Metabolic Risk Factors Worsen Continuously across the Spectrum of Nondiabetic Glucose Tolerance: The Framingham Offspring Study

Methods

Participants

Participants in the Framingham Offspring Study, a long-term, community-based, prospective observational study of risk factors for cardiovascular disease, are the children and spouses of the children of the original Framingham Heart Study cohort. The design of the Framingham Offspring Study has been reported elsewhere [37]. From January 1991 to September 1993, during the fifth cycle of the Framingham Offspring Study, 2874 consecutive participants had a standardized medical history and physical examination at the Framingham Heart Study Clinic. Height and weight were measured, and body mass index was calculated. Persons were classified as obese if they had a body mass index of 27.3 kg/m² or more in women or 27.8 kg/m² or more in men; these values correspond to approximately 120% of the ideal body weight [38]. Waist circumference was measured at the umbilicus while the participant was standing, hip girth was determined, and the ratio of waist-to-hip circumference was calculated. Participants were classified as having an elevated waist-to-hip ratio if the ratio was more than 0.9 for women and more than 1.0 for men. Blood pressure was measured after the participant had been sitting for at least 5 minutes. Participants were classified as hypertensive if the diastolic blood pressure was more than 90 mm Hg or the systolic blood pressure was more than 140 mm Hg on two measurements or the participant reported use of antihypertensive medications [39]. Participants who reported smoking at least one cigarette per day during the year before the examination were classified as current smokers.

Fasting plasma samples were drawn for glucose, insulin, and lipid measurements. A 75-g oral glucose tolerance test was administered according to WHO standards [1] to participants not known to have diabetes, and levels of glucose and insulin were measured 2 hours after oral challenge was given. Participants were identified as having previously diagnosed diabetes if the fasting plasma glucose level was 7.8 mmol/L or more at any two previous examinations or if current or past hypoglycemic drug therapy was reported. Participants were classified as having normal, impaired, or diabetic glucose tolerance on the basis of results of an oral glucose tolerance test using 1980 WHO criteria. Participants were classified as having previously undiagnosed diabetes (fasting plasma glucose level ≥ 7.8 mmol/L or 2-hour postchallenge glucose level ≥ 1.1 mmol/L), impaired glucose tolerance (fasting plasma glucose level < 7.8 mmol/L and 2-hour postchallenge glucose level > 7.8 mmol/L and < 11.1 mmol/L), or normal glucose tolerance (fasting and 2-hour postchallenge plasma glucose levels < 7.8 mmol/L). Participants were classified as having hyperinsulinemia if the fasting insulin level was greater than the 90th percentile of its distribution among Framingham Offspring Study participants with normal glucose tolerance (corresponding to a fasting serum insulin level > 114 pmol/L). We used this criterion to improve the specificity of the hyperinsulinemia classification. Hyperinsulinemia reflects insulin resistance: A single fasting insulin sample has been shown to be inversely correlated with insulin sensitivity or whole-body glucose uptake as measured by using the euglycemic-hyperinsulinemic clamp method [40, 41]. Participants were classified as having a low HDL cholesterol level if the value was less than 1.16 mmol/L in women or less than 0.91 mmol/L in men; triglyceride levels were considered elevated if the fasting value was 2.26 mmol/L or more in men and women. We calculated an index of the total burden of metabolic risk-the metabolic sum-according to the presence of none to all of the following six features: obesity, elevated waist-to-hip ratio, hypertension, low HDL cholesterol level, elevated triglyceride level, and hyperinsulinemia.

Beginning in February 1992, 1711 consecutive samples were collected for hemoglobin A_1c measurement. Hemoglobin A_1c, a time-integrated measure of glycemia over the preceding 2 to 3 months that tracks consistently over at least 4 to 6 years [42, 43], was used to assess chronic glycemic control in each category of glucose tolerance. The distributions of sex and glucose tolerance categories were similar among participants with and participants without hemoglobin A_1c measurements.

