Elevated C-Reactive Protein Is a Risk Factor for the Development of Type 2 Diabetes in Japanese Americans

Study subjects were Japanese Americans enrolled in medical surveys conducted from 1988 to 2000 in Hawaii and Los Angeles, California. Subjects were free of infectious symptoms, autoimmune disease, or any other acute conditions as assessed by medical interview. The surveys have been previously described in detail elsewhere (11–13). The study population consisted of 396 men and 551 women who did not have diabetes, as ascertained by a 75-g glucose tolerance test (GTT) at baseline. Among all repeating participants, 43.6% were examined twice, 24.2% three times, 17.3% four times, 14.8% five times, and 0.1% six times. Family history (FH) of type 2 diabetes, use of hormone replacement therapy (HRT) for women, and smoking status were assessed using standard interviewing procedures. The relationship between CRP and the development of diabetes during the follow-up period was analyzed. During each follow-up examination, participants who were free of diabetes at baseline had a 75-g GTT performed after an overnight fast. Incident diabetes was diagnosed as fasting glucose of ≥7 mmol/l or 2-h glucose of ≥11.1 mmol/l (14). Blood was centrifuged, and the obtained serum was immediately frozen at −80°C. Serum samples were subsequently brought back to Japan. Insulin was measured by double-antibody radioimmunoassay. CRP was measured using latex-enhanced immunonephelometric assays (15) on a BNII analyzer (Dade Behring, Tokyo, Japan). Insulin resistance was evaluated with homeostasis model assessment (HOMA) (16). This study was approved by the ethics committee of the Hiroshima University and the Council of the Hiroshima Kenjin-Kai Association in Hawaii and Los Angeles.

In this study, data on men and women aged 20–93 years were analyzed independently. Results were expressed as the means ± SEM. Because triglycerides, HOMA, and BMI did not show normal distributions, the data were analyzed after logarithmic transformation. Baseline CRP concentration was divided at the 50th percentile of the sex-specific distributions (>0.68 mg/l for men and >0.76 mg/l for women), and relationships between CRP and the development of diabetes were evaluated using Kaplan-Meier plots and log-rank tests. To test the significance of CRP as a predictor of incidence of type 2 diabetes, CRP concentration was divided into quartiles based on its population (among men, <0.39, 0.39–0.67, 0.68–1.43, and ≥1.44 mg/l; among women, <0.37, 0.37–0.75, 0.76–1.66, and ≥1.67 mg/l); quartile specific hazard ratios were estimated through a Cox proportional hazards model. The numbers of the subjects were 99, 98, 99, and 100, respectively, among men; the numbers were 138, 138, 137, and 138, respectively, among women. The development of diabetes was designated the dependent variable (1, developed diabetes; 0, did not develop diabetes), whereas age, smoking status, FH (1, positive; 0, negative), classification of a 75-g GTT (1, impaired glucose tolerance; 0, normal glucose tolerance), HRT for women, BMI, and HOMA were used as independent variables in addition to CRP quartiles. For proportional hazards, assumption was checked with a graphical plot of logS_t versus log_t for the variables included in the model (S denoted the survival function, and t denoted time) and found to be acceptable. For all analyses, SAS package version 8.2 (SAS, Cary, NC) was used.

In this study, the distribution of serum CRP was skewed to low levels. A total of 57 of 396 men and 65 of 551 women developed diabetes during the follow-up period (6.54 ± 0.10 years). We subsequently investigated the rate of development of the disease according to dichotomized CRP levels. Among men with a high CRP level, 36 of 199 were diagnosed with type 2 diabetes. Among men with a low CRP level, 21 of 197 were diagnosed with the disease. Among women with a high CRP level, 43 of 275 were diagnosed with type 2 diabetes. Among women with a low CRP level, 22 of 276 were diagnosed with the disease. We observed significant differences in rates of incidence between subjects with a high versus low CRP level in both sexes because Kaplan-Meier curves and log-rank tests (P = 0.028 for men; P = 0.004 for women) were statistically significant.

