Fetal Growth Velocity, Size in Early Life and Adolescence, and Prediction of Bone Mass: Association to the GH–IGF Axis


  • The authors state that they have no conflicts of interest.


Poor growth in early life is associated with numerous adverse outcomes later in life. In 123 adolescents 16–18 yr of age, the previous findings of a positive relation between size in early life and later bone mass was confirmed. These associations were mediated by the current height and weight, but it was not confirmed that alterations of the GH–IGF axis cause this.

Introduction: Numerous studies have found associations between low birth weight and disease later in life, including decreased bone mass.

Materials and Methods: A longitudinal cohort of 16- to 19-year-old adolescents (n = 123) with data on third trimester fetal growth velocity (FGV) was assessed by serial ultrasound measurements, birth weight (BW), and weight at 1 yr. A follow-up study included DXA scan, anthropometric measurements, and measurements of the growth hormone (GH) –IGF-I axis in a representative subpopulation (n = 30).

Results: BW and weight at 1 yr were positively associated with whole body BMC (p = 0.02 and p < 0.0001, respectively), lumbar spine BMC (p = 0.001 and p = 0.03, respectively), and lumbar spine BMD (p = 0.04). After correction for adolescent height and weight, no association remained significant. There was no relation between IGF-I and IGF binding protein 3 (IGFBP-3) levels in adolescence and size in early life or bone mass. In the subpopulation, GH secretion (median, 2.58 versus 4.05), GH pulse mass (median, 10.7 versus 19.4 mU/liter), and total GH (median, 74.9 versus 108.8 mU/liter/12 h) were decreased in the small for gestational age (SGA) group compared with the appropriate for gestational age (AGA) group; this did not reach statistical significance. Likewise, there were no differences in IGF-I, IGF-II, and IGFBP-1, −2, and −3 levels between the SGA and AGA groups. A statistically significant positive association between FGV and adolescent IGF-II was found (B = 199.9, p = 0.006). Significant negative associations between GH measurement and BMC, as well as BMD, were found (B = −0.008, p = 0.005 and B = −0.008, p = 0.006, respectively).

Conclusions: This study confirms the previous findings of a positive relation between size in early life and later BMC, an association apparently independent of the distal part of the GH/IGF-I axis. However, this association may be mediated mainly by postnatal growth determining size of the skeletal envelope rather than an effect of fetal programming on bone mass per se.


Poor growth during early life has been linked to a number of adverse health outcomes in adulthood, including decreased adult bone mass. Accordingly, adult BMC was positively correlated to weight at 1 yr of age,(1–4) and in other studies, BMC was positively correlated to birth weight.(3,5–7) Controversy exists on the possible association between size in early life and BMD.(3,6) Recent studies showed that the link between poor growth in early life and decreased bone mass in adulthood reflects changes in bone geometry,(8) and a low growth rate during childhood growth is related to a subsequent increased risk of fractures in adulthood.(9)

The growth hormone (GH)–IGF axis regulates both linear growth and bone remodeling, and two studies concluded that fetal programming of the GH–IGF-I axis may play a role for bone mass in elderly people.(10,11) Programming is a term used to describe the effect of environmental stimuli during a critical period of early life, which may result in permanent changes of the endocrine systems,(12) such as the GH–IGF-I axis.

Fetal growth is generally considered independent of fetal GH levels, whereas IGF-I and IGF-II are important endocrine and paracrine regulators of fetal and placental growth by stimulating cell division and differentiation. IGF-II is considered one of the primary growth factors during the embryonic period in early gestation but not in late pregnancy, and consequently, the majority of studies find no association between IGF-II and size at birth.(13) Fetal IGF-I is important for late fetal growth, and IGF-I concentrations increase with advancing gestational age(13,14) and correlate positively with weight corrected for gestational age at birth.(15–17) IGF-I levels in cord blood are positively correlated to neonatal BMC, but this association disappears when correcting BMC for bone size,(16) and controversy exists.(17) The majority (∼90%) of children born small for gestational age (SGA) exhibit an accelerated postnatal growth within the first 2 yr of life and reach a final height within normal limits.(18,19) Presumably, SGA children have an altered GH secretion with a lower mean GH and total GH secretion,(20) but the mechanisms that induce accelerated growth in some children and not in others remains to be determined.

