Lipid, Lipoproteins, C-Reactive Protein, and Hemostatic Factors at Baseline in the Diabetes Prevention Program

Full details of the DPP protocol have been published (13). The current report includes 3,819 participants in all four treatment arms seen at baseline. Individuals were recruited from a variety of sources based on the perceived risk for development of diabetes. Written informed consent was obtained from all participants before screening, consistent with the Helsinki Declaration and the guidelines of each center’s institutional review board. The initial screening step consisted of a fasting glucose measurement and, if eligible, was followed by a 75-g oral glucose tolerance test. Inclusion criteria included a fasting plasma glucose value of 5.3–6.9 mmol/l (≤6.9 mmol/l for American Indians) and a 2-h plasma glucose of 7.8–11.1 mmol/l following the glucose load, aged ≥25 years, and BMI ≥24 kg/m² (≥22 kg/m² for Asian Americans because of differences in body size in this population). Major exclusions included a recent myocardial infarction, symptoms of coronary heart disease, major illness, prior diagnosis of diabetes, or use of medications known to impair glucose tolerance, as previously detailed (14). Participants with >6.77 mmol/l triglyceride were also excluded. Standardized interviewer-administered questionnaires were used to obtain self-reported data on personal medical history. Self-reported race/ethnicity was classified according to the question employed in the 1990 U.S. Census questionnaire (15). Of 3,819 randomized participants, 197 (5.2%) were receiving lipid-lowering medications and were excluded from this analysis. Overall adiposity was assessed by BMI. Waist circumference was assessed in the standing position midway between the highest point of the iliac crest and the lowest point of the costal margin in the mid-axillary line. Hip circumference was assessed at the level of the femoral greater trochanter. All anthropometric measures reflected the average of two measurements.

All the analytical measurements were performed at the central biochemistry laboratory (Northwest Lipid Research Laboratories, University of Washington, Seattle, WA). Plasma glucose was measured by the glucokinase method. HbA_1c (A1C) was measured by ion-exchange chromatography (Variant; BioRad). Insulin measurements were performed by a polyethylene glycol–accelerated double antibody radioimmunoassay method. The homeostasis model assessment for insulin resistance (HOMA-IR) was calculated as follows (16):

. The ratio of the change in insulin-to-glucose concentrations between 0 and 30 min after oral glucose (incremental insulin-to-glucose ratio [IGR]) was calculated as an index of insulin secretory capacity (17). Measurements of total cholesterol and triglycerides were enzymatically performed using methods standardized to the Centers for Disease Control and Prevention Reference Methods (18). HDL fractions for cholesterol analysis were obtained by the treatment of whole plasma with dextran sulfate Mg²⁺ (19). LDL cholesterol was calculated by the Friedewald equation (20). In participants with triglycerides >4.5 mmol/l, the lipoprotein fractions were separated using preparative ultracentrifugation of plasma by β quantification (21). LDL flotation distribution was determined by a single-density gradient ultracentrifugation procedure, as previously described (22). LDL buoyancy (Rf) is calculated as the LDL peak fraction divided by the total number of fractions collected. This technique uses an Rf cut point to discriminate between LDL phenotypes A (>0.263) and B (≤0.263). CRP and fibrinogen levels in plasma were immunochemically measured using the Behring Nephelometer autoanalyzer. Tissue plasminogen activator (tPA) levels were measured in citrated plasma using an enzyme-linked immunosorbent assay (Asserachrom tPA; Diagnostica Stago), which measures total tPA antigen.

