SEARCH

SEARCH BY CITATION

Keywords:

  • children;
  • fracture;
  • prediction;
  • BMD;
  • puberty

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

This study reports on the association between DXA at age 8 and subsequent fractures in both male and female children. Bone densitometry at the total body and spine (but not hip) is a strong predictor of fracture (especially upper limb) during puberty.

Introduction: The aim of this study was to determine if prepubertal DXA can predict fracture risk during puberty.

Materials and Methods: We studied 183 children who were followed for 8 yr (1460 person-years). Bone densitometry was measured at the total body, hip, and spine by DXA and reported as BMC, BMD, and bone mineral apparent density (BMAD). Fractures were self-reported at age 16 with X-ray confirmation.

Results: There were a total of 63 fractures (43 upper limb). In unadjusted analysis, only total body BMD showed an inverse relationship with upper limb fracture risk (p = 0.03). However, after adjustment for height, weight, age (all at age 8), and sex, total body BMC (HR/SD, 2.47; 95% CI, 1.52–4.02), spine BMC (HR/SD, 1.97: 95% CI, 1.30–2.98), total body BMD (HR/SD, 1.67; 95% CI, 1.18–2.36), total body BMAD (HR/SD, 1.54; 95% CI, 1.01–2.37), and spine BMD (HR/SD, 1.53; 95% CI, 1.10, 2.22) were all significantly associated with upper limb fracture risk. Similar, but weaker associations were present for total fractures. There was a trend for overweight/obesity to be associated with increased upper limb fracture risk (HR, 1.53/category; p = 0.08).

Conclusions: Measurement of bone mass by DXA is a good predictor of upper limb fracture risk during puberty. Although we did not measure true BMD, the constancy of fracture prediction after a single measure suggests bone strength remains relatively constant during puberty despite the large changes in bone size.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

Fracture incidence is bimodal, with a peak in adolescence and a peak in the elderly.(1) Fractures in people <50 years of age are more common than those caused by osteoporosis by a factor of three,(2) with wrist and forearm fractures being the prime reason for hospitalization in children 10–14 yr of age.(3) As well as their high incidence, significant consequences are associated with upper limb injury: cost, hospitalization, immobilisation of the limb, missed school days,(4) developmental sequelae,(5) and possibly osteoarthritis if the fracture is intraarticular. Fracture risk during adolescence increases(6) because of a number of factors including a delay in bone mineralization relative to bone growth. Indeed, a number of case control and cross-sectional studies have suggested BMD is a risk factor for childhood fracture, especially those involving the upper limb.(7–12) Recent short-term prospective studies in children have confirmed this association for total fracture incidence but have not specifically reported on upper limb fractures.(8,10) Furthermore, the wide age range in the latter cohort combined with the major effect of puberty on bone size may alter the apparent association between bone mass and fracture, resulting in a marked weakening of fracture prediction over time. A number of studies have implicated body weight or fat as a predictor of fracture in children,(10,13) although some have suggested it may be television, computer, and video watching rather than body fat per se.(14) One study has also implicated previous fracture.(10) The primary objective of this study was therefore to prospectively examine the predictive value of a single DXA measurement at age 8 on fracture incidence (including those of the upper limb) between 8 and 16 yr. A secondary objective was to examine whether measures of body fat or previous fracture predicted subsequent fracture.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

Subjects

In 1988, there were 6779 live births in Tasmania. Of these, 1411 were identified as being at higher risk of sudden death syndrome by previously published criteria(15) and were invited to take part in a longitudinal study. In southern Tasmania, there were 735 births that met these criteria. Of these, mothers of 696 infants (95%) agreed to an in-hospital interview, and the parents of 581 (80%) agreed to the 1-month follow-up. In 1996, 330 of the original cohort received bone densitometry testing at age 8 as previously reported.(16) These children were traced and asked to participate in a study on bone mass and fracture incidence in 2004–2005.

Incident fractures were defined as any fracture that occurred subsequent to participating in the 1996 study and before taking part in this study and were ascertained by self-report with X-ray confirmation (where possible). Information was collected on the age of the child when the fracture occurred, site, and circumstances of fracture. Upper limb fractures were defined as those involving the upper limb, not including the shoulder girdle. Trauma of the fracture was graded according to Landin trauma grading(17) as follows: slight to moderate, a fall from <3 m or equivalent velocity; severe trauma, >3 m or equivalent velocity. Where a fall was not involved in fracture etiology, a judgement was made as to the equivalent level of trauma.

In 1996, weight was measured to the nearest 0.1 kg (with shoes, socks, and bulky clothing removed), and height was measured to the nearest 0.1 cm (with shoes and socks removed) using a stadiometer. BMI was calculated and classified according to the Cole Criteria as normal, overweight, or obese.(18) All children were assumed to be prepubertal given the young mean age, but this was not specifically assessed.

