Secondary Logo

Journal Logo

Insulin Resistance in HIV-Infected Men and Women in the Nutrition for Healthy Living Cohort

Jones, Clara Y MD, MPH*; Wilson, Ira B MD, MSc; Greenberg, Andrew S MD‡§; Shevitz, Abby MD, MPH*∥; Knox, Tamsin A MD*; Gorbach, Sherwood L MD*∥; Spiegelman, Donna PhD; Jacobson, Denise L PhD*; Wanke, Christine MD*∥

JAIDS Journal of Acquired Immune Deficiency Syndromes: October 1st, 2005 - Volume 40 - Issue 2 - p 202-211
doi: 10.1097/01.qai.0000165910.89462.2f
Epidemiology and Social Science
Free

Objective: We evaluated insulin resistance (IR) in an HIV-infected cohort and compared our results with those of the National Health and Nutrition Examination Survey III (NHANES III).

Methods: Using a cross-sectional study design, we determined the Quantitative Insulin Sensitivity Check Index (QUICKI) in 378 nondiabetic participants in the Nutrition for Healthy Living (NFHL) study and evaluated the association of the QUICKI with demographic, socioeconomic, body composition, lipid, liver function, HIV-associated factors (CD4 cell count, viral load, highly active antiretroviral therapy type, and years infected), and injection drug use. The prevalence of IR (QUICKI <0.350) and the mean QUICKI were ascertained for nondiabetic persons aged 25 to 65 years in the NHANES III and compared with those in the NFHL study.

Results: Protease inhibitor (PI) highly active antiretroviral therapy (HAART) and nonnucleoside reverse transcriptase inhibitor (NNRTI) HAART were associated with worse IR in HIV-infected men. Greater waist circumference, triglycerides, age, and alanine aminotransferase were associated with worse IR, and higher high-density lipoprotein, low-density lipoprotein, and smoking were associated with less IR in the NFHL study; CD4 cell count, viral load, and years HIV infected were not associated with IR. There was no significant difference in the prevalence of IR in the NFHL study versus the NHANES III (51% vs. 47%; P = 0.27). NFHL participants were not more IR than NHANES III participants.

Conclusions: IR in the NFHL study was quite common but not significantly different than in the NHANES III and was associated with similar factors as in the general population. PI HAART and NNRTI HAART were associated with worse IR in men.

From the *Department of Public Health and Family Medicine, Tufts University School of Medicine, Boston, MA; †Institute for Clinical Research and Health Policy Studies, Tufts-New England Medical Center, Boston, MA; ‡Department of Endocrinology, Tufts-New England Medical Center, Boston, MA; §Jean Mayer-US Department of Agriculture Human Nutrition Research Center on Aging at Tufts University, Boston, MA; ∥Geographical Medicine/Infectious Diseases, Tufts-New England Medical Center, Boston, MA; and ¶Departments of Epidemiology and Biostatistics, Harvard School of Public Health, Boston, MA.

Received for publication November 18, 2004; accepted March 30, 2005.

Supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant 1p01DK45734-06, NIDDK grant 1p01DK45734-01A2, General Clinical Research Center grant M01-RR00054, National Institutes of Health (NIH)/National Institute of Allergy and Infectious Diseases grant 1 K24 AI055293 (C. Wanke), and NIH grant 5 P30 AI42853 (Center for AIDS Research).

Reprints: Clara Y. Jones, Division of Nutrition and Infectious Disease, Department of Public Health and Family Medicine, Tufts University School of Medicine, 200 Harrison Avenue, Posner 4, Boston, MA 02111 (e-mail: clara.jones@tufts.edu).

Many studies have shown that HIV-infected populations have high rates of insulin resistance (IR), a condition characterized by decreased responsiveness of muscle, liver, and adipose tissue to the action of insulin.1,2 IR may progress to glucose intolerance or frank diabetes mellitus (DM) and is associated with increased risk of atherosclerosis. A higher prevalence of IR and/or diabetes has been reported in HIV-infected persons who are antiretroviral treated versus nontreated,3 those on highly active antiretroviral therapy (HAART) versus non-HAART,4 those treated with protease inhibitor (PI) regimens versus those on nonnucleoside reverse transcriptase inhibitor (NNRTI) regimens,5,6 lipodystrophic versus nonlipodystrophic HIV-infected persons,2,6 and HIV-infected persons versus healthy controls.2,5-9 One group has reported a higher risk of diabetes among HIV-infected patients compared with noninfected patients.10

Some investigators have proposed that IR in HIV-infected persons may be attributable to specific HIV therapies such as indinavir,1 increased accumulation of visceral fat and decreased peripheral fat, cytokines, and even HIV infection per se.6 Indinavir is clearly associated with decreased insulin sensitivity in HIV-negative volunteers,11 and HAART initiation in HAART-naive participants has been associated with worsened insulin sensitivity.12

We undertook a cross-sectional analysis of the prevalence of IR and the associations of IR with sociodemographic factors, body composition, and HIV-specific factors in nondiabetic HIV-infected male and female participants enrolled in the Nutrition for Healthy Living (NFHL) study and compared the findings of the NFHL study with cross-sectional data from the National Health and Nutrition Examination Survey III (NHANES III). We had 4 main study questions:

  1. What is the prevalence of IR in the NFHL study?
  2. Is the prevalence of IR similar in men and women in the NFHL study?
  3. Is HAART therapy associated with worsened IR in the NFHL study?
  4. How do clinical characteristics of IR in the NFHL study compare with those in the NHANES III?
Back to Top | Article Outline