In a Cox proportional hazards model including quartiles of CRP, the hazard ratios for developing diabetes by increasing quartiles of CRP were 1.0, 1.99 (95% CI 0.73–5.41), 2.24 (0.83–6.03), and 2.84 (1.09–7.39) among men, respectively (P = 0.035 for trend), and the hazard ratios for increasing quartiles of CRP were 1.0, 2.22 (0.86–5.74), 2.55 (1.00–6.49), and 3.11 (1.25–7.75) among women, respectively (P = 0.036 for trend) adjusted for age, smoking status, FH, classification of a 75-g GTT, HRT for women, BMI, and HOMA (Fig. 1). The adjusted hazard ratio per unit of CRP was 1.00 (1.00–1.001) among all participants, with the same ratio observed among both men and women. The adjusted hazard ratios per unit log (CRP) were 1.22 (1.03–1.45) among all participants, 1.14 (0.89–1.45) among men, and 1.40 (1.09–1.80) among women.

Although the influence of CRP on the development of type 2 diabetes in Japanese populations has not been demonstrated, this study did show that a relationship exists between CRP and the development of type 2 diabetes in Japanese Americans, as well as other races (4,6–8), which is a population thought to be genetically identical to native Japanese populations.

CRP levels followed in this study have been shown to be the same (2,17) or lower (1,4,7,9,18) in Japanese-Americans compared with Caucasians. However, they were elevated compared with those of native Japanese populations (10,19). From an early age, Japanese Americans have lived in a highly westernized environment, with features such as a high-fat diet, for a longer period of time and more intensely when compared with native Japanese (12), resulting in greater insulin resistance. Elevated proinflammatory mediators, such as tumor necrosis factor-α, subsequent to exposure to a high-fat diet (20) might result in a low-grade inflammatory state (21). This might be one explanation of the relationship between CRP and the westernized lifestyle among Japanese Americans.

CRP is thought to be produced in the liver (2) and stimulated by tumor necrosis factor-α, which is itself derived from adipose tissue (22) and predominantly induced by interleukin-6 (23). Accordingly, the possibility exists that CRP is an indirect indicator of insulin resistance via obesity. In this study, the high CRP group manifested the characteristic so-called metabolic syndrome, which is namely a tendency toward elevated blood glucose, insulin, BMI, and triglycerides and decreased HDL cholesterol levels (data not shown). However, in both sexes, the highest quartile group was at significantly higher risk for developing diabetes compared with the lowest quartile group after adjustment for BMI and HOMA. Additionally, the hazard proportion trend was observed as quartiles increased (Fig. 1). Therefore, the low-grade inflammatory state illustrated by CRP increase was considered a risk factor independent of obesity or insulin resistance, which might suggest an unknown mediator involved in the development of diabetes.

The mechanisms of correlation between CRP and the development of diabetes are unclear. However, there are several possible explanations. One of them is the oxidative stress generated hyperglycemia (24). Oxidative stress might be implicated in promoting a state of low-grade inflammation indicated by markers such as CRP with elderly type 2 diabetes (25). On the other hand, oxidative stress was thought to impair insulin endocytosis in endotherial cell lines (26), and it could precede the development of the endotherial dysfunction and insulin resistance (27). This points to the possibility that CRP may be involved in the development of diabetes.

We also have to consider the genetic factors of CRP. Recently, among nonsmoking healthy adult women, FH is thought to be associated with increased plasma CRP levels (28); although in this study, there was no relationship between FH and CRP in either sex. It is hypothesized that there may be a genetic relationship between CRP and diabetes, even though it is unclear whether a common genetic factor really exists. To understand the relationship between CRP and type 2 diabetes, further study is needed.

Several limitations of this study remained. First, anti-inflammatory medication might not have been considered completely in this study, although subjects were free of inflammatory diseases. Second, the definitions of inflammatory diseases tend to be transitional. Accordingly, whether the subjects were excluded by exactly the same criteria is uncertain.

CRP levels in nondiabetic Japanese Americans were slightly lower compared with Caucasians but higher than those in native Japanese populations. This suggests that westernization of lifestyle may cause a high CRP level. This report demonstrated that among Japanese Americans, CRP is considered a risk factor for diabetes independent of obesity or insulin resistance. It is suggested that the mechanism of diabetes development may be related to inflammation.

The authors thank the members of the Hiroshima Kenjin-Kai of Hawaii and Southern California for their participation. We are indebted to the late Dr. S. Takashina, former President of Hiroshima General Hospital and J. Suita, Dade Behring, for the assay of samples.