Use of Metabolic Markers To Identify Overweight Individuals Who Are Insulin Resistant

Context Insulin resistance is associated with adverse outcomes, such as cardiovascular disease and type 2 diabetes mellitus. The insulin suppression test, the gold standard method of diagnosing insulin resistance, is cumbersome to administer. A simple method to identify persons with insulin resistance would be useful. Contribution In a group of overweight individuals, 3 easily measured variables (triglyceride levels, the ratio of triglyceride to high density lipoprotein [HDL] cholesterol levels, and insulin concentration) identified insulin-resistant individuals with sensitivities of 57% to 67% and specificities of 68% to 85%. Implications Triglyceride levels, the triglyceride-HDL cholesterol ratio, and insulin concentration are imperfect but practical methods for identifying overweight persons who are insulin resistant and at greatest risk for complications. The Editors Recent reports (1) indicate that more than 50% of the U.S. population is overweight (body mass index [BMI] 25 kg/m2), with approximately 20% designated as obese (BMI 30 kg/m2). Because overweight is important in the genesis of type 2 diabetes mellitus and cardiovascular disease (CVD), the absolute number of Americans in this category is disturbing. The gravity of the problem is accentuated in light of the report that only approximately 50% of physicians polled provided weight loss counseling (2) and that pharmacologic treatment of weight loss is not being used appropriately in overweight persons (3). Reluctance to assign weight control programs a high priority might be decreased if identifying overweight or obese individuals at greatest risk for adverse health consequences were possible, particularly if weight loss would significantly attenuate the risk. In this context, it is necessary to begin by emphasizing that the prevalence of insulin resistance is increased in patients with type 2 diabetes mellitus, essential hypertension, and CVD and that insulin resistance and compensatory hyperinsulinemia have been shown to be independent predictors of all 3 clinical syndromes (4-9). Since obese individuals tend to be insulin resistant and become more insulin sensitive with weight loss (10), an obvious approach to identify individuals who would most benefit from weight loss is to measure insulin-mediated glucose disposal. However, direct measures of insulin-mediated glucose disposal are not clinically practical. On the other hand, overweight persons are also at increased risk for glucose intolerance, and the higher the plasma glucose or insulin concentrations in nondiabetic persons, the more likely that the persons are insulin resistant (4, 11). Thus, differences in fasting plasma glucose or insulin concentrations might be useful to identify insulin-resistant persons. These persons also have a characteristic dyslidemia (4), and measuring these variables might also help identify insulin resistance. For example, plasma triglyceride and high-density lipoprotein (HDL) cholesterol levels are independently associated with insulin resistance (12) and are independent predictors of CVD (13, 14). In addition, the plasma concentration ratio of total cholesterol to HDL cholesterol is well recognized as a predictor of CVD (15) and is also highly correlated with insulin resistance (16). A less commonly considered CVD risk factor is the ratio of triglyceride to HDL cholesterol, despite the observation that the triglycerideHDL cholesterol ratio is as significant a predictor of CVD as are the ratios of low-density lipoprotein (LDL) cholesterol to HDL cholesterol or total cholesterol to HDL cholesterol (17). A more recent study showed that persons in the highest tertile of the triglycerideHDL cholesterol ratio had increased CVD risk in the absence of the 4 conventional risk factors, whereas those in the lowest tertile had decreased risk in the presence of the same 4 risk factors (18). Although obese individuals tend to be insulin resistant, hyperinsulinemic, glucose intolerant, and dyslipidemic, not all overweight or obese individuals are insulin resistant, nor do they all have the characteristic disturbances in glucose or lipid metabolism (19-23). Furthermore, not all CVD risk factors improve with weight loss, and the metabolic benefits associated with weight loss are largely confined to overweight or obese individuals with these abnormalities at baseline (20-23). Given the relative ease of measuring plasma glucose, insulin, and lipid concentrations, and their importance as both CVD risk factors and manifestations of insulin resistance, we attempted to develop a simple clinical approach using these measurements to identify overweight or obese individuals who are both insulin resistant and at greatest risk for CVD. Methods The study sample consisted of 258 persons with a BMI of 25 kg/m2 or greater, classified as overweight or obese by National Institutes of Health (24) and World Health Organization criteria (25). Participants