Paris is not for everyone. People who hate beautiful paintings, beautiful people, beautiful bridges, beautiful clothes, beautiful radio towers, beautiful vistas, and tip-top food, for example, had better stay away from Paris. But for the aesthete, the effete, and the wide-eyed, Paris is one swell town.
Even amidst this glut of pulchritude, though, the casual tourist with a scientific leaning doesn't have to look far for the most sublime beau ideal in the Ile de France. It's strung from the dome of the church of Saint Geneviève, also known as the Panthéon, just behind the Sorbonne, in the fifth arrondissement.
Foucault started off like a lot of people who attended the 5th International Workshop on Adverse Drug Reactions and Lipodystrophy in HIV: practicing medicine. But ennui over things internal propelled him into the physical sciences, where he thrived, inventing the gyroscope, measuring the speed of light, discovering eddy currents, and proving that the Earth rotates on its axis. This last feat is where the pendulum comes in.
No, this remains the era of thesis and antithesis, conjecture and contradiction, axiom and nullity, dead reckoning and the crabwise canter. An illustration:
Do you believe in SREBP-1c genes? Or do you have to be reminded that SREBPg1c-sterol-regulatory element-binding protein 1c-flips on genes that orchestrate lipid metabolism and distribution? The gene that gets SREBP-1c clicking, SREBF1, turns out to have a common variant labeled SREBF1 3'332C>G that Swiss researchers tied to high lipids in people taking antiretrovirals.2 David Katz and Abbott colleagues looked at changes in lipids and SREBF1 in 355 people taking lopinavir/ritonavir (LPV/RTV) or nelfinavir (NFV) in a randomized, double-blind trial [abstract 67]. The Abbott team found no "statistically significant or clinically meaningful relationship" between SREBF1 genotype and total cholesterol or triglycerides. SREBF1 3'332C>G genotype did not predict hyperlipidemia in these treatment-naive people starting their first protease inhibitor (PI). Katz tried three different statistical analyses to confirm a gene link, and he came up empty-handed every time.
The Abbott team aptly observed that their findings "do not provide definitive evidence, positive or negative" about the impact of SREBF1 genotypes on high lipids in people taking PIs. They studied a large, ethnically diverse population taking one of only two PIs, and they did not exclude people with a high body mass index or high pretreatment lipids. The Swiss did make those exclusions in an all-Caucasian population of 67 people tested for cholesterol and 44 tested for triglycerides. So researchers will have to keep studying SREBF1 and other possible predictors of high lipids, genetic and nongenetic.
Do lipoatrophy and insulin resistance have the same antiretroviral roots? Here's another illustration of the mazelike mechanisms of metabolic problems, offered by one of the 5th Lipodystrophy Workshop's organizers, Jacqueline Capeau of Pierre and Marie Curie University in Paris. Surveying the past few years of side-effectology, she conjured a convincingly elaborate web of factors, cofactors, causes, and effects that starts with PIs and nucleoside reverse transcriptase inhibitors (NRTIs), wends its way via 16 arrows through three primary, three secondary, and one tertiary outcome, to two final endpoints -- lipoatrophy and insulin resistance in liver and muscle (Figure 1).
Perhaps the most impressive aspect of Capeau's "hypothetical scheme" is that it pieces together enough primary research to suggest the cause of two problems that bedevil people taking antiretrovirals. Obviously, several of the variables boxed in Capeau's flow chart also figure in other "adverse events" scrutinized at this meeting. But it takes a supple enough mind just to array these 12 inputs and outputs without challenging the limits of credibility, audience attention span, and PowerPoint capacity.
Why is this research so darn complicated? Because it involves a marvelous and messy living machine cobbled together through evolutionary chance and necessity, into which we now invite a relentless retrovirus and admix a slurry of cell-slamming drugs. Even in the nineteenth century, when medicine was much, much simpler, Foucault had the good sense to get out and start stringing cannonballs from church domes. As a result, he ended his career with more than one eureka moment. But the field of antiretroviral (and retroviral) side effects offered no such moments in Paris. Its doughty researchers must keep pecking away at prickly questions like why people taking indinavir (IDV) end up with plenty of plasma adiponectin. (More on that later.)
Yet the lambent light of simplicity did shine at the 5th Lipodystrophy Workshop, yielding easy-to-grasp conclusions and sometimes sturdy advice that can be applied today -- because none of it depends on understanding arcane metabolic mechanisms. More than a half-dozen simple-sounding 5th Lipodystrophy Workshop studies belong to this genre, most of which this article will consider in its concluding section. But for those averse to delayed gratification, here are the half-dozen clinical nuggets:
Down in the right hand corner of Jacqueline Capeau's "hypothetical scheme" of antiretroviral side effects (Figure 1) sits insulin resistance -- the slow meltdown of insulin's regulatory effect on glucose. Back upstream, far above the bends and turns of apoptosis and mitochondrial dysfunction, lie the twin fonts of all these misfortunes, PIs and NRTIs.
Happily, the tongue-twisting thiazolidinediones, or glitazones, can reverse insulin resistance at a precise point in the dysregulatory cascade, by rousing PPAR-g,3 a critical receptor in tissues where insulin works hard -- like liver, skeletal muscle, and fat. A key study by Capeau's group implicated downregulation of PPAR-g in lipoatrophy.4 On top of that, it turned out that glitazones can bulk up lean body mass, boost subcutaneous fat, and trim visceral fat in PI-treated people with diabetes.5 So before even worrying over the hand wringing about mechanisms of toxicity (and there will be lots of that later), clinical researchers realized they had some potent pills already on the shelf.
The trailblazing thiazolidinedione, troglitazone, got yanked from the market when safer glitazones became available. One of its successors, rosiglitazone, turned up in the titles of three 5th Lipodystrophy Workshop reports.
Before the 5th Lipodystrophy Workshop, two small studies put rosiglitazone to the test in people with HIV-related lipodystrophy, and they came up with contradictory results. A 30-person placebo-controlled trial in Finland, with 8 mg of rosiglitazone daily for 24 weeks, found no effect on subcutaneous or intra-abdominal fat.6 Although fasting insulin and liver fat fell, triglycerides and cholesterol both climbed. In Italy a 6- to 12-week nonrandomized study of the same dose in eight people charted a 59 percent jump in insulin-mediated glucose disposal (P=0.02) and -- unlike the Finnish study -- a 21 percent drop in visceral adipose tissue (P=0.04) along with a 23 percent gain in subcutaneous adipose tissue (P=0.05).7
Colleen Hadigan (Massachusetts General Hospital, Boston) came to the 5th Lipodystrophy Workshop with the tie-breaker, a 27-person placebo-controlled trial of rosiglitazone at 4 mg daily for three months, followed by three months of open-label treatment for all study participants at 8 mg daily [abstract 12].
Everyone in the study had fat atrophy while taking a stable antiretroviral regimen. (About two thirds were taking a PI regimen and one third a nonnucleoside.) As in the Italian study, but not the Finnish trial, all study participants had insulin resistance (fasting insulin above 3 µ/mL or a two-hour insulin above 75 µ/mL after an oral glucose tolerance test). No one had diabetes. The 11 people randomized to placebo and the 16 randomized to rosiglitazone had a median CD4 count above 400 cells/mm3, and three quarters had an undetectable viral load.
After three months, two measures of insulin sensitivity -- glucose disposal and glucose/serum insulin -- worsened in the placebo group and improved significantly in the rosiglitazone group (P=0.01 and P<0.05 respectively). Body fat measured by bioelectric impedance analysis fell 5 percent in the placebo group and rose 15 percent in the rosiglitazone group, a significant between-group difference (P=0.04).
DEXA-measured leg fat rose about 4 percent with rosiglitazone while falling about the same amount with placebo, a difference that fell shy of statistical significance (P=0.06). Subcutaneous fat area rose about 8 percent with rosiglitazone but less than 1 percent with placebo, a nonsignificant difference. These fat gains proved large enough for people taking rosiglitazone to notice during the blinded phase of the study. After six months of follow-up, subcutaneous fat had risen 12 percent in the rosiglitazone group.
Adiponectin, an insulin-sensitizing fat cytokine whose antiretroviral-induced decrease Capeau places on the causal pathway to insulin resistance (Figure 1), rose significantly more with rosiglitazone than with placebo (2.3 versus 0.1 µg/mL, P<0.01). The adiponectin gain correlated positively and significantly with increased subcutaneous fat (r=0.59). Free fatty acid, whose increase may contribute to insulin resistance (Figure 1), swelled significantly in the placebo group.
But, as in the Finnish study, rosiglitazone had an unhappy effect on total cholesterol, which climbed among people taking the glitazone (+25 mg/dL) while falling in the placebo group (-16 mg/dL, P<0.01). Reviewing these three rosiglitazone studies at the 2nd IAS Conference on HIV Pathogenesis and Treatment, held a week after the 5th Lipodystrophy Workshop,8 Morris Schambelan (University of California, San Francisco) underlined one salient feature that suggests why Hadigan and the Italian group7 saw improvements in subcutaneous fat and the Finnish team6 did not: The first two studies, but not the third, enrolled people with insulin resistance. When insulin resistance is an entry requirement, Schambelan ventured, rosiglitazone seems to help boost subcutaneous fat.
