Introduction

When William Lewis (Emory University, Atlanta) started studying cardiomyopathy in mice exposed to zidovudine (AZT), his was a lonely pursuit. It was 1989, the tender morn of this antiretroviral age. And while many AIDS clinicians rejoiced to see how quickly anti-HIV therapy had come from bench to bedside, Lewis recalls he chose a road less taken--from bench to cageside.

Working with rats and mice, Lewis advanced the simple hypothesis that long-term AZT caused or contributed to dilated cardiomyopathy in people with AIDS [abstract I-1372*]. He suggested it might be a mitochondrial toxicity.

A decade later the rest of the world caught up with William Lewis. That was when a phalanx of physicians, armed with convincing cohort data, accosted attendees at the first Lipodystrophy Workshop with the news that nucleoside reverse transcriptase inhibitors (NRTIs) play a certain, if recondite, role in some of the body fat changes and metabolic meanderings collectively called lipodystrophy syndrome. At the same meeting three teams argued that mitochondrial disruptions account for these and other NRTI side effects.

Except for the link Lewis forged between mitochondrial depletion and AZT cardiomyopathy, this mitochondrial hypothesis remained unproved when many of these same researchers assembled in Toronto for the second edition of the Lipodystrophy Workshop, then hopped cabs downtown to stew over more side effects news at the 40th ICAAC.

And while mitochondrial anomalies may prove tough to pin down, unceasing study of lipodystrophy has at least limned some broad outlines of this syndrome. As workshop co-chair David Cooper (University of New South Wales, Sydney, Australia) told IAPAC Monthly, NRTIs seem more responsible for the facial and limb wasting, while protease inhibitors (PIs) appear to pack fat around the waist. "And the two" drug classes, he said, "exacerbate each other."

Although several studies at both meetings tied stavudine (d4T) to fat atrophy, higher triglycerides, or higher lactates, other studies flatly contradicted such trends. And attempts to assign blame to particular PIs fell short of assigning causality. But two studies showed convincingly that ritonavir or indinavir promptly incites metabolic mix-ups in people without lipodystrophy and without HIV infection--but they don't incite the same mix-ups.

How widespread is lipodystrophy among people taking antiretrovirals? By now, one may venture, that question should be answered. Indeed, the workshop's second co-chair, Morris Schambelan (University of California, San Francisco), adduced evidence from three "third-generation" prevalence studies¹ indicating that close to half of all persons taking antiretrovirals have lipodystrophy. That three-study concordance contrasts with the panoramic 2 to 84 percent prevalence range depicted in earlier studies.

The narrower 42 to 51 percent range in the cohorts Schambelan cited certainly reflects an emerging consensus on what constitutes lipodystrophy, and more uniform objective measures of treatment-linked abnormalities. But if fat changes seem a lot less common--or a lot more common--in your practice, don't fret. You and your brood may not be lonely outliers. An August 2000 survey of 300 physicians who see, on average, well over 100 HIV-infected people set the prevalence of lipodystrophy at 19 percent.² And a 26-month survey of the CHORUS cohort charted recent onsets of lipodystrophy in only 265 of 4025 people (7 percent) [abstract I-1286, see The Drugs below].

But at least one expert thinks those low numbers, and even the third-generation prevalence studies, grossly underestimate the proportion of people with HIV lipodystrophy. Speaking at ICAAC, endocrinologist Steven Grinspoon (Harvard Medical School, Boston) said he believes "the vast majority" of people taking antiretrovirals have some type of fat redistribution or metabolic disruption [abstract I-1373].

Even as the lipodystrophy plot deepens, other research uncovered two further complications of HIV disease--osteopenia and osteonecrosis. The first turns up in 20 to 50 percent of surveyed populations, and in one study more than 4 percent had asymptomatic osteonecrosis. Although some work implicates antiretroviral therapy in these inauspicious osseous changes, a few studies point a finger at HIV itself (see The Bones below).

Meanwhile, murine maven William Lewis has picked up a few new colleagues. Lipodystrophy Workshop registrants heard the latest about metabolic research on the A-ZIP mouse, an ultra-sleek animal with "virtually no white adipose tissue" [abstract 1], and its hapless cousin the SREBP mouse, a rodent engineered to grow up lipodystrophic [abstract 2]. While these efforts qualify as in vivo--and while much other strictly in vitro work also gropes toward potential mechanisms--the most vivid examples of lipodystrophy surely remain the ones in your waiting room. And both ICAAC and the workshop had plenty to say about them.

Return to Article Contents

The Heart

Every meeting has at least one standout study, one that straightens people in their seats with its clever design, level-headed logic, or crystalline conclusion--sometimes all three. Both the Lipodystrophy Workshop and ICAAC had one such presentation, and it was the same one. Epidemiologist Matthias Egger (University of Bristol, UK) took aim at one of the thorniest questions in lipodystrophy research:

Will HIV treatment-linked fat and metabolic changes substantially raise the risk of heart disease?

And he spelled out his answer at both meetings.

