To examine ethnic and sex differences in the pattern of skeletal maturity from adolescence to adulthood using a novel longitudinal analysis technique (SuperImposition by Translation And Rotation (SITAR)).

Johannesburg, South Africa.

607 boys and girls of black as well as white ethnicity from the Birth to Twenty bone health study, assessed annually from 9 to 20 years of age.

Bone maturity scores (Tanner–Whitehouse III radius, ulna, and short bones (TW3 RUS)) assessed longitudinally from hand-wrist radiographs were used to produce individual and mean growth curves of bone maturity and analysed by the SITAR method.

The longitudinal analysis showed that black boys matured later by 7.0 SE 1.6 months (p<0.0001) but at the same rate as white boys, whereas black girls matured at the same age but at a faster rate than white girls (by 8.7% SE 2.6%, p=0.0007). The mean curves for bone maturity score consistently showed a midpubertal double kink, contrasting with the quadratic shape of the commonly used reference centile curves for bone maturity (TW3).

Skeletal maturity was reached 1.9 years earlier in girls than boys, and the pattern of maturation differed between the sexes. Within girls, there were no ethnic differences in the pattern or timing of skeletal maturity. Within boys, however, skeletal maturity was delayed by 7 months in black compared with white ethnicity. Skeletal maturation, therefore, varies differentially by sex and ethnicity. The delayed maturity of black boys, but not black girls, supports the hypothesis that boys have greater sensitivity to environmental constraints than girls.

Skeletal maturity is influenced by age, gender and environmental factors, but has been little studied in contemporary populations, particularly in developing communities.

The timing (tempo) and rate (velocity) of bone maturation during childhood and adolescence determine the age at which adult bone maturity is reached.

Common modelling techniques used for longitudinal growth data are not applicable to the study of bone maturity from childhood to adulthood.

A novel analysis technique (SuperImposition by Translation And Rotation) is used to create mean curves of bone maturity score by gender and ethnicity.

Bone maturity growth curves differ by sex, but do not differ by ethnicity (black or white) among South African adolescents.

Black as well as white girls showed the same pattern and timing of skeletal maturation, whereas black boys were significantly delayed compared with white boys in skeletal maturity.

Skeletal maturity, or ‘bone age’, is a key indicator of biological maturity and reflects progress towards complete fusion of the epiphyses of the long bones. The rate of skeletal maturation is sensitive to environmental influences at both the individual and population levels.

Few contemporary studies have been able to chart skeletal maturation longitudinally. The aims of this study are to: describe the skeletal maturation of urban South African adolescents by sex and ethnic group; develop mean growth curves of skeletal maturity and compare these curves with the TW3 reference charts.

The Birth to Twenty (Bt20) study included singleton children born in Soweto-Johannesburg, South Africa, in 1990.

Hand-wrist radiographs were taken annually by radiographers at the Johannesburg Academic Hospital following standard procedures.

Primary caregivers gave written informed consent for their child to participate. Participants gave their assent annually and consent from 18 years onwards. Ethical approval was obtained from the University of the Witwatersrand Committee for Research on Human Subjects (#M980810) and Loughborough University (R01-P3).

The median age at maturity by sex and ethnicity was estimated cross-sectionally using mixed effects logistic regression, with the dependent variable maturity (RUS bone score=1000) and independent variables age, sex and ethnicity (plus their interaction) as fixed effects, and subject as a random effect. Log transforming age improved the fit. Median age at maturity was estimated as e^{−a/b} with

Each subject's set of RUS bone scores plotted against age forms a growth or maturation curve. Individual RUS bone scores increase over time until reaching the adult score of 1000. Thus, individual curves differ in terms of the timing and rate of maturation, but not in final size. The SuperImposition by Translation And Rotation (SITAR) method of analysis for longitudinal data

Conventionally, RUS bone scores are analysed cross-sectionally, for example, TW3 centiles.