METHODS

Study Population and Data Collection

Nutrition for Healthy Living Study

The NFHL study is a longitudinal cohort study of nutrition and HIV outcomes in HIV-infected adults (18 years of age or older), with study centers in Boston, Massachusetts and Providence, Rhode Island. Study methods have been published previously.13 At baseline and 6-month intervals thereafter, questionnaires, laboratory studies, bioelectrical impedance analysis (BIA), and anthropometry measurements were obtained. Fasting insulin levels, fasting glucose levels, and waist and hip circumferences were obtained starting in September 2000. The protocol was approved by the Human Investigations Research Committee (HIRC) at the New England Medical Center in Boston, Massachusetts and at Miriam Hospital in Providence, Rhode Island, and written informed consent was obtained from each participant.

We selected the most recent eligible visit between September 1, 2000 and May 31, 2003 for each participant using the following inclusion criteria: no prior history of DM, fasting insulin level was available, fasting glucose level was <126 mg/dL, and no data were missing for important covariates (waist circumference, suprailiac skin fold thickness, lipid levels, CD4 cell count, viral load, and current HAART history). Thirteen persons were missing data on alanine aminotransferase (ALT), 5 were missing data for HAART duration, and 4 were missing data for injection drug use (IDU). The missing indicator method was used to impute values for these participants.14 There were 500 participants without a prior history of diabetes seen in clinics between September 1, 2000 and May 31, 2003, and 435 of these participants had fasting insulin levels available. Four were excluded because of fasting glucose levels ≥126 mg/dL, and 53 were excluded because of missing data. Thus, 378 participants were included in the NFHL study cross-sectional analysis. In these participants, route of HIV infection was ascribed to men having sex with men (MSM) in 46% of cases, intravenous drug use in 23%, MSM with intravenous drug use in 3%, heterosexual transmission in 20%, blood transfusion or hemophilia in 1.5%, and undetermined or other in 7%.

Back to Top | Article Outline

National Health and Nutrition Examination Survey III

Data from NHANES III were downloaded from the Centers for Disease Control and Prevention (CDC) web site to obtain a population-based comparison group.15 In the NHANES III, 29,314 persons had blood work performed. We applied the following inclusion criteria: age between 25 and 65 years; no history of diabetes or report of taking diabetic medications; no missing values for body mass index (BMI), waist circumference, waist/hip ratio (WHR), glucose level, or insulin level; and the participant fasted 5 or more hours before the blood drawing with a fasting glucose level <126. Participants with extreme values for waist circumference (>147 cm) or fasting insulin (>200 μL/mL) were excluded because of concern that inclusion of those outliers would unduly influence the comparison of the NFHL study and NHANES III. These criteria yielded 8312 participants available for analysis.

Back to Top | Article Outline

Analytic Variables

Outcome: Insulin Resistance

We evaluated several indices of IR as possible outcome measures, including the Quantitative Insulin Sensitivity Check Index (QUICKI),16 Homeostasis Model Assessment (HOMA),17 and fasting insulin.7,9 For this study, we use the QUICKI as the measure of insulin sensitivity. The QUICKI decreases with worsened IR and is defined as follows:

Serum glucose and serum insulin values were used to calculate the QUICKI for the NHFL study and NHANES III.

The QUICKI correlates as well or better with insulin sensitivity measured by the euglycemic hyperinsulinemic clamp compared with fasting insulin or the HOMA.16,18-23 In Hrbicek et al's study,24 a QUICKI <0.357 identified a population with significant differences in fasting insulin, glucose, uric acid, triglycerides, high-density lipoprotein (HDL), and BMI compared with those with a QUICKI >0.357. Gokcel et al25 found that among nondiabetic Turkish adults, a QUICKI <0.347 distinguished a population with significantly higher waist circumferences, BMI, triglycerides, fasting glucose, and fasting insulin and significantly lower HDL compared with those having a QUICKI ≥0.347. Uwaifo et al19 studied 31 normal glucose-tolerant black and white children (81% were obese) and found a mean QUICKI of 0.354 ± 0.042. Rabasa-Lhoret et al26 found a mean QUICKI of 0.364 ± 0.005 in their normal controls. Mean QUICKI levels in normal populations can range from 0.366 to 0.389.16,24 For this study, we used a QUICKI ≤0.350 to define IR.

Back to Top | Article Outline

Questionnaire-Based Variables

For linear regression models of NFHL, age was entered as a continuous variable. For the NFHL study-NHANES III comparison, age was grouped into <40 years and ≥40 years. Race was self-categorized as black, Hispanic, white, or other, and gender was classified as male or female. Education was coded 1 through 7 as follows: no schooling, grades 1 through 6, grades 7 through 11, high school/General Educational Development (GED), some college, college degree, and graduate school. Overall, 56% of this cohort had more than a high school/GED education. Personal income was coded 1 through 7 as follows: <$5000 per year, $5000 through $10,000 per year, $10,000 through $20,000 per year, $20,000 through $30,000 per year, $30,000 through $40,000 per year, $40,000 through $50,000 per year, and >$50,000 per year. Overall, 70% of this cohort had a personal income of $20,000 or less. Use of injection drugs, heroin, or cocaine in the past 6 months was assessed by self-report.