A second three-month rosiglitazone study tends to support this suggestion, but with four volunteers it is too small to be decisive [abstract 74]. None of the four men recruited by Fehmida Visnegarwala (Baylor College of Medicine, Houston) had insulin resistance or a history of diabetes. All were taking PIs. After three months of rosiglitazone at the 8-mg daily dose, CT-measured mean abdominal subcutaneous fat rose nonsignificantly from 5,630 to 6,654 mm3, while mean visceral adipose tissue fell nonsignificantly from 16,857 to 13,857 mm3. Mean fasting glucose and insulin both fell significantly in this study (from 101 to 90 mg/dL, P=0.02, and from 13.75 to 8.25 IU/mL, P=0.01).
Yet again, rosiglitazone lifted lipids. Mean cholesterol climbed from 229.7 to 291.3 mg/dL and mean triglycerides from 352 to 643 mg/dL. Though neither of these gains proved statistically significant, the leap in total cholesterol took the group from the "borderline high" bracket to "high." The group started the study well above the safe triglyceride range of 10 to 190 mg/dL and got worse.
Neither Hadigan nor Visnegarwala found evidence of rosiglitazone-induced liver toxicity, the problem that torpedoed troglitazone. Guidelines proposed by Schambelan and others recommend avoiding glitazones with an aspartate or alanine aminotransferase more than 2.5 times the upper limit of normal.9 Together these studies support at least two conclusions about rosiglitazone.
Do these studies suggest rosiglitazone promotes good body fat changes in people with lipodystrophy -- at least in those with insulin resistance? Of the two placebo-controlled trials, the Finnish study was powered to detect only big fat changes. Hadigan powered her study to spot significant changes in glucose and insulin, not body fat, she said. Hence, suggested David Cooper (University of New South Wales, Sydney), no one can say whether rosiglitazone eases lipodystrophy. An answer may come in ACTG 5082, which has randomized people with insulin resistance and excess abdominal fat to take metformin, rosiglitazone, or both.
The 5th Lipodystrophy Workshop yielded one more cautionary word on rosiglitazone: Just 4 mg daily lowers nevirapine (NVP) levels. A study by Mark Oette (Heinrich-Heine University, Düsseldorf) measured levels of NVP, efavirenz (EFV), lopinavir (LPV), NFV, and saquinavir (SQV) before and 28 days after people took rosiglitazone [abstract 115]. The insulin sensitizer had no substantial effect on EFV, LPV, or NFV, but the maximum concentration of NVP dropped 32 percent, the minimum concentration 36 percent, and the area under the curve 31 percent in four people taking the nonnucleoside with rosiglitazone (P=0.032). Saquinavir levels also fell, but Oette could draw no conclusions because only one person was taking that PI.
No one came to the 5th Lipodystrophy Workshop with news on statins, the prime choice for reining in runaway cholesterol. But the 2nd IAS Conference featured two studies with important news on these drugs. Knowing that ritonavir (RTV) slices pravastatin levels, John Gerber (University of Colorado, Denver) gauged the effects of NFV on pravastatin in 14 healthy volunteers and the effects of EFV in 11.10 The study started with three days of the statin, followed by two weeks of the statin plus the antiretroviral. Nelfinavir cut pravastatin exposure by a median of 47 percent (range -65 to +10 percent), and EFV lowered exposure by a median of 40 percent (range -73 to +28 percent). The findings could explain the lackluster response to pravastatin in ACTG A5087.11
If people are taking either of the antiretrovirals, or RTV, Gerber advised clinicians to "consider increasing the maximal dose of pravastatin two-fold." But he warned that physicians taking that course must monitor people closely for toxicity because his study showed that pravastatin concentrations don't drop in everyone taking NFV or EFV.
Clinicians starting a statin in someone responding well to antiretrovirals should know something else about these lipid lowerers: They may blunt CD4-cell spurts, according to results of a case-control study, but apparently only for six months or so.12 Benigno Rodriguez (Case Western Reserve University, Cleveland) and coworkers worried that statins may derail CD4 gains because these antilipid drugs have a mélange of immunomodulatory effects. Their retrospective case-control study involved 73 people who started a statin during the first three years of potent antiretroviral therapy and had at least 12 months of follow-up after that. Rodriguez compared these cases with 146 antiretroviral-treated controls who never took a statin, matching cases and controls for age, gender (more than 90 percent were men), year of first potent regimen, and nadir CD4 count.
And the nadirs were low indeed, averaging 30 cells/mm3 among cases and 33 cells/mm3 among controls. By the time they started statin therapy, the cases had a significantly higher average count than controls, 440 versus 221 cells/mm3 (P<0.0001). But the two groups had equivalent proportions with undetectable viral loads. During the first six months of statin therapy, cases gained a median of 15 cells/mm3 (interquartile range [IQR] -36 to 80 cells/mm3) while controls gained a median of 45 cells/mm3 (IQR 14 to 140 cells/mm3, P=0.008).
Rodriguez could not tie this effect to any baseline trait. The poster did not clarify whether the significantly higher CD4 count in the statin group when they started statins affected the relative gains in the two groups over the next half year. Perhaps not, because Rodriguez traced similar upward CD4 slopes in the two groups about one year after the statins started.
The cause of lipoatrophy -- and the toxic effects of NRTIs and PIs in general -- have entered a decidedly unFoucaultian phase. Listening to some convincing studies at the 5th Lipodystrophy Workshop, one could come away advocating any of several toxicity theses, which do not necessarily mesh:
Another hypothesis, offered at the 2nd IAS Conference, holds that the cause of antiretroviral-related fat abnormalities is all in the mind -- or, more precisely, the brain. Peter Reiss (Academic Medical Center, Amsterdam) outlined this work from Amsterdam's Institute for Brain Research.13 Workers there suggest that certain antiretrovirals selectively influence sets of neurons and so upset the balance between sympathetic and parasympathetic output -- and the results could be loss of subcutaneous fat and gains in visceral fat. Here's something new.
In a murine model the Institute team selectively denervated an intra-abdominal fat pad on one side of the animal, but not on the other. Parasympathetic denervation led to such marked drops in uptake of glucose and free fatty acid into fat cells that these adipocytes could no longer store fat. At the same time, the activity of lipase, responsible for breaking down fat, increased. Furthermore, these workers showed, within sympathetic and parasympathetic nuclei lie discrete neuron sets that project to (a) intra-abdominal fat, and (b) subcutaneous fat.
This research "offers the possibility," Reiss proposed, "that the central nervous system can selectively control adipose tissue in different parts of the body, and it can do so by balancing out output from the sympathetic and parasympathetic nervous system." So the fat abnormalities of lipodystrophy may -- "and this is a question mark," Reiss stressed -- represent what he coined "selective autonomic neuropathy."
Where do the antiretrovirals come in? Perhaps not coincidentally, certain NRTIs, like d4T and didanosine (ddI), have earned a reputation for neurotoxicity and, in d4T's case, lipodystrophy. So, the idea goes, these drugs may rile critical neurons and upset the balance between sympathetic and parasympathetic output and thus promote subcutaneous fat loss and visceral fat gain. The kicker here is that the brain region involved is not protected by a blood-brain barrier and so would seem defenseless against neurotoxic drugs.
If this fascinating thesis proves true (and Reiss noted that the data buttressing this scheme should be published soon), it will only increase questions about the mitochondrial and cytokine theories. David Nolan, Jacqueline Capeau, and others proffered the latest evidence supporting those postulates at the 5th Lipodystrophy Workshop:
The mitochondrial hypothesis of NRTI toxicity has gained currency partly because much of the research involves human beings taking drugs like d4T and AZT, then losing subcutaneous fat as various cells shed mitochondrial DNA (mtDNA). David Nolan, who has spearheaded this work in Simon Mallal's group at the Royal Perth Hospital, spelled out his latest findings in people taking d4T or AZT and in control groups of antiretroviral-naive HIV-infected people [abstract 18].
Measuring mtDNA in biopsied subcutaneous fat, Nolan counted 1,288 copies/cell in 24 treatment-naive controls, 726 copies/cell in 29 people taking AZT (a 44 percent drop, P=0.006), and 240 copies/cell in 28 taking d4T (an 81 percent plummet, P<0.001 versus controls and people taking AZT). Another 11 people taking NRTIs other than AZT or d4T had mtDNA numbers equivalent to those of untreated controls.
PI use and CD4 counts at the time of biopsy did not affect mtDNA levels in subcutaneous fat. But mtDNA quotients did correlate with adipocyte toxicity gauged by confocal microscopy. As mtDNA levels dropped, Nolan saw adipocyte pleomorphism that progressed to "adipocyte loss with marked macrophage infiltration and disordered tissue architecture."
