTW, ASBM, and ASCA were also calculated, as were correlations between percentage of measured adult stature and percentage adult stature from the three different predictions (ASTW, ASBM, ASCA). All calculations were made using SAS and SPSS procedures (18,19).


Age-specific regression coefficients for the prediction of adult stature are given in Table 1 with multiple correlations and SEEs. The multiple correlation increases from R = 0.70 at 12 yr to R = 0.87 at 16 yr, which is reflected in the decrease in the SEE from 4.2 cm at 13 yr to 3.0 cm at 16 yr. Means and SDs for measured adult stature and for predicted adult statures with the Tanner-Whitehouse method (ASTW), the Tanner-Whitehouse method with CA as a substitution for SA (ASCA) and the newly proposed Beunen-Malina procedure (ASBM) are given in Table 2. The percentage of adult stature calculated from the different methods using the predicted adult stature and the observed stature at that age level are also included. The percentage of adult stature is the ultimate maturity indicator that is proposed for further use.

In this sample of boys the AS predictions tend to underestimate adult stature. The predictions based on CA as a substitute for SA are more variable and higher standard deviations are apparent at the younger age levels. The SDs of the ASBM predictions correspond more closely to the SDs of the measured adult statures than the SDs of the predictions based on CA (ASCA). The difference between the percentage of adult stature calculated from the ASBM predictions and the measured percentage of adult stature is also quite small(less than 0.5% difference): the differences are smaller than for the ASTW predictions (0.21-2.26%).

Correlations between measured adult stature and adult stature predicted from the three methods vary between r = 0.60 and r = 0.88. Generally, the coefficients are highest for ASTW predictions and lowest for ASCA predictions with the ASBM predictions intermediate (Table 3). When the percentage of adult stature is taken as a reference, the correlations are higher and range from r = 0.79 to r = 0.87 for both the ASTW and ASBM predictions. The median difference between adult stature and predicted stature for the three techniques and the limits for the errors in prediction (P25-P75 and P5-P95) are also used as an indication of the accuracy of predictions. For example, the adult stature of a 13-yr-old boy is predicted to be 180 cm with the ASTW method. The P25-P75 boundaries at this age are -0.18 and +5.57(Table 4), and the range 179.82 to 185.57 cm will include 50% of the adult statures of boys with identical values for the predictor variables. Similarly 90% of the adult statures will range between P5 and P95, i.e., between 175.8 cm and 187.3 cm.

As noted for the mean differences, the median differences for ASTW indicate that adult stature tends to be slightly underestimated with this method from 13 through 15 yr. The ASBM method tends to overestimate adult stature, but to a lesser extent, while the ASCA predictions are more variable. The P25-P75 and P5-P95 boundaries further demonstrate that even with this systematic under- or over-estimation, the range of the boundaries is strikingly similar between the ASTW and ASBM methods. The range of the error boundaries for the ASCA method is considerably larger, especially at the three youngest age levels.


In the original sample of British children upon which the Tanner-Whitehouse prediction method was developed, the multiple correlations for adult stature prediction in boys 12.5-16.5 yr varied between 0.89 and 0.93(20). In the present sample of Belgian boys of the same age range, the correlations between predicted adult statures (ASTW) and measured adult stature vary between 0.74 and 0.88. The somewhat lower correlations in the present sample for ASTW predictions may be explained by several factors. First, given the sample difference, the observed correlations can be considered as a cross validation of the Tanner-Whitehouse method. Second, errors in the estimation of SA may contribute to the differences although the interobserver agreement between the first author and the originators of the method was quite high (83-84% observer agreement)(2). Third, population differences and secular changes in skeletal maturation of British and Belgian children have been observed(3).

