These data demonstrate that accumulation of visceral fat over 5 years is independently associated with greater risk of incident type 2 diabetes in Japanese Americans. The association of central adiposity with the presence of metabolic disorders is well established, but this longitudinal analysis is the first, to our knowledge, to demonstrate the role of accumulation of abdominal fat over time in the development of diabetes. These findings thus provide a novel argument in support of the hypothesis that visceral fat may be causally linked to diabetes risk.
The results of our analysis help provide additional information about the association between an increase in body fat, and in particular visceral fat, and near-term diabetes risk. We did not find an association between IAF change over 5 years and cumulative diabetes incidence over 10 years on univariate analysis, but a significant association emerged for change in IAF after adjustment for age and other covariates, including other CT measures of adipose depots as well as overall adiposity assessed with BMI. Although the associations between greater adiposity and specifically visceral adiposity and diabetes risk are well established, there is the potential that change in visceral fat may not appear to be associated with diabetes risk because of the strong association between increase in this depot and younger age, as demonstrated previously by our group (18
) and in the Insulin Resistance Atherosclerosis Study (19
The association between baseline IAF and diabetes odds diminished with greater age and in the presence of a positive family history of diabetes. The stronger association between IAF and diabetes risk at younger ages is consistent with previously published findings (20
). It may be explained by genetic factors related to β-cell dysfunction or by the association between age and declining β-cell function, in which case the effect of IAF might be expected to be reduced. We are unaware of a possible biological explanation for the interaction between family history and IAF. No interaction was observed by sex, or between change in IAF and either sex, age, or diabetes family history.
Historically, several hypotheses have been advanced to explain the relationship between visceral fat accumulation and diabetes risk, including deposition of free fatty acids mobilized from visceral fat into the liver’s portal circulation and the presence of adipocyte-associated inflammatory markers (22
). Levels of cytokines and visceral fat have been linked to type 2 diabetes and atherosclerosis in cross-sectional analysis (23
). Interestingly, surgical removal of both visceral (24
) and subcutaneous (25
) fat depots has not been associated with improvement in insulin sensitivity, and some recent evidence suggests that the negative metabolic changes associated with fat in the abdomen may be associated specifically with intrahepatic triglyceride content (26
Our study has some potential limitations. Although we did adjust for known covariates, the potential for confounding by unmeasured factors exists given the observational design. Diabetes status was determined with a single measurement at baseline and again at 5- and 10-year follow-up. Variability in the performance of these tests over time may have introduced error, although the error is probably random and any bias created would have been toward the null value, resulting in reporting of underestimates of associations. Although the definition of diabetes status by a single glucose measurement is not adequate for clinical diagnosis of diabetes, the World Health Organization has consistently supported using a single measurement of hyperglycemia to classify diabetes status for epidemiologic purposes (28
). Although HbA1c
levels were available for this population, they were not used for classification of diabetes because of the significant overlap in A1C distribution between normal subjects and individuals classified as having diabetes based on the oral glucose tolerance test and fasting plasma glucose among Japanese (29
). The reason for this phenomenon is unknown, but may be due to higher rates of isolated elevated postprandial glucose levels in Asian subjects (30
Because we did not measure islet cell antibody status of participants, we cannot exclude the possibility that some individuals may have had type 1 diabetes. However, the incidence of type 1 diabetes in individuals of Japanese ancestry is among the very lowest in the world, at less than 2.5 per 100,000 in children and probably just slightly higher in older individuals (31
). The rarity of the outcome and the fact that the youngest age at enrollment in our study was 34 years makes it very likely that all subjects with incident diabetes in this analysis had type 2 diabetes. Body composition measurements were derived from a single CT-scan slice. However, a high correlation has been demonstrated between a single CT slice and direct measurement of visceral fat volume, which limits the potential for bias (32
The 20% loss to follow-up could have introduced bias if such loss were related both to diabetes odds and to change in IAF. Nevertheless, this rate is low for a prospective study of this type conducted over a prolonged period of time. Although the exclusively Japanese-American cohort limits the ability to generalize, it also decreases the chance of confounding because of genetic admixture. Finally, this analysis could not exclude all incident cases of diabetes that developed in the period before repeat IAF measurement at 5 years. It is possible that the development of diabetes preceded change in IAF in some of these subjects.
This observational study does not provide direct implications for diabetes prevention, but it does suggest opportunities for interventional studies. In the Diabetes Prevention Program, loss of visceral adiposity was associated with lower risk of diabetes in post hoc analyses (8
). In middle-aged and older populations, weight loss has been shown to come from central adipose depots early (33
). Because of these phenomena, future research targeting and quantifying central weight loss may help provide important information pertinent to diabetes prevention efforts. In summary, this analysis adds support for a causal role of IAF in diabetes risk, although additional research will be needed to confirm its value and exclude a chance association.