To assess the contribution of fetal and infant growth on the GH–IGF axis, height, and bone mass in adolescence, we studied a unique longitudinal cohort of adolescents. They were examined by DXA scan and anthropometric data at the age of 16–19 yr, and we also had access to data on third trimester fetal growth assessed by serial ultrasound measurement, birth weight, and 1-yr weight.



One hundred twenty-three healthy young white subjects (52 males) participated in a follow-up study of a larger controlled trial on fetal growth.(21) The original study was performed in 1985–1987, where pregnant women with one or more risks factors for giving birth to a SGA child were recruited.(21,22) Third trimester fetal growth velocity (FGV) based on serial ultrasound measurements was calculated, birth anthropometrics were measured, and one half of the cohort was examined at 3, 6, and 12 mo of age. In the follow-up study performed in 2003–2005 (previously described in detail), a sample of 271 young people was invited, and 123 (52 males) agreed to participate (participation rate, 45%). A DXA scan was performed on all the participants, but four participants were excluded because of incomplete DXA scanning results. Of the remaining 119 participants; 73 (32 males) were born appropriate for gestational age (AGA; 27 [8 males] with verified intrauterine growth restriction [IUGR]) and 46 (20 males) were born SGA (17 [7 males]) with verified IUGR). Calculation of SD scores (SDS) of BMC adjusted for area was not possible in six participants because the values were above the normal reference curve. In addition, a study on the overnight pulsatile GH secretion was performed. All 52 male participants were contacted again and asked to attend this study, and 32 agreed to participate; however, 2 young men did not attend on the days scheduled for the examination, leaving 30 subjects to complete the overnight pulsatility study.


All participants were asked to fill out a validated questionnaire including information on smoking and physical activity.(23) A validated food frequency questionnaire (FFQ) was filled out, and dietary calcium intake was calculated. Anthropometric measurements through childhood were collected retrospectively and merged with the previously measured values.


Height was measured to the nearest 0.1 cm using a calibrated wall-mounted stadiometer (Force Institute), and weight was measured on a digital weight scale with a precision of 0.01 kg (Lindeltronic 8000). SDS for height and weight were calculated using Danish national reference material.(24) Pubertal stages were determined by Tanner's classification, and testicular size was estimated using Prader's orchidometer. Target height (TH SDS) was calculated as maternal height (SDS) + paternal height (SDS)/2 and was available in 95 subjects. Height (H SDS) corrected for target height was calculated as H SDS − TH SDS. Change in H SDS (ΔH SDS) was calculated as H SDS − birth length (SDS).

Bone mineral assessment

Whole body BMC measured in grams hydroxyapatite, bone size expressed as anterior-posterior projected bone area (BA) measured in square centimeters, and BMD (BMC/bone area) was determined by DXA using a Hologic QDR-1000/W scanner (Hologic, Waltham, MA, USA). The software version 5.61 for whole body was used for the analyses. The subjects wore underwear and a cotton t-shirt during the scan. The CV for BMC and BA measurements was 0.37% and 0.28%, respectively, calculated from spine phantom scans.(25) BMC, BMD, and BA of the lumbar spine were calculated from the whole body DXA. The entrance radiation dose level was 15 μSv, with an effective dose of 10 μSv equivalent to the background radiation 1 day in Denmark. SDS for height for age, BA for height, and BMC for BA were calculated from age- and sex-specific reference materials based on local data collected from the same DXA scanner.(25)

Blood sampling

A blood sample was drawn from an antecubital vein after an overnight fast for determination of total IGF-I and IGF binding protein 3 (IGFBP-3) in all subjects (n = 119).