Clinically, these changes tracked with levels of subcutaneous fat in the leg, which measured 24 percent in the treatment-naive controls, 17 percent in the people taking AZT, and 11 percent in the people taking d4T. And when people swap d4T for another nucleoside, adipocytes regain mtDNA copies. That conclusion came from a 16-person substudy of an open-label trial in which 118 people traded d4T for abacavir (ABC) or AZT/lamivudine (3TC) [abstract 90]. Grace McComsey (Case Western Reserve University, Cleveland) counted an average 194 mtDNA copies/cell before the switch and 430 copies/cell 48 weeks after the switch (P=0.01). As she reported earlier, the nucleoside tradeoff yielded gains in arm, leg, and subcutaneous adipose fat, and a drop in visceral adipose fat. People switching to ABC had better fat changes than those switching to AZT/3TC.
What happens to mtDNA in other cells when people start taking NRTIs? Ulrich Walker (University of Freiburg) learned that mtDNA also disappears from liver cells of people taking d4T, ddI, or zalcitabine (ddC), a loss that could contribute to high lactates in people taking the d-drugs [abstract 17]. He used Southern blotting to measure mtDNA in liver cells from people biopsied because of diffuse liver pathology, high liver enzymes, or active hepatitis C virus (HCV) infection.
Compared with HIV/HCV-coinfected people not taking antiretrovirals, those taking a d-drug had 40 percent less mtDNA in liver cells (P<0.001). HIV/HCV-coinfected people taking AZT, 3TC, or ABC, but not a d-drug, had mtDNA levels equivalent to those of the untreated group. Among d-drug takers, those using only d4T had greater mtDNA depletion than those taking ddI or ddC. Lactate levels proved significantly higher in the d-drug group (P=0.017) and in the non-d-drug group (P=0.042) than in the untreated people with HIV/HCV coinfection.
Yet work at the University of Barcelona Hospital Clinic by Sònia López showed that it doesn't take d-drugs or other nukes to deplete mtDNA -- it just takes HIV, at least in peripheral blood mononuclear cells (PBMCs) [abstract 54]. López compared PCR-measured mtDNA in PBMCs from 22 antiretroviral-naive people with HIV infection and 27 age- and gender-matched seronegative controls. The HIV group had 68 percent of the mtDNA per cell measured in the control group (P<0.01). This work reflected results already published by a team at the University of British Columbia,14 who found that antiretroviral-treated people with symptomatic hyperlactatemia had a 68 percent lower mitochondrial-to-nuclear DNA ratio in blood cells than did uninfected controls.
David Nolan's work documented mtDNA depletion in fat cells with both thymidine nucleoside analogs, but a significantly greater loss with d4T than with AZT (see preceding section). Using different yardsticks, Martine Caron in Jacqueline Capeau's group at Pierre and Marie Curie University also showed that the thymidine analogs stand out as adipocyte disrupters [abstract 10]. She used cultured adipocytes to gauge the effects of AZT, ddI, d4T, 3TC, and ABC on cell differentiation, lipid accumulation, and response to insulin.
Caron noted that the dose of each drug equaled the maximum concentration given for at least 11 days, levels that would never be sustained in people taking these nucleosides (1 µM AZT, 10 µM d4T, 10 µM ddI, 10 µM 3TC, and 4 µM ABC). Under these experimental conditions, she found that AZT and d4T upset fat cell viability and slowed lipid accumulation in those cells, but ddI, 3TC, and ABC did not (Table 1). Zidovudine but not the other NRTIs induced insulin resistance in fat cells. Adding 3TC or 3TC/ABC to AZT made the insulin resistance worse.
Jacqueline Capeau collared two other potential lipoatrophy culprits, the cytokines TNF-a and IL-6 [abstract 9]. Her study of 26 people taking PIs and NRTIs and 18 uninfected controls linked higher expression of these cytokines to altered differentiation and increased apoptosis of subcutaneous abdominal fat cells.
Most HIV-infected study participants were taking IDV (n=15) or NFV (n=9), usually with d4T and 3TC. Capeau matched them to uninfected controls by age and body mass index. Compared with controls, the HIV group had significantly more fibrosis in fat cells and macrophages per field (P=0.02 for both). The apoptosis rate in fat cells measured 18.4 percent in antiretroviral-treated people versus 5.6 percent in controls (P=0.005). Capeau found a negative correlation between apoptosis and the cell differentiation markers C/EBP-a (r=-0.73, P=0.005) and SREBP-1 (r=-0.66, P=0.01).
Levels of IL-6 mRNA were three times higher in the HIV group than in controls, mirroring the higher level of TNF-a Capeau had found earlier. Expression of those cytokines correlated inversely with the differentiation markers C/EBP-a and SREBP-1 and positively with apoptosis (r=0.694, P=0.01 for IL-6 and r=0.504, P=0.05 for TNF-a).
"Since these two cytokines act at the level of adipose tissue through autocrine/paracrine mechanisms," Capeau concluded, "these results argue for a deleterious role of adipose tissue cytokines not only in altered [fat cell] differentiation but also in increased apoptosis, which could be responsible for clinical lipoatrophy."
Studying 3T3-F442A adipocytes, Simon Jones (University of Liverpool) found that PIs drive IL-6 expression while both PIs and NRTIs stimulate TNF-a [abstract 11]. Jones compared mRNA levels of the cytokines in adipocytes exposed to 20 µM of AZT, d4T, IDV, RTV, and SQV with levels in unexposed adipocytes. Saquinavir boosted IL-6 mRNA expression 28.54-fold and RTV boosted expression 17.85-fold (P<0.001 for both). The other drugs had little effect on IL-6. All drugs studied magnified expression of TNF-a.
As Capeau concluded in a review talk at the 2nd IAS Conference,15 studies like these show that AZT and d4T toxicities do not depend solely on depletion of mtDNA.
Research seeking to explain lipoatrophy, though proceeding full bore, remains in a formative stage. But that hasn't stopped clinicians and people with HIV from deciding which nucleosides they do and do not want to use. In its 2003 draft antiretroviral guidelines, the British HIV Association reflected the growing sentiment against d4T as a first-line NRTI, advising that "combinations including d4T are not recommended because of increasing evidence of its role in the development of lipodystrophy and abnormal lipid profiles."16 But the July 14, 2003, update of US Department of Health and Human Services (DHHS) guidelines sticks with d4T as a first-line choice.17
Clinical concern over d4T-induced toxicity derives not only from cell studies like those reviewed above, but also from several drug-switch trials and cohort studies implicating d4T in lipoatrophy. More research of this ilk appeared at the 5th Lipodystrophy Workshop. But two studies -- one at the 5th Lipodystrophy Workshop and one at the 2nd IAS Conference -- saw d4T in a new (and more flattering) light. And work by d4T's maker, Bristol-Myers Squibb, in collaboration with the University of Colorado, echoed an earlier study in finding a higher risk of lipoatrophy in people who start treatment at lower CD4 counts.
The latest US DHHS guidelines abandon the earlier menu of antiretrovirals in each drug class from which one could concoct a starting regimen.17 Instead, the DHHS experts pick specific combinations as "preferred" first-line options. The pillar of initial therapy, they advise, should be LPV/RTV or EFV. The nucleoside buttresses are:
Emtricitabine (FTC) had not been licensed when the DHHS panel devised this list. Given its superiority to d4T in a randomized, placebo-controlled trial of first-line regimens,18 it will most likely make the "A-team" on the next update. A more notable exception from the DHHS list is the once-familiar pairing of d4T and ddI. That duo lost to AZT/3TC in ACTG 384.19 At the 5th Lipodystrophy Workshop, results of a CPCRA 058 ("FIRST") substudy found a higher rate of fat abnormalities among naive people starting a ddI/d4T regimen than in those starting with 3TC/ABC [abstract 14].
CPCRA researchers randomized 1,400 people to begin therapy with a PI, a nonnucleoside, or both, along with either ddI/d4T or AZT/3TC. Subhasree Raghaven (Harlem Hospital, New York) reported body composition results of the NRTI substudy, which involved 46 people taking ddI/d4T and 50 taking 3TC/ABC. The substudy group was 64 percent African American, 13 percent Latino, and 28 percent female. The two NRTI groups started therapy with similar median CD4 counts (268 cells/mm3 for ddI/d4T and 234 cells/mm3 for 3TC/ABC) and viral loads (4.85 and 5.03 log copies/mL, respectively). Seventy percent in each NRTI group also took a PI.
After a median 32.6 months of follow-up, the ddI/d4T group did worse than the 3TC/ABC group in every morphologic measure recorded (Table 2). Insulin levels climbed more in the ddI/d4T group than in the 3TC/ABC group (5.2 versus 0.3 µM/mL). CPCRA statisticians had not yet correlated these changes with CD4 count, viral load, race, gender, or other variables.
