Longitudinal Assessment of Metabolic Abnormalities in Adolescents and Young Adults with HIV-infection Acquired Perinatally or in Early Childhood

Abstract

Objective

Metabolic complications of HIV pose challenges for health maintenance among young adults who acquired HIV in early childhood.

Materials/Methods

Between 7/2004–7/2009 we evaluated 47 HIV-infected subjects who acquired HIV in early life. Participants completed glucose tolerance testing, insulin, lipid, urine albumin and creatinine determinations and DXA scans. Longitudinal data were available for 39 subjects; duration of follow-up was 26.4±16.8 months.

Results

At baseline, participants were 17.1±3.9y and duration of antiretroviral therapy was 12.7±3.4y. CD4 count was 658±374 cells/mm³ and 55% had undetectable viral load. Impaired glucose tolerance (IGT) was present in 15%; 33% had insulin resistance (HOMA-IR>4.0). Further, 52% had triglycerides ≥150 mg/dL, 36% had HDL-c <40 mg/dl, 18% had LDL-c ≥130 mg/dL and 25% had total cholesterol ≥200 mg/dL. Microalbuminuria was present in 15% of participants and was inversely correlated with CD4% (p=0.001). During follow-up more than one third remained insulin resistant; lipid parameters tended to improve. There were significant increases in BMI (p=0.0002), percent leg fat (p=0.008) and trunk fat (p=0.002).

Conclusions

IGT, insulin resistance, dyslipidemia and microalbuminuria are common among young adults with HIV. Long-term exposure to therapy may translate into substantial persistent metabolic risk.

Introduction

Morbidity and mortality associated with HIV/AIDS decreased dramatically with the advent of potent antiretroviral therapies (ART). However, ART has also been associated with metabolic complications, alterations in body fat distribution and chronic kidney disease [1,2,3]. Relatively fewer data exist to characterize these associations in adolescents and young adults infected with HIV in early childhood.

In adults, the use of ART is associated with insulin resistance, diabetes and cardiovascular disease (CVD) [4–8] . Previous investigation demonstrated that increased exposure to the nucleoside reverse transcriptase inhibitor (NRTI) stavudine, and to a lesser extent didanosine (ddI) and zidovudine (AZT), was associated with an increased risk of developing diabetes in adults [5]. Although long-term effects of insulin resistance are unknown, it may be an important CVD risk factor among HIV-infected adults [9].

There is currently a paucity of data pertaining to metabolic disturbances associated with long-term ART among adolescents and young adults who acquired HIV perinatally or in early childhood. In HIV-infected children, protease inhibitors (PIs) have been shown to contribute to lipid abnormalities and the presence of lipodystrophy [10–14]. HIV-infected children with lipodystrophy have higher fasting insulin compared to children without lipodystrophy [15]. Exposure to ART and inflammation in children with HIV were identified in association with increased carotid intima media thickness, a marker for CVD risk [16, 17]. Insight into the metabolic complications associated with long-term ART will allow development of preventive strategies to reduce CVD risk. Here we characterize metabolic disturbances both cross-sectionally and over time in a group of HIV-infected adolescents and young adults.

Methods

Subjects and Data Collection

We performed a cross-sectional analysis, followed by a longitudinal analysis, of a cohort of subjects who acquired HIV perinatally or in early childhood. Subjects received medical care independent of the study and were evaluated at the National Institutes of Health from 2004–2009. The research was approved by the National Institue for Allergy and Infectious Diseases’ IRB and informed consent was obtained from all subjects and parents or legal guardians, as appropriate.

Thirty-nine of 47 subjects were included in the longitudinal analysis (Figure 1). Six subjects included in the cross-sectional analysis were lost to follow-up and 2 had not yet completed a follow-up oral glucose tolerance test (OGTT). Mean duration of follow-up was 26.4±16.8 months and data were analyzed based on the baseline visit and last follow-up.

Figure 1.