Most HIV experts who dare venture an opinion on this question fudge. And who can blame them? Many of the unwanted changes that accompany antiretroviral therapy--rotund waistlines, high-rise insulin, skyscraping triglycerides, to name but three--forebode heart disease in people without HIV infection. So the gut says they'll do the same in seropositive people. But until the numbers come in--over the decades it can take for disparate risks to conflate in a plugged coronary artery--no one can say how much HIV meds may hasten the crisis.

Hints abound, though. And to complicate things, HIV itself starts stirring the metabolic pot before most people swallow their first PI. Carl Grunfeld (University of California, San Francisco) showed eight years ago that protective high-density lipoprotein (HDL) cholesterol tumbles early in the course of HIV infection; then triglycerides start to climb.³ Baneful low-density lipoprotein (LDL) cholesterol also falls, but "the decrease in LDL is much smaller [than the decrease in HDL] and as a consequence does not completely offset the increased risk" of atherosclerosis with lowered HDL.⁴

Return to Article Contents

PIs, the pump, and the arteries

A small study by Frank Goebel (University of Munich, Germany) amply demonstrated overlapping cardiovascular risks from HIV itself and from highly active antiretroviral therapy (HAART) [abstract 26]. He used positron emission tomography (PET) to track heart rate, rate pressure product, and myocardial blood flow before and after giving atropine to stress the hearts of 11 people taking a PI combo, eight antiretroviral-naive people, and 10 seronegative volunteers. Electron beam tomography measured calcium density in coronary arteries.

Everyone in the HAART group had a cholesterol level above 300 mg/dL for at least two years, though their cholesterol readings were normal before they began taking PIs. People in the treatment-naive group and the healthy volunteers had normal cholesterols during the study. Although the HAART and naive groups matched closely on average CD4⁺ count, about 440 cells/mm³, the untreated people were younger. Their age averaged 36 years, compared with 48 for the HAART contingent. And the healthy volunteers were younger still, averaging 26. So the comparison is less than clean but nonetheless instructive.

Atropine got everyone's heart pounding, but the PI-treated and naive groups both had blunted heart rates and rate pressure products compared with healthy controls. HAART-treated people had a significantly smaller increase in myocardial blood flow than did the naive or control group. Interim electron beam results involving eight HAART patients showed normal calcium density scores in five of them.

Goebel concluded that PI regimens may fiddle with cardiac function, but HIV itself apparently starts scraping those heart strings even in 30-somethings with unremarkable cholesterol counts. Of course this little study says nothing about how, or whether, the sluggish hearts in Goebel's patients augur early coronary disease.

In the abstract on his study, Goebel speculated that "endothelial dysfunction may be discussed as an explanation" of the cardiovascular irregularities he found. Indeed, another workshop study, presented by James Stein (University of Wisconsin, Madison, USA) lent credence to that surmise. This single-site cross-sectional study compared flow-mediated vasodilation of the brachial artery in 22 people taking a PI regimen and 15 taking regimens without PIs [abstract 24]. How well that shoulder-to-elbow artery dilates when the heart beats suggests endothelial cell health in arteries.

People in the PI group had taken a protease drug for an average of 31 months, and their average duration of antiretroviral treatment (70 months) closely matched the non-PI group (68 months). Average CD4⁺ counts for both groups measured in the low 400s. But the PI takers weighed more as a group (body mass index 27 kg/m² versus 24 kg/m², P = 0.069), and they had more ample waist-to-hip ratios. Triglyceride and total cholesterol levels also ranged significantly higher in the PI group (P = 0.007 and P = 0.009 respectively). No one in either group smoked, and age and gender (78 percent men) were equivalent.

These researchers measured flow-mediated vasodilation by ultrasonograms analyzed three times by two technicians blinded to the antiretroviral regimens. Brachial artery diameters were similar in the two groups. But a springy 8.1 percent dilation in the non-PI group significantly exceeded the flaccid 2.6 percent stretch in the PI-treated people (P = 0.005). Stepwise regression analysis fingered four factors that independently predicted flow-mediated dilation (P < 0.001): brachial artery diameter, systolic blood pressure, heart rate, and PI treatment. Variables that didn't predict dilation included CD4⁺ count, viral load, time since HIV infection, and time on treatment.

Stein readily owned up to the study's limitations. The sample is small and the design cross-sectional. The University of Wisconsin team measured dilation only once. Michael Dubé (Indiana University, Indianapolis, USA) added another drawback: Stein didn't measure insulin sensitivity or glucose tolerance, both of which could account for some of the dilation difference between the groups.

Thickness of the carotid artery wall--the intima media, to be specific--offers another hint of atherosclerosis, and that's what Renato Maserati (San Matteo Hospital, Pavia, Italy) measured in another cross-sectional study [abstract 25]. His team used ultrasonograms to size up intima media thickness in 27 people taking PIs for at least 18 months, 16 HIV-infected treatment-naive individuals, and 15 seronegative controls. A single technician read the scans without knowing which group they came from. No one in the PI contingent had taken a nonnucleoside reverse transcriptase inhibitor (NNRTI), and no study participant had a history of coronary artery disease.