Baseline characteristics were described using means, SD, and 95% confidence limits for quantitative variables and the numbers and corresponding percentages for categorical variables. Comparisons among groups were made using ANOVA for quantitative variables and the χ² test of independence for categorical variables. The nominal P values are listed with no adjustment for multiple comparisons. To assess the association of various factors with each risk factor, multiple regression models were fit in stages. This avoids the bias in the estimates of coefficients and tests of significance introduced by stepwise model building. Variables were entered into the model in blocks to allow identification of mediating interrelationships between covariates in the model as follows: demographics (age, female sex, ethnicity), adiposity (BMI, waist-to-hip ratio [WHR]), insulinemia (fasting insulin, HOMA-IR), and glycemia (fasting glucose, A1C). Transformations of independent (x) and dependent (y) variables were used when appropriate.

The mean age of the population was 50.3 years, two-thirds were female, and the mean BMI was 32.0 kg/m² for men and 35.0 kg/m² for women. Caucasians comprised the major ethnic group (55%), followed by African Americans (20%), Hispanics (16%), American Indians (5%), and Asians (4%). Mean fasting and postchallenge glucose were 5.9 and 9.1 mmol/l, respectively, and there was evident fasting hyperinsulinemia (13).

Table 1 shows lipids and lipoprotein cholesterol by ethnic group and sex. Since men and women had differences in their lipid profile, lipids and lipoprotein cholesterol were separately summarized. Women had a more favorable lipid profile than men, with significantly lower levels of triglyceride (1.70 vs. 1.95 mmol/l), VLDL (0.89 vs. 0.80 mmol/l), and LDL cholesterol (3.20 vs. 3.28 mmol/l) and significantly higher levels of HDL cholesterol (1.24 vs. 1.04 mmol/l) and LDL particle size (Rf of 0.270 vs. 0.257). Overall, triglyceride and VLDL cholesterol concentrations were lower and LDL particle size larger in African Americans compared with other ethnic groups, while HDL cholesterol levels were higher in African Americans for both sexes than in the other ethnic groups. Total cholesterol and LDL cholesterol values were significantly lower in American Indians than in other ethnic groups among women (P < 0.001).

Figure 1 characterizes the adverse CVD profile by sex. Forty-nine and 29% of the men and 41 and 21% of women had triglyceride values ≥1.69 and ≥2.26 mmol/l, respectively. The prevalence of hypertriglyceridemia was lowest among African Americans (14% of men and 6% of women had triglyceride levels ≥2.26 mmol/l). LDL cholesterol concentrations ≥3.36 mmol/l were present in 41% of women and in 45% of men, respectively, varying only slightly between ethnic subgroups except for American-Indian women (23%) and African-American men (57%). Low HDL cholesterol concentrations (<1.03 mmol/l) were present in 52% of men, and, using the higher cutoff of <1.29 mmol/l for women, low HDL cholesterol values were present in 60% of women. Overall, 57% of the DPP cohort had low HDL cholesterol values by these criteria, and 31% (35% of men and 30% of women) had both low HDL cholesterol and triglyceride levels ≥1.69 mmol/l. The prevalence of this dyslipidemic combination did not vary much by ethnic group, except for African-American men (18%) and women (13%). Using an Rf cut point of ≤0.263 to define LDL phenotype B, 41% of men and 25% of women had small, dense LDL, respectively. These frequencies were lower among African-American men (29%) and women (15%) and were higher in American-Indian men (53%).

Since 793 women in the study were receiving postmenopausal estrogen therapy (32%), a comparison was made between women receiving estrogens and those who were not. Estrogen users had higher triglyceride (mean 1.85 [95% CI 1.78–1.92] vs. 1.63 [1.59–1.67]; P < 0.001), total cholesterol (5.45 [5.39–5.52] vs. 5.17 [5.12–5.21]; P < 0.001), VLDL cholesterol (0.89 [0.86–0.93] vs. 0.76 [0.74–0.78]; P < 0.001), and HDL cholesterol (1.35 [1.33–1.38] vs. 1.19 [1.18–1.20]; P < 0.001) levels than nonestrogen users; despite this, nonestrogen users had higher HDL cholesterol levels than men (1.04 [1.02–1.05]; P < 0.001). Values for LDL cholesterol (3.20 [3.14–3.26] vs. 3.20 [3.16–3.24]) and mean LDL particle size (0.268 [0.266–0.271] vs. 0.270 [0.269–0.272]) were similar in estrogen compared with nonestrogen users.