The variable of years known to be HIV-positive was determined by self-report. Current HAART group was defined as follows: (1) PI-based HAART: a minimum of 2 PIs or 2 nucleoside reverse transcriptase inhibitors (NRTIs) with 1 PI, (2) NNRTI-based HAART: 2 NRTIs with 1 NNRTI, (3) mixed HAART: a combination of a PI and an NNRTI with an NRTI, (4) triple NRTI HAART: 3 NRTIs, or (5) no HAART: persons not receiving any of these combinations. Ever on HAART was defined as a self-report of having taken a HAART combination during the year before enrollment in the NFHL study or at any time during follow-up, and HAART duration was defined as cumulative months of any reported HAART use.

The association of the following medications with the QUICKI was evaluated in univariate regression: indinavir, saquinavir, nelfinavir, ritonavir, Kaletra, lamivudine, azidothymidine, stavudine, zalcitabine, didanosine, abacavir, delavirdine, efavirenz, nevirapine, and tenofovir. Indinavir approached significance and was evaluated in multivariate models.

Back to Top | Article Outline

Laboratory Variables

All blood work was drawn after at least 5 hours of fasting. Insulin was measured using a radioimmunoassay (RIA) technique (Immulite 2000). The following laboratory variables were analyzed using a chemistry analyzer (Beckman LX-20) in the New England Medical Center clinical laboratory: glucose, total cholesterol, ALT (GPT), HDL, and triglycerides. Low-density lipoprotein (LDL) was calculated using the Friedewald equation. Triglyceride level was entered into models as a three-level variable (≤100, 101-149, or ≥150 mg/dL). CD4 lymphocyte counts were determined using a specific monoclonal antibody and fluorescence-activated cell sorter (FACS) analysis. HIV RNA level was measured with the Roche Amplicor Monitor reverse transcriptase polymerase chain reaction (PCR) assay (Roche Molecular Systems, Somerville, NJ), with a lower detection limit of 400 copies/mL. Log10 transformation of viral load was used in analyses.

Back to Top | Article Outline

Body Composition

BMI was calculated as weight in kilograms per height in square meters. Participants were considered overweight if their BMI was ≥25 kg/m2 and obese if their BMI was ≥30 kg/m2.27

Waist and hip circumferences were measured, and WHRs were calculated. Waist circumference was used in the regression analysis because it is a better indicator of central adiposity than the WHR in a population with peripheral fat atrophy. Four waist circumference groups were generated using gender-specific cutoffs: group 1 included women with a waist circumference ≤81 cm and men with a waist circumference ≤90 cm, group 2 included women with a waist circumference >81 cm but ≤ 88 cm and men with a waist circumference >90 cm but ≤102 cm; group 3 included women with a waist circumference >88 cm but ≤95 cm and men with a waist circumference >102 cm but ≤110 cm; and group 4 included women with a waist circumference >95 cm and men with a waist circumference >110 cm. Participants had truncal obesity if the measurement was >88 cm for women and >102 cm for men.28

Suprailiac skin fold thicknesses were obtained in triplicate,29 and the averaged value was used in the analysis. Fat mass and fat-free mass were determined by BIA using a 2-compartment model (equations by the method of Lukasi,30 with equipment from RJL Systems, Clinton Twp, MI). In a small percentage of participants for whom BIA was not available, fat mass and fat-free mass were determined by anthropometry.31

Back to Top | Article Outline

Dietary Variables

At enrollment, a nutritionist trained in the study protocol instructed each participant on keeping a 3-day food record (FR), including 1 weekend day. Subsequently, a completed FR was reviewed with a trained nutritionist at each study visit. Nutrient calculations for the FR were performed using the Nutrition Data System (NDS) software, version 2.92, developed by the Nutrition Coordinating Center, University of Minnesota (Minneapolis, MN).32 Energy intake was divided by kilograms of body weight and entered into models as kilocalories per kilogram.

Back to Top | Article Outline

Statistical Methods

Insulin Resistance in the Nutrition for Healthy Living Cohort

Demographic and clinical variables were compared within the NFHL cohort: insulin-resistant versus normal QUICKI, women versus men, and ever HAART use versus never HAART use. The χ2 test was used to compare categoric characteristics, and ANOVA (or the Kruskal-Wallis test) was used for continuous characteristics.

Multivariate linear regression models were used to determine the independent correlates of insulin sensitivity as measured by the QUICKI. The covariate of interest was HAART category. The model began with variables that were significantly associated with the QUICKI in univariate analysis or important potential confounders: gender, age, race, waist circumference group, fat mass, suprailiac skin fold, fat-free mass, HDL, LDL, triglycerides, years HIV infected, CD4 cell count, log viral load, current HAART category, cumulative HAART duration, serum ALT, IDU in the past 6 months (yes/no), smoking history, education, income, and energy intake. Initial univariate analysis using gender-specific data sets suggested an interaction of gender with HAART, log viral load, waist circumference, ALT, and IDU; thus, gender interaction terms were also entered into the combined gender multivariate model. This model was successively simplified by eliminating variables with Wald P values >0.05 that were not confounders of the relation of HAART categories to the QUICKI. The final model included HAART group, gender, age, HAART group*gender, waist group, triglycerides, HDL, LDL, ALT, and smoking. For ease of interpretation, results are presented as the estimated difference in the QUICKI for a person on each HAART strategy versus a person not on HAART who is the same with respect to all other covariates.

