Our main finding was that associations of anthropometric measures with metabolic risk indicators were similar to those for full regional adipose tissue volumes from MRI in both HIV-infected and control populations. In particular, WC appeared to be the best anthropometric measure of visceral obesity for HOMA and HDL, whereas, for triglycerides, the best anthropometric measure was WHR. The %BF was consistently weaker than WC, WHR, and most other predictors. After multivariate adjustment for demographic and lifestyle factors, there was no significant difference in adjusted R2 between best-fitting models using anthropometric or MRI measures. Further adjustment for HIV-related factors and use of antiretrovirals did not change this finding. Thus, WC and WHR remain reasonable surrogates, despite the fact that HIV-infected persons have less leg and LT fat, without a compensatory increase in VAT.
The addition of MRI measures to the anthropometric measures significantly improved model fit in 3 analyses in HIV-infected subjects (ie, HOMA and triglycerides in men and trigycerides in women) but in none of the analyses in controls. The improvement in adjusted R2, however, was ≤ 0.05, even when it was significant. Likewise, the addition of anthropometrics to the MRI model offered little improvement in adjusted R2.
Our finding that WC and WHR are strongly associated with metabolic risk indicators is supported by previous work in smaller studies of HIV-infected subjects that used more-limited body-composition measurements. Meininger et al (19
) found WHR to be a stronger predictor of hyperinsulinemia than were other anthropometric measures (eg, WC, HC, and BMI), dual-energy X-ray absorptiometry variables (extremity and trunk fat), and computerized tomography–measured VAT and abdominal SAT in 41 HIV-infected men. Similarly, Dolan et al (20
) found that WHR was a stronger predictor of several metabolic syndrome components (ie, insulin, glucose, triglycerides, and HDL) than were other anthropometric, dual-energy X-ray absorptiometry, and computerized tomography measures in 100 HIV-infected women. Shen et al (3
) reported that %BF did not correlate as well as did WC with health risk indicators in a large cohort of HIV-uninfected persons. Our results also show that %BF is less predictive of metabolic complications than are other anthropometric and MRI measures in both HIV-infected and control subjects. Our MRI measures included depots not traditionally quantified, such as UT and LT, which did enter some models.
Our finding that WC appears to be correlated with health risk indicators at least as strongly as is VAT was previously observed in HIV-uninfected subjects (3
). One explanation for this finding is that even MRI-measured VAT has a fair amount of measurement error (21
), which exceeds that for WC measurements done by trained personnel, and thus the strong associations with WC may reflect greater precision of measurement. It is also possible that, if abdominal size by the supine sagittal diameter or an alternative site for WC were used, this measure may have correlated even more strongly with metabolic outcomes. However, longitudinal studies of weight-loss intervention (22
) and diabetes prevention (23
) have shown a greater percentage reduction in VAT than in WC. The relation of VAT with health risk indicators has been studied (24
), but further work is needed to understand the mechanisms involved, particularly in longitudinal intervention studies.
Correlations were generally stronger in control subjects than in HIV-infected subjects, particularly for HOMA, whose association was significantly higher in control subjects than in the HIV-infected group for nearly every measure of body composition. The MRI + anthropometric combination models were more often stronger than WC, WHR, or VAT in HIV-infected subjects than in control subjects. This difference between HIV-infected and control subject may be due to abnormalities in fat distribution found in HIV infection. The loss of abdominal subcutaneous fat in HIV infection (7
) may decrease the utility of WC as a predictor of the deleterious effects of obesity. Likewise, the presence of severe subcutaneous lipoatrophy may affect the contribution of those depots, especially leg SAT. Of note, 3 of the 4 best MRI and combined models included leg SAT, the depot most affected in HIV infection (7
), which was negatively associated with HOMA and triglycerides in multivariate analysis; negative associations of leg SAT with HOMA and triglycerides were reported previously (9
). Thus, the normal relation between adipose tissue and metabolic health risk indicators may be somewhat weakened or altered in HIV-infected subjects, because of the presence of lipoatrophy. Other reports have found that thigh circumference is associated with improved metabolic indicators (29
), and thus it is possible that the use of a measure of midthigh circumference as a surrogate for leg SAT could improve the predictive ability of the anthropometric measures. However, given the presence of HIV-associated peripheral lipoatrophy (7
), it is likely that thigh circumference would be more reflective of lower limb lean mass. Finally, other HIV-related factors that cause insulin resistance may weaken the relation of adipose tissue with HOMA.
The present study has several limitations. The cross-sectional design limits the ability to determine causality of body composition and metabolic risk indicators. There may have been inadequate control for factors that confounded the association of body composition with metabolic outcomes. The amount of VAT in our cohort is somewhat smaller than that in a previously reported study of HIV-uninfected subjects (21
): medians of 1.9 and 0.9 L in control men and women in this study compared with medians of 2.3 and 1.1 L in corresponding subjects. This difference may explain in part the lower correlation between health risks and VAT seen in the present study. We used a single fasting specimen; insulin is known to be secreted in a pulsatile manner, and thus basal concentrations may be variable.
An important question is whether our findings in HIV-infected subjects and controls will be validated with other obesity-related health risks, development of diabetes, and mortality. A 12-wk study of a small number of obese HIV-infected women (31
) found no improvement in insulin sensitivity, despite the loss of VAT and total body weight. A prospective study of the evolution of metabolic risk indicators in large HIV-infected populations may help address causality concerns.
Simple anthropometric measures had associations with health risk indicators that appeared to be about as strong as MRI-based measures in both HIV-infected subjects and control subjects. For HOMA and HDL, WC appeared to be the best anthropometric correlate of metabolic complications, whereas, for triglycerides, the best was WHR. The addition of MRI depots to the anthropometric models produced only small improvements in model fit. As in control subjects, the effect of adiposity on health risks is better captured by central adiposity measures than by %BF in HIV-infected subjects. A critical question emerging from these observations is that of how best to define and screen for obesity in HIV infection, considering energy stores on the one hand and health risks on the other. Our data suggest that, despite the presence of HIV-associated lipoatrophy, the use of WC and HC is a highly effective screening method, which may be particularly useful in resource-limited and clinical settings. Data are also needed in longitudinally monitored HIV-infected populations, including health risks other than metabolic risk indicators.