After we excluded 182 persons with missing information on glucose tolerance status or other risk variables, 2692 persons with complete information (other than hemoglobin A_1c values) were available for study. Of these 2692 persons, 64 men (5.0%) and 45 women (3.2%) with diagnosed diabetes were excluded, leaving 1221 men and 1362 women as the sample for analysis. Persons with diagnosed diabetes were excluded to avoid the confounding effects of diabetes duration and treatment on metabolic risk variables.

Laboratory Methods

Fasting plasma glucose was measured with a hexokinase reagent kit (A-gent glucose test, Abbott, South Pasadena, California). Glucose assays were run in duplicate, and the intra-assay coefficient of variation ranged from 2% to 3%, depending on the assayed glucose level. Hemoglobin A (_1c) was measured by high-performance liquid chromatography after overnight dialysis against normal saline to remove the labile fraction [44]. The laboratory's nondiabetic mean ±SD for this assay was 5.05 ± 0.65, and the interassay and intra-assay coefficients of variation were less than 2.5%. The assay was standardized against and similar to the glycosylated hemoglobin assay used in the Diabetes Control and Complications Trial [45]. Fasting insulin was measured in plasma as total immunoreactive insulin (Coat-A-Count Insulin, Diagnostic Products Corp., Los Angeles, California) and calibrated to serum levels for reporting purposes. Cross-reactivity of this assay with proinsulin at mid-curve is approximately 40%, the intra-assay and interassay coefficients of variation ranged from 5.0% to 10.0% for insulin concentrations reported here, and the lower limit of sensitivity was 8 pmol/L. The fasting total plasma cholesterol and triglyceride levels were measured enzymatically [46], and the HDL cholesterol fraction was measured after precipitation of low-density and very-low-density lipoproteins with dextran sulfate-magnesium [47]. The HDL₃ cholesterol subfraction was measured by using a double precipitation method [48]. The HDL₂ cholesterol subfraction was estimated indirectly as the difference between the HDL and HDL₃ cholesterol concentrations. Low-density lipoprotein cholesterol values were estimated indirectly for participants with plasma triglyceride levels less than 4.52 mmol/L (n = 2521) [49]. The Framingham Heart Study laboratory participates in the lipoprotein cholesterol laboratory standardization program administered by the Centers for Disease Control and Prevention.

Statistical Analysis

Analyses were performed separately for men and women by using the SAS statistical package [50]. We classified participants with normal glucose tolerance into quintiles of fasting glucose levels to assess trends across the spectrum of glucose tolerance, from normal tolerance to impaired glucose tolerance and previously undiagnosed diabetes. Fasting plasma glucose levels among participants with normal glucose tolerance ranged from 3.33 to 4.83 mmol/L in quintile 1, 4.88 to 5.05 mmol/L in quintile 2, 5.11 to 5.27 mmol/L in quintile 3, 5.33 to 5.61 mmol/L in quintile 4, and 5.66 to 7.77 mmol/L in quintile 5. Log transformation of insulin and triglyceride levels normalized their distributions for use in statistical testing; however, we give arithmetic means in the figures. Three separate analysis-of-variance models were used to calculate crude, age-adjusted, and multivariable-adjusted least-square mean values and 95% CIs for each risk variable. Multivariable models included covariates for age (years), physical activity index score (log score) [51], current cigarette smoking (yes or no), average daily alcohol intake (ounces per day), use of antihypertensive medication (yes or no), use of cholesterol-lowering medication (yes or no), and use of estrogen replacement therapy among postmenopausal women (yes or no). Results were similar for all three models, and we present only results of the multivariable-adjusted models. Using linear regression, we tested the significance of trends across categories from the lowest to the highest quintile of fasting glucose among participants with normal glucose tolerance, from the lowest quintile of normal fasting glucose to impaired glucose tolerance (the range of nondiabetic glucose tolerance), and from the lowest quintile of normal fasting glucose to previously undiagnosed type 2 diabetes. Trends in the latter analysis were similar to those found in the first two analyses and are not presented.

