The Relationship Between Waist Circumference and Biomarkers for Diabetes and CVD in Healthy Non-Obese Women. The Pensacola Study

Kristina Jackson Behan, PhD, MT(ASCP); Justice Mbizo, DrPH


Lab Med. 2007;38(7):422-427. 

In This Article

Materials and Methods


Two hundred women between the ages of 40 and 54 were recruited by newspaper, television, and radio advertisements in the Pensacola, Florida area between November 2004 and May 2005. Participants were interviewed to determine their health status, race, age, and menopause status. Three women had a previous hysterectomy and were not aware of their menopause status; these women were grouped with the postmenopausal women. Participants removed heavy clothing, belts, and shoes before measurements were made. Two pounds (0.91 kg) were subtracted to account for clothing weight. Body mass index was calculated as weight in kg/height in m.[2] The waist measurement was taken above the iliac crest at the natural waistline, and the hip measurement was taken at the largest area of the natural hip line. One inch was subtracted from the waist and hip measurements to account for clothing. Participants had temperature and blood pressure readings taken; readings above 135 systolic or 85 diastolic were repeated in the other arm. Hypertension was defined as blood pressure greater than 135 systolic or greater than 85 diastolic, or a history of hypertension whether or not treated with medication. All measurements were verified with the subjects as a double check to determine if they were expected. All of the subjects had normal body temperature readings the day of the blood collection, and all of them stated that they were healthy.

Participants were excluded if they were outside the specified age range, obese (BMI ≥ 30), had a waist circumference of 39.4 inches (100 cm) or more, were a known diabetic, had an FPG greater than or equal to 126 mg/dL (7.0 mmol/L), or if they took medication to lower their cholesterol. One hundred eighty-two women remained. All of the subjects were in good health. This study was approved by the Institutional Review Board at the University of West Florida; all subjects provided informed consent.

Laboratory Tests and Precision

Subjects were instructed to fast for 10 to 12 hours, and blood was collected in the morning. Blood for glucose and lipid analysis was collected in serum separator tubes and centrifuged within 60 minutes of collection. Automated chemistry analysis was performed on a Dade RXL (Dade Behring, Deerfield, IL), specifically glucose by Hexokinase G-6-PD method, and triglycerides and total cholesterol by enzymatic methods. High-density lipoprotein and LDL cholesterol assays were direct homogeneous methods using accelerator selective detergent methodology. Hemoglobin A1c was collected in ethylendiaminetetraacetate (EDTA) and was performed by high-performance liquid chromatography (HPLC) on a Tosoh A1c 2.2+ (Tosoh Bioscience, San Francisco, CA). High sensitivity C-reactive protein was performed by nephelometry on a BN ProSpec system (Dade Behring, Deerfield, IL). White blood cell count was performed on a Beckman Coulter HMX (Beckman Coulter, Fullerton, CA). All assays were performed at West Florida Hospital, Pensacola, Florida within 1 day of collection by a qualified technical staff member. The between-run imprecision (CV) for glucose, A1c, hs-CRP, total cholesterol, LDL, HDL triglycerides, and WBC were 1.7%, 2.9%, 5.8%, 2.1%, 2.7%, 4.0% 2.9%, and 2.2%, respectively. Non-HDL was calculated as the total cholesterol minus the HDL.


The Statistical Package for Social Sciences (SPSS version 13.0) was used for statistical analysis of the data. Chi square analysis was performed on all dichotomous variables, and a Pearson coefficient was computed. A 0.05 level of significance was established a priori. All P values reported are 2-sided. Variables included postmenopausal status (yes/no), exercise 30 minutes per day (yes/no), current cigarette smoker (yes/no), hypertension (yes/no), and use of hormone replacement therapy (yes/no). Since most of the participants were Caucasian, race was dichotomized as Caucasian (yes/no). Continuous variables were analyzed by general linear model with Tukey as the post-hoc test.

Tests for linear trend across the quartiles of waist circumference were conducted by correlating the median of each variable to its quartile and are reported as the Pearson correlation with 2-tailed significance.