When the accuracy of the Beunen-Malina method is compared with the Tanner-Whitehouse method in this sample of Belgian boys, the BM prediction compares favorably. The correlations between predicted and observed adult stature are of similar magnitude for both prediction techniques, as are the P25-P75 and P5-P95 error boundaries (Tables 3 and 4). Substituting SA with CA in the Tanner-Whitehouse method leads to decrease in the correlations and considerable increase in errors of prediction(Tables 3 and 4). Although in the absence of SA and of the predictor variables used in the Beunen-Malina technique the substitution of SA by CA can be used, its utility can be questioned given the low correlations between measured percentage of adult stature and predicted percentage of adult stature (Table 3) and the considerable errors in prediction (Table 4). This has also been observed by Wainer et al. (23) for the RWT method. When in this prediction CA was substituted for Greulich-Pyle SA, the loss in accuracy was minor except for boys in the age range 12-15 yr, when SA had a higher weight in the prediction equations.

The inclusion of other anthropometric or even performance characteristics, such as isometric strength, may improve the accuracy of the predictions. If six instead of five predictors were selected according to the procedures described in the methods, arm pull performance, a static or isometric strength test, was the most important predictor at all age levels in this sample of adolescent boys. The multiple regression coefficients including six predictors were virtually the same (0.74-0.87) as for five predictors (0.70-0.87). Furthermore, the SEEs were also very similar (3.0-4.0 cm compared to 3.0- 4.2 cm). Considering this small insignificant difference and the burden of including another predictor that has to be precisely measured with standardized procedures, the Beunen-Malina method presented here may be a more optimal solution.

Finally, the percentage of adult stature calculated from the Beunen-Malina prediction was used to classify 14-yr-old (N = 3511) Belgian boys(15) into three contrasting maturity groups: early maturers (N = 599), percentage of adult stature > 94.4% (mean + 1 SD); average maturers (N = 2341), percentage of adult stature varies between 88.1% and 94.3% (mean ± 1 SD); and late maturers (N = 571), percentage of adults stature is < 88.0%. Mean anthropometric and performance characteristics of the three groups were compared with ANOVA and Tukey post-hoc tests (18). Highly significant differences were observed among the maturity groups (Table 5). Expressed in SD units (effect size: ES) of the total sample of 14-yr-old boys, the differences between the means of the early and late maturers exceed 2.0 SDs for stature, sitting height, weight, biacromial breadth, and arm pull. The mean differences exceed 1.5 SDs for arm and calf circumferences, vary between 0.9 and 1.0 SD for the sum of four skinfolds and the vertical jump, and are below 0.5 SD for the shuttle run and bent arm hang. These results are consistent with observations in the same sample and age group when maturity categories were based on skeletal maturity(8).

In adolescent boys and in the absence of skeletal maturation, the Beunen-Malina method is proposed for predicting adult stature using five predictors (CA, stature, sitting height, subscapular skinfold, and triceps skinfold, measured according to procedures described in Ostyn et al.(15). Furthermore, percentage of adult stature calculated in this manner is an efficient maturity indicator and can be used in a variety of settings as a noninvasive substitute for skeletal maturation. The use of sitting height and skinfolds in the prediction takes into consideration the differential timing of the adolescent spurt in body segments and changes in subcutaneous fat that occur during the growth spurt(7,14).


