Abstract
Background Metabolic syndrome is a collection of anthropomorphic and metabolic derangements that, taken together, constitute a strong predictor of stroke, cardiovascular disease, type 2 diabetes, and perhaps other disorders. Diagnosing the metabolic syndrome according to presently accepted definitions requires a mixture of anthropomorphic and laboratory measurements. In indigent patient populations or locations that lack adequate laboratory facilities, a noninvasive low-cost or cost-free alternative would be beneficial as a preliminary screening measure.
Methods To identify potential means for the diagnosis, we evaluated anthropomorphic measures individually and in combination for their predictive value against the standard classification. Receiver operator characteristic curves identified optimal cutoff values for the variables tested, and consideration of sensitivity and specificity were considered concerning the consequences of false negatives and positives.
Results Receiver operator characteristic curves showed that only waist circumference in men gave acceptable specificity and sensitivity. Combinations of measurements did not improve on a simple measurement of waist circumference.
Conclusions Waist circumference was a potentially useful single indicator in men but not in women. Other values and combinations were evaluated as well. Waist measurement alone may serve as sufficient screening in men in locations where laboratory facilities are lacking or funds are severely limited, as in developing countries.
Metabolic syndrome (MS) is characterized by a cluster of derangements, including hypertension, elevated plasma glucose or diabetes, atherogenic dyslipidemia, and central obesity. As defined by the National Cholesterol Education Program (NCEP),1MS is present if a patient meets at least 3 of the following criteria:
Waist circumference, more than 40 inches in men and more than 35 inches in women,
Arterial blood pressure, 130/85 mm Hg or higher, or therapy with an antihypertensive agent,
Fasting plasma glucose, 110 mg/dL or more, or therapy with antidiabetes medication,
Serum triglycerides, 150 mg/dL or more, and
Serum high-density lipoprotein, 40 mg/dL in men and 50 mg/dL in women.
Estimates of the point prevalence of MS in the American population vary from approximately 29%2to approximately 74%,3depending on the populations studied.
Insulin resistance is frequently present in patients with MS and those with type 2 diabetes mellitus (DM-2) and may play a central role in MS and in the pathogenesis of MS4and vascular endothelial dysfunction.5,6The presence of MS in adults and children predicts later development of cardiovascular disease and type 2 diabetes7-10and in patients with type 2 diabetes MS predicts new-onset chronic kidney disease.11
Metabolic syndrome is a significant risk factor for cardiovascular disease in the general population.2,12It is present in 93% of women and 83% of men with DM-2,1322% to 43% of native Americans,14and 9.5% of adolescents according to modified MS criteria proposed by the International Diabetes Foundation.15
Metabolic syndrome itself is associated with poor health and quality of life. A study of employees in a financial services firm reported that persons meeting the MS criterion for waist circumference were significantly more likely to report a perception of poor health (odds ratio, 1.59; P < 0.05) or have a limitation in work attendance.16
Metabolic syndrome is a multifactorial disease of considerable heterogeneity, and its diagnostic criteria are not clearly defined. The NCEP definition is easier to use in clinical practice because more complicated laboratory tests are not required. However, a simple, blood test-free test as a preliminary screening method may be preferable, considering the high prevalence and the ominous consequences of MS. Simple, noninvasive screening is highly desirable (1) in areas where laboratory facilities are inadequate, (2) for patients who cannot afford laboratory work, and (3) in free clinics with inadequate financial resources. Patients with positive test results would go on to receive complete screening with laboratory values, whereas those who tested negative in the preliminary screen would not. The choice of a cutoff point would be made to balance the modest consequences of unnecessary laboratory screening against the grave consequences of a missed diagnosis of MS; the cutoff point would thus be slanted in favor of a greater number of false-positive test results.
This study was undertaken to compare the diagnostic performance of a single measurement of weight, waist circumference (WC), body mass index (BMI), and waist-to-hip ratio (WHR) and their combination to that of the NCEP criteria among high-risk patients in a public teaching hospital.
MATERIALS AND METHODS
A retrospective chart review was conducted on 928 consecutively registered outpatients from 2 primary care clinics in an urban Louisiana area in July 2002. These patients are predominantly women, of African descent, overweight, and indigent. The characteristics of men and women included in this study are shown in Tables 1 and 2. Height and weight were measured with patients wearing one layer of light clothing without shoes; the WC measurement was made at a point midway between the umbilicus and the iliac crest. Data were obtained for 928 patients. One or more laboratory measurements were missing for 208 patients; of these, the calculated low-density lipoprotein (LDL) value of 6 men and 20 women was missing because the triglyceride measurement was greater than 400 and the LDL was therefore not calculable. The data recorded are standard clinical and anthropomorphic data for all patients in the clinic. The diagnosis of MS was made according to NCEP criteria.1This standard was chosen because it is the most widely used definition of MS.
