A total of 89 patients (age 18–65, median: 28
years) including 51 men (57.3%, age range: 18–63
years, median: 32
years) and 38 women (42.7% age range: 19–65
years, median: 21
years) were included in the study. The mean BMI was 23.6 (± 3.9) kg/m2
. The median time from HIV diagnosis was 80.5
months and the median time on antiretroviral therapy was 72.5
months. Sixty five patients (73%) had a current PI-based cART regimen. Median serum levels of adiponectin, leptin, resistin, TNF alpha and IL-6 were 10.2
pg/mL and 26.5
Fifty nine patients (66.3%) were diagnosed with IR based on QUICKI values lower than the cut-off point. The overall prevalence of IR was 62.9%, based on HOMA-IR dichotomization. The median value of HOMA-IR was 3.77 (3.62). The patients were divided into Group 1 (patients without IR) and Group 2 (patients with IR). Group 1 patients had a mean QUICKI value of 0.35 (±0.02) and a median HOMA-IR of 1.84 (0.67) vs. 0.29 (±0.02) and 4.62 (3.48) in Group 2, respectively. The subject characteristics stratified by IR status are shown in Table
. Significant differences were noted in total cholesterol, triglycerides and adiponectin serum levels between the two groups. The prevalence of IR was not significantly different between men and women. No significant differences were found between Group 1 and Group 2 in age, BMI, total time on cART or current protease inhibitor treatment.
Clinical and biological characteristics of study patients by IR
In order to determine the relationship between laboratory parameters and IR, first we assessed the relationship of all individual variables with QUICKI score by fitting a linear regression model. The variables with univariate association to QUICKI were further tested to improve the model predictivity. After all variables were tested in the univariate analysis for the entire population, including BMI and age, only log transformed triglycerides serum levels (LOGTriglycerides) were significantly associated with QUICKI (R=0.24, p=0.01).
Fifity one men were enrolled in the study and 37 (72.5%) were diagnosed with IR. Male subject characteristics stratified by IR status are shown in Table
. Significant between-group differences persisted in triglycerides and adiponectin serum levels. The significant variables selected to build the additive linear regression model in men, the estimate terms for the model and corresponding test statistics are provided in Table
. Of the variables tested only log transformed adiponectin (LOGAdiponectin) and LOGTriglycerides had a significant effect on QUICKI values in our cohort of male subjects. Taken together, the analysis indicates that LOGAdiponectin with LOGTriglycerides were associated with QUICKI index (R=0.43, p=0.007) in males. The additive models that included age and BMI together with LOGAdiponectin and LOGTriglycerides lacked statistical significance.
Biological characteristics of male and female study patients by IR
Multivariate analysis: effect of selected variables (LOGAdiponectin, LOGTriglycerides) on QUICKI (dependent variable) in male subjects
Of 38 women enrolled 22 (57.6%) were diagnosed with IR. Female laboratory characteristics stratified by IR status are shown in Table
. Significant between-group differences were observed in HDL cholesterol and leptin serum levels. In women, of all variables tested in the univariate analysis, only log transformed leptin serum levels (LOGLeptin) were significantly associated with QUICKI (R=0.33, p=0.03) (Table
). Other selected variables, including age and BMI, did not show any improvement of our additive model.
Univariate analysis: effect of selected variables (LOGLeptin) on QUICKI (dependent variable) in female subjects
IR appears when maintaining normal physiological values of glycaemia require insulin levels that exceed normal concentrations [18
]. IR is associated with the development of diabetes and obesity [19
]. Up to 20% of non-HIV western population may have IR [20
]. Several studies reported a two-fold increase of cardiovascular risk compared to non-insulin resistant controls [22
]. In the pre-cART era several cross-sectional studies reported slightly increased [24
] or normal [25
] insulin sensitivity when compared to uninfected controls. cART introduction led to increases in fasting insulin and decreases in insulin sensitivity, effects dependent both on the class of the antiretroviral used and the different antiretrovirals within each class [26
]. Different studies have evaluated the prevalence of IR at 20 to 50%, in HIV-positive patients [28
]. We noted higher levels of IR than reported by previous studies, independent of current treatment with PIs and of BMI, which may be attributed in part to our method of assessment. QUICKI has a very good correlation with the glucose clamp index of insulin sensitivity, which is considered the gold standard method for IR assessment [15
]. IR presence in our population of HIV-positive young adults was not influenced either by the time from HIV diagnosis or to the duration of cART. This finding may suggest that the effect of cART on IR may not have a cumulative or long-term response, a similar result being reported also by other studies [29
]. As IR presence is associated with an increased risk of diabetes, cardiovascular events, stroke and death, the reduction of its prevalence must be considered an essential therapeutic target in HIV-patients undergoing retroviral therapy.