Overnight pulsatile GH secretion

The participants (n = 30 males) were admitted to the department at 7:00 p.m., where an intravenous catheter was inserted into a forearm vein and connected to a constant withdrawal pump by a heparinized catheter. From 8:00 p.m. to 8:00 a.m., blood samples (4 ml) were collected at 20-min intervals for GH determinations. A meal was served between 8:00 and 9:00 p.m., the participants were encouraged to sleep at 12:00 a.m., and room lights were turned off at 1:00 a.m. All IGF-related parameters (see below) were determined on a single fasting blood sample taken at 8:00 a.m.


The frequency of GH receptor (GHR) transcript variants with retention (fl-GHR) or exclusion (d3-GHR) of exon 3 was tested by the multiplex PCR assay described by Pantel et al.(26) The results of this analysis have been reported in detail previously.(27)

Hormone assays

Serum IGF-I was determined by radioimmunoassay (RIA) as previously described.(28,29) Inter- and intra-assay CVs were 9% and 6%, respectively. IGFBP-3 was determined by a RIA as previously described(30) (assay reagents from Mediagnost, Tübingen, Germany). Sensitivity was 0.29 ng/ml; inter- and intra-assay CVs were 10.7% and 7.6%, respectively.(31) GH was determined by a time-resolved immunofluorometric assay (TR-IFMA; DELFIA; Perkin Elmer Lifesciences, Turku, Finland). Detection limit was 0.05 mU/liter; inter- and intra-assay CVs were 4.15% and 4.87%, respectively. In the subpopulation, serum IGF-II was determined after acid-ethanol extraction using noncompetitive TR-IFMA as previously described.(32) Inter- and intra-assay CVs were <10% and 5%, respectively. Serum free IGF-I was determined using ultrafiltration by centrifugation as previously described.(33) The detection limit of free IGF-I in the ultrafiltrates was 20 ng/liter. Including ultrafiltration and immunoassay, the intra-assay CV averaged 18%. Serum IGFBP-1 was determined by an in-house RIA performed as described previously(34); inter- and intra-assay CVs were 16% and 5%, respectively. Serum IGFBP-2 was determined by an in-house TR-IFMA based on reagents from R&D Systems (Abingdon, UK); inter- and intra-assay CVs were 12% and 5%, respectively.(35)

Analysis of pulsatile hormone secretion

A two-component deconvolution analysis was applied to estimate pulsatile GH secretion in each individual from the overnight profiles, using techniques described previously by Veldhuis et al.(36) A mean first component (fast component) half-life of 3.5 min and a relative contribution of the slow component to the total elimination of 0.63 (fraction) were used.(37) The analyst was blinded to the classification of the subjects. Approximate entropy (ApEn) analysis was calculated.(38) ApEn parameters of m = 1 and an r value of 0.35 (35%) of the SD of the individual subject time series were chosen. Higher absolute ApEn denotes greater disorderliness of hormone release and vice versa for lower ApEn.

Statistical analysis

To evaluate males and females together, all variables were presented as SDSs. All variables were tested for normal distribution, and the dependent variables with a skewed distribution were transformed. Multiple linear regression analyses were performed to determine the associations between early growth and bone mass in adolescence and to identify confounding variables. Because of an intentionally skewness in the sampling of the cohort, FGV and BW were included together as independent variables in all multiple regression analyses. Furthermore, when determining the effect of either FGV or BW, we countered for this bias by weighting the under-represented group and applied bootstrapped SE of estimates for the calculation of CIs of the weighted estimates. For clarification, the results are presented as median and interquartile range (IQR) and tested by Mann-Whitney U-test because the results did not alter. All statistical analyses were performed using the statistical package SPSS (version 15; SPSS, Chicago, IL, USA).

Ethical aspects

This study was performed according to the Helsinki II declaration and approved by the local Ethical Committee (KF 01-229/02 and KF 01-065/03) and The Danish Data Protection Agency. Written informed consent was obtained from all participants and from the parents/guardians of the participants <18 yr of age.


DXA results were available for 119 adolescents (52 males) who were mature singletons at birth (37 > GA < 42 wk; Table 1). At the time of follow-up, the mean age was 17.5 ± 0.7 (SD) yr, mean height (SDS) was 0.16 ± 1.0 SDS, and mean weight (SDS) was 0.9 ± 1.7 SDS. In the entire cohort, mean whole body BMC (adjusted for sex and age) was 0.11 ± 1.4 SDS, mean lumbar spine BMC (SDS) was 0.10 ± 1.2 SDS, mean whole body BMD (SDS) was −0.0004 ± 1.23 SD, and mean lumbar spine BMD was −0.19 ± 1.02 SDS. Of the responders to the questionnaire (n = 103), 36% were current smokers, and 76% of those (n = 28) smoked <10 cigarettes/d. Mean daily activity (measured as MET-time) was 46.6 ± 10.3 min. The mean daily calcium intake was 1175 ± 862 mg. Pubertal stage did not differ between those born SGA (±IUGR) and those born AGA (±IUGR) (χ2, p = 0.41).

Table Table 1.. Clinical Characteristics of Boys and Girls
original image

In multiple linear regression analyses, the association between FGV, BW, weight at 1 yr of age, and bone mass parameters in adolescence were tested (Table 2). Third trimester FGV was weakly but significantly associated with whole body BMC, but there was no significant correlation to any of the other variables. We found a positive association between BW and whole body BMC (SDS), lumbar spine BMC (SDS), and lumbar spine BMD (SDS). A very strong association between weight at 1 yr and whole body BMC (SDS) was determined, and weight at 1 yr was also significantly associated with lumbar spine BMC. BW and weight at 1 yr were strongly correlated to height in adolescence, and weight at 1 yr was also strongly correlated to weight in adolescence (Table 2). The association between current height and weight and bone mass in adolescence was highly significant for all the bone variables (data not shown). Adjustment for current height (SDS) altered the significant associations between BW, weight at 1 yr, and bone mass, and these associations were no longer significant except the association between weight at 1 yr and BMC (SDS). When adjusting for both height and weight, none of the associations between size early in life and bone mass remained significant. Adjusting for BMI did not alter any of the results. Furthermore, adjustment for other confounders such as smoking habits, physical activity, and calcium intake in adolescence did not alter the results.

Table Table 2.. Associations Between Early Growth and Bone Mass, Anthropometric Measurements, and Growth Factors
original image

A method of correction for size-related artefacts previously described by Prentice et al.(39) was applied to our dataset. In this model, all variables were transformed by the natural logarithm, and BMC was included as a dependent variable, and BA, height, and weight were included as independent variables in all multiple regression analyses. Thus, this size-adjusted model detected no significant associations between FGV (B = −0.000; 95% CI, −0.03 to 0.03; p = 0.98), BW (B = −0.002; 95% CI, −0.08 to 0.08; p = 0.97), or weight at 1 yr (B = −0.16; 95% CI, −0.35 to 0.02; p = 0.09) and adolescent BMC.

The three-step analysis previously described by Molgaard et al.(25) was also applied to our dataset. The first step of this analysis revealed that height (SDS) was strongly associated to both BW (p < 0.0001) and weight at 1 yr of age (p < 0.0001; Table 1; Fig. 1A). In the next step, BA adjusted for height according to a normal reference was not associated with BW (B = 0.03; 95% CI, −0.17 to 0.23; p = 0.76) but was strongly associated to weight at 1 yr (B = 0.52; 95% CI, 0.21–0.83; p = 0.001; Fig. 1B). In the last step, BMC adjusted for BA according to a normal reference population was negatively associated with both birth weight (B = −0.27; 95% CI, −0.45 to −0.09; p = 0.004) and weight at 1 yr (B = −0.64; 95% CI, −0.94 to −0.34; p < 0.0001; Fig. 1C). No significant correlations were found between FGV and any of the variables in the three-step analysis.

Figure FIG. 1..