STATISTICAL ANALYSIS
Standard epidemiologic measures (sensitivity [SE], specificity [SP], positive predictive value, and negative predictive value) were calculated using 2 × 2 tables for various cutoff values centered around the NCEP cutoffs. The same calculations were also performed for WHR and BMI, calculated as weight (in kilograms) divided by the square of the height (in meters). These values were used to generate receiver operator characteristic (ROC) curves for each NCEP parameter and for WHR and BMI; the area under the curve was calculated for each ROC curve. Receiver operator characteristic curves depict the relationship between true-positive (SE) and false-positive (1 − SP) results, and the curves for different parameters can be compared by means of the area under the curves; the area can fall in the range 0.5 to 1.0, with larger areas indicating a larger fraction of correctly classified cases. The best cutoff points were identified as those having the optimal combination of SE and SP for our study population with its own prevalence of the metabolic syndrome. First, a range of cutoff points of WC, BMI, WHR, and high-density lipoprotein (HDL) covering the recommended cutoff values by NCEP or the World Health Organization were tested as single predictors. Logistic regression was also performed, with the presence of MS as the dichotomous response variable and measurements of the NCEP criteria as continuous independent variables (effects). All statistical analyses were performed using the SAS v.8 software (SAS Institute, Cary, NC).
RESULTS
No combination of parameters performed better than the best single parameter in either men or women. High-density lipoprotein curves for the best performing parameters are shown in Figures 1 and 2. The individual test results are shown in Tables 3 to 6. The most promising result was that in men. A single measurement of WC seemed to give high sensitivity (0.75 at a cutoff of 40 in) with acceptable specificity (0.86). This was the best performance for all parameter cutoffs in men. The various combinations of parameters did not improve the diagnostic performance.
When evaluating the components of a classification scheme by comparing individual factors to the classification determined in the same population, circularity in the analysis can be removed by performing logistic regression, a type of multivariable linear regression in which the outcome is dichotomous and represented by a dummy variable (0 = absent, 1 = present). P values for each factor indicate which factors do not contribute, and regression coefficients indicate the relative strengths of the contributing factors. Logistic regression in our sample identified WC as the most significant factor in both men and women (P < 0.001 in both), with systolic blood pressure a significant contributor in men (P = 0.0042) and systolic pressure (P = 0.0130) and HDL (P = 0.0128) in women. Despite a statistically significant contribution to the diagnosis in women, this parameter did not have an acceptable combination of SE and SP at any cutoff point.
The cutoff for blood glucose concentration was recently changed from 110 to 100 mg/dL. A reanalysis of a portion of the data resulted in 2 men and 13 women being reclassified from MS negative to MS positive. This caused changes in SE and SP that were generally 0.03 or less for men and 0.02 or less for women, and area under the ROC curves changed minimally. Accordingly, neither the cutoff points for the parameters nor our recommendations were altered.
DISCUSSION
Conclusions
In this patient population, WC proved to be a simple, convenient, an cost-effective screening tool for the diagnosis of the metabolic syndrome, particularly in men. Men whose WC exceeded the cutoff point of 40 in completed the screening with laboratory tests, with approximately 10% undergoing blood sampling unnecessarily. Of the patients with MS, 81% tested positive, and 86% of those without MS tested negative. The results shown in Table 4 illustrate the dynamic inverse relationship between SE and SP and the direct relationship between SE and negative predictive value (the probability that the test is negative when the disease is absent) as the cutoff value changes.
Given the elevated risks of DM-2 and cardiovascular disease in patients with MS and emerging evidence that reducing the number of MS features present reduces risk of these disorders, identification of patients at high risk is vitally important. The NCEP guidelines are the most convenient and acceptable criteria. However, because the NCEP criteria still require blood testing, simple and noninvasive screening methods are highly desirable, especially for indigent patients who may not be able to afford laboratory studies and in locations where laboratory facilities are inadequate or absent.
In choosing a cutoff point, gains in SE are more desirable than gains in SP as long as the loss of SP is not too great. Cutoff points allowing high SE and high negative predictive value should be chosen for screening purpose, as the consequences of a false-positive result would only be further and perhaps more invasive and expensive testing, whereas the consequences of a false-negative result would likely be a missed opportunity for treatment and perhaps devastating health consequences. Given these considerations and our results, routine measurement of WC may be indicated for health maintenance. Waist circumference and BMI may be better indicators for men than for women, and HDL may be more useful in women than in men.
Our results show that in men, a single measurement of WC has high SE and acceptable SP to define MS; this measurement should be routinely collected when other patient data such as weight and height are collected. This does not suggest replacing complete profiling of patients when laboratory facilities and adequate funds are available; rather, its suggested use is as a no-cost preliminary screen for men in resource-embarrassed clinics or in developing countries.
Weaknesses of the Study
Since WC is a component of the MS definition, inclusion of WC as a diagnostic test for MS is subject to some recursive error. However, this effect is not completely represented in any patient population; a patient may be classified as having MS without satisfying the WC criterion if he satisfies 3 other criteria, and similarly, a patient with a huge WC may not be classified as having MS if only one of the other criteria is satisfied. There are thus a potentially large number of patients in any MS group without recursive influences. We believe that our analysis explored the contribution of MS criteria taken one at a time and therefore contributes useful information.
Generalizability
Similar analyses could be undertaken in medical practices with different anthropomorphic and clinical characteristics. Cutoff points would vary in different populations as would SE, SP, and predictive values. Before use in any clinic setting, our findings should be validated for the individual patient population.
The variation of anthropomorphic parameters should be taken into consideration among different ethnic and sex groups in setting the thresholds and determining their diagnostic values. We did not study ethnic differences in our population because controlling for both sex and race would have compromised statistical power. However, epidemiologic evidence suggests different relationships of obesity to MS between persons of African and Caucasian descent.17This should be studied in a larger sample of patients.