| |
Low body weight mediates the relationship between HIV infection and low bone mineral density: a meta-analysis
|
| |
| |
weight loss bones hiv.pdf
Journal of Clinical Endocrinology & Metabolism, Epub Oct 9, 2007 doi:10.1210/jc.2007-1660
Mark J. Bolland MBChB*, Andrew B. Grey MD, Greg D. Gamble MSc, and Ian R. Reid MD
Department of Medicine, University of Auckland, Auckland, New Zealand
"In summary, we found that HIV-infected patients in studies that report BMD data are lighter than controls. Most studies reported lower BMD in HIV-infected patients than controls, but after adjusting for the body weight differences between the groups, we found only small residual differences in BMD that were unlikely to be of clinical importance. We conclude that low body weight in HIV-infected patients can largely account for the reported high occurrence of low BMD in HIV-infected patients, and that HIV infection in the absence of low body weight is not a risk factor for low BMD. The stability of BMD in longitudinal studies of HIV-infected cohorts lends support to this conclusion. Therefore, routine assessment of skeletal status is not justified in HAART-treated HIV-infected patients, and decisions around the need to investigate and treat low bone mass in such patients should be made using guidelines available for the general population (54, 55).
differences in BMD of this magnitude (approximately 0.25 standard deviations) [found in this study] would be predicted to confer a 1.1 to 1.2 fold increased risk of osteoporotic fracture, which is considerably smaller than other recognised risk factors such as smoking, low body weight, family history, previous fracture or glucocorticoid treatment (all > 1.5 fold increased risk) (47). Consistent with these results, Arnsten et al. reported that HIV infection was not an independent risk factor for incident fractures in men
The cause of the weight differences between the HIV-infected groups and the control groups is likely to be multifactorial. It may be in part due to the loss of body weight that occurs with chronic illness and advancing HIV infection, in part due to the higher prevalence of smoking12 in the HIV-infected groups, since smoking exposure is inversely related to body weight (48), and in part due to the predominance of homosexual men in the HIV-infected groups. Homosexual men have been reported to have lower body weight (49), be more concerned with body shape and appearance (50), and have higher rates of smoking (51) than heterosexual men."
Abstract:
Context: HIV infection has been associated with low bone mineral density (BMD) in many cross-sectional studies, although longitudinal studies have not demonstrated accelerated bone loss. The cross-sectional studies may have been confounded by the failure to control for low body weight in HIV-infected patients.
Objective: To determine whether low body weight might explain the association of HIV infection with low BMD.
Data sources: MEDLINE and EMBASE were searched for English language studies published from 1966 to March 2007, and conference abstracts prior to 2007 were hand-searched.
Study selection: All studies reporting BMD and weight or body mass index in adult patients with HIV and a healthy age- and sex-comparable control group were included. 9/40 identified studies and 1/68 identified abstracts were eligible.
Data synthesis: We adjusted for the between-groups weight differences using regression coefficients from published cohorts of healthy men and women.
On average, HIV-infected patients were 5.1 kg (95%CI -6.8,-3.4; P<0.001) lighter than controls.
At all skeletal sites, unadjusted BMD was lower, by 4.4-7.0%, in the HIV-infected groups than the controls (P<0.01).
After adjustment for body weight, residual between-groups differences in BMD
were small (2.2-4.7%) [lumbar spine: -0.02 (95%CI -0.05,0.01) g/cm2, P=0.12; total hip: -0.02 (-0.04,0.00) g/cm2, P=0.031; femoral neck: -0.04 (-0.07,-0.01) g/cm2, P= 0.013; totalbody: -0.03 (-0.07,0.01) g/cm2, P=0.11).
Conclusion: HIV-infected patients are lighter than controls and low body weight may largely account for the high prevalence of low BMD reported in HIV-infected patients. However, in the setting of current treatment practice, HIV infection per se is not a risk factor for low BMD.
Discussion
We found that HIV-infected patients are on average 5.1 kg lighter than controls, and that unadjusted BMD was 4.4% to 7.0% lower in HIV-infected patients than controls. However, after taking into account this weight difference, the between-groups differences in BMD ranged from 2.2% to 2.8% at the lumbar spine, total hip and total body. At the femoral neck the differences were greater, in part due to an outlying study, and after we excluded this study or reduced its weighting in the analyses, the between-groups differences were also 2-3%.
These between-groups differences in BMD were statistically significant at the femoral neck, but of borderline statistical significance at the total hip, and not significant at the lumbar spine or total body. We and others have previously reported that changes in body weight are associated with greater changes in BMD at the spine, radius and total body than at the femoral neck (45, 46). This may in part explain why, when we adjusted for the weight differences between the groups, the greatest persisting differences in BMD were at the femoral neck.