Flow diagram of subject recruitment and participation

Participants completed a clinical evaluation, including a complete history of all ART exposure. Data collected included Tanner stage, fasting lipid profile, fasting and 2 hour glucose and insulin following 75-g OGTT. Impaired glucose tolerance (IGT) was defined as a 2-hour glucose ≥140 mg/dL but <200 mg/dL. Laboratories included CD4 T lymphocyte counts, CD4%, and HIV-1 RNA by RT-PCR. Abnormal fasting lipid levels were defined based on the National Cholesterol Education Program criteria[18]. Hypertriglyceridemia was defined as triglyceride >150mg/dL, hypercholesterolemia as total cholesterol >200mg/dL, abnormal LDL-c as ≥130mg/ dL, and abnormal HDL-c as <40mg/dL. Microalbuminuria was defined as the geometric mean urine albumin-to-creatinine (urine A/C) ratio of 17–250 mg/g (males) and 25–355 mg/g (females) using all available time points [19, 20]. Thirty-five subjects provided 2–5 samples, four subjects provide one.

The homeostatic model of insulin resistance (HOMA-IR= fasting insulin (μU/m) × fasting glucose (mmol/L)/22.5) was used to evaluate insulin resistance. Subjects with a HOMA-IR >4.0 were classified as insulin resistant [21,22,23]. HOMA-IR >3.16 was also used as an alternative classification for insulin resistance [11, 22) as both have been used in prior investigations. Duration of ART was calculated at the subjects’ corresponding visit date for each individual antiretroviral agent and each class of agents. Ritonavir exposure was assessed by total years of exposure irrespective of the dose used.

Body Measurements and DXA Scan Characteristics

Standardized anthropometric measurements included height, weight, hip and waist circumferences and body mass index (BMI). Whole body dual energy X-ray absorptiometry (DXA, Hologic QDR4500A, Bedford, MA) was performed annually to measure regional and total fat and lean masses. All scans were analyzed by a single radiologist.

Statistical Analyses

Clinical characteristics are summarized using mean ± standard deviation; non-normally distributed variables are presented as a median (interquartile range) or log transformed to approximate a normal distribution. Comparisons were made using Kruskal-Wallis tests and chi-square statistics. Univariate regression analyses were performed to identify potential associations between metabolic parameters and clinical characteristics. Factors identified as significant on univariate analyses (p<0.05) were entered into multivariate regression models to identify independent predictors. Paired t-tests were used to compare longitudinal metabolic parameters between baseline and last follow-up. Statistical analyses were performed using SAS JMP Statistical Software (Version 7.0, SAS Institute Cary, NC), using a two-tailed alpha level of 0.05.

Results

Subject Characteristics

Forty-two subjects acquired HIV perinatally and 5 acquired HIV from transfusions in early childhood. Sixty-seven percent (31/46) of participants were Tanner stage 4 or greater at baseline (Table 1). At baseline, 55% of subjects had an undetectable viral load; the median CD4 T-cell count was 650 cells/mm³ (IQR 402–897). All subjects but one were ART-experienced with a duration of 12.7±3.4 years. All ART experienced subjects also had a history of exposure to a PI and 94% were currently receiving ART. Thirty-four percent of subjects were currently on a non-nucleoside reverse transcriptase inhibitor (NNRTI), whereas 85% were currently on a PI. One subject was on an integrase inhibitor, and four were on a fusion inhibitor. At entry 2 subjects were on a fibrate; no one was receiving a statin or antihypertensive agent. One participant had a diagnosis of polycystic ovary syndrome and was on metformin at the time of evaluation.

Table 1.