were drawn from a large database of 490 healthy volunteers who have participated in research studies in the past 10 years. These studies typically used newspaper advertisements to identify persons without known disease to participate in our efforts to define the relationship between insulin resistance and metabolic abnormalities. According to their medical histories, study participants did not have major chronic medical illnesses, including CVD, and were not taking any medication known to influence insulin resistance or lipid metabolism (such as corticosteroids and lipid-lowering drugs). No clinically significant abnormalities were found during physical examination; participants were not anemic, had normal liver and kidney function, and were nondiabetic on the basis of plasma glucose concentrations in response to a standard oral glucose challenge (26). The 258 individuals included 127 men and 131 women with a mean age (SD) of 50 16 years (range, 19 to 70 years) and a mean BMI (SD) of 29.2 3.2 kg/m2 (range, 25.0 to 39.1 kg/m2). Most participants were white (87%); the remaining participants were Asian American (9%), Hispanic (3%), or African American (1%). Insulin-mediated glucose disposal was estimated by a modification (27) of the insulin suppression test introduced and validated by our research group (28, 29). We have used this approach for more than 35 years to measure insulin action, and results are highly correlated (r > 0.9) with the more commonly used euglycemic, hyperinsulinemic clamp approach (29). After an overnight fast, intravenous catheters are placed in each of the patient's arms. A 180-minute infusion of somatostatin (250 g/h), insulin (179 mol/m2 per min 1), and glucose (13.3 mmol/m 2 2 per min) is administered into 1 arm. Blood samples are collected from the other arm every 30 minutes initially and at 10-minute intervals from 150 to 180 minutes of the infusion to determine the steady-state plasma insulin and glucose concentrations. Since steady-state plasma insulin concentrations are similar for all participants, the steady-state plasma glucose concentration directly measures the insulin's ability to mediate disposal of the infused glucose load; the higher the steady-state plasma glucose concentration, the more insulin resistant the patient. Blood samples were obtained before the insulin suppression test to measure plasma glucose (30), insulin (31), and lipid and lipoprotein (32-34) levels by methods that were identical during the period of study. We have found that insulin's ability to stimulate glucose disposal varied continuously in a sample of 490 healthy persons (35), precluding an objective definition of an individual as being insulin sensitive or insulin resistant. However, in 2 prospective studies (8, 9), we showed that CVD and glucose intolerance or type 2 diabetes developed to a statistically significantly greater degree in one third of the healthy sample that was the most insulin resistant (that is, the tertile with the highest steady-state plasma glucose concentrations). On the basis of these considerations and for the purposes of this analysis, we used as an operational definition of insulin resistance a steady-state plasma glucose concentration in the upper tertile of the distribution of the original 490 healthy volunteers. Because of possible interaction between metabolic markers, sex, and menopausal status of women, we performed logistic regression analysis for predicting insulin resistance that included the best metabolic marker, sex, menopausal status, and all interaction terms. Since there were no significant interactions, men and women, regardless of their menopausal status, were considered together in subsequent analyses. Clinical utility of metabolic markers to identify individuals in the most insulin-resistant tertile was evaluated by constructing receiver-operating characteristic (ROC) curves, which depict the relationship between true-positive (sensitivity) and false-positive (1 specificity) test results for each diagnostic marker. Markers for which a relative increase in sensitivity is matched by a similar increase in false-positive results are represented by a diagonal line and are of less clinical use. Metabolic markers considered were fasting plasma concentrations of glucose, insulin, triglyceride, cholesterol, and HDL cholesterol, as well as the cholesterolHDL cholesterol ratio and the triglycerideHDL cholesterol ratio. Areas under the ROC curves were compared using the method of Hanley and McNeil (36). The metabolic markers of insulin resistance that were statistically significantly better performers were selected for cut-point analysis to identify specific values that would be useful in predicting insulin resistance. The cut-points diagnostic of the top tertile of steady-state plasma glucose were based on the formula M = ws + (1 w) p, where w = prevalence of disease (top tertile steady-state plasma glucose), s = sensitivity, and p = specificity (37). According to this equation, the cut-point identi