A Longitudinal Study of Bone Gain in Pubertal Girls: Anthropometric and Biochemical Correlates


  • Joanna Cadogan,

    1. Centre for Human Nutrition, University of Sheffield, Sheffield, United Kingdom
    Search for more papers by this author
  • Aubrey Blumsohn,

    1. Department of Human Metabolism and Clinical Biochemistry, University of Sheffield, Sheffield, United Kingdom
    Search for more papers by this author
  • Margo E. Barker,

    1. Centre for Human Nutrition, University of Sheffield, Sheffield, United Kingdom
    Search for more papers by this author
  • Richard Eastell

    Corresponding author
    1. Department of Human Metabolism and Clinical Biochemistry, University of Sheffield, Sheffield, United Kingdom
    • Address reprint requests to: Prof. Richard Eastell, Division of Clinical Sciences (NGHT), Section of Medicine, Bone Metabolism Group, University of Sheffield, Northern General Hospital, Herries Road, Sheffield S5 7AU, U.K.
    Search for more papers by this author


The aim of this longitudinal study was to investigate the factors associated with bone mineral acquisition in pubertal girls. Subjects were 37 healthy, Caucasian girls aged 12.1 years (SD 0.3). Measurements were made at 6-month intervals over a period of 18 months and included total body bone mineral content (TBBMC), total body bone mineral density (TBBMD), lean mass, and fat mass by dual-energy X-ray absorptiometry, anthropometry, lifestyle factors, four biochemical markers of bone turnover, hormonal status, and fractional calcium absorption. In multiple regression analysis, correlates of relative gain in TBBMC were gain in lean mass (p < 0.001) and estradiol (p = 0.008). For TBBMD, correlates were gain in lean (p < 0.001) and fat mass (p = 0.003), estradiol (p < 0.001), dietary energy intake (p = 0.003), and parathyroid hormone (p = 0.023). Statural growth and gain in bone mass were unrelated; both height velocity and bone turnover peaked ∼20 months prior to menarche, whereas gain in bone mass peaked at menarche. Bone turnover markers correlated with height velocity (0.40 < r < 0.62), but not with bone gain. Estradiol was independently and negatively associated with all markers of bone turnover (−0.67 < r < −0.80). We conclude that estradiol is an important determinant of bone mineral gain in pubertal girls and is probably responsible for the reduction in bone turnover in late puberty; lean mass was the body composition parameter most closely associated with bone gain; height gain and bone gain are dissociated during the period of rapid growth at puberty; and bone turnover markers are modestly related to height gain, but are not predictive of bone gain.


THE PERIOD OF ACCELERATED growth during puberty is crucial for bone mineral acquisition in girls, contributing ∼50% of peak bone mass.1 Peak bone mass is largely achieved by the end of the second decade,2,3 with about 5–10% of total body bone mass achievable in the third decade. In adolescent girls, bone gain at sites of clinical interest, such as the spine and femoral neck, may be complete by the end of longitudinal growth.4–8 Peak bone mass is probably determined by genetic, hormonal, medical, and lifestyle factors.9,10 Since maximizing peak bone mass may protect against osteoporotic fracture risk, an understanding of the endogenous and exogenous factors associated with pubertal bone gain has implications for preventive health measures.

The aim of this longitudinal study was to identify the factors that determine bone mineral accumulation during puberty in healthy Caucasian teenage girls. Clinical characteristics, body composition, diet, lifestyle, bone turnover, and endocrine factors were considered. The relationship between mineral accumulation and linear growth, and the time courses of clinical, densitometric, and biochemical parameters throughout puberty were investigated. In addition, although biochemical markers of bone turnover are moderately predictive of growth velocity in a clinical setting,11,12 few prospective studies have been carried out in healthy children,13 and none have examined the relationship among indices of bone turnover, statural growth, and total body bone mineral acquisition in normal pubertal girls.



The study sample comprised 37 healthy, 12-year-old Caucasian girls, recruited from schools in the city of Sheffield. The subjects were the control group from a milk intervention trial, reported previously.14 None of the subjects had any history of bone disease or were taking any medications known to influence calcium metabolism, and none of the subjects were taking calcium supplements. Written informed consent was obtained from all volunteers and their parents. The study was carried out with the approval of the North Sheffield Local Ethics Committee.