Demographics and Clinical Characteristics

	Cross-Sectional Study	Longitudinal Study
	Total Cohort (n=47)	Baseline (n=39)	Last Visit (n=39)
Age (y)	17.1 ± 3.9	17.5 ± 3.7	19.7 ± 3.7
Duration of follow-up (mo)	-	-	26.4 ± 16.8
Range of follow-up (mo)			8.4–56.4
Sex – number (%)	-	-	-
Male	26 (53%)	20 (51%)	20 (51%)
Female	21 (47%)	19 (49%)	19 (49%)
Tanner Stage – number (%)	-	-	-
<4	15 (33%)	10 (26%)	1 (3%)
>4	31 (67%)	28 (74%)	35 (97%)
Race – number (%)	-	-	-
White	23 (49%)	20 (51%)	20 (51%)
Black	18 (38%)	14 (36%)	14 (36%)
Mixed race	3 (6.5%)	3 (8%)	3 (8%)
Hispanic ethnicity	1 (2%)	0 (0%)	0 (0%)
Native American	2 (4.5%)	2 (5%)	2 (5%)
HIV Characteristics
HIV Viral Load <50 copies/mL – no. (%)	26 (55%)	22 (56%)	24 (62%)
CD4 T cell count – cells/mm3	658 ± 374	626 ± 370	725 ± 475
CD4 T cell %	29.5 ± 10.2	28.9 ± 10.5	30.3 ± 10.3
Current ART use – number (%)	44 (94%)	37 (95%)	37 (95%)
Current PI use – number (%)	40 (85%)	33 (85%)	30 (76.9%)
Current or past PI use – number (%)	46 (98%)	39 (100%)	39 (100%)
Current d4T use – number (%)	21 (44%)	17 (44%)	7 (17.9%)
Current or past d4T use – number (%)	41 (87%)	35 (90%)	35 (90%)
Current ABC use – number (%)	4 (9%)	4 (10%)	8 (21%)
Current or past ABC use – number (%)	13 (28%)	10 (26%)	18 (46%)
Current TDF use – number (%)	9 (19%)	7 (18%)	13 (33%)
Current or past TDF use – number (%)	15 (32%)	12 (31%)	20 (51%)

Note. Data are mean ± standard deviation, except as indicated.

Mean BMI was 21.6±3.4 kg/m²; 15% were overweight (BMI >25kg/m²) and one subject had a BMI >30kg/m². Waist-hip ratio was 0.92±0.07, and 17% had a waist-hip ratio ≥1.0. Percent body fat was 19.4±8.0%; percent trunk fat was 18.2±8.7%; and percent leg fat was 20.8±9.7%

Baseline Evaluations of Metabolic Parameters

At baseline, IGT was present in 15% of subjects, but no subject had type 2 diabetes. Thirty-three percent of subjects had a HOMA-IR > 4.0, which is considered insulin resistance in adolescents (21, 23). Using a less stringent cut-off of HOMA-IR > 3.16 to define insulin resistance (11, 22), 46% were insulin resistant. Dyslipidemia was common: 52% of subjects had hypertriglyceridemia, 36% low HDL-c, and 25% elevated total cholesterol. The mean total:HDL cholesterol ratio was 4.1±1.1(Table 2), and 18% of participants had suboptimal LDL-c. In univariate regression analyses, duration of AZT exposure (r=−0.31, p=0.03) was inversely correlated with HOMA-IR. Years of ddI and stavudine (d4T) did not correlate with HOMA-IR, however subjects with exposure to ddI (p=0.0008) or d4T (p=0.0005) had significantly higher HOMA-IR compared to ddI- d4T-naïve subjects. Subjects with a HOMA-IR >4 were more likely to have received ddI (χ²=7.1, p=0.008) or d4T (χ²=5.2, p=0.02). Subjects with a HOMA-IR >4 had higher triglycerides (p=0.01), waist circumference (p=0.02) and waist-hip ratio (p=0.009). There were no significant differences in HOMA-IR based on sex, Tanner stage, PI or NNRTI use. In a multivariate logistic regression model including significant factors identified in univariate analysis, only waist-hip ratio (p<0.05) remained an independent predictor of HOMA-IR >4.

Table 2.

