The natural history of HIV-associated lipodystrophy in the changing scenario of HIV infection

Authors


Abstract

Objectives

In long-term HIV-infected patients, peripheral lipoatrophy (LA) and central lipohypertrophy (LH) appear to be related to the same insults (virus and antiretroviral drugs), but are likely to be associated with different fat depot physiologies. The objective of this study was to describe the natural history of lipodystrophy assessed using dual energy X-ray absorptiometry (DEXA) and computed tomography (CT) in a large HIV out-patients metabolic clinic.

Methods

An observational retrospective study was carried out including HIV-infected patients recruited at the Metabolic Clinic of Modena, Modena, Italy, who were assessed for lipodystrophy and had at least two anthropometric evaluations using DEXA for leg fat per cent mass and abdominal CT for visceral adipose tissue (VAT). Factors associated with leg fat per cent and VAT changes were analysed using multivariable generalized estimating equation (GEE) regression models.

Results

A total of 6789 DEXAs and 7566 CT scans were evaluated in the observation period. A total of 1840 patients were included; the mean age was 45.2 ± 7.2 (standard deviation) years, 621 (34%) were women, and the median HIV infection duration was 176 (interquartile range 121–232) years. According to the GEE multivariable regression analysis, leg fat per cent evaluated with DEXA appeared to increase over calendar years (ß = 0.92; P < 0.001); moreover, a progressive increase in VAT was observed in the cohort (ß = 5.69; P < 0.001). No association with antiretroviral drugs was found.

Conclusions

In our study, neither LA nor LH appeared to be associated with antiretroviral drug exposure. We observed a progressive increase in LH in HIV-infected patients over calendar years. This anthropometric change, together with loss of appendicular lean mass, could describe a physiological aging process in HIV-infected patients.

Introduction

In the late 1990s, reports of unusual changes in body fat distribution, named “lipodystrophy” (LD), in HIV-infected patients began to appear, dampening enormous enthusiasm about the improvement in survival and quality of life provided by highly active antiretroviral therapy (HAART), which had just become available at that time [1-4].

The idea that the most distinctive aspect of LD is lipoatrophy (LA) was supported by data provided by the Fat Redistribution and Metabolic Changes in HIV Infection (FRAM) study. In this cohort study, peripheral LA was found to occur commonly in both men and women with HIV infection and was not associated with reciprocally increased abdominal visceral adipose tissue [5, 6].

Men were more likely than HIV negative controls to have higher triglycerides and lower high-density lipoprotein (HDL) and low-density lipoprotein (LDL) cholesterol, while in men and women more visceral adipose tissue (VAT) and upper trunk fat (subcutaneous adipose tissue) were independently associated with insulin resistance [7, 8].

The pathophysiology of LA is multifactorial: contributing factors are CD4 lymphocyte cell count, HIV clinical stage, race, sex, exercise level, and age at the start of antiretroviral therapy (ART) [9], but the driving force underlying LA is undoubtedly cumulative exposure to thymidine analogue (TA) drugs. These drugs, in particular stavudine and to a lesser extent zidovudine, antagonize mitochondrial DNA polymerase function, producing apoptosis of fat cells [10, 11].

Earlier detection and treatment of HIV infection [12], as well as the use of antiretroviral drugs with less deleterious effects on body fat, mean that it is reasonable to hypothesize that there will be a decrease in the prevalence of LD in the next few years.

The most common site for abnormal fat accumulation [lipohypertrophy (LH)] in treated HIV-infected patients is the VAT compartment, but excess fat has also been detected in the dorsocervical, hepatic, cardiac, intrathoracic, and subcutaneous regions, which may also contribute to metabolic abnormalities [8, 13-15]. Increased fat also occurs in the intermuscular [16] and intramyocellular [17] compartments, affecting glucose homeostasis.