Back to Top | Article Outline

Insulin Resistance in the National Health and Nutrition Examination Survey III Cohort

Means from the NHANES III data set were obtained using SAS Proc SurveyMeans (SAS version 9.1; SAS Institute, Cary, NC), and frequencies were obtained using SAS Proc SurveyFreq using weights to obtain the correct estimate and accounting for the complex sampling methodology to obtain the correct variances.15 The weighted values thus obtained constitute nationally representative estimates for similar persons in the civilian noninstitutionalized adult US population.

Back to Top | Article Outline

Comparison of the Nutrition for Healthy Living Study and National Health and Nutrition Examination Survey III

Statistical comparison of the NFHL study and NHANES III used SAS Proc SurveyReg. The 2 studies were combined by assigning a unique stratum number for the NFHL study and setting NFHL participant identification to the primary sampling unit (PSU) with a weight of 1 with a binary indicator for “study” (1 = NFHL study, 0 = NHANES III) as the covariate of interest. Because information on HIV infection in the NHANES III subjects is not available and the population prevalence of HIV infection among civilian noninstitutionalized US adults was quite low during the time frame of the NHANES III, all NHANES III subjects were considered to be HIV-negative. Differences in QUICKI levels between the 2 studies were adjusted for age (>40 or <40), gender, race, waist circumference group, HDL, LDL, triglycerides, ALT, and smoking (never, prior, or current). Thus, the P value for the study covariate assesses the statistical significance of any difference observed in relation to HIV infection status. Analysis was also conducted using the specific HAART therapy categories and adjusting for the same confounders to assess whether subjects in the NFHL study on a particular type of HAART were more IR compared with subjects in the NHANES III. Analysis was repeated using Proc Survey Logistic (SAS Institute) for the dichotomous outcome of IR (1 = QUICKI ≤0.350, 0 = QUICKI >0.350). Statistical analyses used SAS version 9.1.

Back to Top | Article Outline

RESULTS

Nutrition for Healthy Living Study

Prevalence of Insulin Resistance by Different Criteria

The prevalence of IR in the NFHL study varied by the criteria used to define it (21% of participants were IR by a fasting insulin level ≥15 μU/mL, 51% by a QUICKI level ≤0.350, and 17% by a HOMA index >3.5).

Back to Top | Article Outline

Stratification by Quantitative Insulin Sensitivity Check Index Status (insulin resistant vs. normal)

To look for independent predictors of the QUICKI, we stratified by IR status (Table 1). A higher percentage of IR participants were female (28% vs. 20%) and Hispanic (15% vs. 8%). BMI, waist circumference, WHR, fat mass, glucose, and triglycerides were significantly higher and HDL and LDL were significantly lower in the IR subjects. There was no significant difference in CD4 count, log viral load, years known to be HIV infected, cumulative months of HAART, current indinavir use, age, percentage with IDU in the past 6 months, lean body mass, education, or income between the IR and non-IR participants.

TABLE 1

TABLE 1

Back to Top | Article Outline

HAART Use in Relation to Insulin Resistance

Fifty-four persons reported never having been on a HAART regimen, and 324 were currently (n = 289) or had previously taken HAART (n = 35) (Table 2). The mean (SD) HAART duration in those ever using HAART was 42 ± 22 months. Those never on HAART had a lower WHR as well as lower cholesterol, LDL, and triglyceride levels and a higher log viral load (3.8 vs. 2.8) compared with those currently on HAART. The mean fasting insulin and QUICKI levels were not significantly different between the 3 groups, however. When we compared those currently on HAART (n = 289) with those previously but not currently taking HAART (n = 35), glucose values were slightly higher in those currently on HAART, although still within the normal range (81 vs. 77 mg/dL; P = 0.02). There was no significant difference in mean fasting insulin (12 vs. 10 μU/mL, P = 0.37), QUICKI (0.349 vs. 0.359; P = 0.10), or HOMA (2.5 vs. 1.9; P = 0.26) values. Cholesterol was significantly higher in those currently on HAART versus those previously on HAART (P = 0.005). There was no significant difference in LDL, HDL, or triglycerides in those currently versus previously on HAART (P = 0.10, P = 0.09, and P = 0.06, respectively).

TABLE 2

TABLE 2

The 28 participants currently taking indinavir had a QUICKI of 0.341 vs. 0.351 among those not on indinavir (P = 0.14), and there was no significant difference in the HOMA (2.56 vs. 2.46; P = 0.87) or fasting insulin (13 vs. 12; P = 0.70) values. Other PI, NNRTIs, and NNRTIs tested were not significantly associated with the QUICKI.

Difference in the mean QUICKI in those in each HAART category versus those not currently on HAART were adjusted for the following independent predictors of QUICKI level: waist circumference group, age, triglycerides, HDL, LDL, ALT, interaction terms of gender with PI HAART and NNRTI HAART, and smoking (men are shown in Table 3). In men, PI HAART and NNRTI HAART were significantly associated with worsened IR compared with non-HAART users, with the most marked association seen with NNRTI-based HAART. None of the HAART strategies was significantly associated with worsened IR in women, but power was low, with only 32, 20, 6, and 6 women in each of the 4 HAART groups.

TABLE 3

TABLE 3

To address the possibility that inclusion of lipids in the model might attenuate results because changes in lipids could be involved in the causal pathway of an effect of HIV medication on IR, this model was run without including the lipid covariates (triglycerides, HDL, or LDL) and reintroducing CD4 count, log viral load, and years of known HIV infection. In men, PI HAART and NNRTI HAART were again associated with worse IR; in women, none of the HAART types was significantly associated with worse IR compared with no HAART. There was no significant association of CD4 count, log viral load, and years HIV-positive with the QUICKI (data not shown).