Glucose Homeostasis, Lipid, Anthropometric Assessment

	Cross-Sectional Study	Longitudinal Study		P-value
	Total Cohort (n=47)	Baseline (n=39)	Last Visit (n=39)
Glucose	-	-	-	-
Fasting glucose – mg/dl	86.3 ± 10.0	85.9 ± 10.2	90.9 ± 12.5	0.06
2-hr OGTT glucose – mg/dl	108.1± 31.4	109.7 ± 31.5	104.3 ± 29.2	0.3
Fasting insulin - μIU/ml	19.7 ± 18.5	21.2 ± 19.9	20.3 ± 22.7	0.8
2-hr OGTT insulin - μIU/ml	94.5 ± 119.7	99.6 ± 128.1	78.1 ± 70.6	0.1
HOMA-IR	4.4 ± 4.5	4.8 ± 4.9	5.1 ± 7.8	0.8
Lipids	-	-	-	-
Total Cholesterol (TC) mg/dL	177 ± 42	174 ± 43	165 ± 45	0.2
HDL Cholesterol mg/dL	44 ± 10	45 ± 10	43 ± 12	0.3
LDL Cholesterol mg/dL	106 ± 30	103 ± 27	95 ± 38	0.2
TC/HDL ratio	4.1 ± 1.1	3.9 ± 0.9	4.0 ± 1.5	0.7
Triglycerides mg/dL	248 ± 266	226 ± 216	177 ± 151	0.07
Anthropometrics	-	-	-	-
Height (cm)	161.2 ± 11.9	162.2 ± 10.0	164.7 ± 8.9	0.003
Weight (kg)	56.5 ± 12.7	57.8 ± 11.8	63.7 ± 12.4	0.0001
Body-mass index (kg/ m2)	21.6 ± 3.4	21.9 ± 3.4	23.5 ± 4.4	0.0002
Waist-hip ratio	0.92 ± 0.07	0.92 ± 0.08	0.90 ± 0.15	0.3
Percent body fat (%)	19.4 ± 8.0	19.9 ± 8.1	22.5 ± 9.8	0.0006
Percent trunk fat (%)	18.2 ± 8.7	19.0 ± 8.9	22.1 ± 11.1	0.002
Percent leg fat (%)	20.8 ± 9.7	20.8 ± 9.6	23.2 ± 11.2	0.008

Note. Data are presented as mean ± standard deviation, except as indicated.

Compared to those without IGT, participants with IGT had significantly higher total cholesterol (215±46 mg/dL vs. 169±38 mg/dL, p=0.04), triglycerides (377±212 mg/dL vs. 226±274 mg/dL, p=0.005) and total:HDL cholesterol ratio (5±1 vs. 4±1, p=0.03). Participants with IGT had greater total percent body fat (p=0.009), greater percent leg fat (p=0.02) and were more likely to be male (χ²=9.6, p=0.002). There was no significant relationship between Tanner stage and IGT.

Higher viral load (r=−0.41, p=0.009) and greater years of abacavir use (r=−0.31, p<0.05) were associated with lower HDL-c, whereas NNRTI use (r=0.35, p=0.02) was related to higher HDL-c. In multivariate regression, only viral load remained a significant independent predictor of HDL-c (p=0.03). Additionally, total years of ritonavir was positively correlated with LDL-c (r=0.34, p=0.03). There were no significant relationships identified between total or LDL-c and other HIV parameters, waist-hip ratio, BMI, trunk fat or limb fat.

Microalbuminuria

Microalbuminuria was present in 15% (6/39) of subjects (95% CI, 7–30%) during follow-up; median urine A/C ratio was 5.4 (range, 1.6–125.8 mg/g). Five of 6 subjects with microalbuminuria were male and 4 were African American. African American subjects had elevated baseline A/C ratio compared to other subjects (A/C ratio 39±49 vs. 8±13 mg/g, p=0.008). No subject had macroalbuminuria. Baseline urine A/C ratio was inversely correlated with CD4% (r=−0.53, p=0.001) and positively correlated with systolic blood pressure (r=0.35, p=0.04). While no other ART medication correlated with urine A/C ratio, years of tenofovir (TDF) was highly positively correlated with baseline urine A/C ratio (r=0.53, p=0.001). In a multivariate model, years of TDF (p=0.003) and African American race (0.008) remained significant independent predictors of urine A/C ratio. Though not statistically significant, subjects with microalbuminuria tended to have higher viral load (368 vs. 100 copies/mL) and lower median CD4 counts (389 vs. 747 cells/mm³) compared to those without microalbuminuria. There was no significant difference in HOMA-IR or IGT in relation to microalbuminuria.

Longitudinal evaluation of metabolic parameters

For longitudinal analysis the mean number of visits per subject was 2.9±1 (n=39). HIV viral load was below the limit of detection in 56% at entry and 62% at last follow-up; median CD4 T-cell counts were unchanged over time. Two subjects were not on ART throughout the entire duration of follow-up. Eighty-five percent of subjects were on a PI at baseline versus 76.9% at follow-up and 44% were on stavudine at baseline versus 17.9% at last follow-up.