However, any persisting differences at all sites were small, and given the findings of stable or increasing BMD in longitudinal studies of BMD in HAART-treated HIV-infected patients (6-16), they are unlikely to be of clinical importance. For example, differences in BMD of this magnitude (approximately 0.25 standard deviations) would be predicted to confer a 1.1 to 1.2 fold increased risk of osteoporotic fracture, which is considerably smaller than other recognised risk factors such as smoking, low body weight, family history, previous fracture or glucocorticoid treatment (all > 1.5 fold increased risk) (47). Consistent with these results, Arnsten et al. reported that HIV infection was not an independent risk factor for incident fractures in men (42) and Prior et al. reported that HIV infection was not an independent risk factor for lifetime fragility fractures in women, after adjusting for possible confounders (44). 11
The failure to adjust for body weight differences between groups in many of the published studies may also explain a discrepancy between the findings of cross-sectional and longitudinal studies of BMD in HIV-infected patients taking HAART. At least 30 cross-sectional studies have reported a higher than expected rate of low BMD in HIV-infected subjects (1, 2), whereas 11 longitudinal studies have reported stable or increasing BMD over time (6-16). Weight loss occurs with advancing untreated HIV-infection, but there is steady regain of this lost weight following the commencement of HAART. Weight is a major determinant of BMD and changes in weight are positively correlated with changes in BMD in
various populations (5, 45, 46). Therefore, it would be predicted that BMD would decrease with advancing untreated HIV-infection and increase following the commencement of HAART, although there may be time lags between initiation of HAART, regain of body weight and regain of BMD (16). If this was the case, cross-sectional studies performed shortly after the commencement of HAART would report higher than expected rates of low BMD because of low body weight.
However, because of HAART-related weight gain, longitudinal studies would report stable or increasing BMD, and cross-sectional studies performed after a
longer time interval following the commencement of HAART would not report increased rates of low BMD. Consistent with this hypothesis, we found that the year of publication of the individual studies was positively correlated with the duration of HAART, and that the greatest between-groups differences in BMD occurred in the earliest published studies. In addition, there is only one published study with a HIV-infected group and a comparable weight-matched control group (30) and this reported no between-groups difference in BMD.
The cause of the weight differences between the HIV-infected groups and the control groups is likely to be multifactorial. It may be in part due to the loss of body weight that occurs with chronic illness and advancing HIV infection, in part due to the higher prevalence of smoking12 in the HIV-infected groups, since smoking exposure is inversely related to body weight (48), and in part due to the predominance of homosexual men in the HIV-infected groups. Homosexual men have been reported to have lower body weight (49), be more concerned with body shape and appearance (50), and have higher rates of smoking (51) than heterosexual men.
The regression coefficient used to adjust BMD data for the weight differences between the groups determines the magnitude of the weight-adjustment. We used coefficients generated by two different methods, but obtained similar results using either. First, we used gender- and densitometer manufacturer-specific coefficients derived from regression of individual data from published cohorts of healthy men and women. Second, we used the mean data for BMD and weight from the studies in the meta-analysis to calculate coefficients specific to these
data. The coefficients obtained in this way were similar to the published coefficients for the femoral neck but higher elsewhere, although the 95% confidence intervals for both sets of coefficients overlapped. In our previous study of HIV-infected men, the regression coefficients were also in the similar range (lumbar spine 0.005 g/cm2/kg, total hip and femoral neck 0.004 g/cm2/kg, and total body 0.003 g/cm2/kg) (4), and other studies of HIV-infected
participants have reported similar values (40-42, 52).
There are several limitations to our analyses. There were only 10 studies that met the criteria for inclusion in our review. Although we followed established methods for identification of eligible studies, it is possible that we overlooked some. Publication bias may have also influenced our findings. If such a bias was present, it is likely to favour publication of studies reporting significant reductions in BMD in HIV-infected populations, and therefore to accentuate rather than diminish the between-groups differences in the meta-analysis. Because individual-level data for weight and BMD were not available from the studies in our metaanalysis, we used summary data from other healthy cohorts and from the studies in our meta-analysis to adjust for body weight. If the individual-level data were available, we would expect to obtain similar results although the exact magnitude of the weight-adjustment may be different from those obtained using the summary data (53).
In summary, we found that HIV-infected patients in studies that report BMD data are lighter than controls. Most studies reported lower BMD in HIV-infected patients than controls, but after adjusting for the body weight differences between the groups, we found only small residual differences in BMD that were unlikely to be of clinical importance. We conclude that low body weight in HIV-infected patients can largely account for the reported high occurrence of low BMD in HIV-infected patients, and that HIV infection in the absence of low body weight is not a risk factor for low BMD. The stability of BMD in longitudinal studies of HIV-infected cohorts lends support to this conclusion. Therefore, routine assessment of skeletal status is not justified in HAART-treated HIV-infected patients, and decisions around the need to investigate and treat low bone mass in such patients should be made using guidelines available for the general population (54, 55).
|
|
| |
| |
|
|
|