The pathophysiology of LH is complex, including factors related to pre-HIV exposure body composition, HIV effects and ART-associated sequelae. Protease inhibitors (PIs) and nonnucleoside reverse transcriptase inhibitors (NNRTIs) may activate adipocyte nuclear transcription factors with downstream effects leading to LH [18]; nevertheless, a direct impact of HAART on VAT is yet to be demonstrated and no clinical studies have ever shown a benefit of switching from any antiretroviral (ARV) drug classes. The proinflammatory environment, linked to ART, might be an important pathophysiological factor for LH.

Aging also is associated with a physiological change in fat redistribution. Previous studies in uninfected subjects have found that younger and middle-aged adults gain 0.5–1.0 kg per year [19]. In the general population, a 60−85% increase in fat mass, predominantly represented by VAT, is expected between 25 and 65 years of age; in the same period there is a 20% decline in skeletal muscle mass. Moreover, HIV-infected patients display features of premature aging affecting the bones, brain, vascular wall, muscles, kidney and liver, which result collectively from long-term HIV infection, immune depletion and the toxicity of some ARVs.

Little is known about what happens to LH over the long term in HIV-infected patients. Presumably the body fat changes currently observed in HIV-infected patients are the net result of competing phenomena: on one side the natural history of LH as a result of the impacts of HIV and HAART and on the other the physiological body fat changes observed in the aging population.

Capeau et al. postulated that peripheral LA and central LH result from the same insults (virus and ARV drugs), but are likely to be related to different fat depot physiologies. Lipoatrophy is linked to severe mitochondrial dysfunction, oxidative stress and inflammation. By contrast, LH might be related to mild mitochondrial dysfunction and cortisol activation promoted by inflammation. Both LA in the lower part of the body and abdominal LH are involved in insulin resistance and metabolic disorders, as observed in genetic lipodystrophies and the metabolic syndrome [18], a constellation of abnormalities that leads to an increased risk of cardiovascular disease and diabetes in the general population.

The objective of this study was to describe the natural history of LD, assessed using dual energy X-ray absorptiometry (DEXA) and computed tomography (CT), in a large, well-characterized HIV-infected cohort. This will help to contextualize the impact of drugs on LD in the wider scenario of aging HIV-infected patients.

Methods

Subjects

This observational retrospective study included 1840 HIV-infected patients recruited at the metabolic clinic of the University of Modena and Reggio Emilia in Italy between 2005 and 2012. Patients from HIV clinics throughout Italy are referred to this metabolic clinic or have direct access to the multidisciplinary treatment available at the clinic, where they undergo comprehensive metabolic and anthropometric diagnostic and therapeutic assessments for the presence of LD and noninfectious comorbidities [20].

Inclusion criteria were serologically documented HIV-1 infection, age > 18 years, at least 18 months of ART exposure, clinical assessment for LD using the Multicentre AIDS Cohort Study (MACS) criteria [21] definition and at least two anthropometric evaluations of LA with DEXA for leg fat percentage mass and of LH with abdominal CT for VAT. A form providing signed informed consent to participate in this study was obtained from each patient. The study was approved by the local institutional review board (Comitato Etico Provinciale di Modena).

Demographics

Demographic and clinical data, including duration of HIV infection, ART history and lifestyle, were obtained by medical chart review. Smoking status was categorized as follows: nonsmoker (< 1 cigarette per day) or smoker (more or less than 10 cigarettes per day). Alcohol consumption was categorized as no alcohol consumed (< 10 g of ethanol per day) or alcohol consumed (≥ 10 g of ethanol per day). Physical activity when present, was defined as mild or intense when < 4 or ≥ 4 h per week of exercise, respectively, was reported. HIV and ART histories, including cumulative exposure to nucleoside reverse transcriptase inhibitors (NRTIs), non-nucleoside reverse transcriptase inhibitors (NNRTIs), proetase inhibitors (PIs) and integrase inhibitors (INIs).