A possible limitation of this study is its heavy reliance on anthropometric measurements -- skinfolds and body circumference -- rather than on DEXA or CT scans. Even when personnel are centrally trained, as in this trial, results can vary from one technician to the next. In the already-mentioned study of 118 people who traded d4T for ABC or AZT/3TC, Grace McComsey found no correlation between 48-week DEXA-measured gains in arm or leg fat and physical measurements of arms and legs [abstract 90]. Still, the lock-step consistency of the CPCRA data on every variable assessed is hard to ignore.
The MITOX study was the first nucleoside switch trial to chart improvements in peripheral fat among people who traded d4T or AZT for ABC.20 A two-year update of that study by Andrew Carr (St. Vincent's Hospital, Sydney) showed a continuing slow accrual of limb fat in switchers [abstract 16]. Two years after the randomized switch, people who had lost about half of their limb fat had gained back about one third of that loss. It took about six years of NRTI therapy for these 106 study participants to lose half of their limb fat, Carr observed, and at this pace it could take that long to get it back.
MITOX randomized people -- nearly all of them white men and 85 percent of them taking d4T -- to continue d4T or AZT or to switch to ABC. At that point the ABC switch group averaged 3.54 kg of limb fat and the nonswitch group averaged 3.75 kg -- about half of the 7 or 8 kg that most men have. After 24 weeks the ABC group had gained 0.39 kg of limb fat, significantly more than the 0.09 kg in the d4T-or-AZT group (P=0.016). But gains were so modest that neither the men nor their clinicians noticed them. After 24 weeks everyone could switch to ABC.
In an intent-to-treat analysis at 108 weeks, people originally randomized to ABC had gained 1.26 kg in limb fat (+36 percent), compared with 0.49 kg (+13 percent) among those originally randomized to stay with d4T or AZT, a significant difference (P=0.008). In a 108-week on-treatment analysis, people who originally switched to ABC gained 1.29 kg (+36 percent), those who switched to ABC by week 24 gained 0.55 kg (+15 percent), and those who stayed with d4T or AZT gained 0.16 kg (+4 percent). A multivariate model found three independent predictors of greater limb fat gains:
Carr added that the limb fat gains -- which plumped up legs more than arms -- varied considerably from person to person. Visceral adipose tissue, CD4 counts, viral loads, and glycemic variables did not change significantly through 108 weeks of follow-up.
Although MITOX now offers the longest and strongest evidence that stopping a thymidine analog -- usually d4T -- can restore peripheral fat, some observers want more proof. In his 2nd IAS Conference review of 5th Lipodystrophy Workshop studies,8 Morris Schambelan cautioned that "whether [fat gains] will continue on an upward trend remains to be determined."
Tracking DEXA-measured leg fat changes in 72 white men starting a regimen containing d4T or AZT, David Nolan saw a significantly greater loss in those taking d4T than in those taking AZT through 36 months of treatment (P=0.02) [abstract 95]. He estimated a 45 percent fat loss in the d4T group versus 30 percent in the AZT group. PIs, used by more than 70 percent in each group, did not affect the results; nor did baseline CD4 count or concomitant treatment with ddI.
Men who traded d4T for AZT saw no further drop in leg fat through an average 22 months (range 3 to 40 months). But they did not regain leg fat, as people did in McComsey's study of switching to ABC or AZT [abstract 90] or in Carr's study of switching to ABC [abstract 16]. In Nolan's analysis, controls matched to d4T-AZT switchers for duration of treatment and taking d4T the whole time continued losing fat (P<0.01 compared with switchers).
A Thai study of 54 men and 26 women with lipoatrophy while taking d4T registered fat gains after cuts in the d4T dose.21 Although this strategy proved safe -- no one had a viral breakthrough from below 50 copies/mL -- gauging the precise impact of lower dosing is difficult because the investigators relied almost solely on clinical judgment to rate fat gains.
Everyone had a sub-50-copy viral load for at least six months when the study began, though CD4 counts ranged from a scary 35 cells/mm3 to a hefty 922 cells/mm3 (mean 334 cells/mm3). Mattana Hanvanich (Chulalongkorn University, Bangkok) and colleagues at Bumrungrad Hospital classified lipoatrophy as mild in 22 people, moderate in 41, and severe in 17. They trimmed the d4T dose according to the following scheme:
Improvement appeared to be faster in people with more moderate lipoatrophy. Fifty-six study participants (70 percent) needed the additional reduction to 50 percent less of the standard dose. Overall weight gains averaged 0.78 kg in the mild group, 2.2 kg in the moderate group, and 1.1 kg in the severe group. Abdominal girth (not further defined) dropped during the study. All viral loads stayed under 50 copies/mL through a median 93 weeks of follow-up (range two to 309 weeks). Eight people (10 percent) had traded d4T for ABC at the time of the analysis.
Because HIV Outpatient Study (HOPS) investigators linked nadir CD4 count -- but no specific antiretroviral -- to development of lipoatrophy,22 clinicians in Connecticut and New York decided to see if d4T may be safer in people who started it with more T cells [abstract 76]. Their retrospective analysis compared 21 people who took d4T as part of their first regimen with 23 who took it after trying AZT.
Gary Blick (Circle Medical Group, Norwalk) reported a mean age of 46 years in both groups, and about 40 percent in each group were women. The first-line d4T group had a higher percentage of nonwhites (80 versus 42 percent) and a higher percentage of nongays (80 versus 58 percent). The nadir CD4 percent and nadir CD4 count were substantially higher in the first-line group, and the highest viral load substantially lower in that group (Table 3). Despite a longer duration of d4T therapy in the first-line group, those people had a much lower incidence of clinically judged lipodystrophy and lipoatrophy (Table 3).
Blick concluded that current DHHS guidelines, designed to avert or hold off side effects by delaying treatment until the CD4 count falls below 350 cells/mm3, "may be placing patients at increased risk of ... lipoatrophy and peripheral neuropathy." But, as results of other cohort studies suggest, that may be true only if people start therapy with d4T. Blick and colleagues promised a complete analysis of 85 people treated with d4T as a first or second thymidine analog. The results will be interesting if the authors offer statistical correlations between CD4 nadirs and the later emergence of lipoatrophy. The current analysis did not specify the nadirs of people in whom side effects developed, only the averages and ranges for each group.
Nadir CD4 count independently predicted lipoatrophy in a database analysis by Christopher Dezii (Bristol-Myers Squibb) and Kenneth Lichtenstein (University of Colorado, Denver) [abstract 78]. They calculated that, for every 100-cell increment in the nadir count, the risk of fat wasting dropped 18 percent. The correlation looked most convincing in CD4 brackets below 350 cells/mm3 (Table 4), because more than 90 percent of people in those strata were taking antiretrovirals.
Dezii analyzed fat trends in 7,980 HIV-infected people in the Cerner Corporation's HIV Insight database. To ensure that he counted only new cases of fat build-up or loss, he included only people with no reported fat abnormalities for six months before such a report. As in any database analysis of this size, the lipodystrophy diagnosis depended on a clinician's subjective impression of "lipodystrophy without fat wasting" or "fat wasting" of the limbs, hips, buttocks, face, or neck.
The cohort had a mean age of 38.9 + 8.45 years, a mean follow-up of 3.67 + 2.99 years, and a mean HIV infection duration of 7.77 years. Most (85 percent) were men, 58 percent were Caucasian, and 29 percent were African American. Dezii charted a significant increase in the incidence of fat wasting -- but not of lipodystrophy without wasting -- in groups with lower CD4 nadirs (Table 4). Logistic regression modeling controlling for age, race, HIV duration, body mass index nadir, and CD4 nadir -- but not antiretroviral exposure -- found a higher risk of fat wasting with a lower CD4 nadir, a lower body mass index nadir, longer HIV duration, male gender, or white race. The CD4 result confirms Lichtenstein's similar finding in the HOPS cohort.22
Avoiding or retiring the more toxic antiretrovirals and starting certain antiretrovirals earlier may trim the risk of lipoatrophy, as the foregoing studies suggest. But many people don't get their HIV diagnosis until their T cells fall into double digits, and many others are already paying the toxic price of drug therapy. So wouldn't it be nice if you could just take a pill that would reverse the damage already done? The University of Freiburg's Ulrich Walker thinks such a medicine may already be for sale in Europe, a dietary supplement called NucleomaxX with Mitocnol,23 at least if the drug toxicities are of the mitochondrial kind. Cell studies in his lab suggest that Mitocnol's essential ingredient, uridine, can correct mitochondrial miscues [abstract 19].
After exposing liver cells (HepG2 hepatocytes) to 177 nM of ddC, Walker recorded severe mtDNA depletion (to 8 percent of normal levels), a steep drop in cell proliferation (to 20 percent of normal), severe intracellular steatosis (fat buildup), and up to a 350 percent jump in lactates. Uridine restored mtDNA levels to about 65 percent of normal and thereby "fully normalized" cell proliferation and intracellular lactate and lipid levels. Uridine worked this magic even when Walker continued treating the cells with ddC. The best results came at the highest uridine dose, 200 µM. Similar experiments with d4T and AZT -- but not ddI -- had similar results. Walker also found that uridine does not alter the 50 percent or 90 percent inhibitory concentrations of current nucleosides.