Of the subjects followed longitudinally, 13% met criteria for IGT at entry versus 10% at last follow-up. Mean HOMA-IR remained constant; 14 subjects had HOMA-IR >4 at entry and last follow-up (Figure 2). Increases in fasting glucose between entry and last follow-up approached statistical significance (85.9±10.2 vs. 90.9±12.5 mg/dL, p=0.06). In general, lipid values tended to improve over time, but not significantly (See Table 2). Change in lipids did not differ between those who started or discontinued a PI.

Figure 2.

Prevalence of metabolic abnormalities among HIV-infected participants in longitudinal follow-up at baseline and last visit (n=39). Abbreviations: NRTI = nucleoside reverse transcriptase inhibitor; PI = protease inhibitor; NNRTI = non-nucleoside reverse transcriptase inhibitor; BMI = body mass index (kg/m²); TC = total cholesterol (mg/dL); TG = triglycerides (mg/dL); HDL = high density lipoprotein (mg/dL). HOMA-IR, homeostasis model of insulin resistance.

Changes in body composition were characterized by an increase in total percent body fat and BMI. There were significant increases in BMI (21.9±3.4 vs. 23.5±4.4 kg/m², p=0.0002), percent body fat (19.9±8.1 vs. 22.5±9.8%, p=0.0006), leg fat (20.8±9.6 vs. 23.2±11.2%, p=0.008) and trunk fat (19.0±8.9 vs. 22.1±11.1%, p=0.002). There was a significant positive correlation between change in BMI and change in fasting glucose (r=0.37, p=0.02). Individuals who remained on a d4T, a medication associated with lipoatrophy, had smaller increases in BMI compared to those who stopped or who had not been on d4T (0.23±0.96 vs. 1.95±2.64, p=0.007). Additionally, subjects with BMI >25 kg/m² increased from 15% to 26% (p=0.0001). There were no significant changes in total lean mass or waist-hip ratio.

Discussion

Few studies have described metabolic disturbances in relation to lifelong exposure to ART among HIV-infected adolescents and young adults. In this study, we identified insulin resistance, IGT, dyslipidemia, and microalbuminuria among a substantial proportion of HIV-infected adolescents and young adults with extensive ART exposure. Of interest, those individuals with IGT also had more dyslipidemia, which may translate into a considerable increase in the risk of CVD.

The PACTG 1045 study reported data from a large cohort of vertically-infected children using OGTT and found fewer than 5% with abnormal results [24]. We identified a 15% prevalence of IGT which may be attributable to the older age of our cohort. Pediatric studies have also shown higher fasting insulin levels among HIV-infected children as a marker of insulin resistance [11, 24, 25, 26]. Chantry et al. found that 48 weeks after initiating or changing antiretroviral therapies, insulin resistance (HOMA-IR > 3.16) increased from 1% to 8% among a cohort of HIV-infected children[11], whereas 46% of our participants had HOMA-IR >3.16. Again this may be due to the relatively older age, longer duration of therapy and higher rates of PI use in the present study.

Several clinical factors were identified as associated with insulin resistance. Subjects with stavudine and ddI exposure were more likely to have elevated HOMA-IR, whereas years of AZT exposure was associated with lower HOMA-IR. Waist-hip ratio was an important predictor of insulin resistance after adjusting for ART exposure. A study by Rosso et al. [27] showed that BMI was a significant predictor of insulin resistance among HIV-infected youth, while Chantry et al.[28] concluded that greater adiposity may not be linked to an increase in HOMA-IR in HIV-infected children. We found that distribution of adiposity as measured by waist-hip ratio was more closely related to insulin resistance than BMI per se.

Similar to the D.A.D. findings on diabetes[5], we found that stavudine and ddI were related to insulin resistance. However, stavudine and ddI did not remain significant when waist-hip ratio was taken into account, suggesting that alterations in body fat distribution may underlie the observed drug effects on insulin resistance.

Dyslipidemia is frequently reported among HIV-infected patients [10, 29, 30]. We identified abnormal lipid values among a considerable proportion of our cohort and the rates are similar to those reported in the PACTG 1045 cohort [24]. Both studies also identified abnormally low HDL-c levels in a large percentage of subjects who acquired HIV in infancy or early childhood and who had extensive ART experience. PI use is frequently implicated in hyperlipidemia in adults [30] and the extensive use of these agents in our cohort likely contributed to the observed dyslipidemia.