Study design

The study design was a prospective cohort study. Subjects were observed for 18 months, and densitometric, anthropometric, and biochemical measurements were performed at 6-month intervals.


Total body bone mineral content (TBBMC), total body bone mineral density (TBBMD), lean mass, and fat mass were measured by dual-energy X-ray absorptiometry on a Hologic QDR-1000W densitometer (Hologic Inc., Waltham, MA, U.S.A.). This method has a precision error (CV percentage) of 0.9–1.0% for TBBMD in children.15 A daily quality assurance test was performed using a manufacturer-supplied spine phantom. The reproducibility of the phantom measurement over the duration of the study was 0.4%.

Anthropometry and pubertal staging

Height was measured in light clothing and bare feet to the nearest millimeter with a stadiometer (Holtain Ltd., Crymych, Dyfed, U.K.) and weight to the nearest 100 g with a set of upright balance scales (Seca 220; Hallamshire Scales, Ltd., Sheffield, U.K.). All measurements were made in the morning by the same observer at each time point. Pubertal staging was ascertained by self-assessment, using line drawings and written descriptions of the five stages of puberty, according to Tanner's definitions.16 This method has been validated in adolescents.17 Menstrual status was determined by interview, including poststudy interview for those girls who did not reach menarche during the observation period, enabling calculation of months postmenarche for data analysis.

Diet and activity

Dietary intake was assessed at baseline and at 18 months using the 7-day weighed intake method. Nutrient intakes were calculated from the diet records using FOODBASE dietary software (Institute of Brain Chemistry and Human Nutrition, London, U.K.). Habitual physical activity levels were assessed using a questionnaire designed for this age group.18


Samples of nonfasting blood (drawn between 11.00 and 12.00 h) and urine (collected between 09.00 and 11.00 h) were obtained at each visit. For each subject, repeat sample collections were as closely timed as possible, to minimize the effects of diurnal variation. Blood samples were allowed to clot for 30 minutes, centrifuged, and serum stored at −80°C. Urine samples were stored at −20°C. All samples were stored until the end of data collection, so that all samples from an individual were included in the same assay run. Serum levels of osteocalcin (OC) were measured using a two-site immunoradiometric assay (IRMA) (Elsa-Osteo; CIS BioInternational, Gif-sur-Yvette, France); the intra-assay coefficient of variation (CV) was 1.7%, the interassay CV was 2.5% at 23.16 ng/ml, and the detection limit was 0.4 ng/ml. Serum immunoreactive bone-specific alkaline phosphatase (iBAP) was measured using an IRMA (Tandem-R Ostase; Hybritech Europe SA, Liège, Belgium); the intra-assay CV was 3.5%, the interassay CV was 5.7% at 84.5 μg/l, and the detection limit was 2.0 μg/l. Cross-linked N-telopeptides (NTx) of type I collagen were measured in urine using a competitive ELISA (Osteomark; Ostex International, Inc., Seattle, WA, U.S.A.); the intra-assay CV was 5.7%, the interassay CV was 11.1% at 396 nM BCE, and the detection limit was 20 nM BCE. Urinary immunoreactive free deoxypyridinoline (iFDpd) cross-links were measured by competitive ELISA (Pyrilinks-D; Metra Biosystems, Inc., Mountain View, CA, U.S.A.); the intra-assay CV was 6.0%, the interassay CV was 9.6% at 81.5 nM/mmol creatinine (Cr), and the detection limit was 3 nM. All urinary data are expressed as a ratio to urinary Cr excretion.

Serum parathyroid hormone (PTH) levels were measured using an IRMA (Nichols Institute, San Juan Capistrano, CA, U.S.A.); the intra-assay CV was 6.5%, the interassay CV was 5.4% at 35.8 pg/ml, and the detection limit was 1.0 pg/ml. Serum estradiol (E2) was measured by radioimmunoassay (Diagnostic Products Corp., Los Angeles, CA, U.S.A.); the intra-assay CV was 9.1%, the interassay CV was 10.4% at 214 pmol/l, and the detection limit was 1.4 pmol/l. Serum insulin-like growth factor I (IGF-I) was measured by radioimmunoassay (Medgenix Diagnostics SA, Fleurus, Belgium) after acid-ethanol extraction, to prevent interference from the binding proteins. The intra-assay CV was 5.0%, the interassay CV was 5.9% at 467 ng/ml, and the detection limit was 0.25 ng/ml. Fractional calcium absorption (FA) was measured at baseline and at 18 months using stable strontium incorporated in a standard test breakfast (method described in Blumsohn19). Strontium was measured by electrothermal atomic absorption using a PE5000 spectrophotometer (Perkin Elmer, Beaconsfield, U.K.); the intra-assay CV was 2.6%.