Given at a dose of three sachets per day for four days, NucleomaxX with Mitocnol reversed cellular toxicities in a person taking d4T. Walker and his colleagues have taken up to twice the recommended dose of this supplement, and no side effects emerged in this preliminary test. (The recommended dose is three sachets daily for three days, then no treatment for the rest of the month.) He plans formal tests to see whether this agent can help people with suspected mitochondrial toxicities such as lipoatrophy and polyneuropathy.
Much of the basic research laid out at the 5th Lipodystrophy Workshop involved NRTIs, but PIs did not escape scrutiny. Probably the most revealing study -- because it plugs a big hole in the lipodystrophy databank -- examined the effects of PI-based therapy in a large cohort of American women. Other work tied IDV to endothelial dysfunction and -- in healthy volunteers -- to higher hepatic glucose production. Stopping PIs for 96 weeks barely improved glycemic dysregulation in people with severe lipodystrophy. In the nonnucleoside arena, two teams disagreed on how fast toxicity forced people to stop EFV or NVP.
The FRAM study identified lipoatrophy as the main feature of lipodystrophy in HIV-infected men.24 Compared with uninfected men enrolled in a heart disease cohort, FRAM men with HIV had less peripheral and central fat. But as Kathleen Mulligan (University of California, San Francisco) observed at the 5th Lipodystrophy Workshop, the HIV-infected men in FRAM tended to be lean at baseline, while HIV-infected women in the United States tend to be overweight or obese. So what happens when these women start taking antiretrovirals?
To find out, Mulligan and colleagues in the Women's Interagency HIV Study (WIHS) ran a cross-sectional comparison of four groups [abstract 15]:
They excluded pregnant or lactating women, women on hormone replacement therapy or steroids, and women weighing more than 120 kg. Nearly half in the HIV and non-HIV groups were African American, and 40 percent in each group were Hispanic. More than half in each group were overweight or obese. The women with HIV were older (41 versus 36 years, P<0.001), and more women with HIV had one or more pregnancies (85 versus 53 percent, P<0.0001). Fewer HIV-infected women smoked cigarettes, though most in both groups smoked (63 percent with HIV versus 76 percent without HIV, P=0.04).
Body mass index was significantly lower in the women taking a non-PI regimen (about 26 kg/m3) than in the HIV-infected untreated group (about 28 kg/m3, P=0.03) or in the HIV-uninfected group (about 30 kg/m3, P=0.01). The women taking PIs averaged about 28 kg/m3 but did not differ significantly from the other groups. Total fat was also significantly lower in the non-PI treated group than in the HIV-infected untreated women (P=0.006) or the HIV-uninfected women (P<0.01). The PI group was closest to the untreated group in total fat but again did not differ significantly from the other groups.
Compared with the HIV-uninfected women, leg fat was lower in both the PI group (P=0.01) and the non-PI treated group (P<0.001). But the women taking a PI didn't differ from the non-HIV group or the untreated group in trunk fat, whereas the non-PI treated women had significantly less trunk fat than the untreated (P=0.02) or uninfected women (P<0.001).
In a multivariate model controlling for age, race, smoking, exercise, number of live births, CD4 count, viral load, and d4T treatment, five factors proved significant predictors of lower leg fat -- not being African American, having a CD4 count below 200 cells/mm3, taking d4T, cigarette smoking, and exercising more than six hours a week. Except for race and d4T, the same factors correlated with less trunk fat.
Mulligan advanced the following conclusions:
Mulligan added that the study's cross-sectional design "mandates caution in drawing conclusions from these associations."
Earlier work by researchers at the University of California, San Francisco, documented glucose intolerance and lower insulin-mediated glucose disposal and storage in healthy volunteers taking IDV for just four weeks.25 Analysis of hepatic glucose metabolism in nine of these volunteers, presented at the 5th Lipodystrophy Workshop by Grace Lee (University of California, San Diego), showed that a month of IDV moderately boosts fasting glucose production by the liver and blunts insulin-mediated suppression of hepatic glucose output [abstract 7].
The study used the standard, three-times-daily dose of IDV. Fasting glucose climbed from 4.9 mmol/L at baseline to 5.1 mmol/L after week four, a 7 percent jump that landed just short of statistical significance (P=0.06). Fasting hepatic glucose production rose from 12.6 mmol/kg/min at baseline to 13.5 mmol/kg/min after week four (P<0.03). Over the same period, fasting insulin bolted from 62.1 pmol/L to 80.7 pmol/L (P=0.055). HOMA-calculated insulin resistance increased significantly. Insulin's ability to suppress glycogenolysis and gluconeogenesis waned after treatment with IDV.
Once PI therapy unglues glucose metabolism, patching things up takes time. Even after 96 weeks without PIs, glucose metabolism improves only "modestly," 5th Lipodystrophy Workshop attendees learned from Marc van der Valk (International Antiviral Therapy Evaluation Center, Amsterdam) [abstract 8]. Sidelining PIs did yield slow improvements in subcutaneous and visceral adipose tissue.
The study involved eight men with severe lipoatrophy who traded their PIs for ABC without losing viral control for 96 weeks. Van der Valk chalked up these changes after 96 weeks of non-PI treatment:
Reviewing this study at the 2nd IAS Conference,13 van der Valk's collaborator Peter Reiss spelled out the bottom line: Modestly improved glucose production and lipolysis did not approach levels seen in healthy volunteers after 96 weeks without PI therapy, a result "suggesting persistence of insulin resistance both in adipose tissue and muscle or in the liver."
Earlier work blamed PI inhibition of the glucose transporter GLUT4 for insulin resistance during PI therapy.26 The sluggish improvement in glucose metabolism charted by van der Valk led him to suggest PIs may permanently damage GLUT4. "Something must be broken," he proposed. An alternative hypothesis, mentioned in his study abstract, is that PIs gum up other (or additional) go-betweens that insulin exploits to regulate glucose. Grace Lee endorsed this second option, noting that IDV's effects in her study cannot be explained by GLUT4 alone.
Despite traumatized glucose transport in people taking PIs, a diabetes epidemic has not descended on people taking these drugs. Or is it just a matter of time? A comparison of 288 men without HIV infection and 339 infected men in the Multicenter AIDS Cohort Study (MACS) logged a higher incidence of "pre-diabetes" (hyperglycemia) and of diabetes itself in antiretroviral-treated men [abstract 43]. The incidence of hyperglycemia (fasting glucose between 110 and 125 mg/dL) or diabetes (fasting glucose 126 mg/dL or higher) proved highest among men taking antiretrovirals, lower in the HIV-seronegative group, and lowest in an HIV-infected group not taking antiretrovirals.
The analysis included men with an initial fasting glucose at or below 105 mg/dL and no history of diabetes. Todd Brown (Johns Hopkins University, Baltimore) reckoned the overall incidence of hyperglycemia at 6.6 cases per 100 person-years and the diabetes incidence at 2.5 cases per 100 person-years (Table 5). Compared with the uninfected men, the antiretroviral-treated group had a 40 percent higher chance of hyperglycemia and a 71 percent higher chance of diabetes.
When Brown determined the risk of hyperglycemia or diabetes according to specific drugs, he found the highest risk among people taking ddI/d4T or EFV. Compared with the uninfected group, hazard ratios adjusted for age, body mass index, and alcohol intake measured 1.3 for any PI, 1.38 for any antiretroviral regimen, 1.48 for d4T, 2.31 for EFV, and 2.55 for ddI/d4T. The MACS team counted 11 diagnoses in people taking EFV and 24 in people taking any PI, but it was unclear if some of those taking EFV had just switched to the nonnuke after running into trouble with a PI. Earlier studies did not find that EFV causes hyperglycemia or insulin resistance. Brown noted that the study may underestimate the overall incidence of high glucose and diabetes in people taking antiretrovirals because everyone's first visit came during or after April 1999, three years into the HAART era.
The healthful hormone adiponectin enthralls lipodystrophy mavens because it plays roles in two arenas where PIs also perform -- the circulatory system and glucose metabolism. On top of that, as its name implies, adiponectin comes from fat cells. So when fat cells vanish, as they do in people with lipoatrophy, adiponectin levels fall. A cottage industry of adiponectin research in people with HIV has already cranked out the following findings:
The consistency of these findings from one cross-sectional study to the next clearly implies, as one team writes, that "changes in adiponectin may contribute to the metabolic dysregulation" in people with HIV lipodystrophy.30 Researchers at Boston's Brigham and Women's Hospital propose that "NRTI use may worsen insulin resistance by decreasing adiponectin levels."31 And, perhaps inevitably, people have started pondering adiponectin replacement therapy as "a potential treatment option to ameliorate the metabolic changes observed in this patient population."31
But some people taking antiretrovirals may already be pumping up their adiponectin payloads -- to compensate for PI-induced endothelial dysfunction. That unexpected possibility came to light in a four-week study of 10 healthy, nonobese volunteers without HIV infection or hypertension [abstract 1]. Because earlier work tied PIs to impaired endothelial function,34 Sudha Shankar (Indiana University, Indianapolis) gauged leg blood flow after infusion of methacholine before and after four weeks of thrice-daily IDV. A blood flow surge in response to methacholine means the endothelium and vascular smooth muscle have not lost their elasticity. Decreased blood flow means damage.