Bone mass was assessed in 1996 as previously reported(16) using DXA at the total body, spine, and right femoral neck (QDR2000 densitometer; Hologic, Waltham, MA, USA). Bone mass was examined as BMC, BMD, and bone mineral apparent density (BMAD; g/cm3), which is an approximation of the volumetric density of bone and is calculated by dividing site-specific BMD by the square root of the area at that site.(19) Body fat was also available from this scan. Precision estimates in vivo were not available in our subjects for ethical reasons. However, the longitudinal CV for our machine during 1996 using daily measurements of a spine phantom was 0.54%.

Statistics

Student t-tests were used for comparisons of means. The longitudinal relationship between study factors and childhood fracture was analyzed using Cox proportional hazards model for both total and upper limb fractures. Insufficient fractures were present for analysis of other fracture subgroups. Time was considered as months from clinic appointment in 1996 to first subsequent fracture (or to follow-up visit in 2004–2005 if no fracture occurred). All results were adjusted for age at baseline, weight, height, and sex. A p value <0.05 (two-tailed) or a 95% CI not including the null point was considered statistically significant. All statistical analyses were performed on Intercooled Stata 9.0 for windows (StataCorp LP).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

A total of 183 children were followed up in 2004–2005 (mean age, 16.3 ± 0.4 [SD] yr) representing ∼1460 patient-years (55% of those for whom bone densitometry was performed in 1996). Total first fractures numbered 63, with 53 being slight to moderate trauma and 10 being high trauma. Forty-three children sustained an upper limb fracture; of these, 20 were forearm/wrist, 8 were finger, 5 were elbow, 5 were hand, and 5 were arm. Slight to moderate Landin trauma grade was recorded for 40 of the upper limb fractures; the remaining 3 were severe Landin grade. There were 12 lower limb fractures, 4 fractures of the nose, 2 clavicle fractures, 1 skull fracture, and 1 spine fracture. No significant differences between those studied and those lost to follow-up were observed for explanatory factors or bone densitometry variables (Table 1). There were 16 of these subjects who had two fractures, 4 who had three fractures, and 1 who had four fractures. Girls tended to have a lower age at first upper limb fracture than boys, but this was not statistically significant (11.6 versus 12.7 yr, p = 0.17). There was no difference in age for any fracture (data not shown).

Table Table 1.. Characteristics of Subjects at 8 Years of Age in the Study and Those Lost to Follow-Up
Thumbnail image of

No differences in sex, age at densitometry, height, or weight were seen between the upper limb fracture group and those without upper limb fracture (Table 2). However, children who had an upper limb fracture had lower total body BMD and a trend for lower total body BMAD and spine BMC in unadjusted analysis over the 8-yr follow-up period (Table 2; Fig. 1). The association appeared continuous for total body BMC and BMD but appeared more threshold in nature for spine BMC and BMD and total body BMAD. In multivariate analysis after adjustment for height, weight, sex, and age at bone densitometry, upper limb fracture was predicted by (in rank order) total body BMC, spine BMC, total body BMD, total body BMAD, and spine BMD, but not hip measurements (Table 3). Associations were similar in rank but consistently lower in magnitude when total fractures (including those of the upper limb) were analyzed.

Table Table 2.. Characteristics of Subjects at Age 8 Years With and Without Upper Limb Fractures and Fractures at Any Site
Thumbnail image of
Table Table 3.. Bone Mass at Age 8 Years and Fracture Prediction
Thumbnail image of
thumbnail image

Figure Figure 1. Percentage of children sustaining upper limb fractures between 8 and 16 years of age by bone densitometry quartiles: total body areal BMD (TBBMD); total body BMC (TBBMC); total body BMAD (TBBMAD); spine BMC (LSBMC); spine areal BMD (LSBMD). This figure is suggestive of a continuous relationship between fracture incidence and total body BMD and a threshold type relationship for spine BMD, spine BMC, and total body BMAD.

Download figure to PowerPoint

Figure 1 suggested a threshold effect rather than a continuous effect for spine sites and upper limb fracture risk. Analysis of the lowest quartile of bone densitometry versus the other three quartiles revealed hazard ratios for spine BMD of 2.05 (95% CI, 1.01–4.20) and spine BMC of 2.40 (95% CI, 1.14–5.06), which were statistically weaker than when examined as a continuous variable with adjustment. However, the goodness of fit was similar for both the continuous and threshold models. Log likelihood for spine BMD for continuous and threshold models were similar (p > 0.05), indicating that neither model was superior.

There was no association between percentage body fat and fracture in unadjusted (Table 2) or adjusted analysis (data not shown). Category of overweight (normal, overweight, obese) did not predict total fractures in unadjusted analysis (Table 2) or adjusted analysis (data not shown), but there was a trend for upper limb fracture (adjusted HR, 0.53/category; p = 0.08). A total of 18 children had fractures at baseline. However, fracture at baseline did not predict further fracture in this sample (HR, 0.50; p = 0.34)

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

This is the first prospective study to report that a single bone densitometry scan at age 8 can predict fracture risk in boys and girls (especially those involving the upper limb) during puberty. This ability for long-term fracture prediction (an average of 4.3 yr later) is important, especially given the magnitude is similar to short-term studies. This suggests that BMD tracks throughout adolescence, as has been documented in short-term studies this far,(20) and that bone strength may remain relatively constant through this period despite the rapid change in bone size.