Associations between (A) height corrected for sex and age, (B) bone area corrected for height (SDS), (C) BMC corrected for bone area (SDS), and (top row) birth weight (SDS) and (bottom row) weight (SDS) at 1 yr.

There was no significant association between IGF-I and IGFBP-3 and FGV, BW, or weight at 1 yr in the entire cohort (Table 2). In the study of overnight GH secretion (n = 30 young men), the GH profiles (Fig. 2) showed a trend toward higher mean GH secretion (median, 4.05 versus 2.58 mU/liter), higher GH pulse mass (median, 19.4 versus 10.7 mU/liter), and higher overall GH secretion (108.8 versus 74.9 mU/liter) in the group born AGA compared with those born SGA, but this difference did not reach statistical significance (Table 3). Likewise, there was no difference in total and free IGF-I, IGF-II, or IGF binding proteins (IGFBP-1, −2, and −3) between the AGA and SGA groups (Table 3). In a multiple regression analysis with BW (SDS) and FGV as independent variables, there was no association to IGF-I or IGFBP-1, −2, or −3. However, a significant positive association was found between third trimester FGV and IGF-II in adolescence (B = 200; 95% CI, 62–339; p = 0.006). Measurements on total GH secretion were negatively associated with both BMC (B = −0.008; 95% CI, −0.01 to −0.003; p = 0.005; Fig. 3) and BMD (B = −0.008; 95% CI, −0.01 to −0.002; p = 0.006), and this was also the case for mean GH secretion and GH pulse mass.

Table Table 3.. GH Secretion and IGF Axis According to BW
original image
Figure FIG. 2..

Twelve-hour overnight profiles of GH in the group born AGA (black lines are those with verified IUGR) and the group born SGA (black lines are those with verified IUGR).

Figure FIG. 3..

Association between BMC (SDS) and total GH secretion.

The presence of the d3-GHR allele did not influence BMC or BMD; all comparisons were nonsignificant, with p values >0.05 (data not shown).

There was a near-significant association between GH and change in height (ΔH SDS; B = −47.2; 95% CI, −97.2 to 3.6; p = 0.067), and in the entire cohort, BMC was positively associated with ΔH SDS (B = 0.41; 95% CI, 0.16–0.66; p = 0.001).


In this unique longitudinal cohort of healthy young people with data on third trimester FGV, birth anthropometrics and DXA scans in adolescence, we found a significant association between size in early life and whole body and lumbar spine BMC in adolescence. There was an association between BW and adolescent lumbar spine BMD, but no other associations between infant size and BMD were found. These findings are in line with previously published studies. However, when correcting for the strong association between size in early life and adolescent height and weight, the association between BW, weight at 1 yr, and bone mass no longer remained significant. There was a near significant trend toward decreased mean GH secretion, GH pulse mass, and overall GH secretion in those born SGA compared with those born AGA. A significant negative association between GH measurements and both BMC and BMD was found. There were no differences in IGF-I, IGF-II, and the IGFBPs between the SGA and AGA group, but surprisingly, a positive association between FGV and adolescent IGF-II was found.

The GH–IGF-I axis regulates both linear growth and bone remodeling; thus, programming of this endocrine axis has been proposed to influence acquisition of bone mass.(10,11) Intrauterine bone accumulation of calcium primarily occurs during third trimester. Fetal growth is regulated mainly by substrate supply and the glucose–insulin–IGF-I axis, and programming of IGF-I may have effects throughout childhood and adulthood,(40,41) but controversy exists.(42,43) In this study, we found no associations between IGF-I and IGFBP-3, respectively, and size in early life in the entire cohort. In the overnight profile study (n = 30), there were no differences in total or free IGF-I or any of the binding proteins between children born SGA and AGA. There were no associations between either BW or FGV and the IGF axis apart from a strong significant association between third trimester FGV and IGF-II in adolescence. IGF-II is an important growth factor in utero, with increasing levels throughout gestation, but in contrast to the strong link discovered between IGF-I and BW, results on the association between IGF-II and birth weight are inconsistent.(15,44,45) To our knowledge, programming of IGF-II from fetal life to adolescence has not been shown in any previous study. Thus, our study did not support a fetal programming of IGF-I, but this could reflect the small number of participants.

