Systematic Review of Antiretroviral-Associated Lipodystrophy: Lipoatrophy, but Not Central Fat Gain, Is an Antiretroviral Adverse Drug Reaction

Background Lipoatrophy and/or central fat gain are observed frequently in patients on antiretroviral therapy (ART). Both are assumed to be antiretroviral adverse drug reactions. Methods We conducted a systematic review to determine whether fat loss or gain was more common in HIV-infected patients on ART than in uninfected controls; was associated with specific antiretrovirals; and would reverse after switching antiretrovirals. Results Twenty-seven studies met our inclusion criteria. One cohort study reported more lipoatrophy, less subcutaneous fat gain, but no difference in central fat gain in HIV-infected patients on ART than in controls. Randomised controlled trials (RCTs) showed more limb fat loss (or less fat gain) with the following regimens: stavudine (versus other nucleoside reverse transcriptase inhibitors (NRTIs)); efavirenz (versus protease inhibitors (PIs)); and NRTI-containing (versus NRTI-sparing). RCTs showed increased subcutaneous fat after switching to NRTI-sparing regimens or from stavudine/zidovudine to abacavir/tenofovir. There were no significant between-group differences in trunk and/or visceral fat gain in RCTs of various regimens, but results from efavirenz versus PI regimens were inconsistent. There was no significant between-group differences in central fat gain in RCTs switched to NRTI-sparing regimens, or from PI-containing regimens. Conclusions There is clear evidence of a causal relationship between NRTIs (especially thymidine analogues) and lipoatrophy, with concomitant PIs possibly having an ameliorating effect or efavirenz causing additive toxicity. By contrast, central fat gain appears to be a consequence of treating HIV infection, because it is not different from controls, is not linked to any antiretroviral class, and doesn't improve on switching.


Introduction
Fat redistribution, also called lipodystrophy, is frequently observed in patients on long term antiretroviral therapy (ART) [1]. Some patients develop subcutaneous fat loss, or lipoatrophy; others gain fat, particularly in the breasts, dorsocervical fat pads, and viscerally. Individuals with mixed phenotypes of fat loss and fat gain also occur commonly. Fat redistribution is also associated with metabolic abnormalities, notably dyslipidaemia and insulin resistance, which increase the risk of cardiovascular disease [2].
Lipoatrophy has been associated with exposure to thymidine analogue nucleoside reverse transcriptase inhibitors (NRTIs) [3]. Central fat gain is also assumed to be an adverse drug reaction [4]. However, there is evidence that visceral abdominal fat in HIVinfected patients on ART is not increased relative to healthy controls [5]. Untreated HIV infection eventually results in wasting, including loss of adipose tissue. Fat gain, which is widely prevalent in the general population and increases with age, may in part be the result of effective ART reversing fat loss due to HIV infection. It is important to determine whether lipodystrophy is an adverse drug reaction to avoid unnecessary drug substitutions which may result in risks of virologic failure, new toxicities, and undermining patient confidence if the lipodystrophy does not improve. Treatment adherence is compromised when patients believe they have lipodystrophy from antiretrovirals [6].
If fat loss and fat gain were adverse antiretroviral drug reactions they would occur more commonly in HIV-infected patients on ART than in HIV-uninfected controls. Second, fat loss and/or fat gain would be associated with specific antiretroviral drugs or drug classes. Third, fat loss and/or fat gain would reverse after switching the identified antiretroviral drugs. We conducted a systematic review to test those three assumptions.

Eligibility criteria
Types of studies. To answer the question 'Does fat loss and/ or fat gain occur more commonly in patients on ART than in HIV-uninfected controls?' we included prospective cohort studies comparing HIV-infected patients with ART exposure to population controls either known or presumed to be HIV-uninfected. To answer the questions 'Is fat loss and/or fat gain associated with specific antiretroviral drugs?' we included randomised controlled trials comparing antiretroviral regimens. To answer the question 'Is fat loss and/or fat gain reversed after switching antiretroviral drugs?' we included studies where participants with virologic suppression were randomised to continue their current ART regimen or switch to an alternative regimen.
Participants. We included both ART-naïve and ARTexperienced HIV-infected patients who were at least 12 years old. For the cohort studies we included control participants who were presumed to be HIV-uninfected. We excluded studies with fewer than 20 participants in any arm.
Interventions. We included studies that used any antiretroviral regimens, given for at least 24 weeks, with the exception of those containing hydroxyurea.
Outcome measures. We included studies with at least one objective measure of fat distribution done at baseline, and repeated at least once, at a minimum of 24 weeks after baseline. Objective methods of measuring fat distribution included: dualenergy x-ray absorptiometry (DEXA), computerized tomography (CT), or magnetic resonance imaging (MRI). We included measures done both as primary or secondary outcomes, and in the whole study population, or within a sub-study. Specific outcomes included: To assess fat loss:

Search strategies
We searched two electronic journal databases, PubMed and EMBASE, for articles published between 1 January 1990 and 7 July 2011. We hand-searched electronic databases for the Conferences on Retroviruses and Opportunistic Infections and the International AIDS Society conferences, from 2001. There was no language restriction, provided that an English translation of the abstract was available. The PubMed search strategy terms were as follows: HIV

Data collection
Two authors (RdW and KC) independently reviewed all study abstracts identified by the search strategy, using a specially designed eligibility form. We obtained the full articles, conference abstracts or conference posters for all studies that met the inclusion criteria. An independent translator reviewed articles that were published in languages other than English. We resolved disagreements as to study eligibility through consensus and discussion with the third author (GM) if necessary. One author (RdW) extracted data using a data extraction form; another author (KC) checked the extracted data. Two authors (RdW and KC) assessed the risk of bias of all included studies [7].

Results
We identified a total of 27 studies for inclusion in the review: one cohort study comparing HIV-infected patients with controls [8]; 18 randomised controlled trials comparing antiretroviral regimens [3,4,[9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24]; seven switching studies [25][26][27][28][29][30][31]; and one study that fulfilled the criteria for both randomised controlled trials and switching studies [32]. Our search of PubMed and Embase  databases identified 3031 potential abstracts. We identified a further 67 articles through conference databases or other sources. After removing duplicate records, we screened 3042 abstracts and excluded 2966 as they did not meet our inclusion criteria. We retrieved and assessed the full text articles for the remaining 76, and excluded 49 that did not meet our inclusion criteria (see Figure 1 and Table S1). We assessed two articles that were published in Spanish through means of a translator. We included the full text articles for five studies that we identified through conference abstracts.
HIV-infected patients compared with healthy controls. The FRAM2 study compared fat distribution in HIV-infected patients with healthy controls, using MRI at two time-points separated by about five years [8]. The control participants were recruited from the Visceral Fat and Metabolic Rate in Young Adults sub-study of the Coronary Artery Risk Development in Young Adults study. They were selected as they had previous experience of fat distribution investigations, and had a similar age and ethnic distribution to most HIV-infected patients in the United States [35]. Although the analyses included HIVinfected patients who had never been on ART (11.8% at baseline and 5.7% at year 5), we included the study as only a small minority were not on ART. Sub-group analyses by ART status were not done.
There was clear evidence of fat loss in HIV-infected people, 53% of whom had lipoatrophy (defined as leg SAT below the 10 th percentile of controls) after five years of observation. Longer duration of stavudine use was associated with less leg SAT. After five years there was significantly less SAT at all sites in HIVinfected men, but only in the limbs in HIV-infected women. Multivariable analysis showed that increase in SAT over five years was less in HIV-infected people at all sites except the lower trunk.  By contrast, there was no evidence of regional fat accumulation in HIV-infected people. After five years the amount of trunk SAT and VAT was similar in HIV-infected and control women, while HIV-infected men had less fat at all sites than control men. The gains in VAT over five years were similar in HIV-infected people and controls.  Tables 2, 3, 4, 5, and 6, and the incidences of peripheral lipoatrophy (defined as $20% loss in limb fat) are summarised in Tables 7,8,9,10,and 11. In general, participants who received NRTI-sparing regimens gained more (or lost less) SAT or limb fat over time than those on NRTI-containing regimens. However one study that compared an abacavir-containing regimen with an NRTI-sparing regimen found no significant between-arm differences in change in subcutaneous thigh fat or SAT [32]. Participants who received PI-containing regimens also gained more (or lost less) SAT or limb fat over time than those on efavirenz-based regimens [3,13,19,24,33]. The incidence of peripheral lipoatrophy (defined as $20% loss of limb fat) was significantly lower in participants on NRTI-sparing regimens compared with those on NRTI-containing regimens, and in those on PI-containing regimens compared with those on EFV-based regimens [9,11,13,19].
