In this prospective analysis of a population-based sample of older adults, we found that measures of overall and central adiposity were strongly associated with the risk of incident diabetes in both men and women. Using longitudinal measures of weight from midlife, at study entry, and over follow-up, we were able to demonstrate that weight gain during midlife (after 50 years of age) and in late life (after 65 years of age) is an important risk factor for diabetes among older adults. Although the risk associated with adiposity appeared to wane with age, individuals in the highest category of BMI remained at twice the risk of diabetes compared with those in the lowest category among participants 75 years of age and older.
In the current analysis, simple anthropometric measures such as BMI, body weight, and waist circumference were as strongly associated with the risk of diabetes as were fat mass estimates derived from bioelectrical impedance measures. The 2 composite measures, waist-hip ratio and waist-height ratio, had RR estimates for diabetes that were similar to those of waist circumference alone. While in certain populations, waist circumference6,20,21
or waist-to-hip ratio22
may offer better predictive power for diabetes risk than BMI, our findings are generally consistent with the findings of a meta-analysis of 32 population-based studies that found that RRs for diabetes were equivalent for standardized differences in BMI, waist circumference, and waist-to-hip ratio.23
Age modified the risk of diabetes associated with adiposity in this analysis. For each of the various adiposity measures evaluated at baseline, RR estimates associated with higher adiposity were appreciably lower in individuals 75 years of age and older compared with those 65 to 74 years of age. The presence of effect modification by age has been noted for the relationship between body composition measures and outcomes such as mortality24-26
and coronary heart disease risk,27
but we are unaware of previous studies reporting an age interaction of the association of adiposity with diabetes risk.
There are a number of potential explanations for a weaker association of body composition measures with diabetes risk among individuals 75 years of age and older. Among older adults, standard anthropometric measures may not adequately quantify body fat due to age-related changes in body composition, including decreases in skeletal muscle mass and height.28
However, the RRs associated with fat mass estimates from bioelectrical impedance measures showed reductions in magnitude similar to those of anthropometric measures among those 75 years of age and older. A second possibility is that regional fat distribution is more important in the etiology of diabetes than absolute fat mass. Visceral fat and intermuscular thigh fat are associated with impaired glucose tolerance and diabetes in older adults independent of total adiposity.29,30
Waist circumference is highly correlated with visceral fat in older adults in some31,32
but not all33
studies, and the strength of association may vary according to age, sex, race, overall adiposity, and cardiorespiratory fitness.34,35
This raises some uncertainty about whether waist circumference is a reliable surrogate for direct measurement of visceral fat. Intermuscular thigh fat can be measured reliably only through an imaging modality such as computed tomography. To the extent that these fat depots become more important to diabetes etiology as individuals age, the inability of the adiposity measures included in our study to measure such depots could explain the observed effect modification by age.
Another explanation for the effect modification by age is that the pathophysiology of diabetes in older adults differs from that of young and middle-aged adults. If, for instance, defects in insulin secretion played a larger role than insulin resistance in the development of diabetes in older individuals, one might expect there to be less of an association with adiposity. While possible, we know of no data that support a different pathophysiology of diabetes in older adults.
In addition, the observed age-adiposity interaction with diabetes risk may result from selective survival among older adults.36
Individuals who are more susceptible to adverse health outcomes associated with adiposity may be less likely to survive into old age. Such an effect would be consistent with the uniform attenuation of all adiposity measures among the older age group observed in the current analysis.
We did not observe a reduction in diabetes risk associated with measured weight loss over a 3-year period. This contrasts with the results of studies in younger populations that found weight loss to be associated with a decreased risk of diabetes.37,38
Older adults may lose proportionately more muscle mass with weight loss than younger ones,39
decreasing the accuracy of weight loss as a surrogate for loss of adipose tissue in older adults. Furthermore, the loss of skeletal muscle mass may decrease insulin sensitivity,40
negating the benefit derived from fat loss. Alternatively, weight loss associated with insulin resistance41
that preceded the onset of clinical diabetes may have obscured an association between weight loss and decreased diabetes risk. Because of these complexities, our results do not preclude the possibility that voluntary weight loss reduces the risk of diabetes in older adults.
Our analysis showed an association between waist circumference and diabetes risk in individuals with a BMI of less than 25, suggesting that measurement of waist circumference may add important information beyond BMI regarding diabetes risk in normal-weight individuals. The observation that waist circumference was less strongly associated with diabetes risk at higher BMIs may reflect the fact that waist circumference is a better measure of visceral fat at a low BMI.35
This study has several strengths. We used data from a well-characterized population-based cohort with long-term follow-up. Aside from self-reported weight at 50 years of age, all our anthropometry was based on direct measurement rather than self-report. We were able to examine both incident diabetes and body composition changes prospectively, analyzed both men and women, and had extensive covariate data. We examined multiple measures of adiposity—particularly valuable in light of heterogeneity among previous studies of adiposity, the marked changes in body composition that occur with aging, and the paucity of studies of this type in older adults.
Potential limitations should also be noted. The measurement of fasting glucose at limited time points may have resulted in misclassification of participants with untreated diabetes. Such misclassification would likely be non-differential and result in attenuation of risk estimates. Despite the wealth of covariate data that allowed us to adjust for well-known confounders, residual confounding due to unknown factors related to both adiposity and diabetes may be present.
Results of this study affirm the importance of maintaining optimal weight during middle age for prevention of diabetes and, while requiring confirmation, suggest that weight control remains important in reducing diabetes risk among adults 65 years of age and older.