Back to Top | Article Outline

Comparison of Nutrition for Healthy Living Study With National Health and Nutrition Examination Survey III

Overall, there was no significant difference in the prevalence of IR between the NFHL study and NHANES III (51% vs. 47%; P = 0.27). Table 4 shows demographic characteristics and laboratory measurements for HIV-infected adults in the NFHL study and for adults in the NHANES III by gender. Of note, in the NFHL study, male participants were slightly older but had a lower BMI and lower mean waist circumference compared with male participants in the NHANES III. In the NFHL study, both genders had lower HDL, lower LDL, and higher ALT compared with participants in the NHANES III. Mean insulin was higher but mean glucose was lower for both genders in the NFHL study versus the NHANES III, and there was no significant difference in the mean QUICKI. Although women in the NFHL study were significantly more likely to have a QUICKI ≤0.350 compared with women in the NHANES III in univariate analysis, there was no significant difference after adjusting for race.

TABLE 4

TABLE 4

In multivariate analysis, after adjusting for gender, age, race, smoking, and waist circumference, there was no significant difference in the mean QUICKI for male or female participants in the NFHL study versus the NHANES III. When ALT was added to the model, men in the NFHL study tended to have less IR than men in the NHANES III (this was significant for black men as well as for “other” men). If the model is additionally adjusted for HDL, LDL, and triglycerides, participants of both genders had less IR in the NFHL study than in the NHANES III (this was significant for white men, black men and women, and other men; Table 5).

TABLE 5

TABLE 5

This analysis was repeated comparing the mean QUICKI of NFHL study participants in each HAART category (no HAART, PI, NNRTI, mixed PI/NNRTI, and triple NRTI) with that of those in the NHANES III. Overall, NFHL study participants with no HAART use had less IR than NHANES III participants, and those on PI, NNRTI, mixed PI/NNRTI, or triple NRTI were not significantly different than those in the NHANES III. After adjusting for gender, waist circumference, age, race, smoking, HDL, LDL, triglycerides, and ALT, NFHL study subjects had less IR or were not significantly different from those in NHANES III in every gender-race-HAART category stratum.

Back to Top | Article Outline

DISCUSSION

There was a high prevalence of IR among HIV-infected participants. This prevalence did not exceed the high prevalence found in nondiabetic persons in the general population, however. These results differ from those of several other investigators who have compared HIV-infected persons with HIV-negative controls. Carr et al6 compared HIV-infected men with healthy controls and found that HIV-infected PI users had significantly higher fasting insulin and HOMA values (although HIV-infected PI-naive men had similar HOMA and insulin values compared with controls). They did not describe how controls were selected. Walli et al5 compared HIV-infected persons (PI users and PI-naive persons) with controls using an intravenous insulin tolerance test and found lower insulin sensitivity and higher basal glucose among the PI users compared with the PI-naive HIV-infected persons and the controls. The median age and BMI of the PI users and the controls were not comparable (44 vs. 30 years and 24.0 vs. 22.3 kg/m2 for PI users and controls, respectively), however. Hadigan and her colleagues published 2 studies comparing HIV-infected persons with healthy controls. In the first, HIV-infected women were matched by weight with premenopausal controls. The HIV-infected women had significantly higher insulin and insulin-to-glucose ratios compared with the controls.33 In the second, HIV-infected men had higher fasting insulin and higher HOMA values than healthy controls.9 The HIV-infected persons in both studies were chosen for a history of wasting, however, and there is no description of how the controls were recruited. Bruno et al8 compared HIV-infected persons on HAART with age- and gender-matched blood donors (but did not match for BMI). They found that fasting insulin, fasting glucose, and HOMA values were all significantly higher in the HIV-infected persons. Blood donors tend to be healthier than the general population. In addition, the mean waist circumference of 86 cm in their predominantly male controls suggests that these blood donors were quite lean. In a third study, Hadigan et al2 compared HIV-infected persons with and without lipodystrophy with healthy control subjects matched for age, sex, and BMI. These controls were selected at a 3:1 ratio from the Framingham Offspring Study and thus represent a more population-based control group. They found that the HIV-infected patients with lipodystrophy had significantly higher fasting insulin and HOMA values as well as a higher prevalence of impaired glucose tolerance on oral glucose tolerance testing compared with their healthy controls. The HIV-infected persons without lipodystrophy did not differ significantly from controls in prevalence of hyperinsulinemia or impaired glucose tolerance.

Our results may differ from these published studies for several reasons. Selection of appropriate controls is vital to answering the question as to whether HIV infection is associated with increased prevalence of IR compared with that in the general population. In most of the studies cited previously, there was either no description of how controls were selected, or they were selected from a pool that is likely to be healthier than the general population (eg, blood donors), which could bias a comparison. In addition, these published studies have shown that HIV-infected persons with lipodystrophy or those taking PI-based HAART regimens had significantly worse IR compared with controls, whereas those without lipodystrophy or those who were PI naive did not. Our NFHL study cohort may have a lower prevalence of lipodystrophy. Only 17% of our participants reported body shape changes within the prior 3 years that suggest lipoatrophy or lipoaccumulation. Nevertheless, anthropometric analysis by Jacobson et al (manuscript submitted for publication) suggests that the baseline prevalences (male vs. female) of lipoatrophy, central fat deposition, and combined lipoatrophy-fat deposition in our cohort are 38% versus 26%, 40% versus 53%, and 15% versus 13%, respectively. These prevalences are in the range of those seen in other populations that have not been specifically selected for lipodystrophy.34-40