There were no significant differences in average gains in SAT or limb fat over time in the four studies that compared abacavir-with tenofovir-based regimens [3,14,16,23]. One study found that participants who received an abacavir-containing regimen gained limb fat over time, in contrast to participants who received a stavudine-containing regimen who lost limb fat over time (gain of 913 g versus loss of 1578 g, p,0.001) [22]. Another study found that participants who received a zidovudine-lamivudine-containing regimen gained limb fat over time, in contrast to participants who received a stavudine-didanosine-containing regimen who lost limb fat over time (gain of 1.7% versus loss of 26.4%, p value not reported) [33].
Unboosted atazanavir was associated with significant reduction in limb fat at 96 weeks, while there was no significant change in    limb fat in the ritonavir-boosted atazanavir arm (both arms were on stavudine and lamivudine) [15]. The proportion of participants with $20% loss in limb fat was significantly greater in the unboosted atazanavir arm at 96 weeks [15]. However, there were no significant between-group differences in absolute change from baseline in SAT or limb fat with ritonavir-boosted versus unboosted atazanavir [15]. There were no significant betweengroup differences in absolute change from baseline in SAT or limb fat in a study comparing ritonavir-boosted tipranavir (at the registered dose) with ritonavir-boosted lopinavir [17].
In the one study that compared raltegravir with efavirenz, average gains in limb fat over time and the incidence of lipoatrophy were similar in both groups [10].
Fat gain: changes in trunk fat or VAT with different antiretroviral regimens. In general, participants randomised to different ART regimens gained similar amounts of trunk fat or VAT over time. The changes from baseline in trunk fat and VAT are summarised in Tables 12,13,14,15,16. There were no significant between-group differences in changes from baseline in trunk fat or VAT in the six studies that compared NRTI-sparing and NRTI-containing regimens [4,9,11,12,21,32], and in the four studies that compared different NRTI-containing regimens [3,22,23,33]. Similarly, there were no significant differences in the incidences of lipohypertrophy (defined as .20% gain in trunk fat) [9,11,19].
The results of studies that compared PIs and NNRTIs were not consistent. One study found that participants who received ritonavir-boosted atazanavir had significantly greater increases in trunk fat at week 96 than those who received efavirenz (36.5% versus 21.1% respectively, p = 0.028) [3]. Another study found no significant changes at week 48 in VAT between those who received unboosted atazanavir compared with those who received efavirenz (15.3 cm 2 versus 14.1 cm 2 respectively, 95% confidence interval for the difference: 210.4 to 12.6 cm 2 ) [24]. There were no significant between-group differences in changes from baseline in trunk fat, or in incidence of lipohypertrophy in one study that compared ritonavir-boosted lopinavir and efavirenz [19]. One study found that participants who received efavirenz gained more trunk fat over time on average than those who received nelfinavir, however it was not reported whether or not the difference was statistically significant [33].
There were no significant between-group changes from baseline in trunk fat or VAT in those who received ritonavir-boosted atazanavir compared with unboosted atazanavir in one study [15]. A study that compared ritonavir-boosted tipranavir and ritonavirboosted lopinavir found small, but statistically significant betweengroup differences in changes from baseline in both trunk fat and VAT: those who received ritonavir-boosted tipranavir lost trunk fat over time, while those who received ritonavir-boosted lopinavir gained trunk fat over time [17].
In the one study that compared raltegravir with efavirenz, gains in trunk fat over time were similar in both groups [10].
Changes in limb fat and SAT after switching antiretroviral regimens. In general, participants who were switched away from NRTI-containing, or more specifically thymidine analogue-containing, regimens gained limb fat over time, when compared with participants who continued NRTI-or thymidine analogue-containing regimens, who generally lost limb fat [27][28][29][30]32,34].
There were no significant between-group differences in changes from baseline in limb fat or SAT in studies that switched to NNRTI-from PI-containing regimens [31], to ritonavir-boosted atazanavir from other ritonavir-boosted PI regimens [25], or to raltegravir from PI regimens [26]. Changes from baseline in limb fat, subcutaneous thigh fat and SAT are summarized in Tables 17, 18, 19, 20, and 21.
Changes in trunk fat or VAT after switching antiretroviral regimens. In general, participants who were switched away from NRTI-containing regimens, or from thymidine analoguecontaining regimens, had similar increases in trunk fat over time to those who continued NRTI-or thymidine analogue-containing regimens [29,30].
There were no significant between-group differences in changes from baseline in trunk fat or VAT in studies that switched to NNRTI-from PI-containing regimens [31], to ritonavir-boosted atazanavir from other ritonavir-boosted PI regimens [25], or to raltegravir from PI regimens [26].
Changes from baseline in trunk fat and VAT are summarised in Tables 22, 23, 24, 25, and 26.