Data are expressed as means ± SD unless otherwise indicated. Rates of change in TBBMC, TBBMD, height, weight, lean mass, and fat mass were derived by calculating the slope of the regression line of each against time, using all four measurements for each individual subject, and values were expressed as fractional change per year. Changes over each separate 6-month study period were calculated as a two-point difference divided by the time difference between measurements (actual number of days between visits). Data are expressed as both absolute (Δ) and percentage (%) annual increments. Slopes were also fitted to the biochemical data, for graphical representation of their relationship to puberty (months postmenarche), but for univariate and multivariate analyses, the mean values from measurements made at four time points were entered. Pearson's correlation coefficients were calculated between ΔTBBMC and ΔTBBMD and the other variables, between height and the biochemical indices, and between hormones and biochemical markers. Stepwise multiple linear regression analysis was performed to explore the associations between all measured variables and bone mass and height, and also between hormonal status and bone turnover. Lean mass and fat mass were entered instead of weight, since the latter incorporates the weight of the skeleton itself. Adjusted R2 values are presented, and a significance level of 0.05 accepted. All data were checked for normality using the Shapiro-Wilks test, and log transformations of skewed data performed (iFDpd, NTx, PTH, E2). Data were analyzed using SPSS software for Windows Version 6.0 (SPSS, Chicago, IL, U.S.A.).


The baseline characteristics of the study cohort are shown in Table 1. At study entry, all girls assessed themselves as either prepubertal (Tanner stage I) or in Tanner stages II and III; 7 of the 37 girls (19%) had reached menarche at baseline. By 18 months, 89% of the girls assessed themselves as Tanner stages III-V, and 25 (68%) had reached menarche. Table 2 shows the annualized rates of change of the anthropometric and bone mineral measures, both during each of the three 6-month study periods and during the 18 months overall. As expected from the pubertal staging data, the greatest height gain (mean velocity of ∼6 cm/year) occurred during the first 12 months of the study, when most of the cohort were in Tanner stages I-III. A similar pattern emerged for weight and lean mass. However, for fat mass there was a much larger population variance and no clear pattern was seen. The annualized rate of gain in TBBMC was highest during the first 6 months, approximating almost 1 g of bone mineral gained/day. Similarly, gain in TBBMD during this early part of the study was almost double that in months 6–18.

original image
original image

With respect to menarche, height velocity was significantly lower in post- compared with premenarchal girls during the periods 6–12 and 12–18 months (p < 0.001, unpaired t-test), but there was no difference during the first 6 months when relatively few of the cohort had reached menarche. There were no significant differences in ΔTBBMC and ΔTBBMD between pre- and postmenarchal girls during the first 12 months of the study. During the final 6-month study period, the differences reached significance (ΔTBBMC, 281 ± 152 vs. 160 ± 73 g/year, p = 0.003; ΔTBBMD, 58.4 ± 38.7 vs. 33.0 ± 21.8 mg/cm2/year, p = 0.016, for pre- and postmenarche, respectively).

In univariate analysis, ΔTBBMC was positively related to baseline weight (r = 0.36, p < 0.05) and fat mass (r = 0.44, p < 0.05), but unrelated to baseline height. ΔTBBMD was unrelated to any of the baseline measures. The rates of change in weight and lean mass were strongly positively related to ΔTBBMC (r = 0.67, p < 0.001 and r = 0.68, p < 0.001, respectively), whereas Δheight and Δfat mass were not. No significant correlations emerged between ΔTBBMD and rates of change in any parameter. The relationship between gain in height and gain in bone mass is shown in Fig. 1.

Figure FIG. 1.