The geometric mean (95% CI) CRP level was 3.50 mg/l (3.38–3.62) and was significantly higher in women than in men (4.64 vs. 1.92 mg/l). Asian Americans had significantly lower CRP values overall (1.86 mg/l) and in women (2.90 mg/l) compared with other ethnic groups. The mean ± SD fibrinogen concentration was 11.3 ± 2.5 μmol/l (383 ± 86 mg/dl) overall and was higher in women (11.7 μmol/l or 396 mg/dl) than in men (10.4 μmol or 354 mg/dl), lowest in Asian Americans (10.5 μmol/l or 358 mg/dl), and highest among African Americans (11.8 μmol/l or 402 mg/dl) and American Indians (11.8 μmol/l or 404 mg/dl). The mean tPA value overall was 11.3 ± 4.2 ng/dl, was significantly higher in men than in women (12.5 vs. 10.8 ng/dl), and differed among ethnic groups. CRP values were significantly higher in estrogen users than female nonusers (mean 5.42 mg/l [95% CI 5.08–5.78] vs. 4.30 [4.10–4.52]; P < 0.0001); whereas tPA levels (10.4 ng/dl [10.1–10.7] vs. 11.0 [10.8–11.2]; P < 0.001) and fibrinogen concentrations (11.5 μmol/l [11.3–11.6] vs. 11.7 [11.6–11.9]; P = 0.05) were significantly lower among those receiving estrogen treatment.

The relative effects of demographic, body fat, insulin resistance, and glycemia measures on lipids, lipoproteins, and inflammatory variables were assessed using multiple regression analysis, and covariates with significance levels of P < 0.001 are shown in Table 2. These four groups of factors explained 4.8% of the variation of LDL cholesterol. For HDL cholesterol, the group of factors explained 25.4% of the variation, with 4.1% of the variation attributable to being a woman and 3.4% to HOMA-IR. Being African American (versus Caucasian), as well as having higher levels of HOMA-IR, were associated with lower triglyceride values. The four factors explained 11.9% of the variation in LDL size, with HOMA-IR (3.1%) being the most significant. Sex and BMI were the most important determinants of the variation in both CRP and fibrinogen levels, with no influence of WHR; A1C had a small impact of 2.2% on fibrinogen values. The total effect of these determinants explained 29.3 and 19.0% of the variation in CRP and fibrinogen concentrations, respectively. Finally, HOMA-IR had the strongest association with tPA (R² = 1.9%), while BMI and sex contributed approximately equally to its variation, with the combination of all four factors explaining 15.2% of the variation.

Since HOMA-IR was a significant determinant of triglyceride, HDL cholesterol, LDL particle size, and tPA, and correlated significantly in univariate analysis with CRP and fibrinogen (data not shown), a comparison between those in the upper half of the HOMA-IR distribution was compared with those in the lower half. In addition, to explore whether β-cell deficiency modulated the effects of HOMA-IR on risk-factor levels, the comparison was extended to include four groups based on median splits of both HOMA-IR and the IGR (Table 3). Triglyceride, VLDL cholesterol, CRP, fibrinogen, and tPA were all significantly higher and HDL cholesterol and LDL size significantly lower in the high–HOMA-IR than the low–HOMA-IR groups. There were no differences in any of the values between the high- versus low-IGR groups within either the low–or particularly in the high–HOMA-IR categories.