Anthropometric measurements

All patients underwent physical examinations on the day on which fasting blood was obtained. Waist circumference (WC) was measured at the narrowest point midway between the lowest rib and the iliac crest at the end of expiration with the subject standing.

All circumference measures were calculated as the average of three measurements. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in metres.

Lipodystrophy definition

Clinical assessment of LD was performed according to the MACS definition [21]. For the purpose of the study, we categorized patients as having either LA or LH, including in the latter category both patients with central fat accumulation and those with mixed forms.

Body composition

Whole-body DEXA for body composition was assessed with Hologic Discovery (Hologic, Inc., Waltham, MA). The percentage of fat tissue in the legs (leg fat percentage) was used as the anthropometric endpoint for LA.

A single CT image at the level of the L4 vertebra was taken for quantification of VAT and subcutaneous adipose tissue (SAT) using a 64-multislice CT scanner (LightSpeed VTC; General Electric Medical System). Each CT image was analysed using a software application based on the Advantage Windows 4.4 GE medical system.

Cardiometabolic risk factors

Levels of total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides, glucose and insulin were measured after an overnight fast. Insulin resistance was calculated using the homeostasis model assessment equation (HOMA-IR) [= fasting insulin (mU/ml) × fasting glucose (mmol/L)/22.5].

Statistical analyses

Baseline characteristics were compared among LD phenotypes. Categorical variables were analysed using the χ2 test, while continuous variables were compared using analysis of variance (ANOVA) or the Kruskal−Wallis test where appropriate. Post hoc analyses were performed using Bonferroni adjustment.

Lowess smoothing curves were drawn to describe the average distribution of VAT and leg fat per cent in the study population during the follow-up period. Different lines were drawn for each LD phenotype.

Each patient had at least two clinical assessments and we evaluated factors associated with leg fat per cent and VAT changes using generalized estimating equation (GEE) regression models for panel data. We considered the change between two consecutive measurements as the outcome variable and we included in the model, as independent variables, characteristics of interest assessed the first evaluation of each pair of observations. Changes in VAT and leg fat per cent were normally distributed.

Variables of interest included in the models were: baseline VAT, LD phenotypes, age at initiation of ART, sex, CD4 nadir, time between observations, and cumulative exposure to single class regimen (in months).

Statistical significance was defined as a P-value < 0.05.

All statistical analyses were performed using stata 12.1 Intercooled version for Mac (StataCorp, College Station, TX).

Results

A total of 1840 HIV-infected patients were included in the study, of whom 621 (34%) were female. At baseline, the mean age of the cohort was 45.2 ± 7.2 (SD) years, and most of the patients had a long history of HIV disease (mean ± SD 14.2 ± 5.8 years).

Each patient contributed multiple clinical and anthropometric evaluations, with a median number of 4 (range 2 to 11) DEXA assessments for LA and abdominal CTs for LH. A total of 6789 DEXAs and 7566 CTs were evaluated in the observation period.

Table 1 shows the demographic, anthropometric and metabolic characteristics of the cohort.