There may also be an effect of shifting medication regimens over time. Our participants were taking a broad range of different antiretroviral regimens, with less than 10% currently taking indinavir. Although it did not reach statistical significance, those on indinavir did have a lower QUICKI. This is consistent with the considerable literature linking indinavir to worsened IR in both HIV and HIV+ individuals. Many of these prior studies were conducted during a period when indinavir use was more prevalent.2 NNRTI-based regimens have not been associated with worsened IR to the same degree as indinavir-based regimens. Nevertheless, our data suggest that currently used NNRTI therapies are associated with greater IR than currently used PI therapies in men. As a cross-sectional study, it is not possible to tell if the association between NNRTI use and IR in men is causal or a consequence, because HIV providers may be increasingly taking family histories or personal predispositions to diabetes into account in prescribing PI-sparing regimens. It is clear that it is no longer appropriate to lump all HAART treatments together in epidemiologic analyses.5,6,8,9

The estimated prevalence of IR varied substantially depending on the specific epidemiologic measure used. In several studies that have performed direct comparisons in non-HIV-positive populations,16,19-23 the QUICKI has had better correlation with the “gold standard” hyperinsulinemic euglycemic clamp than the HOMA index or fasting insulin level. Other studies have shown equivalent correlation of the HOMA and QUICKI with the euglycemic clamp.18 It is not clear whether one of the epidemiologic indices (QUICKI or HOMA) or the fasting insulin level best correlates with the hyperinsulinemic euglycemic clamp among HIV-infected populations.

IR in the HIV-infected cohort was associated with similar factors as those reported in the general population (eg, waist circumference, triglyceride level, HDL, ALT) as well as with smoking. Characteristics of HIV disease, namely, years of known HIV infection, CD4 cell count, and viral load, were not significantly associated with IR. Although HAART therapy was significantly associated with IR in men, with NNRTI-based HAART more adversely associated than PI-based HAART, it was not significantly associated with IR in women. Further study is required to ascertain whether there is a true gender difference in the effect of medications or whether this reflects lack of power to detect the effect of HAART among HIV-infected women, given a smaller sample size.

Race, surprisingly, was not a significant predictor of IR among the NFHL study participants, although it was a strong predictor for the NHANES III participants. Although the racial makeup of the 2 cohorts differed, comparisons between the NFHL study and NHANES III were adjusted for race. Residual confounding by race would only make the NHFL study seem to have more IR and would thus not explain the lack of significant difference.

Level of exercise is known to be an independent predictor of IR; thus, systematic differences in the level of exercise between the NFHL study and NHANES III must be considered. The NFHL study collects self-reported information on physical activity without discriminating between work-related versus leisure time physical activity. Only 42% of NFHL study participants reported any regular physical activity that would make them sweat, make the heart thump, or make them short of breath for at least 15 minutes at least once a week, and only 31% report doing so 3 or more times per week. Direct comparisons cannot be made between the NFHL study and NHANES III because of differences in how exercise questions were asked (the NFHL study questionnaire asked about usual activity in the past week but did not restrict to leisure activity, whereas the NHANES III study questionnaire asked about leisure time activity only in the past month). Among the NHANES III participants included in this study, 30.5% reported leisure activity (jogging or running, biking, swimming, dancing, calisthenics or exercise, or garden/yard work) an average of 3 or more times per week. Thus, there is no reason to suspect a systematically higher level of exercise in NFHL study participants compared with NHANES III participants.

Women included in the NFHL study were somewhat more overweight than those in the NHANES III (BMI of 27.3 vs. 26.4 kg/m2, 52% vs. 43% had truncal obesity, and the mean waist circumference was 89.5 vs. 88.0 cm). Hadigan et al7 found that HIV-infected women had more central adiposity compared with weight-matched healthy female controls, regardless of PI use. Galli et al,36 reporting on the LipoICoNA Study of adipose tissue alterations in HIV-infected patients started on first-line antiretroviral therapy, noted that female gender was independently associated with increased risk of morphologic alterations. Thus, HIV-infected women may be more likely to experience visceral fat accumulation compared with HIV-infected men. Despite the increased prevalence of truncal obesity in women compared with men in the NFHL study and NHANES III, women were not significantly more likely to have IR compared with men.

We found an association of elevated ALT with IR. Other studies have also shown an association of ALT and viral hepatitis with IR. Mehta et al41 found an increased risk of hyperglycemia in patients coinfected with hepatitis C virus (HCV) and HIV compared with HIV-infected patients. Data from the NHANES III show that elevations of ALT not explained by alcohol, viral hepatitis, or iron overload were significantly associated with fasting insulin levels.42 The mean ALT in men and women in the NFHL study was significantly higher compared with that of men and women in the NHANES III. Given the increased prevalence of intravenous substance abuse in our cohort compared with the general population (32.4% of the men and 39.6% of the women in the NFHL cohort included in this study reported ever using intravenous drugs), the higher mean ALT, and the known increased prevalence of HCV infection in other HIV populations, one could reasonably suspect a substantial prevalence of HCV in NFHL study participants. We do not have information on HCV, although future studies in the NFHL cohort may shed some light on this matter. Intravenous drug use and ALT were, however, a part of the multivariate analysis, and intravenous drug use was not retained in the final model. ALT was a significant independent predictor as well as a confounder of the relation of HAART to IR. One might expect the cohort with the higher prevalence of hepatitis to have more IR. This makes the lack of a significant difference in IR in our population compared with the NHANES III cohort even more striking.