These results demonstrate that significant heterogeneity exists within the DPP IGT population, and this heterogeneity is likely to influence the risk for future CVD. Demographic factors had significant influences on the lipoprotein profile. Men had higher triglyceride and lower HDL cholesterol concentrations and smaller LDL particle size than women, as in the general population and in those with diabetes (23,24). Among women, postmenopausal estrogen users had higher triglyceride and total and HDL cholesterol values compared with women not receiving estrogen treatment. Of interest was the fact that LDL cholesterol levels were not reduced compared with nonusers, as has been reported in some studies (25). In this respect, the results are similar to those in the cohort with diabetes from the National Health and Nutrition Examination Survey III (26), suggesting heterogeneity in the LDL cholesterol response to hormone replacement therapy. Despite the differences between sexes, approximately one-third of both men and women had triglyceride values of ≥1.69 mmol/l together with reduced HDL cholesterol levels, which comprise two of three factors required to diagnose the metabolic syndrome (27). A similar proportion had the small dense LDL size phenotype. Almost 30% of men and 21% of women had frankly elevated triglyceride levels (≥2.26 mmol/l). Approximately 25% of the DPP population may therefore be eligible for treatment approaches aimed at lowering non-HDL cholesterol values, as proposed for participants with triglyceride levels ≥2.26 mmol/l by the National Cholesterol Education Program (27).

As has also been documented for diabetic and normoglycemic populations, African-American men and women had lower triglyceride and higher HDL cholesterol levels than did other ethnic groups (28,29). While this may be expected to have an impact on heart disease risk, evidence suggests that the incidence of CVD is not decreased in African Americans with glucose intolerance (30,31). Increasing age was associated with an elevation in total, LDL, and HDL cholesterol levels, as has been reported for the general population (32), but not with triglyceride or the total-to-HDL or LDL-to-HDL cholesterol ratios. Age-related increases in LDL cholesterol in participants with IGT will likely be accompanied by an increasing need for lowering of LDL cholesterol in older participants with IGT. Overall, 14% of the DPP population had LDL cholesterol values that merited drug therapy according to National Cholesterol Education Program criteria irrespective of other risk factors (>4.9 mmol/l), while another 40% with LDL cholesterol values >3.36 mmol/l would be eligible for drug therapy if they had two major risk factors. These findings indicate that a significant proportion of the subjects in DPP with IGT had lipid abnormalities associated with increased risk for CVD.

Among the metabolic factors examined, obesity, especially abdominal adiposity, is known to influence lipids and lipoproteins (33) and also contributes to the deterioration of the lipid profile with development of IGT (34). However, in the regression analysis, anthropometric variables were found to only modestly contribute to the lipid or lipoprotein measures. These results do not necessarily contradict the importance of the effect of increasing body weight or abdominal obesity on the lipid profile but rather that within the DPP cohort of overweight and obese participants, the effect of increasing body weight and WHR on the profile was limited. The severity of obesity and the lack of difference in BMI between Hispanic and non-Hispanic participants may also have explained the absence of triglyceride and HDL cholesterol differences between these two ethnic groups, as has been previously described (35). There was also little impact of glycemic measures on lipids and lipoproteins, probably for the same reason. Stronger associations were demonstrated between HOMA-IR and HDL cholesterol, triglyceride levels, and LDL particle size, as has been typical of previous studies in normoglycemic (36) as well as glucose-intolerant (37) populations, pointing to the importance of insulin resistance in the genesis of dyslipidemia in IGT. The significance of this result is strengthened by the finding that participants with HOMA-IR values in the upper half of the distribution had a significantly more adverse lipid profile than those in the lower half, and this did not appear to be influenced by whether they were in the higher or lower half of the range of an index of insulin secretion (IGR).