We identify higher viral load and cumulative exposure to abacavir as factors associated with lower HDL-c and increased exposure to NNRTIs in association with greater HDL-c. These results suggest a potential protective effect of NNRTI exposure on lipids and are supported by recent studies of HIV-infected children, in which NNRTI therapy was associated with increases in HDL-c (10). In addition, NNRTIs are often used in ART regimens to replace a PI for their more favorable side effect profile with regard to lipids[1].

Microalbuminuria is a risk factor for CVD among otherwise healthy individuals [31, 32]. In this HIV-infected adolescent population microalbuminuria was present in 15% of subjects, a rate similar to what Szczech et al. [33] describes in HIV-infected adults (11%). While the confidence interval of our result (7–30%) encompasses the prevalence of 8.9% of microalbuminuria seen in healthy adolescents [34], our results are slightly higher than the estimated 10% observed in obese adolescents [35]. Our findings further highlight the importance of African American race in the risk of development of kidney injury in the context of HIV. Similar to previous findings in adults [33], we found that CD4% was inversely related to A/C ratio and there was a trend for subjects with microalbuminuria to have a greater HIV viral burden. These data implicate HIV and immune function in the development of microalbuminuria and HIV associated kidney disease.

The prevalence of insulin resistance was stable over time, whereas there were modest improvements in lipids. While this may represent a lowering of underlying CVD risk, low HDL-c remained highly prevalent and significant increases in BMI were also noted. Increases in BMI are anticipated with growth, but disproportionate increases in body fat were seen without increases in lean mass. This could reflect restoration of fat stores previously depleted with lipoatrophy, but may also indicate unhealthy gains in adiposity. Leptin, an adipokine involved in glucose homeostasis, particularly in the context of lipoatrophy [36, 37], was not measured, and its role in youth with HIV is less clear [38, 39]. As previously suggested by Blass et al.[40] and seen in the present study, emergent obesity in HIV cohorts may be driving the increase in fasting glucose levels, an indicator of pre-diabetes.

The strength of our study is the study cohort, which represents some of the most treatment-experienced HIV-infected adolescents and young adults. Additionally, our study longitudinally evaluated insulin resistance and dyslipidemia combined with detailed body composition assessments as markers for CVD risk. Although we lack a control group for direct comparison, have a relatively small sample size and did not directly measure adipokines, this is a well-characterized cohort with respect to ART exposure, metabolism, body composition and pubertal development. Socio-economic status, dietary and life-style factors that contribute to metabolic disturbances may be enriched in this cohort relative to the general population. Another potential limitation is that the subjects were not characterized by clinical criteria of lipodystrophy; however, objective assessment methods, such as DXA scans were used. Additionally, the very high prevalence of PI use and complex regimen modifications limit our ability to distinguish effects from individual drugs or classes of drugs.

In summary, our findings demonstrate that a substantial proportion of the study population have persistently altered glucose metabolism, dyslipidemia, and microalbuminuria. While modest improvements in certain lipid parameters were noted over time, insulin resistance was stable and the prevalence of overweight was on the rise. This study has significant global implications, as HIV-infected children continue to survive into adulthood with expanded access to ART and indicates the need for careful consideration of antiretroviral toxicities when treating children with HIV. Alterations in lipids and glucose metabolism may confer significant risk of future diabetes and CVD, and therefore, continued longitudinal follow-up evaluations are warranted to better understand the relationship between HIV infection, life-long ART exposure and CVD risk in this growing and unique population.

Abbreviations

ART

antiretroviral therapy

CVD

cardiovascular disease

NNRTI

non-nucleoside reverse transcriptase inhibitor

NRTI

nucleoside reverse transcriptase inhibitor

PI

protease inhibitor

ddI

didanosine

AZT

zidovudine

OGTT

oral glucose tolerance test

IGT

impaired glucose tolerance

HOMA-IR

homeostatic model of insulin resistance

BMI

body mass index

DXA

dual energy X-ray absorptiometry

urine A/C

urine albumin-to-creatinine ratio

TDF

tenofovir