There are several limitations of our study. We used a well-known and validated epidemiologic measure of IR, but we did not measure IR using a gold-standard method such as the hyperinsulinemic euglycemic clamp. In addition, this was a cross-sectional analysis; thus, we could not assess the influence of time-varying variables on IR. We adjusted for covariates that are independently associated with IR but may also be in the causal pathway of an effect of HAART on IR. This is an important consideration with lipid covariates such as triglycerides, which have been shown to increase in persons on HAART therapy and have also been shown to be associated with worsened IR in non-HIV-infected persons. When the final model was run without triglycerides, HDL, or LDL, however, the associations of HAART therapies with IR were essentially unchanged and the impact of CD4 count, log viral load, and years HIV-positive remained insignificant.

Back to Top | Article Outline

CONCLUSION

IR was quite common in our cohort but no more so than in the general population as estimated in the NHANES III. Factors associated with IR in the general population (eg, waist circumference, lipids, ALT) are also associated with IR in HIV-infected persons. HAART therapy was significantly associated with IR in men but not in women in multivariate analysis. PI-based HAART and NNRTI-based HAART were associated with IR in HIV-infected men. Other HIV-associated factors (CD4 cell count, log viral load, and years HIV infected) were not significantly associated with IR.

Back to Top | Article Outline

REFERENCES

1. Grinspoon S. Insulin resistance in the HIV-lipodystrophy syndrome. Trends Endocrinol Metab. 2001;12:412-419.
2. Hadigan C, et al. Metabolic abnormalities and cardiovascular disease risk factors in adults with human immunodeficiency virus infection and lipodystrophy. Clin Infect Dis. 2001;32:130-139.
3. Smith C, Levy I, Sabin C, et al. Cardiovascular disease risk factors and antiretroviral therapy in an HIV-positive UK population. HIV Med. 2004;5:88-92.
4. Brown TT, Cole SR, Li X, et al. Prevalence and incidence of pre-diabetes and diabetes in the Multicenter AIDS Cohort Study [oral session]. Presented at: 11th Conference on Retroviruses and Opportunistic Infections. San Francisco, CA, February 8-11, 2004.
5. Walli R, Herfort O, Michl G, et al. Treatment with protease inhibitors associated with peripheral insulin resistance and impaired oral glucose tolerance in HIV 1 infected patients. AIDS. 1998;12(Suppl):F167-F173.
6. Carr A, Samaras K, Burton S, et al. A syndrome of peripheral lipodystrophy, hyperlipidaemia and insulin resistance in patients receiving HIV protease inhibitors. AIDS. 1998;12(Suppl):F51-F58.
7. Hadigan C, Miller K, Corcoran C, et al. Fasting hyperinsulinemia and changes in regional body composition in human immunodeficiency virus-infected women. J Clin Endocrinol Metab. 1999;84:1932-1937.
8. Bruno R, Gazzaruso C, Sacchi P, et al. High prevalence of metabolic syndrome among HIV-infected patients: link with the cardiovascular risk. J Acquir Immune Defic Syndr. 2002;31:363-365.
9. Hadigan C, Corcoran C, Stanley T, et al. Fasting hyperinsulinemia in human immunodeficiency virus-infected men: relationship to body composition, gonadal function, and protease inhibitor use. J Clin Endocrinol Metab. 2000;85:35-41.
10. Delorenze GN, Horberg M, Karter AJ, et al. DM among HIV-infected patients in northern California's largest HMO [oral session]. Presented at: 10th Conference on Retroviruses and OI; 2003.
11. Noor M, Seneviratne T, Aweeka F, et al. Indinavir acutely inhibits insulin-stimulated glucose disposal in humans: a randomized, placebo-controlled study. AIDS. 2002;16(Suppl):F1-F8.
12. Noor M, Lo J, Mulligan K, et al. Metabolic effects of indinavir in healthy HIV-negative men. AIDS. 2001;15:11-18.
13. Knox T, Spiegelman D, Skinner S, et al. Diarrhea and abnormalities of gastrointestinal function in a cohort of men and women with HIV infection. Am J Gastroenterol. 2000;95:3482-3489.
14. Jones M. Indicator and stratification methods for missing explanatory variables in multiple linear regression. J Am Stat Assoc. 1996;91:222-230.
15. NHANES III Laboratory Data File. US Department of Health and Human Services (DHHS). National Center for Health Statistics. Third National Health and Nutrition Examination Survey, 1988-1994, NHANES III Laboratory Data File (CD-ROM) Public Use Data File Documentation Number 76200. Hyattsville, MD: Centers for Disease Control and Prevention; 1996. Available from National Technical Information Service (NTIS), Springfield, VA. In PDF format; includes access software: Adobe Systems, Acrobat Reader 2.1.2004.
16. Katz A, Nambi S, Mather K, et al. Quantitative Insulin Sensitivity Check Index: a simple, accurate method for assessing insulin sensitivity in humans. J Clin Endocrinol Metab. 2000;85:2402-2410.
17. Matthews D, Hoskier J, Rudenski A, et al. Homeostasis Model Assessment: insulin resistance and B-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia. 1985;28:412-419.
