values derived from graph; c

values derived from graph; c. 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 sn-Glycero-3-phosphocholine 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 HIV-infected Hmox1 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. Methods 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 sn-Glycero-3-phosphocholine 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-na? ve and sn-Glycero-3-phosphocholine ART-experienced 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 sn-Glycero-3-phosphocholine 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: dual-energy 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: Change from baseline in limb fat Change from baseline in subcutaneous adipose tissue (SAT) Proportion with 20% loss in limb fat Proportion with 20% loss in SAT To assess fat gain: Change from baseline in trunk fat Change from baseline in visceral adipose tissue (VAT) Proportion with 20% gain in trunk fat Proportion with 20% gain in VAT. 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.