We assessed interaction between obesity and glucose tolerance status by introducing a multiplicative term for overall obesity and glucose tolerance status into each model. We found no effect modification by obesity for any of the metabolic risk variables. We assessed the possibility that the rate of change of the multivariable linear regression slope for each metabolic risk variable increased near the upper end of the range of nondiabetic glucose tolerance by comparing the slope across categories from the first to the third quintile of normal fasting glucose to the slope across categories from the fourth quintile to impaired glucose tolerance. To assess the possibility that our results might be an artifact of categorizing participants by WHO criteria for glucose tolerance, we repeated the analyses, using deciles of fasting glucose levels, 2-hour postchallenge glucose levels, or hemoglobin A_1c levels to represent increasing levels of glucose intolerance. Results of these analyses were essentially identical to those presented here. Odds ratios and 95% CIs for categorical outcomes were calculated by using logistic regression [52], with the lowest quintile of normal fasting glucose as the referent category. We assessed the significance of the trend across odds ratios with logistic regression. Statistical significance was defined as a two-tailed P value less than 0.05.

Results

The mean age of study participants was 54 years (range, 26 to 82 years); 52.7% of participants were women. After exclusion of persons with known diabetes, oral glucose tolerance test results classified 5.8% of men (95% CI, 4.5% to 7.1%) and 4.0% of women (CI, 2.9% to 5.0%) as having previously undiagnosed type 2 diabetes and classified 11.2% of men (CI, 9.4% to 13.0%) and 14.1% of women (CI, 12.2% to 15.9%) as having impaired glucose tolerance (Figure 1). Of the 2129 remaining participants with normal glucose tolerance, more men than women fell into the top two quintiles of fasting glucose levels (P < 0.001). Hemoglobin A_1c levels increased from the lowest quintile of normal fasting glucose level to impaired glucose tolerance (P < 0.001 for trend) (Figure 1). Thus, although participants with impaired glucose tolerance had a broader range and a lower mean fasting plasma glucose level than participants in the highest quintile of normal glucose tolerance (5.7 ± 0.1 mmol/L and 5.9 ± 0.1 mmol/L, respectively; P = 0.003), hemoglobin A_1c levels revealed a progressive linear increase across the nondiabetic spectrum of glucose tolerance. Participants with previously undiagnosed type 2 diabetes had much higher hemoglobin A_1c levels than did nondiabetic participants, reflecting the gross decrease in glucose tolerance that is characteristic of diabetes.

View larger version:

• In this window
• In a new window

Figure 1. Mean hemoglobin A levels and 95% CIs (error bars) among women (□s and dotted lines) and men (black squares and solid lines) are given for each glucose tolerance category from the lowest quintile (N1) to the highest quintile (N5) of fasting plasma glucose level among participants with normal glucose tolerance, impaired glucose tolerance (IGT), and previously unrecognized diabetes mellitus (DM). The fasting plasma glucose range and number and proportion of men and women in each category are given below the figure. Means are multivariable adjusted. *  < 0.001 for trend from the lowest quintile of normal fasting glucose level to impaired glucose tolerance. Distribution of hemoglobin A_1c levels and of the study population by glucose tolerance categories.1cP