Table 1. Demographic, anthropometric and metabolic characteristics of the cohort
 No LDLipoatrophyLipohypertrophyOverall P-value
  1. APO, apolipoprotein; ATPIII, adult treatment plan III; SAT, subcutaneous adipose tissue; TAT, total adipose tissue; CDC, Centers for Disease Control and Prevention; LD, lipodystrophy; SD, standard deviation; IQR, interquartile range; BMI, body mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; HOMA, homeostasis model assessment of insulin resistance; IDU, injecting drug user; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, nonnucleoside reverse transcriptase inhibitor; PI, protease inhibitor; VAT, visceral adipose tissue; VL, viral load.
  2. Post hoc analyses were carried out using Bonferroni adjustment. Significant differences between groups are indicated by symbols (Ш and Ø). Significant P-values are indicated in bold.
Number298708834 
Women [n (%)]85 (28.52)197 (27.82)339 (40.65)< 0.001
ШØШ Ø
Age (years) [mean (SD)]43.13 (± 9.9)44.7 (± 6.3)46.3 (± 7.47)< 0.001 (all)
Physical activity [n (%)]   < 0.001
None162 (54.79)344 (60.55)495 (69.57) 
Mild84 (28.77)175 (25.47)187 (23.23)
Intense48 (16.44)96 (13.97)58 (7.2)
ШØØ
Smoking [n (%)]   < 0.001
Nonsmoker162 (53.82)344 (49.5)495 (59.93) 
< 10 cigarettes/day49 (16.28)128 (18.42)131 (15.86)
> 10 cigarettes/day81 (26.91)214 (30.79)180 (21.79)
Alcohol [n (%)]   < 0.001
None159 (52.82)370 (53.24)414 (50.12) 
< 10 g/day130 (43.19)306 (44.03)387 (46.85)
> 10 g/day3 (1)10 (1.44)4 (0.61)
BMI [mean (SD)]23.7 (± 3.075)21.29 (± 2.36)24.9 (± 3.83)< 0.001 (all)
Waist circumference (cm) [mean (SD)]84.8 (± 9.25)79.6 (± 6.51)90.6 (± 9.59)< 0.001 (all)
VAT (cm3) [median (IQR)]97 (62–145)99.6 (59–127)137 (97–195)< 0.001 (all)
SAT (cm3) [median (IQR)]139 (93–195)70 (39–108)163 (107–234)< 0.001 (all)
TAT (cm3) [median (IQR)]246 (180–325)170 (119–230)308 (240–401)< 0.001 (all)
Leg fat percentage [mean (SD)]19.4 (± 10.4)9.4 (± 5.5)15 (± 9.1)< 0.001 (all)
Systolic blood pressure (mmHg) [mean (SD)]119 (14)119 (15)122 (15)< 0.001
ШØШ Ø
Diastolic blood pressure (mmHg) [mean (SD)]76 (10)77 (12)79 (12)< 0.001
ШШ ØШ Ø
Triglycerides (mg/dL) [median (IQR)]129 (82–191)155 (101–222)(105–201)< 0.001
Ш ØШ(115–250)
Ø
Total cholesterol (mg/dL) [mean (SD)]187 (40.8)187 (47.07)193 (47.8)0.027
ШШ
HDL cholesterol (mg/dL) [mean (SD)]44 (14)44 (14)47 (16)0.001
Ш ØШØ
LDL cholesterol (mg/dL) [mean (SD)]113 (34)112 (36)117 (39)0.07
APO1 lipoprotein (mg/dL) [mean (SD)]144 (28)140 (29)143 (27)0.14
APOB lipoprotein (mg/dL) [mean (SD)]94 (26)99 (28)104 (28)< 0.001 (all)
Metabolic syndrome (ATPIII criteria) [n (%)]32 (10.63)102 (14.68)228 (27.6)< 0.001
ШØШ Ø
Glucose (mg/dL) [median (IQR)]91 (85–97)90 (83–98)94 (86–104)< 0.001
ШØШ Ø
HOMA [median (IQR)]2.29 (1.4–4.01)2.62 (1.7–4.2)3.7 (2.3–5.8)< 0.001 (all)
Insulin (ng/mL) [median (IQR)]10.6 (6.4–16.7)12.1 (7.9–18)16.1 (10.5–23.95)< 0.001 (all)
HIV risk factors [n (%)]   < 0.001
IDU57 (18.94)249 (35.83)224 (27.12) 
Homosexual115 (38.21)210 (30.22)182 (22.03)
Heterosexual100 (33.22)193 (27.77)335 (40.56)
Other29 (9.63)43 (6.19)85 (10.29)
Ш ØШØ
CDC classification group C [n (%)]58 (19.86)183 (26.87)219 (27.41)0.03
Ш ØØШ
HIV exposure (months) [median (IQR)]122 (58–205)184 (133–239)182 (130–232)< 0.001
Ш ØШØ
VL undetectable [n (%)]230 (77.2)558 (78.8)666 (79.86)0.613
CD4 count nadir (cells/μL) [median (IQR)]221 (132–317)156 (63–261)169 (61–265)< 0.001
Ш ØШØ
Current CD4 count (cells/μL) [median (IQR)]548 (395–906)500 (372–687)525 (378–700)0.09
Cumulative exposure to NRTIs (months) [median (IQR)]55 (27–98)95 (30–145)121 (62–142)0.17
Cumulative exposure to NNRTIs (months) [median (IQR)]3 (0–35)8.5 (1–25)17 (0–57)0.59
Cumulative exposure to PIs (months) [median (IQR)]34 (3–50)51 (17–100)30 (7–52)0.2