Developmental age is assumed to be linearly related to chronological age, and the timing (or

The effect of adjusting for tempo is to shift individual curves left/right on the age scale, so as to match the mean curve. Similarly, adjusting for velocity involves shrinking/stretching the age scale to match the mean curve and this has the effect of making the curve steeper or shallower. The net effect of translating and rotating the individual curves is to superimpose them on the mean curve. To measure the goodness-of-fit of the SITAR model, its residual SD is compared with that for the baseline model omitting the random effects.

The data were initially analysed separately by sex and ethnic group (black as well as white), and then joint models were fitted combining first the ethnic groups by sex, and then the sexes. To represent group differences, fixed effects for sex and/or ethnicity were included in the model terms for tempo and velocity, which with their SEs allowed the equality of the group growth patterns to be tested. Log transforming age improved the fit. Thus, the random effects for both tempo and velocity are in fractional units, which multiplied by 100 can be viewed as percentage differences from the mean.

For comparison, the SITAR mean curves are plotted against the TW3 centiles for RUS bone score, which were generated from the RUS bone z-scores produced using software of Tanner

Four thousand five hundred and sixty-five radiographs from 607 subjects aged 9–20 years were scored, with 1–10 scores per subject (median 9, IQR 6–10). The logistic regression model showed that the median age of skeletal maturity was, on average, 1.9 SE 0.2 years earlier in girls than boys (p<0.001); and that girls of black as well as white ethnicity matured at the same age (p=0.7), while black boys matured 6.0 SE 3.2 months later than white boys (p=0.06).

The SITAR analyses used all the serial bone scores by age.

Individual and mean curves of radius, ulna and short bones (RUS) bone maturity score in boys by ethnicity analysed by SuperImposition by Translation And Rotation. (A) White boys: individual unadjusted curves. (B) White boys: individual curves adjusted for tempo and velocity, with the mean adjusted curve. (C) Black boys: individual unadjusted curves. (D) Black boys: individual curves adjusted for tempo and velocity, with the mean adjusted curve.

The SITAR adjustments accounting for differences in tempo and velocity explained 89%–92% of the variance (

Median age at maturity, and summary of SITAR analyses of RUS bone score on age fitted to the four groups by sex and ethnicity

White boys | Black boys | White girls | Black girls | |
---|---|---|---|---|

N of subjects/measurements | 106/575 | 214/1803 | 101/615 | 186/1572 |

Median age at maturity (years) | 16.5 | 17.0 | 15.0 | 15.0 |

Spline degrees of freedom | 5 | 7 | 7 | 6 |

Residual SD (bone score units) | 29 | 30 | 22 | 23 |

SD of tempo random effect (%) | 6.1 | 7.9 | 7.6 | 7.2 |

SD of velocity random effect (%) | 20 | 20 | 22 | 15 |

Tempo–velocity correlation | 0.2 | 0.2 | –0.2 | –0.2 |

Variance explained (%) | 89 | 92 | 92 | 91 |

RUS, radius, ulna and short bones.

The fitted mean curves are shown superimposed on the adjusted individual curves (

Individual and mean curves of radius, ulna and short bones (RUS) bone maturity score in girls by ethnicity analysed by SuperImposition by Translation And Rotation. (A) White girls: individual unadjusted curves. (B) White girls: individual curves adjusted for tempo and velocity, with the mean adjusted curve. (C) Black girls: individual unadjusted curves. (D) Black girls: individual curves adjusted for tempo and velocity, with the mean adjusted curve.

Given the similarity of the models by ethnicity within each sex, combined models were fitted to the boys and girls separately, allowing the mean values for tempo and velocity to differ by ethnicity, and they fitted well (BIC for boys 24 844 combined vs 24 871 separate; for girls 21 742 combined vs 21 780 separate). However, a model combining the sexes fitted poorly (BIC 46 919 combined vs 46 586 separate), showing that the two mean curves were different in shape and should not be combined.