Among men and women, levels of all metabolic risk factors for type 2 diabetes and for cardiovascular disease increased continuously across categories of increasing glucose intolerance (Figure 2 and Figure 3). Body mass index, waist-to-hip ratio, systolic and diastolic blood pressure, fasting insulin levels, and triglyceride levels increased and HDL cholesterol levels decreased; significant linear trends were seen from the lowest quintile of normal fasting glucose to impaired glucose tolerance. Participants with previously undiagnosed type 2 diabetes had even higher levels of all these risk factors. Subfractions of HDL₂ and HDL₃ cholesterol showed trends similar to those for HDL cholesterol, but total and low-density lipoprotein cholesterol levels did not statistically significantly differ across categories of glucose tolerance (data not shown). Two-hour postchallenge insulin levels increased steadily across quintiles of normal fasting glucose and increased dramatically in participants with impaired glucose tolerance (P < 0.001 for men and women for difference in regression slopes, comparing a regression slope across the first three quintiles of normal fasting glucose level to a slope from the fourth quintile to impaired glucose tolerance), only to decrease again in participants with previously undiagnosed type 2 diabetes. The regression slope at higher levels of nondiabetic fasting glucose in the distribution of triglyceride levels among men also increased significantly (P = 0.002 for difference in regression slopes). Although it was barely perceptible, a change in slope for waist-to-hip ratio (P = 0.02) (Figure 2) and HDL cholesterol (P = 0.015) (Figure 3) in men was statistically significant. However, apart from 2-hour insulin levels and triglyceride levels in men, no clear evidence was found for a distinct threshold, or point at which metabolic risk begins to increase more rapidly, for any of the other measures of risk factors across categories of increasing glucose intolerance.

View larger version:

• In this window
• In a new window

Figure 2. Mean body mass index ( ), waist-to-hip ratio ( ), systolic blood pressure ( ), and diastolic blood pressure ( ) and 95% CIs (error bars) among women (white squares and dotted lines) and men (black squares and solid lines) are given for each glucose tolerance category from the lowest quintile (N1) to the highest quintile (N5) of fasting plasma glucose level among participants with normal glucose tolerance, impaired glucose tolerance (IGT), and previously undiagnosed diabetes mellitus (DM). Means are multi-variable adjusted. * < 0.001 for trend from the lowest quintile of normal fasting glucose level to impaired glucose tolerance. Distribution of obesity and blood pressure by glucose tolerance category.top lefttop rightbottom leftbottom rightP

View larger version:

• In this window
• In a new window

Figure 3. Mean fasting and 2-hour post-oral glucose challenge levels of serum insulin ( ), plasma triglyceride ( ), and plasma high-density lipoprotein cholesterol ( ) and 95% CIs (error bars) among women (white squares and dotted lines) and men (black squares and solid lines) are given for each glucose tolerance category from the lowest quintile (N1) to the highest quintile (N5) of fasting plasma glucose level among participants with normal glucose tolerance, impaired glucose tolerance (IGT), and previously undiagnosed diabetes mellitus (DM). Means are multivariable adjusted. * < 0.001; † = 0.003 for trend from the lowest quintile of normal fasting glucose level to impaired glucose tolerance. Distribution of insulin and lipid levels by glucose tolerance category.top left and top rightbottom rightbottom leftPP

We further evaluated these relations by assessing trends in measures of metabolic risk across quintiles of fasting glucose levels among participants with normal glucose tolerance, excluding participants with impaired glucose tolerance or undiagnosed type 2 diabetes. These trends were similar to trends across the range of nondiabetic glucose tolerance, with two exceptions (Table 1): Among men, the trends in HDL cholesterol and triglyceride levels across quintiles of normal fasting glucose were not statistically significant (P = 0.10 and 0.06, respectively).

To examine the possibility that obesity could explain the association between glycemia and increasing measures of risk, we added a term for body mass index to each multivariable analysis-of-variance model. Controlling for body mass index did not substantially diminish trends across the nondiabetic spectrum of glucose tolerance for most risk variables, with the exception of waist-to-hip ratio in men and HDL cholesterol levels in men and women. Adjustment for body mass index diminished the decrease in HDL cholesterol levels across categories of increasing glucose intolerance (P for multivariable trend including body mass index = 0.02 for men and 0.10 for women), suggesting that obesity accounts for much of the decrease in HDL cholesterol levels commonly seen in impaired glucose tolerance. Figure 4 shows this diminished trend and the magnifying effect of obesity on measures of risk factors: In each glucose tolerance category, obese participants had dramatically lower HDL cholesterol levels than nonobese participants. A similar magnifying effect was seen for each of the metabolic risk factors. In addition, the relation between glycemia and HDL cholesterol level (Figure 4), as well as other risk factors, was essentially the same for obese and nonobese participants; this showed a lack of interaction by level of obesity.