1. Bayley, N. and S. R. Pinneau. Tables for predicting adult height from skeletal age: revised for use with the Greulich-Pyle hand standards. J. Pediatr. 40:423-441, 1952.
2. Beunen, G. and N. Cameron. The reproducibility of TW2 skeletal age assessment by a self-taught assessor. Ann. Hum. Biol. 7:155-162, 1980.
3. Beunen, G., J. Lefevre, M. Ostyn, R. Renson, J. Simons, and D. Van Gerven. Skeletal maturity in Belgian youths assessed by the Tanner-Whitehouse method (TW2). Ann. Hum. Biol. 17:355-376, 1990.
4. Beunen, G., J. Lefevre, A. L. Claessens, et al. Age-specific correlation analysis of longitudinal physical fitness levels in men. Eur. J. Appl. Physiol. 64:538-545, 1992.
5. Beunen, G. and R. M. Malina. Growth and biological maturation: relevance to athletic performance. In: The Encyclopedia of Sports Medicine: The Child and the Adolescent athlete, O. Bar-Or (Ed.). Oxford: Blackwell, 1996, pp. 3-24.
6. Beunen, G., R. M. Malina, J. Lefevre, et al. Size, fatness and relative fat distribution of males of contrasting maturity status during adolescence and as adults. Int. J. Obes. 18:670-678, 1994.
7. Beunen, G., R. M. Malina, M. A. Van't Hof, et al.Adolescent growth and motor performance: a longitudinal study of Belgian boys. Champaign, IL: Human Kinetics, 1988, pp. 6-10.
8. Beunen, G., M. Ostyn, R. Renson, J. Simons, P. Swalus, and D. Van Gerven. Skeletal maturation and physical fitness of 12 to 15 year old boys. Acta Paediatr. Belg. 28:221-232, 1974.
9. Beunen, G., M. Ostyn, J. Simons, R. Renson, and D. Van Gerven. Chronological and biological age as related to physical fitness in boys 12 to 19 years. Ann. Hum. Biol. 8:321-331, 1981.
10. Claessens, A., G. Beunen, and J. Simons. Stability of anthroposcopic and anthropometric estimates of physique in Belgian boys followed longitudinally from 13 to 18 years of age. Ann. Hum. Biol. 13:235-244, 1986.
11. Claessens, A., G. Beunen, and J. Simons. Anthropometric principal components and somatotype in boys followed individually from 13 to 18 years of age. Humanbiol. Budapest. 16:23-36, 1985.
12. Khamis, H. J. and A. F. Roche. Predicting adult stature without using skeletal age: the Khamis-Roche method. Pediatrics 504-507:1994.
13. Malina, R. M. Physical growth and biological maturation of young athletes. Exerc. Sport Sci. Rev. 22:389-433, 1994.
14. Malina, R. M. and C. Bouchard. Growth, maturation, and physical activity. Champaign, IL: Human Kinetics, 1991, pp. 60-63, 202, 251-301.
15. Ostyn, M., J. Simons, G. Beunen, R. Renson, and D. Van Gerven. Somatic and motor development of Belgian secondary schoolboys. Norms and standards. Leuven: Leuven University Press, 1980, pp. 1-48.
16. Roche, A. F., F. Tyleshevski, and E. Rogers. Non-invasive measurement of physical maturity in children. Res. Quart. Sport Exerc. 54:364-371, 1983.
17. Roche, A. F., H. Wainer, and D. Thissen. The RWT-method for the prediction of adult stature. Pediatrics 56:1026-1033, 1975.
18. Sas. SAS/STAT user's guide. Cary, NC: SAS Institute, 1988, pp. 547-640.
19. Spss. SPSS reference guide. Chicago, Il.: SPSS Inc., 1990.
20. Tanner, J. M., K. W. Landt, N. Cameron, B. S. Carter, and J. Patel. Prediction of adult height from height and bone age in childhood. Arch. Dis. Child. 58:767-776, 1983.
21. Tanner, J. M., R. H. Whitehouse, W. A. Marshall, M. J. R. Healy, and H. Goldstein. Assessment of Skeletal Maturity and Prediction of Adult Height (TW2 Method). London: Academic Press, 1975, pp. 21-29, 1975.
22. Tanner, J. M., R. H. Whitehouse, N. Cameron, W. A. Marshall, M. J. R. Healy, and H. Goldstein. Assessment of skeletal maturity and prediction of adult height (TW2 method). London: Academic Press, 1983, pp. 22-37.
23. Wainer, H., A. F. Roche, and S. Bell. Predicting adult stature without skeletal age and without parental data. Pediatrics 61:569-572, 1978.


©1997The American College of Sports Medicine