Figure 1 shows the Lowess curve for leg fat percentage, depicting LA changes with calendar year according to baseline LD phenotype. What can be observed is that both patients with LA and those with LH appeared to gain leg fat over calendar years. The gain in leg fat in patients with LH is expected, given that, per protocol, these patients included those with the mixed form of LD. This means that they may also have had appendicular fat loss. The same phenomenon did not seem to be present in patients with no LD.

Figure 1.

Lowess curve for leg fat percentage showing lipoatrophy changes over calendar years according to baseline lipodystrophy (LD) phenotype.

Table 2 shows determinants of LA changes over time. It can be observed that the “slow reversibility” of LA is supported by the statistical model, being associated with time between observations. Older patients, as expected, had higher leg fat. Male sex and previous LD diagnosis were negatively associated with leg fat increase. What is more interesting is that cumulative exposure to NRTIs, NNRTIs and PIs did not appear to be associated with leg fat increase; in contrast, we found a protective effect associated with cumulative INI exposure.

Table 2. Factors found to be associated with leg fat per cent changes using generalized estimating equation (GEE) regression models for panel data
 Change in leg fat per cent
β95% CIP-value
  1. ARV, antiretroviral; CI, confidence interval; LD, lipodystrophy; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, nonnucleoside reverse transcriptase inhibitor; INI, integrase inhibitor; PI, protease inhibitor.
Sex (male)−0.49–0.67; −0.32 < 0.001  
Lipodystrophy phenotypes
No LD   
Lipoatrophy   
Mixed forms   
Age at ARV initiation0.010.004; 0.02 
Time between abservations (per 1 year)0.920.80; 1.04< 0.001
NRTI cumulative exposure0.0005−0.001; 0.001 
PI cumulative exposure  
NNRTI cumulative exposure0.0007−0.001; 0.0030.517
INI cumulative exposure0.050.03; 0.06< 0.001
Baseline leg fat per cent−0.06−0.08; −0.05< 0.001
CD4 count nadir0.0004–0.0001; 0.001  

Figure 2 shows the Lowess curve for VAT, depicting LH changes with calendar year according to baseline LD phenotype. What can be observed is that both patients with LA and those with LH appeared to gain leg fat over calendar years. As expected, this fat gain was more pronounced in patients with a baseline LH phenotype.

Figure 2.

Lowess curve for visceral adipose tissue (VAT) during the follow-up period. LD, lipodystrophy.

Table 3 shows determinants of LH changes over time. It can be observed that a progressive increase in VAT was observed in the cohort, being associated with time between observations.