View larger version:

• In this window
• In a new window

Figure 4. Mean plasma HDL cholesterol levels and 95% CIs (error bars) among female ( ) and male ( ) participants who were obese (body mass index ≥ 27.3 kg/m in women or ≥ 27.8 kg/m in men [black squares and solid lines]), nonobese (white squares and dotted lines), and obese and nonobese combined (slashed squares and dashed lines) in each glucose tolerance category from the lowest quintile (N1) to the highest quintile (N5) of fasting plasma glucose level among participants with normal glucose tolerance, impaired glucose tolerance (IGT), and previously undiagnosed diabetes mellitus (DM). Means are multivariable adjusted. The values indicate the significance of trends from the lowest quintile of normal fasting glucose level to impaired glucose tolerance. Distribution of high-density lipoprotein (HDL) cholesterol levels by obesity and glucose tolerance category.leftright22P

The relation between glucose intolerance and worsening metabolic risk was also apparent when we classified participants into risk factor groups. Among men and women, odds ratios for obesity, elevated waist-to-hip ratio, hypertension, low HDL cholesterol levels, and hyperinsulinemia showed continuous increases across categories of nondiabetic glucose tolerance (Figure 5). Among men, odds ratios for elevated triglyceride levels showed the increased rate of change at the upper end of the normal glucose tolerance category, a finding similar to that observed for mean triglyceride levels. Participants with undiagnosed type 2 diabetes had especially high odds of increased metabolic risk.

View larger version:

• In this window
• In a new window

Figure 5. Odds ratios and 95% CIs (error bars) for obesity (body mass index ≥ 27.3 kg/m in women or ≥ 27.8 kg/m in men [top left]), elevated waist-to-hip ratio (>0.9 for women and >1.0 for men [top middle]), hypertension (two measurements with diastolic blood pressure > 90 mm Hg, systolic blood pressure > 140 mm Hg, or any use of antihypertensive medication [top right]), low high-density lipoprotein (HDL) cholesterol level (<45 mg/dL in women or <35 mg/dL in men [bottom left]), elevated triglyceride level (>200 mg/dL [bottom middle]), and hyperinsulinemia (fasting insulin level > the 90th percentile of its distribution among Framingham Offspring Study participants with normal glucose tolerance [bottom right]) among women (white squares and dotted lines) and men (black squares and solid lines) are shown for each glucose tolerance category from the lowest quintile (N1) to the highest quintile (N5) of fasting plasma glucose level among participants with normal glucose tolerance, impaired glucose tolerance (IGT), and previously undiagnosed diabetes mellitus (DM). Odds ratios are multivariable adjusted. * < 0.001; † = 0.002; * = 0.008 for trend from the lowest quintile of normal fasting glucose level to impaired glucose tolerance. Multivariable odds ratios for metabolic risk factors by glucose tolerance category.22PPP

The total risk burden conferred by these six risk factors, expressed as the metabolic sum, also increased steadily across the spectrum of glucose tolerance (Figure 6). The proportion of participants with four or more metabolic risk factors increased steadily from the lowest quintile of normal glucose tolerance (3.8% of men and 2.5% of women) to impaired glucose tolerance (29.2% of men and 17.7% of women; P < 0.001 for trend in men and women) to undiagnosed type 2 diabetes (43.7% of men and 38.9% of women). Impaired glucose tolerance occupied a metabolic risk state between normal glucose tolerance and type 2 diabetes. Although Figure 6 may be interpreted as showing a subtle increase in the rate of change in proportions at about the fourth quintile of normal fasting glucose level (P < 0.001 for difference among men in the regression slope of mean metabolic sum), increasing metabolic risk associated with increasing glucose intolerance was continuous across the spectrum of glucose tolerance; increasing risk was apparent even among normal glucose-tolerant participants.