Shankar found that adiponectin levels rose 30 percent, from 14.2 to 18.4 µg/mL, after four weeks of IDV (P<0.05). At the same time, as in other PI studies, HOMA-measured insulin resistance worsened (P<0.05). And leg blood flow ebbed by 60 percent from the baseline measure to week four (P<0.01). This decreased flow correlated with increased adiponectin (r=0.853, P<0.005), and increased adiponectin correlated with the degree of endothelial dysfunction (r2=0.585, P<0.05).
What does it all mean? The jump in adiponectin, Shankar suggested, may be the body's way of fighting back after endothelial damage. Animal studies show that adiponectin lessens the neointimal thickening and vascular smooth muscle proliferation that result from vascular damage. So, strange as it may sound, IDV-induced endothelial damage may protect people from even worse insulin resistance by boosting adiponectin levels -- though mauling your endothelium is probably not the best way to ward off diabetes. Shankar did not suggest the other, even more bizarre, corollary: Indinavir-induced jumps in adiponectin may ameliorate the metabolic and morphologic mix-ups tied to antiretrovirals.
Of course it's way too early to read much meaning into this intriguing experiment. Shankar noted that the study does not prove that adiponectin levels rose in response to vascular damage, only that the two correlate. The adiponectin uptick could be a direct result of IDV, she suggested, or some other mechanism may be involved. And whether other PIs have the same effect remains to be seen.
Another study of blood vessels in people taking antiretrovirals yielded another uncanny result: After 12 months of antiretroviral therapy, carotid intima media thickness -- a harbinger of atherosclerosis -- increased most in people who started the study with the highest CD4 counts [abstract 2]. Perhaps, suggested Patrick Mercié (University Hospital, Bordeaux), treatment-induced immune restoration takes a toll on the arteries. A player in this plot may be CD40 ligand, "which is widely implicated in the progression of atherosclerosis" and which CD4 cells spew.
The study involved 346 members of the Aquitaine cohort who had ultrasonography to gauge intima media thickness two times, 12 months apart. Mean thickness inched significantly upward from 0.57 mm at month 0 to 0.59 mm at month 12 (P<0.0001). Those measures fall within the normal range for the general population.
The increase correlated with several classic cardiovascular risk factors -- older age (P<0.0001), male gender (P=0.02), and smoking (P=0.05). But were the cohort's carotid arteries narrowing just because they were getting older and smoking too much (70 percent smoked at month 12)? Or did their high CD4 counts (447 cells/mm3 at baseline and 490 cells/mm3 at month 12) have more to do with it? The baseline CD4 count, Mercié found, was the only variable that correlated independently with thicker intima media at month 12 (P=0.0145). Splitting the cohort into baseline CD4 quartiles, he charted dramatically higher intima media thickening from the first through the third quartile (Table 6).
These findings offer an interesting contrast with a study of 148 people with HIV infection, 79 of them tracked for one year [abstract 35]. Priscilla Hsue (University of California, San Francisco) reported two striking differences from Mercié's findings:
But a third finding by Hsue may reflect one of Mercié's results: A nadir CD4 count at or below 200 cells/mm3 correlated with thicker artery walls at baseline (P=0.072) and with the one-year progression rate (P=0.049). The median nadir in the group measured 110 cells/mm3, and they had been taking a PI regimen for a median of 3.3 years. At that point the median CD4 count had climbed to 354 cells/mm3. So the immune restoration mechanism proposed by Mercié could also be at work in Hsue's cohort. But the gain from nadir to baseline CD4 count did not predict intima media thickness in Hsue's analysis. Instead she stressed the chronic inflammation of untreated HIV infection -- reflected in the low CD4 nadirs -- as a potential explanation of the thickening artery walls.
Why did Hsue measure much thicker artery linings at baseline, and much faster progression over one year, than Mercié did? Smoking is not the answer, because 64 percent of the Aquitaine cohort smoked at baseline compared with 56 percent of Hsue's group. Neither is age, with a median of 45 years in Hsue's cohort and a mean of 42 in Mercié's.
So other factors must explain these differences. Hsue reported that 34 people (23 percent) had hypertension, and eight (5 percent) already had coronary artery disease. Mercié did not report vascular disease rates in his cohort. Smoking (P=0.001) and hypertension (P=0.091) correlated with baseline intima media thickness in a multivariate analysis by Hsue, as did being Latino (which usually means Mexican in California) (P=0.047). Mercié did not define the ethnic breakdown of his cohort, though it probably included few if any Mexicans.
But the most important difference between the studies probably lies in the methods used to gauge artery wall thickness. Aquitane researchers measured only common carotid intima media thickness, whereas Hsue measured 12 segments -- six on the left and six on the right. "We find the majority of disease in the bifurcation of the common carotid into the external and internal carotid arteries," Hsue wrote to IAPAC Monthly. "This would be missed if one only measured one segment of the vessel." That difference could explain both the higher baseline thickness in Hsue's study and the faster progression.
Other variables that independently correlated with baseline intima media thickness in Hsue's study were older age (P<0.001), higher low-density lipoprotein cholesterol (P=0.002), and (as already noted) nadir CD4 count at or below 200 cells/mm3 (P=0.072). When Hsue rated predictors in an analysis that included 63 gender- and age-matched seronegative controls, HIV infection also independently predicted narrower carotids (P=0.001). Compared with controls, the HIV group had significantly more intima media thickening at baseline (P=0.0001), even though the median baseline thickness of Hsue's control group was well above the 12-month thickness in Mercié's HIV group (0.70 mm versus 0.59 mm).
In a multivariate analysis of one-year intima media thickness progression, Hsue found three independent predictors: older age (P=0.01), being Latino (P=0.024), and CD4 nadir (P=0.049).
A metabolic substudy of the international trial comparing atazanavir (ATV) with EFV (plus AZT/3TC) in treatment-naive people found similar rates of fat changes in the two treatment groups after 48 weeks [abstract 14]. Joseph Jemsek (Jemsek Clinic, Huntersville, North Carolina) reported that 111 people taking ATV and 100 taking EFV had modest fat gains in all body compartments. While average weight rose 1.2 kg, three ratios of fat distribution did not change significantly with either regimen through 48 weeks -- visceral-to-total adipose tissue, visceral-to-subcutaneous adipose tissue, and subcutaneous-to-total adipose tissue. Noting that body fat differences between groups taking EFV or NFV in ACTG 5005 did not appear until week 80, Michael Dubé (Indiana University, Indianapolis) suggested that follow-up in the ATV study may be too short to support conclusions.
Analysis of two trials comparing ATV with LPV in PI-experienced people confirmed the sleek lipid profile of ATV [abstract 119]. In a study pitting standard-dose LPV against 400 mg of ATV once daily for 24 weeks, mean low-density lipoprotein (LDL) cholesterol rose 5 percent with LPV while falling 6 percent with ATV (P<0.05). Triglycerides soared by a mean 56 percent in the LPV arm while falling 2 percent with ATV (P<0.0001). In the second 24-week study, comparing three PI rescue regimens after triple-class failure, triglycerides climbed 31 percent with LPV while falling 2 percent with ATV/RTV (300/100 mg once daily) and falling 14 percent with ATV/SQV (400/1,200 mg once daily) (P<0.0001).
More interesting than the comparison of ATV with LPV was the comparison of unboosted ATV (400 mg once daily) with RTV-boosted ATV (300/100 mg once daily) in the two studies. Kenneth Lichtenstein (University of Colorado, Denver) reported that both regimens effectively lowered total cholesterol, LDL cholesterol, and triglycerides (Table 7). Because trough levels of unboosted ATV can fall dangerously low in some people,35 more than one expert has asked whether ATV should always be boosted, regardless of whether a person is taking the drug as a first PI or in a backup regimen. Since 100 mg of RTV appears not to mar ATV's limpid lipid record, that argument sounds sensible.
Abbott researchers led by Scott Brun fed body fat data from four LPV trials into a Cox proportional hazard model to isolate risk factors for fat gains and losses through 96 weeks of treatment [abstract 99]. Four independent predictors emerged:
Among the factors that did not favor lipodystrophy were male gender, baseline weight, baseline lipids or lipid gains, baseline CD4 count, smoking, and hypertension.
The choice between EFV and NVP as first-line therapy rests only partly on how one interprets the conflicting efficacy data.36-40 Equally important is the relative safety of these drugs. Two reports at the 5th Lipodystrophy Workshop reached divergent conclusions about nonnucleoside toxicity, but a third report at the 2nd IAS Conference saw more merit in EFV.