Table 3. Factors found to be associated with visceral adipose tissue (VAT) changes using generalized estimating equation (GEE) regression models for panel data
 Change in VAT
β(95% CI)P-value
  1. ARV, antiretroviral; CI, confidence interval; LD, lipodystrophy; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, nonnucleoside reverse transcriptase inhibitor; INI, integrase inhibitor; PI, protease inhibitor.
Sex (male)13.2310.47; 15.99 < 0.001  
Lipodystrophy phenotypes5.85; 13.09  
No LD  
Lipoatrophy0.333  
Mixed forms< 0.001  
Age at ARV start0.510.36; 0.67< 0.001 
Time between abservations (per 1 year)5.683.69; 7.66 < 0.001  
NRTI cumulative exposure0.02–0.006; 0.050.120  
PI cumulative exposure  
NNRTI cumulative exposure0.02−0.02; 0.060.282
INI cumulative exposure0.05−0.25; 0.370.733
Baseline VAT−0.28−0.30; −0.26< 0.001
CD4 count nadir0.0004–0.007; 0.008  

VAT increase was also associated with male sex, age and baseline LD phenotype. VAT change also did not appear to be associated with cumulative exposure to NRTIs, NNRTIs, PIs or INIs.

Discussion

This observational study, describing a large well-characterized population, clearly demonstrates that regional fat changes are an evolving phenomenon in HIV-infected patients on ART. We observed a slow reduction in LA, by means of a progressive gain in leg fat, and a significant increase in LH, by means of a progressive gain in VAT. The description of this pattern across calendar years is quite innovative.

Only a few previous studies have described the natural history of LD in HIV-infected patients. The most frequently quoted is FRAM2, which demonstrated, in a 5-year follow-up period, that gains in adipose tissue were similar in HIV-infected and control participants [22]. HIV-infected participants had significant subcutaneous LA at the baseline examination and, 5 years later, relative LA persisted compared with controls, even in those who discontinued ARV drugs associated with LA, such as stavudine, during this period. It should nevertheless be considered that these data may not be reproducible in European Union (EU) cohorts; in fact, in the FRAM2 cohort, which mainly consisted of African Americans, baseline BMI was 24.5 (interquartile range 22.1–27.2), higher than that expected in EU HIV-infected cohorts.

A second study from the USA was performed in the MACS cohort [23]. Subcutaneous LA was observed in HIV-infected patients, even those without clinical evidence of LD, compared with age-matched HIV-uninfected men. Despite markedly lower BMI, HIV-infected men with LD had a similar amount of VAT to HIV-uninfected men and tended to have more rapid increases in waist circumference over 6 years of follow-up. These longitudinal increases in WC may contribute to the development of cardiovascular risk in HIV-infected patients with LD.

According to a European study, carried out in the Swiss HIV cohort, LA has become less frequent from 2000 to 2006 [24]. A weight gain of more than 5 kg was associated with the use of atazanavir and lopinavir. Unfortunately, no anthropometric measurements were used to support this clinical observation.

All the above-mentioned studies described LD in the pre- or early-HAART period, when a considerable proportion of patients were receiving ARV regimens that included thymidine analogues and first-generation PIs.

In our study, neither LA nor LH appeared to be associated with ARV exposure. This observation is somewhat reinforced by the findings of randomized clinical trials in treatment-naïve and -experienced patients which included drugs brought to the market after 1 January 2006. They showed a neutral or sometimes beneficial impact of these drugs on anthropometric determinants of LA or LH.

In particular, we were unable to identify studies showing a negative impact of ARV drugs or regimens on limb fat loss. This leads us to hypothesize that the risk of long-term mitochondrial toxicity, responsible for facial and appendicular fat loss, no longer exists for drugs brought to the market after 1996, and that LA will progressively become less prevalent in the “metabolic context” of HIV disease. We identified a protective effect of INIs. The observational nature of the study means that it cannot be determined whether there was a causal association between INI exposure and LA, and we may argue this depicts contemporary cohort observations.

We observed a progressive increase in LH in HIV-infected patients over calendar years, which was not associated with drug class exposure. It has been postulated that this is an anthropometric change that, together with loss of appendicular lean mass, describes the physiological aging process and predicts the frailty phenotype [19].

Future studies will clarify the impact of drugs, rather than HIV- and host-specific factors, on LD progression in the context of the aging process affecting people living with HIV.

Ancillary