18. Abbasi F, Reaven G. Evaluation of the Quantitative Insulin Sensitivity Check Index as an estimate of insulin sensitivity in humans. Metabolism. 2003;51:235-237.
19. Uwaifo G, Fallon E, Chin J, et al. Indices of insulin action, disposal, and secretion derived from fasting samples and clamps in normal glucose-tolerant black and white children. Diabetes Care. 2002;25:2081-2087.
20. Bastard J, Robert J, Jardel C, et al. Is Quantitative Insulin Sensitivity Check Index a fair insulin sensitivity index in humans? Diabetes Metab. 2001;27:69-70.
21. Chen H, Sullivan G, Yue L, et al. QUICKI is a useful index of insulin sensitivity in subjects with hypertension. Am J Physiol Endocrinol Metab. 2003;284:E804-E812.
22. Katsuki A, Sumida Y, Gabazza E, et al. QUICKI is useful for following improvements in insulin sensitivity after therapy in patients with type 2 diabetes mellitus. J Clin Endocrinol Metab. 2003;87:2906-2908.
23. Kirwan J, Huston-Presley L, Kalhan S, et al. Clinically useful estimates of insulin sensitivity during pregnancy: validation studies in women with normal glucose tolerance and gestational diabetes mellitus. Diabetes Care. 2001;24:1602-1607.
24. Hrbicek J, Janout V, Malincikova J, et al. Detection of insulin resistance by simple Quantitative Insulin Sensitivity Check Index (QUICKI) for epidemiological assessment and prevention. J Clin Endocrinol Metab. 2002;87:144-147.
25. Gokcel A, Baltali M, Tarim E, et al. Detection of insulin resistance in Turkish adults: a hospital-based study. Diabetes Obes Metab. 2003;5:126-130.
26. Rabasa-Lhoret R, Bastard J, Jan V, et al. Modified Quantitative Insulin Sensitivity Check Index is better correlated to hyperinsulinemic glucose clamp than other fasting-based index of insulin sensitivity in different insulin-resistant states. J Clin Endocrinol Metab. 2003;88:4917-4923.
27. Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults. NIH Publication No. 98-4083. National Institutes of Health, National Heart, Lung and Blood Institute; 1998.
28. Ford E, Giles W, Dietz W. Prevalence of the metabolic syndrome among US adults: findings from the Third National Health and Nutrition Examination Survey. JAMA. 2002;287:356-359.
29. Harrison G, Buskirk E, Lindsay Carter J, et al. Skinfold thicknesses and measurement technique. In: Lohman T, Roche A, Martorell R, eds. Anthropometric Standardization Reference Manual. Champaign, IL: Human Kinetics Books; 1988:55-70.
30. Lukasi H. Use of bioelectrical impedance analysis to assess human body composition: a review. In: Livingston G, ed. Nutritional Status Assessment of the Individual. Food & Nutrition Press, Trumbull, CT. 1989:189-204.
31. Durnin J, Wormersley J. Body fat assessed from total body density and its estimation from skinfold thickness: measurements on 481 men and women aged from 16 to 72 years. Br J Nutr. 1974;32:77-97.
32. Schakel S, Sievert Y, Buzzard I. Sources of data for developing and maintaining a nutrient database. J Am Diet Assoc. 1988;88:1268-1271.
33. Hadigan C, Corcoran C, Piecuch S, et al. Hyperandrogenemia in human immunodeficiency virus-infected women with the lipodystrophy syndrome. J Clin Endocrinol Metab. 2000;85:3544-3550.
34. Carr A, Samaras K, Thorisdottir A, et al. Diagnosis, prediction, and natural course of HIV-1 protease-inhibitor-associated lipodystrophy, hyperlipidaemia, and diabetes mellitus: a cohort study. Lancet. 1999;353:2093-2099.
35. Thiebaut R, Daucourt V, Mercie P, et al. Lipodystrophy, metabolic disorders, and human immunodeficiency virus infection: Aquitaine Cohort, France, 1999. Clin Infect Dis. 2000;31:1482-1487.
36. Galli M, Cozzi-Lepri A, Rodriguez W, et al. Incidence of adipose tissue alterations in first-line antiretroviral therapy. Arch Intern Med. 2002;162:2621-2628.
37. Miller J, Carr A, Emery S, et al. HIV lipodystrophy: prevalence, severity and correlates of risk in Australia. HIV Med. 2003;4:293-301.
38. Lichtenstein KA, Ward DJ, Moorman AC, et al, and the HIV Outpatient Study Investigators. Clinical assessment of HIV-associated lipodystrophy in an ambulatory population. AIDS. 2001;15:1389-1398.
39. Lauenroth-Mai E, Schlote F. HIV-associated lipodystrophy syndrome: LioN-HAART Cohort: (Lipodystrophy in Patients on Nucleoside-Based HAART). J Acquir Immune Defic Syndr. 2002;31:253-255.
40. Heath KV, Hogg RS, Chan KJ, et al. Lipodystrophy-associated morphological, cholesterol and triglyceride abnormalities in a population-based HIV/AIDS treatment database. AIDS. 2000;15:231-239.
41. Mehta S, Moore R, Thomas D, et al. The effect of HAART and HCV infection on the development of hyperglycemia among HIV-infected persons. J Acquir Immune Defic Syndr. 2003;33:577-584.
42. Clark J, Brancati F, Diehl A. The prevalence and etiology of elevated aminotransferase levels in the United States. Am J Gastroenterol. 2003;98:955-956.
Keywords:

insulin resistance; HIV; nutrition for healthy living; NHANES III

© 2005 Lippincott Williams & Wilkins, Inc.