View larger version:

• In this window
• In a new window

Figure 6. The proportion of women (W) and men (M) with 2 to 3 (white bars) or 4 to 6 (diagonally striped bars) of the following features: obesity (body mass index ≥ 27.3 kg/m in women or ≥ 27.8 kg/m in men), elevated waist-to-hip ratio (>0.9 for women and >1.0 for men), hypertension (two measurements of diastolic blood pressure > 90 mm Hg, systolic blood pressure > 140 mm Hg, or any use of antihypertensive medication), low high-density lipoprotein cholesterol level (<45 mg/dL in women or <35 mg/dL in men), elevated triglyceride level (>200 mg/dL), or hyperinsulinemia (fasting insulin level greater than the 90th percentile of its distribution among Framingham Offspring Study participants with normal glucose tolerance) are displayed for each glucose tolerance category from the lowest quintile (N1) to the highest quintile (N5) of fasting plasma glucose level among participants with normal glucose tolerance, impaired glucose tolerance (IGT), and previously undiagnosed diabetes mellitus (DM). *P< 0.001 for trend from the lowest quintile of normal fasting plasma glucose level to impaired glucose tolerance. Age-standardized distribution of the metabolic sum by glucose tolerance category.22

Discussion

The Framingham Offspring Study, an unselected, community-based sample of predominantly white men and women, had a prevalence of impaired glucose tolerance and diagnosed and previously undiagnosed diabetes mellitus that was almost identical to that reported in a broad U.S. population sample of similar age distribution [3]. Measures of metabolic risk for type 2 diabetes or for cardiovascular disease worsened continuously across categories of increasing glucose intolerance. Most risk factors showed no clear evidence of a threshold for increased risk at higher levels of fasting glucose within the nondiabetic range. Participants with previously undiagnosed type 2 diabetes had very high measures of metabolic risk.

Our observations are consistent with other evidence of associations between glycemia and risk for cardiovascular disease among nondiabetic participants [25, 29-32, 53]. We extend the findings in these reports by examining metabolic risk across a full spectrum of nondiabetic glucose tolerance in a large population in whom glucose tolerance status has been clearly defined by using standardized oral glucose-tolerance testing. A continuous gradient of increasing measures of metabolic risk across increasing fasting glucose levels within the normal range of glucose tolerance has not been demonstrated previously.

Although the glucose levels that define diabetes are based on empirical evidence that diabetes-specific complications occur only above these thresholds [2], the glucose values that separate normal from impaired glucose tolerance have not had as clear a basis in pathophysiology. Recently, the American Diabetes Association proposed decreasing the fasting glucose levels that serve as criteria for defining type 2 diabetes and impaired glucose tolerance [54]. Although the continuous nature of metabolic risk belies risk thresholds implied by categorical definitions, the newly proposed glucose criteria reflect increased metabolic risk associated with fasting glucose levels that have previously been considered normal.

Whereas body mass index increased steadily with increasing glucose intolerance, the associations between most other measures of metabolic risk and glycemia were independent of overall obesity. Accounting for obesity attenuated the effect of increasing glucose intolerance only for HDL cholesterol levels. These observations are consistent with previous evidence that improved glucose tolerance was associated with reduced metabolic risk independent of changes in obesity [55]. The gradient of increasing risk was similar for nonobese and obese participants; at every level of glucose tolerance, however, the presence of overall obesity magnified measures of risk factors. This effect may reflect the global mechanism by which obesity confers risk for type 2 diabetes [56, 57] or for cardiovascular disease [58].