A EuroSIDA study of 1,738 people starting NVP and 1,635 starting EFV found that a higher proportion stopped NVP, usually because of toxicity [abstract 24]. Liver failure proved uncommon with either drug. Nina Friis-Møller (Hvidovre University Hospital, Denmark) reported several factors that may have favored EFV in this analysis. The EFV group had a significantly lower median viral load (3.4 logs versus 3.8 logs with NVP, P=0.0001) and a higher CD4 count (319 cells/mm3 versus 288 cells/mm3 with NVP, P=0.004). Significantly more people starting EFV had taken no earlier antiretrovirals, though the proportion of naive people was small in each group (6.7 percent for EFV and 3.7 percent for NVP, P=0.006).
Time to stopping a nonnucleoside for any reason proved significantly shorter with NVP (P<0.0001). After 40 months of follow-up, about 60 percent wanted no more NVP and about 50 percent forsook EFV. In the first month of treatment, more people stopped NVP because of rash, and more stopped EFV because of central nervous system side effects (P<0.05 for both comparisons). After 12 months of therapy, dyslipidemia accounted for about 5 percent of EFV withdrawals and central nervous system toxicity about 10 percent, both significantly more than with NVP (P<0.05). But after a year of treatment, NVP had failed virologically in about 50 percent compared with about 35 percent taking EFV (P<0.05).
About 6 percent abandoned NVP because of liver toxicity versus about 3 percent for EFV, but that difference lacked statistical significance. The incidence of liver failure proved nonsignificantly higher with EFV -- 0.49 per 100 person-years (95 percent confidence interval [CI] 0.19 to 0.86) versus 0.29 per 100 person-years (95 percent CI 0.11 to 0.62) with NVP. More people taking NVP than taking EFV had HCV coinfection (85.7 versus 77.8 percent) or hepatitis B virus (HBV) coinfection (37.5 versus 22.2 percent).
A much smaller study, involving 249 people taking EFV and 165 taking NVP at Madrid's 12 de Octubre Hospital, counted more safety failures with EFV [abstract 141]. As in the EuroSIDA study, a significantly higher proportion started EFV as part of their first regimen (22 percent versus 10 percent with NVP, P<0.001). But Miguel Torralba reported no other baseline differences between the groups.
Although rash proved more common with NVP, the difference from EFV was not significant (15.3 percent versus 12.5 percent). But a higher NVP-linked incidence of grade 3 or 4 rash approached significance (6.1 percent versus 2.4 percent with EFV, P=0.059), and a significantly higher proportion stopped NVP because of rash (11.5 percent versus 5.8 percent, P=0.044). Eighty-four people taking EFV had central nervous system toxicity, and 35 of them (42 percent) dumped the drug as a result. A Cox regression analysis did not link NVP to higher rates of hepatotoxicity, after controlling for HCV, HBV, gender, age, use of RTV, CD4 count, and viral load.
The overall discontinuation rate measured 29 percent with EFV and 20 percent with NVP after one year of follow-up, though this difference lacked statistical significance (P=0.101). But in another Cox analysis, after adjustment for age, gender, HCV, HBV, number of antiretrovirals, use of RTV, CD4 count, and viral load, people were 1.66 times more likely to quit EFV than NVP (P=0.034). One factor that may explain the difference between these two studies is the high proportion of injecting drug users among Spanish people with HIV compared with the larger European population represented by EuroSIDA. Drug users could be more sensitive than others to the central nervous system insults of EFV.
A third cohort study, reported by Chelsea and Westminster Hospital's Gail Matthews at the 2nd IAS Conference, mirrored the EuroSIDA finding that more people dropped NVP sooner than EFV.41 Unlike the two cohorts reported at the 5th Lipodystrophy Workshop, the London group consisted entirely of treatment-naive people starting either EFV or NVP. After 292 patient-years of follow-up, 35 of 287 (12 percent) taking NVP had to switch because of toxicity, compared with 22 of 328 taking EFV (7 percent). Kaplan-Meier analyses graphed a longer time to virologic failure (P=0.003) and a longer time to virologic plus toxicity failure (P=0.0324) with EFV.
Although the two groups did not differ in gender, proportions with an AIDS diagnosis, or baseline CD4 count or viral load, the NVP group tended to take ddI/d4T, while the EFV group usually opted for AZT/3TC. And in a multivariate analysis, ddI/d4T raised the risk of failure 1.92 times (P=0.008). Starting with EFV lowered the failure risk 40 percent (P=0.012). Switching only NRTIs did not count as a regimen failure in Matthews' analysis, but she noted that some people ditching the d-drugs because of toxicity may have dumped NVP at the same time.
Leptin was the watchword at last year's Lipodystrophy Workshop; this year leptin's cytokine cousin adiponectin took its turn in the spotlight. Early evidence shows that these two agents straddle mechanistic pathways to miscarried metabolics.
But whether low levels of leptin or adiponectin are cause or consequence of specific disorders is not always clear. And whether fiddling with levels of these hormones can ease lipodystrophy and related problems remains an even tougher call.
Despite these many unknowns (and the merely jocular suggestion of this article's subtitle), no one really thinks HIV researchers should forget adiponectin in the struggle to understand fat morphs and metabolic meanderings. But neither did anyone inflate hopes about therapeutic prospects for leptin or adiponectin replacement when certain simpler tactics can be tried today -- and when some of them will certainly help ease the toxicities considered at this 5th Lipodystrophy Workshop. Like exercising, and stopping smoking, and screening women with HIV for osteoporosis.
Such tactics can be called Foucaultian in their humble, nearly hit-yourself-in-the-head transparency. Just set that big pendulum swinging, stand around for a few hours, and watch the world wag. This view of clinical research does not mean that any good study is easy in its conception, planning, or execution. What most impresses about Foucault's pendulum is not how painfully obvious it seemed in the end, but how this elegant proof eluded all until Foucault had the thought, planned the experiment, and got it done. So here are some nominations for the Order of Jean-Bernard-Léon Foucault, bestowed on those who uncover antitoxicity tactics that sound obvious but remain obscure until someone does the work.
1. Manage PI-induced diarrhea with measures that work for other diarrhea.
"A straightforward, effective way to resolve [PI-induced] diarrhea and promote gut health is necessary," suggest Carla Heiser (Indiana University, Chicago) and colleagues at other sites [abstract 105]. Now who would argue with that?
Yet NFV has been on the market for six years, and the main pre-Heiser help proffered was Imodium (loperamide). Heiser randomized 28 people to dietary supplementation and seven to standard of care. In the intervention group, 22 were taking NFV and six LPV. All study participants had noticeable fat redistribution and two of the following seven problems: diarrhea despite loperamide, more than two episodes of watery or loose stools daily, total cholesterol above 200 mg/dL, LDL cholesterol above 130 mg/dL, triglycerides above 200 mg/dL, fasting glucose above 115 mg/dL twice, insulin above 20 mIU/L. In the treatment group, only one person qualified for the study because of hyperlipidemia. The dietary regimen included two or three courses:
Diarrhea resolved completely in 13 of 28 people after four weeks of treatment. Adding L-glutamine reduced diarrhea in 11 of the remaining 15 people by week 12 (Table 8). At the end of the study, Heiser reported, participants "were highly motivated to continue with these simple strategies."
Honorable mention: At the 2nd IAS Conference, Canadian researchers reported some success in treating NFV-induced diarrhea with different combinations of fiber (psyllium), loperamide, and calcium carbonate.42
In a study not covered by this reporter, 18 people took 1,250 mg of calcium carbonate twice daily. If that didn't help within 48 hours, they doubled the dose. After nine weeks the average number of bowel movements dropped from three to two daily, and diarrhea decreased significantly. A full report of this study is online.43
2. Get people with wide waists and insulin resistance to exercise.
A randomized study of 37 people with fat redistribution and hyperinsulinemia showed that exercise plus metformin outdoes metformin alone in correcting insulin imbalances and trimming central fat [abstract 4]. As a bonus, strength and aerobic capacity improved in the exercise group.
Susan Driscoll (Massachusetts General Hospital, Boston) randomized 18 people to take metformin at a dose of 850 mg twice daily and 19 to take metformin and start a supervised exercise program. Men had a waist-to-hip ratio above 0.9 and women a ratio above 0.85. The exercise group kept food diaries to ensure that their diets didn't change after they started to work out. For 12 weeks they completed three weekly exercise sessions consisting of 30 minutes of aerobic activity at 75 percent of maximum capacity and three sets of 10 resistance repetitions at 80 percent of the maximum weight wielded in a single repetition.