The risk factor total for diabetes was determined as the sum of these risks: age ≥ 45, BMI ≥ 25, family history of diabetes, history of gestational diabetes or birth of a baby over 9 lbs, hypertension whether treated or not, fasting triglycerides > 250 mg/dL or HDL < 35 mg/dL, race other than Caucasian, fasting glucose ≥ 100 mg/dL, and CRP > 6.1 mg/L5. Proportions were compared by chi-square analysis, and a P value ≤ 0.05 was considered significant.

Correlations between waist circumference and biomarkers were computed with Pearson coefficients, and are presented as crude correlations, adjusted for menopause status (model 1), or adjusted for menopause, HRT use, exercise, smoking, and presence of hypertension (model 2).


One hundred eighty-two middle aged women qualified for the study, with an average age of 47.6 years. They ranged in weight from 87 to 196 lbs, had a BMI between 16.5 and 29.9, and had a waist-to-hip ratio (WHR) of 53.7% to 88.8%. They were predominantly Caucasian (93.4%), with 2.2% Asian American, and 6.6% African American. Almost half reported exercising at least 30 minutes per day (45%); 40.1% were post menopausal or reported a past hysterectomy. About a quarter of them had hypertension (21.4%), 7.6% were current smokers, and 14.8% used some form of hormone replacement therapy (HRT). None of the subjects were pregnant at the time of the study. Waist circumference ranged from 22 to 36 inches (55.9 to 91.4 cm). Waist circumference correlated strongly to BMI (r = 0.827, P = 0.000). Participants were categorized by quartile of WC as shown in Table 1 . The WC quartiles were significantly different for BMI, WHR, the percent that exercise regularly, and the presence of hypertension. They were similar with respect to age, menopausal status, race, smoking status, and the use of HRT.

Risk Factors for Diabetes

Participants were tested for biochemical markers of glycemic status and inflammation, shown in Figure 1. Fasting plasma glucose showed a steady increase across the quartiles (P for trend 0.008). Hemoglobin A1c showed no difference between the groups (P = 0.122). High-sensitivity CRP increased across the quartiles (P for trend 0.03), and all other quartiles were significantly different than Q4. White blood cell count was not different between the groups (P = 0.833).

Figure 1.

Waist circumference correlated with glucose and with CRP (shown in Table 2 ), and the correlation remained significant after adjustment for menopause alone (model 1) or in combination with the use of HRT, exercise, smoking status, and the presence of hypertension (model 2). Waist circumfrence did not correlate with hemoglobin A1c or WBC. The WC quartiles showed a statistically-significant difference with respect to the total risk factors for diabetes, shown in Figure 2 (c2 = 69.146, P = 0.000). At least half of the subjects in quartiles 1, 2, and 3 were at low risk (total of 0 to 1) for diabetes, but only 5% of the subjects in quartile 4 were low risk.

Figure 2.

Risk Factors for Cardiovascular Disease

Triglycerides, total cholesterol, LDL, and HDL were measured directly and each showed a statistically-significant difference between the WC quartiles (Figure 3). Quartile 4 was significantly different from every other quartile for triglycerides level, and almost double the value of quartile 1. The average cholesterol level was below the recommended level of 200 mg/dL only for quartile 1. While quartiles 2 and 3 showed similar total cholesterol values, the composition of that cholesterol was different. Low-density lipoprotein increased with increasing WC quartiles (P for trend 0.034). High-density lipoprotein cholesterol was significantly lower for quartile 3 compared with quartile 2, which resulted in a 23% difference in the average cholesterol/HDL ratio between quartiles 2 and 3. Quartile 4 had the lowest HDL of all of the quartiles and the highest results for the remaining lipids and calculations.

Figure 3.

The correlation between waist circumference and lipids was strong and did not change significantly when adjusted for menopause or other confounders ( Table 2 ). The strongest correlations were found between WC and triglycerides (r = 0.465, P = 0.000) and between WC and cholesterol/HDL (r = 0.321, P = 0.000).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.