Adiponectin is exclusively secreted from adipose tissue and has an important role as a mediator of peroxisome proliferator-activated receptor gamma action [30
]. Adiponectin may increase insulin responsiveness of the glucose transport system in adipocytes through increased GLUT4 gene expression and increased GLUT4 recruitment to the plasma membrane [31
]. Several studies have reported that adipocyte differentiation was perturbed both in vitro and in vivo as a direct effect of cART and the consecutive decrease in serum adiponectin levels was associated with insulin resistance [32
]. We found an inverse correlation between adiponectin and IR in men, but not in females which partially concurs with previous studies. Vigouroux et al. (2003) have also reported that adiponectin was positively correlated with insulin sensitivity in a cohort of 131 HIV-infected men with PI-based cART regimens [9
]. In a study on 237 Afro-Caribbean HIV-1-infected patients, published in 2011, Deloumeaux et al. found also an inverse correlation between adiponectin and IR in both men and women. After adjusting for clinical and metabolic parameters, the correlation between adiponectin and insulin resistance was observed only in women [34
], a conclusion that contradicts our results and the previously mentioned study. Alterations of serum adiponectin concentration in relation to insulin resistance was also observed in a cohort of HIV-infected youths (aged 5.0 - 19.4
years), in a study published by Viganò et al. in 2011, which did not take into account sex differences but evaluated lipodystrophy. Low adiponectin concentration was associated to central fat and mixed lipohypertrophy and to signs of IR [35
]. Other studies have also shown that adiponectin was inversely associated with IR, type 2 diabetes and obesity in HIV and non HIV-patients in without adjusting for gender differences [36
Leptin is a 167-amino-acid adipokine secreted mainly by white adipose tissue. The levels of leptin are positively correlated with the amount of body fat [37
]. Leptin regulates energy homeostasis and various neuroendocrine functions [38
]. Our data showed higher levels of leptin in women with IR vs women without IR and a weak correlation between IR presence and leptin, observed in the linear regression model, with QUICKI as dependent and LOGLeptin as independent variable. Leptin did not contribute to an improvement of our linear model of IR in men. Other studies investigating this association have reported inconclusive results but very few have based their analyses on gender differences. Azzoni et al. in 2010 observed that serum leptin levels are inversely associated with HIV replication, independent of disease progression. In the same study, the authors reported a weak association between leptin levels and the Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) [39
]. Mynarcik et al. (2002) found that levels of leptin did not correlate with insulin sensitivity, in HIV-infected subjects with body fat redistribution [37
]. In other study investigating adipokines and antiretroviral therapy related lipodystrophy, Tao et al. (2009) observed that leptin was positively correlated with the levels of total body fat, limb fat, and trunk fat, both in lipodystrophic and patients without body fat redistribution, but did not correlate with IR, and might serve only as a biomarker for total body fat mass [40
]. Nagy et al. (2003) found a significant association between leptin concentrations and IR in patients with different morphological phenotypes. The lowest mean leptin levels were observed in patients with lipoatrophic changes. Hypoleptinemia was independently associated with IR in patients with lipoatrophy. Highest leptin concentrations were observed in patients with lipohypertrophy. Leptin levels in between the two extremes were observed in subjects with no body shape changes [41
]. The role of leptin alterations in IR pathogenesis remains to be clarified by future long-term and prospective studies.
Despite the relative low number of patients, our study has several strengths. This is the first study in Romania to investigate adipokine dysregulation in association with HIV metabolic abnormalities during cART. The young age of our cohort is a distinct feature for the HIV epidemiology in Romania, compared with western European countries. The patients were probably infected parenterally in their childhood and adolescence between 1980–1990. QUICKI was used in order to diagnose IR, which is one of the most accurate surrogate indexes for determining insulin sensitivity in humans. Very few studies have evaluated untill now IR risk factors based on sex differences. Factors associated with IR, which have been lacking in previous studies, including the time from HIV diagnosis and the time on cART were measured and evaluated in our analyses.
Our study has some limitations. Correlations were small to moderate in magnitude. The association between IR and adiponectin/triglycerides in men and IR and leptin in women disappeared after including age and BMI in the regression models. This may be caused by BMI variations and changes in adiponectin with age that may contribute to IR pathogenesis and also by the low number of enrolled patients, with loss of statistical power as more variables were added. QUICKI was used to diagnose IR, a surrogate method, which has some limitations, such as the absence of a definite threshold for IR in Caucasian populations. QUICKI requires the quantification of fasting plasma insulin. Similar observations in clinical practice may be predicted with the use of International Diabetes Federation (IDF) or Adult Treatment Panel III (ATPIII) definitions of metabolic syndrome, which may provide cheaper ways for detecting patients at risk of diabetes and cardiovascular events. As leptin may be positively correlated with the levels of total body fat, limb fat, and trunk fat [41
], the predictive value of our model might have been improved if we had included body fat redistribution in our analysis.