Discussion
We found overwhelming evidence that lipoatrophy is an antiretroviral adverse drug reaction. Subcutaneous fat volumes are considerably lower in patients on ART than in controls, subcutaneous fat loss progresses on ART, is associated with stavudine and zidovudine use, and partially reverses after switching to abacavir, tenofovir or an NRTI-sparing regimen. By contrast, central fat gain does not appear to be an antiretroviral adverse drug reaction. Visceral and trunk fat volume is no different in women on ART compared with control women, and is less in men on ART than in control men. Visceral fat accumulates at the same rate in patients on ART and controls. Finally, central fat gain generally occurs at similar rates in HIV-infected patients randomised to different ART regimens, is not associated with any specific antiretroviral drug or drug class, and does not reverse on switching antiretrovirals. We believe that this evidence indicates that central fat gain is a consequence of treating HIV infection, which normalizes the concentrations of inflammatory markers such as TNF-a (tumour necrosis factor alpha) that are known to cause wasting [36]. Lipoatrophy occurring together with central fat gain results in an unusual appearance, which may have persuaded clinicians that the fat gain is an antiretroviral adverse drug reaction. The fact that diet and exercise have been shown to improve central fat gain in patients on antiretroviral therapy  provides some support for our conclusion that it is a consequence of lifestyle [37][38][39].
Efavirenz is associated with a higher risk of limb fat loss than PIs when combined with NRTIs that cause fat loss. A possible explanation for this observation is that the anti-apoptotic properties of PIs partially ameliorate the loss of adipocytes by increased apoptosis that is induced by NRTIs [40]. The observation that unboosted atazanavir, which is a non-peptidomimetic PI that does not have anti-apoptotic properties, is associated with more limb fat loss than ritonavir (a peptidomimetic PI)-boosted atazanavir supports this hypothesis. Alternatively, efavirenz may increase the adipocyte toxicity of thymidine analogue NRTIs. Efavirenz has been shown to be more toxic to adipocytes and to release more inflammatory cytokines than lopinavir-ritonavir [41]. Furthermore, efavirenz also displays mitochondrial toxicity in hepatocytes [42]; although we could find no data to support this, it is possible that it has a similar effect on adipocytes. Mitochondrial toxicity is thought to be a key mechanism of thymidine analogue-induced lipoatrophy [3].
Our study has several limitations. First, many of the studies that reported objective measures of fat redistribution by ART regimen were convenience sub-studies of randomised controlled trials, therefore patients undergoing DEXA and/or CT scans were not randomised. However, fat distribution sub-studies generally reported that the characteristics of the patients undergoing imaging of fat tissue were not different from the parent trial population. Many RCTs were powered according to efficacy endpoints, and not for objective measures of fat distribution. Second, many studies reported only summary statistics and different outcome measures were reported, which prevented us from pooling data from different studies. Third, some studies did not report statistical analyses of changes in body fat distribution by study arm. Fourth, we were unable to assess the role of older protease inhibitors (other than nelfinavir) or nevirapine in fat redistribution as no studies with those antiretrovirals fulfilled our inclusion criteria. Fifth, we found only one study that compared fat changes over time in patients on ART and HIV-infected controls that was conducted in the United States. This limits our ability to generalise findings to other populations. Finally, we cannot exclude the causative role of specific antiretrovirals in focal forms of fat gain, such as buffalo humps, as the included studies reported trunk or visceral fat changes only.
In conclusion, our systematic review supports the hypothesis that peripheral lipoatrophy, but not central fat gain, is an antiretroviral adverse drug reaction. Lipoatrophy can be avoided and at least partially reversed by avoiding thymidine analogue nucleoside reverse transcriptase inhibitors. Central fat gain appears to be a consequence of treating HIV-infection, and reflects patterns of fat gain seen in the HIV-uninfected population.