Summary of SITAR fixed effects for black ethnicity compared with white ethnicity, by sex

Regression coefficient | SE | p Value | |
---|---|---|---|

Boys | |||

Tempo (months) | 7.0 | 1.6 | <0.0001 |

Velocity (%) | −1.0 | 3.0 | 0.7 |

Girls | |||

Tempo (months) | 0.0 | 1.8 | >0.9 |

Velocity (%) | 8.7 | 2.6 | 0.0007 |

Mean curves by sex and ethnicity of radius, ulna and short bones (RUS) bone maturity score in the Birth to Twenty subsample compared with the Tanner–Whitehouse III (TW3) reference (25th, 50th and 75th centiles).

This study presents unique longitudinal data on skeletal maturation in adolescents from a middle-income country. Moreover, the study provides mean growth curves for RUS bone scores which permit the analysis of sex and ethnic differences in patterns of skeletal maturation. The results show that black boys mature later than white boys whereas black as well as white girls mature at the same age. A longitudinal analysis clarifies how this pattern emerges; skeletal maturation in black girls starts to accelerate later than in white girls, but develops faster, whereas in black boys, skeletal maturation starts to accelerate later and then develops at a similar rate.

The differing patterns of skeletal maturity in the two ethnic groups by sex are intriguing. In all populations, girls are more skeletally mature than boys from birth onwards and reach adult bone maturity, on average, 2 years earlier than boys (1.9 years here).

This is the first study to produce mean growth curves for skeletal maturity by sex and ethnicity. The mean curves are closely similar in the two ethnic groups by sex, indicating that the curve shape is robust. The curves are notably different from longitudinal growth curves of height and weight in puberty

What does this pattern mean? The curve shapes are likely to be genuine for two reasons; first, the same pattern emerged independently in all four groups, implying that this may be universal; and second, the complexity of each curve's shape was data-driven—the number of degrees of freedom for the regression spline was chosen to minimise the BIC, which penalises the deviance (goodness-of-fit) for each extra degree of freedom used. Thus, the models all fitted appreciably better with the extra degrees of freedom, implying that the

All four curves are based on the TW3 scoring system, which converts the radiographic appearance of the radiographs to RUS bone scores. Unlike height or weight, the scores are on a constructed scale, which will not necessarily represent growth as a smoothly changing process.

The findings also cast light on individual variation in skeletal maturation. The SITAR models fitted consistently better when using log age rather than age. This indicates a multiplicative age scale, where individuals whose maturation is advanced have a foreshortened (or shrunken) scale, while delayed maturers have an extended (stretched) scale. This corresponds to individuals starting to mature at birth and continuing to mature relatively quickly or slowly through childhood until reaching maturity. This is analogous to Peto's ‘horse-racing effect’,

A particular strength of this study is the availability of longitudinal bone maturity scores in two ethnic groups of children and adolescents with all ratings undertaken by a single observer.

In conclusion, the study has shown that the pattern of skeletal maturation in South African adolescents is similar in white and black girls, but delayed by 7 months in black boys compared with white boys. The fact that skeletal maturity was delayed in black boys, but not in black girls, supports the hypothesis that boys are more susceptible to delays in growth and maturation in unfavourable environments.

The authors would like to thank the Birth to Twenty staff, Bone Health Study staff and the adolescents and parents of the Bt20 study for their participation.

JMP, NC and SAN designed and led the Bone Health Study; JMP, SAN, NC and NLH undertook data collection. NLH rated radiographic materials and calculated bone scores. NLH and EKR collated data and conducted preliminary analyses. TJC fitted the SITAR models and produced the growth curves. All authors contributed to the interpretation of data and writing of the manuscript. All authors have read and approved the final manuscript.

The Birth to Twenty cohort and Bone Health study were funded by the Wellcome Trust, South African Medical Research Council, National Research Foundation and the University of the Witwatersrand.

TJC was funded by Medical Research Council grant MR/J004839/1.

University of the Witwatersrand Committee for Research on Human Subjects.

Not commissioned; externally peer reviewed.

_{e}scale simplify the presentation of log transformed data