Fasting insulin levels increased steadily across categories of glucose tolerance, reflecting progressive insulin resistance. Two-hour postchallenge insulin levels, a function of peripheral insulin resistance and pancreatic β-cell secretory capacity, increased steadily across the range of normal glucose tolerance but were dramatically elevated in participants with impaired glucose tolerance; this result is consistent with a delayed β-cell response to hyperglycemia [59]. The decrease in 2-hour postchallenge insulin levels in participants with type 2 diabetes is consistent with the failing β-cell response characteristic of type 2 diabetes. Despite the high variability associated with single measures of fasting or 2-hour postchallenge insulin levels [8], the distribution of these values across the spectrum of glucose tolerance reflected patterns of insulin response seen in highly controlled metabolic ward studies [60]. Our data support the proposition that single insulin values can be informative for epidemiologic analyses when glucose tolerance status is known [40, 41].

The high intraperson variability of oral glucose-tolerance test results may have introduced some misclassification bias into our analysis. One would expect that on repeated testing, some persons classified into the upper quintiles of normal glucose tolerance would have impaired glucose tolerance or some persons with undiagnosed type 2 diabetes or impaired glucose tolerance would have more normal glucose tolerance. Misclassification may account for the relatively small difference in hemoglobin A_1c values between participants in the highest quintile of normal fasting glucose and participants with impaired glucose tolerance. Alternatively, these hemoglobin A_1c values probably reflect the chronic level of ambient glycemia more accurately than do either fasting or postchallenge blood glucose levels, as shown by the clear linear trend in hemoglobin A_1c values across the spectrum of nondiabetic glucose tolerance categories, with a dramatic increase in participants with type 2 diabetes. Other analyses done by using deciles of fasting or 2-hour postchallenge glucose levels or hemoglobin A_1c values instead of glucose tolerance categories gave similar results. These observations argue against a substantial bias caused by errors in classification made on the basis of results of a single oral glucose-tolerance test.

Impaired glucose tolerance and undiagnosed type 2 diabetes were common in this primarily white population. These conditions affected 18% of middle-aged Framingham Offspring Study participants and accounted for about 29% of participants who had at least two additional risk factors. Although variation in the relation among metabolic risk factors across ethnic groups [61, 62] may limit the generalizability of our findings, our data reinforce the concept that asymptomatic glucose intolerance is not a benign metabolic condition. The cross-sectional study design does not allow the conclusion that worsening glucose tolerance causes increasing metabolic risk; however, our findings are supported by prospective studies that found associations between hyperglycemia and development of type 2 diabetes [15] or cardiovascular disease among nondiabetic persons [29-31]. The relation between glycemia and cardiovascular disease is less clear among persons with diagnosed diabetes [63]. Definitive experimental evidence may be provided by the Diabetes Prevention Program, a large randomized trial testing the hypothesis that detection and treatment of impaired glucose tolerance can prevent or delay the development of type 2 diabetes and its associated complications [64]. Although normalization of blood glucose in type 2 diabetes may be cost-effective [65], screening for diabetes cannot be widely recommended until evidence shows that early detection improves outcomes [66]. However, strong experimental evidence shows that treatment of other risk factors (particularly hypertension and hyperlipidemia) reduces cardiovascular mortality, even among persons with diagnosed type 2 diabetes [67, 68]. Clinical interventions to reduce metabolic risk should focus on hypertension and low HDL cholesterol levels as well as obesity and the sedentary lifestyle that promote insulin resistance. Sustained reduction in body mass index, in particular, is likely to have global beneficial effects on metabolic risk.

From Massachusetts General Hospital, Harvard Medical School, Boston University School of Public Health, and Boston University School of Medicine, Boston, Massachusetts; and the National Heart, Lung, and Blood Institute, Framingham, Massachusetts.

Dr. Nathan: Diabetes Unit, Bulfinch 408, Massachusetts General Hospital, Boston, MA 02114.

Dr. Wilson: Framingham Heart Study, 5 Thurber Street, Framingham, MA 01701.

Dr. Cupples: Boston University School of Public Health, Department of Epidemiology and Biostatistics, 715 Albany Street, Boston, MA 02118.