This regimen proved too taxing for some. Eight people dropped out of the exercise arm, most because they couldn't keep up with the workout schedule. Four people quit the metformin-only arm. One person in each group had a mild increase in lactates, a side effect of metformin. Among people who completed the three-month study, strength and aerobic capacity surged in the exercise-metformin group compared with the metformin-alone group (P=0.045). People who exercised also enjoyed significant improvements in:
Several small studies establish that aerobic or resistance exercise can cut total or central fat,44-46 along with triglycerides45, 47 and cholesterol45 in people with HIV infection. As Driscoll and other workers48 show, it also helps get glucose metabolism clicking again. IAS-USA guideline writers recommend resistance training to ward off bone loss.9 For people taking antiretrovirals, the only apparent drawback of aerobic exercise is the risk that subcutaneous fat may fade faster.46
For lots of HIV-infected people in rich countries -- injecting drug users, the homeless -- exercise will never be a priority. But clinicians should at least try to motivate all people with HIV who can exercise to start sweating. As two New York teams showed, workout routines are feasible even among poor inner-city dwellers.49, 50
3. Nag people to stop smoking.
Cigarette smoking turned up on the wrong side of the risk ledger in multivariate analyses from several 5th Lipodystrophy Workshop studies. Lighting up independently predicted lower leg fat in American women with HIV infection [abstract 15 above] and signaled thicker carotid artery walls in French and US populations [abstracts 2 and 35 above]. But the biggest indictment of tobacco came from APROCO cohort investigators, who found a nearly negligible rate of new cardiovascular disease in PI takers who didn't smoke [abstract 34].
Besides smoking, the only other variable that raised the risk of heart disease in these 1,281 people is a problem no one can correct -- getting older. Eight other factors in a Poisson regression model didn't matter: body mass index, gender, CD4 count, viral load, PI prescribed, dual NRTIs prescribed, alcohol consumption, and cannabis consumption. Indeed, reported Geneviève Chêne (INSERM U593, Bordeaux), cannabis connoisseurs had a nonsignificantly lower rate of cardiovascular disease (3.2 per 1,000 person-years, 95 percent confidence interval [CI] 0 to 7.7) than people who never inhaled the stuff (6.2, 95 percent CI 3.2 to 9.3).
From May 1997 through March 2003, Chêne counted 28 cardiovascular "events" (18 of them coronary heart disease) in 22 people for an incidence of 5.3 per 1,000 person-years (95 percent CI 3.0 to 7.3). Comparing people who smoked more than 10 cigarettes a day with those who smoked fewer than 10 and those who smoked none, she figured the following rates per 1,000 person-years and 95 percent CIs (P=0.003):
Only one nonsmoker had a new heart disease diagnosis. Among light and heavy smokers, the incidence of new diagnoses rose with advancing age (Table 9). Given these results, one might countenance a blunter conclusion than Chêne's suggestion that "smoking cessation should be seriously considered" in PI takers.
4. Watch for preeclampsia in pregnant women taking triple therapy.
Preeclampsia (pregnancy-induced hypertension and proteinuria) and stillbirths proved more common among antiretroviral-treated women at Barcelona's University Hospital Clinic than among pregnant women not infected with HIV [abstract 22]. Among the 71 women with HIV, 61 were taking a three-drug regimen, while seven others were taking one or two antiretrovirals. In the past two years, Oriol Coll and collaborators began treating HIV-infected pregnant women during all three trimesters. Coll suggested that if others confirm the heightened risk of preeclampsia and stillbirths with potent antiretroviral therapy, clinicians may want to reconsider antiretroviral recommendations for pregnant women.
The study involved all 7,720 pregnant women who delivered a child at the University Hospital Clinic from January 2001 through March 2003. Coll reported 104 cases of preeclampsia (1.3 percent) and 53 stillbirths (0.7 percent). The relative risk (RR) of both proved much higher among women with HIV than in the uninfected women:
A preliminary analysis suggested a correlation between AZT (but no other individual antiretroviral) and preeclampsia or stillbirth, Coll added. But the correlation may just reflect frequent use of prophylactic AZT among pregnant women with HIV infection.
5. Screen HIV-infected women for osteoporosis.
All women risk waning bone mineral density as they pass menopause, and dwindling bone density poses a threat to everyone with HIV infection. So no one professed surprise when two recent studies charted high rates of osteopenia and osteoporosis in HIV-infected women.51, 52 A comparison of 84 women with HIV and 63 age-, weight-, and race-matched controls confirmed the threat of weakening bones in women with HIV so strongly that routine screening of older or at-risk HIV-infected women seemed to many an inescapable conclusion [abstract 23].
Steven Grinspoon (Massachusetts General Hospital, Boston) reported no significant differences between the HIV group and controls in age (41 + 1 years in both groups), body mass index (26 and 27 kg/m3), lowest adult weight (57 and 59 kg), endocrine parameters, or daily calcium and vitamin D intake. The study excluded anyone taking a medication or suffering from a condition that would lower bone mineral density, anyone with an opportunistic infection or diabetes, any pregnant women, or anyone taking an anabolic steroid or estrogens.
DEXA-calculated t scores of the lumbar spine, total hip, and femoral neck proved significantly lower in the women with HIV than in controls. A significantly higher percentage of HIV-infected women had osteopenia or osteoporosis of the lumbar spine, total hip, and hip or spine (Table 10).
The women with HIV and the controls had similar levels of osteocalcin, a marker of bone formation. But the HIV group had significantly higher levels of urine N-telopeptide, a marker of high bone turnover (39.6 versus 29.2 nM/mM creatinine, P<0.05), and significantly higher levels of osteoprotegerin, which inhibits differentiation of bone-wasting osteoclasts (4.76 versus 3.39 pmol/L, P<0.0001). One would expect high osteoprotegerin quotients to protect against osteopenia. In these women with HIV, Grinspoon suggested, the high levels may indicate a compensatory response to faster bone turnover.
Grinspoon traced positive correlations between lumbar spine density and body mass index (r=0.38, P=0.00001), lowest adult weight (r=0.38, P<0.001), total fat (r=0.37, P<0.001), and total lean mass (r=0.4, P<0.001) and a negative correlation between lumbar spine density and N-telopeptide (r=-0.28, P=0.01).
In a multivariate analysis controlled for age, body mass index, menstrual function, and race, women with HIV were 2.6 times more likely to have osteopenia than controls (95 percent CI 1.2 to 5.8, P=0.02). Not one study of antiresorptive therapy found that it didn't work, Grinspoon noted. And there is no reason to assume it won't work in women with HIV infection. To be sure, some early results point in the right direction: A 48-week randomized trial involving 27 men and four women with HIV infection and osteopenia found significantly greater gains in lumbar spine density with alendronate plus vitamin D and calcium than with vitamin D and calcium alone.53
Because women with HIV have a magnified risk of osteopenia or osteoporosis, because diagnosis is accurate and noninvasive, and because effective therapy exists, Grinspoon proposed, bone mineral density should probably be measured even in infected women in whom you might not suspect osteopenia, and certainly in women in whom you do suspect it. William Powderly (Washington University, St. Louis) phrased that advice a little differently, saying all HIV-infected women 40 years old or older should be screened for low bone density. No one at the 5th Lipodystrophy Workshop disagreed.
6. Be aware of slow bone growth in HIV-infected children.
Women aren't the only HIV-infected people who run a high risk of thin bones. Girls and boys do too. Confirming an earlier study,54 Alessandra Viganò (L. Sacco Hospital, Milan) learned that children with HIV have significantly lower bone mineral density than healthy children of the same age [abstract 148]. And, unlike adults, children seem not to return toward normal bone densities when treated with potent antiretrovirals.
Viganò's study involved 32 HIV-infected, Caucasian children from 6.3 to 17.7 years old and 381 healthy controls from 5.7 to 19.2 years old. All 32 children with HIV had a viral load below 50 copies/mL and an average CD4 percent of 31.6 percent at baseline and 34 percent after one year of follow-up while taking d4T, 3TC, and a protease inhibitor. None of these children or any controls were taking sex hormones, corticosteroids, vitamin D, or calcium. The controls fell between the third and 97th percentile for anthropometric measures in their age group, none played competitive sports, and none had ever broken a bone.
Compared with controls, at baseline the children with HIV had significantly lower bone mineral density of the lumbar spine (0.083 versus 0.875 g/cm3, P=0.001), arms (0.647 versus 0.655 g/cm3, P<0.001), legs (0.923 versus 0.980 g/cm3, P<0.001), and total body (0.913 versus 0.933 g/cm3, P<0.0001). After one year of continued successful PI therapy, bone density in the children with HIV had not caught up with control bone density for arms, legs, or total body, though it did for the lumbar spine.
When Viganò figured annual changes in bone mineral density, the spinal growth spurt in the HIV group did not differ significantly from the annual change in uninfected controls. Growth rates of leg bone density also proved similar in the two groups. But annual bone growth for children with HIV significantly lagged normal growth in the arms and total body (Table 11).
Viganò proposed two conclusions:
During a 5th Lipodystrophy Workshop discussion of bone mineral density in women with HIV, Perth's David Nolan observed that the most important time to avoid osteopenia is during maximum bone formation -- in other words, during childhood and adolescence.
Mark Mascolini writes about HIV infection (firstname.lastname@example.org).