Scatter plots showing the relationship between height velocity and gain in total body bone mineral content (ΔTBBMC) and total body bone mineral density (ΔTBBMD) in pubertal girls.

Figure 2 shows the relationship between Δheight, ΔTBBMD, and ΔTBBMC, and pubertal status (months postmenarche). The peak in height velocity preceded the period of observation in this cohort; as expected, height velocity slowed dramatically after menarche, but the slope of the regression line remained positive throughout the study. In contrast, the peaks in TBBMD and TBBMC accumulation occurred at menarche, at least 2 years after the peak in height velocity. The slope of the regression lines for TBBMD and TBBMC again remained positive for the duration of the study, thus bone growth was not complete by 30 months postmenarche.

Figure FIG. 2.

Relationship between gains in (A) height (Δheight), (B) total body bone mineral content (ΔTBBMC), (C) total body bone mineral density (ΔTBBMD), and pubertal status (months postmenarche). Data points were derived by calculating the slope of the regression line for each individual using all four time points in the study.

Figure 3 depicts the same analysis for iBAP (formation marker) and iFDpd/Cr (resorption marker). The peak increment in these markers occurred, as with height velocity, up to 20 months prior to menarche. The same pattern emerged for OC and NTx/Cr (R2 = 0.58 and 0.63, respectively, data not shown). Subsequently, the slopes became negative, reaching a nadir around the time of menarche, indicating a fall in marker levels. The regression lines suggest a plateau (no change in marker levels) approximately 2 years postmenarche, although there are few data points. Figure 4 shows changes in hormone levels in relation to months postmenarche. For PTH, the slope is negative throughout most of the observation period, indicating a decline in serum PTH levels. In contrast, the slope for E2 is positive throughout the study. Changes in serum IGF-I are positive until ∼6 months postmenarche (Tanner stages III-IV), indicating maximum serum levels at around this time.

Figure FIG. 3.

Relationship between markers of bone turnover and months postmenarche. (A, B) iBAP, (C, D) iFDpd. Graphs (A) and (C) show the actual data points for each subject throughout the study. Graphs (B) and (D) were plotted by calculating the slope of the regression line for each subject using all data points then fitting a summary regression line.

Figure FIG. 4.

Relationship between hormone levels and months postmenarche. (A, B) PTH, (C, D) E2, (E, F) IGF-I. Graphs (A), (C), and (E) show the actual data points for each subject throughout the study. Graphs (B), (D), and (F) were plotted by calculating the slope of the regression line for each subject using all data points then fitting a summary regression line.

Significant positive correlation coefficients were observed between mean concentrations of markers of bone turnover throughout the study period and height velocity (Table 3). However, there was no relationship between indices of bone turnover and gain in either TBBMC or TBBMD. Mean serum E2 levels were significantly correlated with ΔTBBMD. Both baseline and mean E2 and IGF-I levels were inversely related to height velocity. E2 and IGF-I levels were correlated (r = 0.50, p = 0.002). In multivariate analysis, regressing biochemical marker and hormone levels on relative increases in bone mass or height, E2 was significantly positively associated with the increase in TBBMD and negatively with the increase in height. iBAP was the only marker associated with bone mineral gain (TBBMD and TBBMC). iFDpd and OC were positively associated with the gain in height.

original image

The estimate of FA was unrelated to increments in TBBMC and TBBMD, but a significant negative relationship emerged with height velocity (Table 3). FA increased significantly between baseline and 18 months (p = 0.021, paired t-test). Accordingly, months postmenarche was significantly associated with FA (r = 0.52, p = 0.011) and was the only significant predictor of FA in stepwise multiple regression analysis controlling for serum E2, IGF-I, and calcium intake (p = 0.012, adjusted R2 = 23%).