No single bone densitometry measure has consistently proven to be the best predictor of fracture risk in children. Total body BMC after adjustment for height, weight, age, and sex had the strongest association with fracture risk, which is consistent with the previously published study of Clark et al.,(8) who reported an 89% increase in fracture risk for each SD decrease in adjusted BMC, but contrasts with the study by Goulding et al.,(10) who found BMD and BMAD but not BMC related inversely to fracture risk, and our previous case control study where BMC was not associated with fracture risk.(11) Both this study and the study by Clark et al. included children of the same age, whereas the latter two included a wide age range of children; thus, this discrepancy may be explained by the effect of a greater size variation on BMC leading to a weakening of fracture prediction. Whereas there was size adjustment in these studies, the wide size variation still leaves open the possibility of residual confounding when the correlation between size and BMD is greater than the correlation between fracture and BMD. In this study, the BMC measures performed better after size adjustment with weight accounting for the majority of this change. Given this, it may be that total body BMD is the best practical measure to assess fracture risk, because it requires no adjustment or manipulation.

We found weaker, yet significant, associations were present for spinal sites, with a 2-fold increase in upper limb fracture risk for each SD decrease in adjusted BMC. Goulding et al.(10) found a similar association for spinal BMAD and found low spinal BMAD predicts all fracture in children, but not forearm fracture alone. This is in contrast with our previous case control study, where spine sites were consistently, but not significantly, stronger than total body sites for predicting wrist and forearm fracture.(12) No associations were seen for hip measures in this study. Our previous study was the only study to report an association between hip sites and upper limb fracture in children, which is in contrast to adult studies that have reported that the hip is the best predictor of hip and total fractures.(21) The reasons behind this discrepancy are unclear, although most likely reflect the different fracture distribution in adults and children, different fracture etiology, the measurement of hip BMD may be less reproducible in children compared with adults, and possibly sample size considerations.

A number of studies have implicated body weight or fat as a predictor of fracture in children,(10,13) although some have suggested it may be television, computer, and video watching rather than body fat per se.(14) This study suggests that body fat may be a predictor of upper limb fracture only, but this result did not reach statistical significance and requires confirmation in larger studies. The hazard ratio was 1.53 per increasing category; thus, an overweight child had 1.5 times more risk and an obese child had 2.3 times more risk in comparison with a normal weight child. Relevant to this, our sample was selected because they were at higher risk of sudden death syndrome. As a result, they were of lower mean birth weight than normal children as previously described.(16) It is difficult to compare BMI distribution in different samples because of the rapid secular changes that are occurring in overweight in children. In randomly selected fracture free controls studied from 1998 to 2002 from our location, the prevalence of overweight/obesity was 29% overall but varied greatly by which fracture they were matched to (32% for hand fracture, 25% for forearm fracture, and 38% for upper limb fracture).(14) This predominantly reflects the different age structure of the controls and the increase in true body mass index with increasing age. In this study, the prevalence of overweight/obesity according to the Cole criteria in fracture-free subjects was 21% at age 8 in 1996 but increased to 29% at age 16 in 2004, consistent with this. Overall, this suggests that the BMI distributions are similar but may be slightly lower in this sample. This is somewhat surprising because low birth weight is a predictor of obesity in later life(22) but gives this study marginally less power to examine body fat as a predictor of fracture. Previous fracture was not a predictor of subsequent fracture in this sample. This most likely reflects sample size limitations, with only 18 children having had a previous fracture.

This study has a number of potential limitations. Although we had sufficient power for upper limb and total fractures, we could not examine the other fracture subtypes, which would require much larger studies. Second, a sizeable percentage were lost to follow-up, raising the potential for bias, but reassuringly, there were no statistically significant differences between the two groups, suggesting this is unlikely. Because we measured the children at age 8 and age 16, we do not have valid data on peak height velocity. However, fracture incidence in our location does not provide strong support for this hypothesis, with wide variations in peak fracture incidence that, with few exceptions, do not coincide with the timing of peak height velocity in Australian children. Indeed, peak fracture incidence of forearm fractures in both sexes appears likely to be considerably before peak growth velocity.(2) Furthermore, height velocity was not a predictor of upper limb fractures in our large fracture case control study (G Jones, unpublished data).

In conclusion, measurement of bone mass at age 8 by DXA is a good predictor of upper limb fracture risk during puberty. Although we did not measure true BMD, the constancy of fracture prediction after a single measure suggests bone strength remains relatively constant during puberty despite the large changes in bone size.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

This work was supported by the National Health and Medical Research Council of Australia (1996 and 2004/5), Blundstone Pty Ltd, Royal Hobart Hospital Acute Care Program, Lions Club of Australia, Coca-Cola Amatil, Tasmanian Dairy Authority, and Talays (in 1996). Data collection in 1988 was supported by the Tasmanian Government. Special thanks also to Carole Goff (Research Coordinator, 1996), Jenny Cochrane (Data Manager), Denise Kaye, and the staff of the Medical Imaging Department at Royal Hobart Hospital, Jack Allan and Pip Boon (Research Coordinator, 2004/5).

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES