Circulating adiponectin levels are lower in Latino versus non-Latino white patients at risk for cardiovascular disease, independent of adiposity measures

Subjects in the Latinos Using Cardio Health Actions to Reduce Risk (LUCHAR) study were included in this cross-sectional study. The overall aim of LUCHAR is to improve cardiovascular disease (CVD) prevention in Latinos, taking into account a higher prevalence of cardiovascular risk factors, lower treatment rates, and unique social, cultural, language, and access-to-care barriers when compared to non-Latino white patients. The LUCHAR cohort included in the present study was recruited to investigate surrogate markers of CVD in Latinos. Laboratory measures specific to the present study were measured from stored blood samples obtained from the LUCHAR project. The study protocol was approved by the Colorado Multiple Institutional Review Board (COMIRB) and written consent was obtained from all participants. All study procedures were conducted according to the principles expressed in the Declaration of Helsinki. The study was conducted at Denver Health Medical Center (DHMC), an integrated community health and safety-net hospital system in Denver, Colorado, United States of America. Potential subjects were identified using the DHMC electronic hypertension registry comprised of patients with at least one diagnostic (ICD-9) code for hypertension, seen in the DHMC system between July 1, 2002 and June 30, 2007. Health information, including past medical history and medication use, were obtained from an initial chart review and from patient self-report. Patients were eligible for the study if they had a prior diagnosis of hypertension, were ≥18 years of age, attended a clinic in the DHMC system, and were of either self-reported Latino or non-Latino white race/ethnicity. Subjects were also required to have self-reported history or diagnosis of at least one other cardiovascular risk factor including diabetes, dyslipidemia, obesity, chronic kidney disease/microalbuminuria, current smoking, or older age (men >55 years, women >65 years). Patients were excluded if they had a known history of cardiovascular disease (myocardial infarction or coronary revascularization), history of stroke, cerebrovascular revascularization, peripheral arterial disease, chronic heart failure, valvular heart disease, end-stage renal disease, inflammatory disease or vasculitis. Eligible participants were contacted via telephone and invited to schedule a study visit. All individuals meeting eligibility criteria and consenting to participate in the study were included in the LUCHAR cohort. Subject recruitment was continued until the predetermined number of participants needed for the analyses of CVD outcomes was achieved. All 179 subjects in the final LUCHAR cohort were included in the present analysis.

Eligible participants were contacted via telephone and scheduled for a study visit. Participants were asked to fast for the 8 hours prior to their study visit. At the study visit, subjects completed a personal history questionnaire and reviewed a list of current medications. Height, weight, waist circumference, and blood pressure were measured. Fasting blood samples were drawn for basic chemistry panel, fasting lipids, complete blood count, HbA_1c, and high sensitivity C-reactive protein (hsCRP). Insulin, total adiponectin, and HMW adiponectin were measured from stored plasma samples from this study visit.

Insulin and total adiponectin were measured by RIA (Diagnostic Systems Laboratory, Inc., Webster, TX; and LINCO Research, Inc.; St. Charles, MO respectively). Intra-assay precision for the total adiponectin assay is 1.78-3.59% CV, and inter-assay precision is 6.90-9.25% CV. HMW adiponectin was measured by ELISA (Millipore, Billerica, MA); intra-assay precision is 2.5-4.7% CV and inter-assay precision is 5.8-6.9% CV. All other labs were measured by the Denver Health Core Laboratory using standard procedures.

The homeostasis model insulin resistance index [29] was calculated from fasting insulin and glucose levels using the formula: HOMA-IR = (fasting insulin in mU/l × (fasting plasma glucose in mg/dl)/18))/22.5. Body mass index (BMI) was calculated as weight divided by squared height. Estimated GFR (eGFR) was calculated using the 4-variable abbreviated MDRD formula [30]: eGFR = 186 × (serum creatinine)^-1.154 × (age)^-0.203 × (0.742 if female) × (1.210 if black).

SAS version 9.2 (SAS Institute, Inc., Cary, NC) was used for all statistical analyses. Means, medians, and standard deviations were generated for each variable, stratified by gender and ethnicity. Group differences were assessed by Students t-test. Wilcoxon's test was used for non-normally distributed variables. A two-sided p value < 0.05 was considered statistically significant. Kendall's Tau bivariate analyses were performed between total or HMW adiponectin and variables thought likely to predict adiponectin level, both within each ethnic group and in the entire population. In addition to known associates of adiponectin levels including age, gender, BMI, waist circumference, diagnosis of diabetes, renal function, and use of TZD, we included use of medications associated with adiponectin in other populations (ACEI and ARB, acetylsalicylic acid, and statins) as exploratory variables. The correlation between total adiponectin and HMW adiponectin was also measured in order to verify an expected association between these two variables.

To identify independent determinants of adiponectin, multivariate linear regression analyses were performed with total or HMW adiponectin as the dependent variable. An interaction term for ethnicity x waist circumference was used to assess a possible moderating effect of waist circumference on the relationship between ethnicity and adiponectin.

We measured adiponectin's contribution to the observed ethnic difference in insulin resistance by performing multivariate regression analyses with logarithmically transformed HOMA-IR as the outcome variable. Initially, ethnicity, confounders (age, gender, TZD use, Statin use, estimated GFR, and diabetes status), and either BMI or waist circumference were entered into the model. Next, total adiponectin was added to each model. Finally, HMW adiponectin was substituted for total adiponectin.

Clinical and biochemical characteristics are shown in Table 1, stratified by ethnicity and gender. Our subject cohort was made up of 119 Latinos (81 women, 38 men) and 60 non-Latino whites (39 women, 21 men). Though individuals from any Latino sub-group were eligible to participate, >97% of participants in the Latino cohort were of Mexican-American descent, reflective of the make-up of the Latino population in the Denver area. Average age was 61.4 +/- 9.7 years and mean BMI was 33.1 +/- 6.7 kg/m². Overall, 49% of subjects had type 2 diabetes. There were no ethnic group differences in BMI, but mean waist circumference was lower in the Latina women compared to the white women. White women had lower mean estimated GFR than Latina women, but mean eGFR was within the normal range for all groups. In regards to medication use, TZD use was higher and ASA use was lower in Latino men compared to white men, but use of ACEI/ARB and statin was not different in the two ethnic groups. No ethnic group differences were observed in other demographic or clinical characteristics (age, gender representation, fasting lipids, systolic blood pressure, or hs-CRP).

Median total and HMW adiponectin was lower in Latinos compared to non-Latino whites (Table 1 and figure 1). Total adiponectin in Latinos and non-Latinos was 7.0 (4.9-11.4) versus 11.4 (7.3-17.1) μg/ml in women, p = 0.001 and 4.6 (3.8-6.5) versus 6.0 (5.0-10.6) μg/ml in men, p = 0.02 (figure 1). Similarly, HMW adiponectin was lower in the Latino versus non-Latino group (2.8 (1.7-5.0) versus 4.4 (2.8-7.9) μg/ml in women, p = 0.009 and 2.3 (1.2-3.7) versus 2.9 (2.0-3.9) μg/ml in men, p = 0.1). HMW/total adiponectin ratio was not different between the two groups.

In order to identify additional correlates of adiponectin levels, bivariate associations between total or HMW adiponectin and variables thought likely to influence adiponectin levels were assessed (Table 2). Strength and magnitude of correlations with total and HMW adiponectin were similar, and an expected strong association between total and HMW adiponectin was observed. Older age, and female gender, were positively associated with adiponectin levels in both ethnic groups. TZD use was significantly associated with adiponectin in the Latino group only. ACEI/ARB, aspirin, and Statin use was not associated with adiponectin levels in the population as a whole. Adiponectin levels were not correlated to diabetes status or estimated GFR. Though adiponectin levels were not correlated with BMI in either ethnic group or in the population as a whole in the bivariate analyses, we observed a negative association between adiponectin levels and waist circumference in the non-Latino group but not in the Latino group.

To further explore our observation of an association between waist circumference and adiponectin in the white participants but not the Latino group, we used an interaction term for ethnicity x waist circumference. As demonstrated by the plotted regression lines of waist circumference versus adiponectin by ethnicity (Figure 2) we found that waist circumference was negatively associated with adiponectin in the non-Latino group, whereas no association was found between these two variables in the Latino group (p value for interaction <0.05). When waist circumference and adiponectin values are plotted by ethnicity and gender (Additional file 1: Figure S1), it becomes apparent that the observed ethnic difference in the waist circumference and adiponectin association is driven by the men, with Latino men having low adiponectin regardless of waist circumference. Thus, at any given waist circumference, adiponectin is lower in Latinos than in non-Latino whites. The only possible exception to this observation is at the higher waist circumference measures where the regression lines for the men cross and men in both ethnic groups have low adiponectin. The relationship between adiponectin and BMI (Figure 3; Additional File 2: Figure S2) is similar, with lower adiponectin in Latinos versus non-Latino whites at any given BMI.

In the multivariate analysis examining the relationship between ethnicity and adiponectin levels (Table 3 and Additional File 3: Table S1), Latino ethnicity was the strongest negative predictor of both total (-4.7; 95% CI -6.7, -2.8; p < 0001) and HMW (-1.8; 95% CI -2.9, -0.8; p = 0.001) adiponectin, independent of all other variables included in the model (age, gender, TZD use, Statin use, estimated GFR, diabetes status, and waist circumference/BMI).

In order to measure the contribution of adiponectin levels to ethnic-group differences in insulin resistance, we performed multivariate linear regression analyses in the entire cohort, with logHOMA-IR as the outcome variable (table 4 and Additional File 3: Table S2). Though Latino ethnicity showed a strong positive association with logHOMA-IR when ethnicity and confounders (age, gender, TZD use, statin use, estimated GFR, waist circumference/BMI) were included in the model (Model 1), with the addition of total adiponectin level (Model 2), the association with Latino ethnicity was no longer significant. Similarly, including HMW instead of total adiponectin and adjusting for confounders and ethnicity (Model 3), HMW adiponectin level was inversely associated with logHOMA-IR, and the association between Latino ethnicity and logHOMA-IR was no longer significant.