In univariate analysis, strong negative correlations were observed between changes in markers of bone turnover through puberty, and serum E2 (−0.67 < r < −0.80) and IGF-I levels (−0.45 < r < −0.66). In multivariate analysis regressing E2, IGF-I, PTH, and months postmenarche on to biochemical markers (Table 4), E2 was independently and negatively associated with all four markers of bone turnover.

original image

Regression models were derived to explore the associations between percentage changes in TBBMC, TBBMD, and height (dependent variables), and body composition, height gain (for bone models), pubertal status, hormonal status, diet, and physical activity (Table 5). The increment in lean mass and serum E2 concentration were the only predictors of gain in TBBMC (R2 = 64%). These two variables also showed the strongest associations with the increase in TBBMD. The increase in fat mass, and dietary energy intake, were also positively associated with TBBMD gain, while PTH was negatively associated (model R2 = 67%). Finally, 86% of the variance in height gain was explained by a positive association with gain in lean mass and negative associations with gain in fat mass and mean IGF-I level.

original image


This prospective study of bone growth in adolescent girls has confirmed the rapid acceleration in bone mineral accretion associated with puberty. During the transition from a median Tanner stage of II (early puberty) to a median of IV (mid-late puberty) over the course of the 18-month observation period, the gain in TBBMC was 255 ± 69 g/year (15.6 ± 4.0%/year). The changes in TBBMC were paralleled by those in TBBMD, although the annualized percentage gain was lower (5.5 ± 1.8%/year) for the more size-independent areal density. A number of cross-sectional4,7,15,20,21 and longitudinal5,8,22 studies have established the importance of puberty to bone mineral acquisition.

In this study, we have observed a dissociation between statural growth and gain in bone mass of the total body. There was no correlation between the two parameters, and height gain was not predictive of bone gain in multivariate analysis. Height velocity peaked more than 2 years prior to menarche, whereas the increment rate in bone mass peaked at menarche. These are consistent with previous observations at appendicular23,24 and axial6,8 sites in pubertal subjects. It has been suggested that this phenomenon may underlie the susceptibility of the immature skeleton to fracture. In both genders, there is a large increase in fracture incidence coinciding with the timing of the adolescent growth spurt, principally caused by only slight trauma rather than sporting or play activities.24,25

Furthermore, we found that the peak in bone turnover occurred 1–2 years in advance of peak mineral accretion. Changes in marker levels closely parallel changes in growth velocity during puberty, with maximum concentrations obtained in Tanner stages II to III.26 Accordingly, we found positive associations between Δheight and markers of bone turnover. However, there was no association between marker levels and ΔTBBMC or ΔTBBMD, with the exception of a weak association with bone alkaline phosphatase, but it is possible that this may be explained by its putative role in the mineralization process.27

Whether the association between marker levels and height velocity could have any clinical relevance is uncertain. The correlations were relatively weak (0.40 < r < 0.62), and only two of the markers (OC and iFDpd) were associated with height velocity in multivariate analysis controlling for puberty. While other authors have also found that biochemical markers are predictive of height velocity,11–13 this has only been demonstrable when subjects with disease are included in the analysis, or when the age range of subjects studied is very wide (ensuring that any pubertal “event” will correlate with almost any other). That caveat aside, it is apparent that statural growth is to a limited extent reflected in indices of bone turnover at puberty, but not the rate of mineral accretion. Our analysis showed that marker levels reached a nadir around the time of menarche and beyond, despite continued bone accumulation. Similar findings were reported by others for serum alkaline phosphatase28 and OC.7

Parfitt25 attributes both phenomena—the delayed mineral gain relative to height gain and the dissociation between bone turnover indices and bone mass gain—to a late-pubertal decline in the rate of intracortical remodeling and subsequent decrease in cortical porosity. He suggests that a temporary increase in cortical porosity is a functional response to the increased demand for calcium required for the rapidly growing metaphyses of the long bones during the growth spurt. Whether or not such a temporary “mineral redistribution” process is wholly or partly responsible for the dissociation between height and mineral gain is not known, but alternative explanations are considered. First, the inability of the mineralization process to keep pace with the growth in length of the long bones may be the inevitable consequence of the magnitude of the sex steroid-driven growth spurt. Bone modeling at the epiphyses/metaphyses may be so active that skeletal volume is expanding at a faster rate than the mineralization process. The bone thus formed would consist primarily of unmineralized growth plate and relatively undermineralized metaphyseal bone. Furthermore, Parfitt25 states that during growth, bone laid down beneath the periosteum may be resorbed at the endocortical surface within a few months, long before there has been time for attainment of maximal mineral density.

Second, artifacts of scaling could explain the dissociation between statural growth and bone mineral gain. In simple geometric terms, skeletal calcium content would not be expected to parallel height velocity, but would be expected to parallel the growth in skeletal volume, which scales with the third power of its linear dimensions. The volume of the skeleton, rather than its length, is therefore correlated with body weight29 (hence strong correlations between gain in weight and gain in bone mass). If growth was proportionate, then the surface area of the body would increase as the square of linear dimensions, and body volume would increase as the cube of linear dimensions. Thus, if all dimensions of one body are 10% greater than another, then the larger body has 10% greater height, but 21% more surface area and 33% greater weight. Even in the face of proportionate growth, the ratio of “height velocity” to “volume velocity” would be expected to change.

Growth is not proportionate, however, confounding the relationship between statural growth and bone mineral acquisition still further. Body proportions change with age throughout growth, as do the shapes (not just size) of different bones. Growth does not take place at a uniform rate throughout the skeleton. For example, during puberty the peak velocity for limb length precedes the peak for the trunk by approximately 1 year, and the magnitude of the growth spurt in trunk length is also greater than that in the lower limbs.10 Because of the differential rates of growth of bones in different regions of the body, height velocity would not be expected to show a direct relationship with the mineral content of individual bones or of the whole skeleton. In addition, Wales and Gibson30 stress the complexity and nonlinearity of the growth processes in both the whole body and individual bones, which includes “mini growth spurts,” periods of “negative” growth, and seasonality of growth.

In this study, there was a strong inverse relationship between mean serum E2 and the change in markers of bone turnover, confirming observations made by Blumsohn and associates in a previous cross-sectional study.26 The effect of estrogens on growth in healthy girls is biphasic.26 At low levels (early puberty), E2 stimulates growth, while at higher levels in later puberty it has a potent inhibitory effect on longitudinal growth and accelerates epiphysiodesis. It is therefore probable that E2 is responsible, either directly or indirectly via its inhibitory effect on the growth plate, for the reduction in bone turnover in late puberty.

Serum E2 concentration also emerged as a highly significant determinant of gain in bone mass. This could be related in part to the above-mentioned effects on bone turnover and inhibition of statural growth in late puberty, resulting in the continued mineralization of the physis and newly formed metaphyseal bone. Others have demonstrated a positive association between measures of estrogen exposure and areal bone density in adolescent girls and young women.31,32 Several studies have related Tanner stage to bone mass measurements, using dual-energy X-ray absorptiometry (e.g., studies in Refs. 4, 15, 21, and 33) and quantitated computerized tomography.20 Clinically, low levels of endogenous estrogen are associated with menstrual dysfunction in adolescent girls and young women, resulting in low bone density or the prevention of normal rates of accretion.34 In normal adolescents, girls with earlier menarche or regular menses had higher bone mass than those with later menarche or irregular menses.35

It is likely that hormonal status acts synergistically with body weight and nutritional status with respect to skeletal status, as exemplified by the bone loss observed in anorectics.9 In this cohort of healthy girls, dietary energy intake was a significant correlate of bone accumulation in multivariate analysis controlling for measures of body size including height. Other factors were unrelated to bone acquisition in this cohort, including calcium intake and physical activity, which could relate to the small sample size and the observational study design. Intervention trials in children and adolescents have demonstrated positive effects of calcium or dairy supplementation on bone mass acquisition.14,36–38

We observed a significant reduction in PTH levels from early to late puberty, evaluated in terms of months postmenarche. Few longitudinal data are available, but in boys serum concentration of PTH did not change during puberty; however, values in children and adolescents did tend to slightly exceed normal adult values.39 We found no effect of PTH levels on changes in levels of markers of bone turnover in these subjects. PTH showed a significant inverse association with gain in TBBMD, which could relate to the reduction in PTH levels at the time when the gain in TBBMD is at its peak (at menarche).

FA increased significantly over 18 months, from early to late puberty. This accords with the work of others who have found modest increases in absorptive efficiency in pubertal compared with prepubertal subjects40,41 and is consistent with the peak in calcitriol concentrations in Tanner stages III-IV.42 Since fractional absorption was increasing at a time when height velocity was decreasing, we noted a negative correlation between these two parameters and a positive correlation with months postmenarche. No relation with gain in bone mass was seen. However, in a large study in healthy pubertal girls, Illich and associates42 reported a significant positive correlation between serum calcitriol concentrations and bone mass accumulation.