Recent evidence has implicated CRP levels as a predictor of cardiovascular events independent of other risk factors (8). However, the question as to whether this applies equally to obese, insulin-resistant, and glucose-intolerant populations is confounded by observations demonstrating strong associations between CRP levels and measures of obesity and insulin resistance (38,39). CRP was strongly influenced by BMI (but not independently by WHR) in this cohort, whereas measures of insulin resistance and the narrow range of glycemia in DPP had no independent effect. Furthermore, we found mean CRP levels to be lowest among the Asian-American subgroup, which had the lowest mean BMI among ethnic groups. In addition to body weight, female sex had a powerful effect on CRP levels, which were twice as high in women compared with men, as reported in the general population (40). High-risk CRP levels (>3 mg/l) were present in 57% of the DPP cohort and were far more prevalent in women (69%) than in men (32%). A significant portion of this sex difference appears to be related to body weight as well, since women were significantly more overweight than men in the DPP. In addition, 32% of women in the DPP were receiving estrogen treatment, and this was associated with increased CRP values, as has previously been reported (41). The clinical significance of this is unknown. An important unanswered question is whether the body weight/CRP relationship has any bearing on the predictive value of CRP for cardiovascular events in overweight and obese participants.

Like CRP, fibrinogen levels were strongly influenced by BMI but less strongly by sex, as has been reported before (42). This association likely reflects the fact that fibrinogen is known to be an inflammatory reactant (43). However, fibrinogen was also increased in relation to glucose levels, as has been previously noted (44). These influences are important because of evidence that fibrinogen levels are independently predictive of CVD (45). Aside from a modest relationship between A1C and LDL cholesterol levels, HDL cholesterol and tPA were the only risk factors studied here that appeared to be associated with increasing glucose levels within the IGT range. Levels of total tPA in cohort studies have fairly consistently been demonstrated to predict cardiovascular events (46), since they estimate the level of antifibrinolytic activity in plasma predominantly reflecting the tPA in tPA–plasminogen activator inhibitor-1 complexes (47). These levels were independently related to both total and abdominal adiposity and to HOMA-IR in the multiple regression model. Each of these factors has been proposed to increase the level of plasminogen activator inhibitor-1, the major inhibitor of fibrinolysis (6,48). Overall, participants with HOMA-IR values in the upper half of the distribution had significantly higher values of these novel risk factors. Estrogen treatment, on the other hand, was associated with reduced fibrinogen and tPA levels, as has been previously described (49).

Collectively, this analysis indicates that in a large multiethnic cohort with IGT, dyslipidemia is common, and while the prevalence varies by sex and ethnicity, it is also significantly influenced by the degree of insulin resistance and less by the level of obesity, insulin secretory reserve, or glucose intolerance in the population. Women with IGT have higher CRP and fibrinogen levels but slightly lower tPA values than their male counterparts, with the degree of excess body fat and of glucose intolerance emerging as the most important metabolic determinants of these novel risk factors. In summary, these findings reveal that participants with IGT exhibit significant heterogeneity in the prevalence of lipid, hemostatic, and inflammatory cardiovascular risk factors. Whether those with a more adverse pattern of CVD risk factors are more likely to become diabetic or develop CVD remains to be examined.

Funding was provided by the National Institutes of Health/National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute of Child Health and Human Development, and the National Institute on Aging; the Office of Research National Center on Minority Health and Health Disparities, Office of Women’s Health; the Indian Health Service; Centers for Disease Control and Prevention; the General Clinical Research Program, National Center for Research Resources; the American Diabetes Association; Bristol-Myers Squibb; Lipha Pharmaceuticals; and Parke-Davis.

LifeScan, Health-O-Meter, Hoechst Marion Roussel, Merck-Medco Managed Care, Merck and Company, Nike Sports Marketing, Slim Fast Foods, and Quaker Oats donated materials, equipment, or medicines for concomitant conditions. McKesson BioServices, Matthews Media Group, and the Henry M. Jackson Foundation provided support services under subcontract with the coordinating center.

We thank the thousands of volunteers in this program for their devotion to the goal of diabetes prevention. This article was prepared by Ronald B. Goldberg (chair), Marinella G. Temprosa, Steven M. Haffner, Trevor J. Orchard, Robert Ratner, John M. Lachin, and Santina M. Marcovina.