Programming of GH was suggested by a trend toward lower GH secretion in children born SGA compared with those born AGA. Surprisingly, GH was inversely correlated to BMC and BMD indicating that those with a low GH secretion had a high bone mass. Accelerated growth during childhood was also negatively associated with GH secretion but positively associated with BMC. There may not be a causal relation but the finding that those with a slower growth, and thereby a decreased adult height and a decreased BMC, had an increased GH secretion may indicate a compensatory mechanism or a degree of GH resistance.

The presence of the d3-GHR allele has previously been reported to influence postnatal growth, but there was no association between genotype and BMC or BMD in this study, which confirms the study by Kenth et al.(46)

In this study, we found strong associations between BW, weight at 1 yr, and bone mineralization in adolescence. Furthermore, a weak but significant association between third trimester FGV and BMC (p = 0.048) was determined. However, after adjustment for current size, all associations became insignificant. This confirms the previous studies of associations between BW and BMC at various skeletal sites(3,5–7) and associations between weight at 1 yr and BMC.(1–4) The link between early weight and BMD was weaker and nonsignificant in the majority of the previous studies. Correction for adult size by adjusting for current height and weight was performed in some of the studies,(1,2,4,6) and the link between size early in life and bone mass no longer remained significant. In some studies,(3,7) correction for size was made by adjusting for BMI, which did not alter the results. In a large twin study (n = 1411 pairs), the intrapair differences in BW were significantly linked to both BMC and BMD, but when adjusting for current height and weight in the entire cohort, these associations disappeared.(5) However, when dividing the twins into monozygotic and dizygotic pairs, the association remained significant in the monozygotic twins. This observation points to an influence of the intrauterine environment. Furthermore, a study on children born prematurely concluded that bone mass at 8–12 yr of age related to current bone and body size and that this persisted throughout childhood.(47) Our findings and previous findings suggest that the link between size in early life and bone mass is mediated primarily by the size of the skeletal envelope obtained during postnatal growth rather than prenatal growth.

BMC determined by DXA is a 2D measurement reflecting BA and not bone volume, and this method may artificially overestimate relative BMC in tall people with larger bones (higher BA) and underestimate BMC in short people with smaller bones (lower BA). Adjusting for current height and weight is one way of correcting for size; another way is the three-step analysis of bone mineralization(25) that was applied to our data. This analysis showed that the decreased BMC in those with low weight at 1 yr was caused by “shorter” and more “narrow” bones; correction of BMC for BA according to a normal reference showed that the bones were actually “heavier,” However, this means that low weight in infancy was associated with a compensatory mechanism of accumulating an increased amount of calcium per square centimeter of BA. This finding was somewhat surprising considering previous studies, but this may be one of the explanations for the lack of association between size in early life and BMD. Large epidemiological studies have shown that decreased BMC and BMD in adulthood are strong predictors of osteoporosis and thereby of increased risk of fragility fractures,(48) and slow childhood growth has been linked to a subsequent increased risk of fractures in adulthood.(9) Our study confirms the relation between slow growth and decreased BMC because we found a positive association between change in height SDS from birth to adolescence and BMC. This study points toward a compensatory mechanism of increasing BMC for BA in the short bones, but this does not exclude that low weight early in life, slow growth in childhood, and subsequent low bone mass predict an increased risk of osteoporotic fractures later in life.

In conclusion, this study confirms the previous findings of a relation between size in early life and later BMC, which was not influenced by GHR genetic polymorphisms or serum IGF-I. However, serum GH correlated inversely with BMC and BMD. The associations between size in early life and later BMC may be mediated mainly by the postnatal growth determining the adult size of the skeletal envelope rather than an effect of fetal programming on bone mass per se.