We observed a negative correlation between serum IGF-I and height velocity. The role of IGF-I as a mediator of the effects of growth hormone, together with its stimulatory effect on longitudinal bone growth via the clonal expansion of chondrocytes in the growth plate, is established.43 Nevertheless, previous authors have also observed the lack of correlation between the pubertal growth spurt and IGF-I concentrations.26,44 In our study, IGF-I levels progressively increased until about 6 months postmenarche, corresponding with pubertal stages III-IV. Juul et al.44 studied 877 healthy Caucasian children and adolescents and observed that IGF-I levels were maximal at 14.5 years in girls and 15.5 years in boys—almost 2 years later than the average peak height velocity. In addition, IGF-I levels were positively related to bone turnover in prepubertal children and early puberty, but after peak height velocity, markers of bone turnover decreased dramatically, whereas IGF-I levels decreased only slightly.26 The above findings explain the absence of a correlation between IGF-I and growth velocity in pubertal subjects; our finding of a negative association likely relates to the number of subjects actually reaching their peak growth velocity within the observation period, resulting in maximal dissociation between the two variables. The absence of a continued stimulatory effect of IGF-I on the growth plate, despite persistently high serum concentrations in later puberty, is likely related to the high estradiol concentration at this time, which as noted above has a potent inhibitory effect on continued longitudinal growth.

In addition to its chondrocyte mitogenic effects, IGF-I is anabolic to bone itself; it stimulates osteoblast proliferation and differentiation, and matrix formation including the synthesis of type I collagen and other protein components.43 Despite the late pubertal peak in IGF-I concentrations, there was no association between IGF-I and gain in bone mass in this study. Whether serum levels of IGF-I reflect bone tissue concentrations in children is unknown, but in adults there is a concomitant decline in circulating and skeletal IGF-I with age.45 In addition, the relationship between IGF-I and its binding proteins, especially IGFBP3 (the predominant circulating IGFBP) remains to be fully elucidated. The binding proteins are probably involved both in IGF-I transport, and in the regulation of its action at the cellular level.46 Thus, the total serum concentration of IGF-I may not accurately reflect the level of bioavailable IGF-I. Nevertheless, the relationship between IGFBP3 and pubertal status is similar to that of IGF-I, with significantly higher levels observed in late puberty, and in postmenarchal versus premenarchal girls.26

The two major components of body weight, lean and fat mass, were correlated with gain in TBBMD, although the gain in lean mass showed the strongest relationship, and lean mass alone was related to gain in TBBMC. This finding with respect to lean mass is consistent with the work of others in studies among adolescents and young women.22,47–49 Several mechanisms for the strong association between lean mass and bone mass have been postulated, including the effect of muscular forces applied to bone,50 a shared genetic component for BMD and muscle mass,51 and common hormonal mediators (e.g., growth hormone, androgens, or insulin) driving anabolism in both lean and bone tissue.52 Additionally, Millward53 suggests that one of the key physiological stimuli for muscle growth involves a “passive” effect of muscle stretching by bone lengthening during growth. If this is the case, then a close association between lean mass (of which skeletal muscle is the largest component) and bone growth seems logical, at least up until the time of epiphyseal closure.

We conclude that, during puberty in girls, estrogen status and changes in lean mass are important correlates of bone accumulation. Height gain is not associated with bone accumulation during this period of rapid growth; peak height gain is an early pubertal event, whereas maximal bone mineral accumulation occurs later in puberty, around the time of menarche. This could be due to changes in cortical porosity during growth,25 rapid expansion of skeletal volume and a consequent delay in mineralization, disparate growth rates throughout the skeleton, or simple artifacts of scaling. Biochemical markers of bone turnover were modestly correlated with statural growth but showed very little predictive value for gain in bone mass.


The authors thank Dr. Diane Baldwin, Kings College Hospital, London, for carrying out the strontium assay. This study was supported by grants from the European Community (EEC grant Reg 116/92) and the Nutritional Consultative Panel of the U.K. Dairy Industry (scholarship for J.C.).