Carotid walls proved significantly thicker in the PI group (mean 0.63 mm) than in the naive individuals (0.45 mm) or the healthy volunteers (0.5 mm) (P = 0.0003). Triglycerides were significantly higher in the PI patients (P = 0.005), and HDL levels significantly lower (P = 0.001), than in the naive and control groups.

Maserati concluded that the results "confirm that risk factors for coronary artery disease are markedly elevated in PI-exposed patients." But one attendee observed that the intima media measures, though higher in the PI group, still fell within the normal range. As in Stein's study, the cross-sectional design and small size necessitate independent confirmation of these findings in larger populations studied over time. Such a longitudinal design would tell whether artery walls continue thickening as PI treatment extends past two years.

The findings of these three studies, though worrisome, hardly prove that HAART will touch off an epidemic of heart disease. But one big retrospective study published just before the workshop and ICAAC did find a higher rate of myocardial infarction (MI) since potent regimens came into play.⁵ Volker Rickerts (J.W. Goethe University, Frankfurt, Germany) split the HIV clinic's 4993-person cohort into four time blocks coinciding roughly with preantiretroviral times, the monotherapy era, dual nucleoside days, and the age of HAART (Table 1). The incidence of MI per 1000 patient-years jumped in the last time bracket (P = 0.002). Multiple regression analysis linked age over 40 years and HAART with more MIs.

Interpreting these findings is not easy. First, this trawl through patient records did not look into nondrug risk factors like family history, hyperlipidemia, and smoking. Second, the Frankfurt team reported that the median age of cohort members increased significantly from the first to the second observation period (P < 0.001), and from the third to the fourth (P < 0.05). People taking HAART were significantly older than those taking other regimens (P = 0.0001).

Speaking at ICAAC, Judith Currier (University of California, Los Angeles) noted that age can be critical in such a retrospective analysis [presentation I-1877]. Before PIs' advent in 1997, and especially before NRTIs hit the market, more people died much younger with AIDS. Since older age correlated with more MIs in the Frankfurt analysis, as it typically does, those younger pre-HAART deaths may well have kept MI incidence artificially low before the days of potent antiretrovirals. Indeed, overall mortality in the cohort fell in the last two observation periods. The Frankfurt authors cite Currier's point as a possible confounder and add that the average age of the people who died of MIs (47 years) was significantly older than the age of the entire population (34.7 years, P < 0.0001). Finally, not everyone in the HAART era actually took HAART.

Return to Article Contents

How many can you treat before harming one?

Is there a better way to reckon the impact of potent antiretrovirals on heart disease? Matthias Egger thinks so, and he explained why twice in Toronto [abstracts 23 and I-1374]. He assumes, as most in this field do, that the metabolic and body shape changes that accompany treatment could spell big trouble for cardiovascular health as people live longer with HIV infection. But this epidemiologist is brasher than most in voicing his opinion. "We must assume that lipodystrophy is associated with increased CHD [coronary heart disease] risk," he said at ICAAC. "It would be unreasonable not to."

But how to gauge that risk? What's lacking in earlier antiretroviral-CHD risk calculations, according to Egger, is a tool that balances risk against the life-prolonging benefits of HAART. Epidemiologists have two such tools, called NNT_B and NNT_H.

NNT_B = the number of persons needed to be treated for one person to benefit

NNT_H = the number of persons needed to be treated for one person to be harmed

For example, among the Multicenter AIDS Cohort Study members with the highest risk of HIV disease progression,⁶ Egger calculated an NNT_B of 1 to 2. In other words, you need treat only one or two such people for one of them to benefit, a fine rate of return. But for Swiss HIV Cohort members with CD4⁺ counts between 300 and 500 cells/mm³ and viral loads above 10,000 copies/mL, Egger figured an NNT_B of 50. In other words, you must treat 50 such people with antiretrovirals to yield a life-prolonging benefit in one. Meanwhile, all of the treated 50 eventually endure drug side effects.

Egger then took the cardiovascular risk numbers from the Australian cohort described by Andrew Carr⁷ (St. Vincent's Hospital, Sydney). That cohort displayed a range of risks in PI-treated and PI-naive individuals, including high triglycerides and total cholesterol, low HDL cholesterol, central fat mass, and age. Egger plugged those numbers into the Framingham Heart Study risk equation, and into a Framingham equation modified by French epidemiologists to reflect the lower rates of heart disease in France.

That allowed Egger to tally both the absolute five-year risk of coronary heart disease and the NNT_H (Table 2) for US and French men and women stratified by age (30 or 50 years), by whether they smoke, and by whether they have lipodystrophy. The lessons of the five-year risk exercise, Egger reported, were:

1. Don't smoke.
   
2. Don't get lipodystrophy.
   
3. Don't get old.
   
4. Move to France.
   
5. Become a woman.
   

Sadly, all of these risk factors may not be modifiable. But a few comparisons make the point:

• For a nonsmoking 30-year-old US man with lipodystrophy, the absolute risk of heart disease in five years is 1.9, compared with 0.5 for the same man without lipodystrophy. So lipodystrophy in such men raises the risk of heart disease 3.8 times.
  
• For a smoking 30-year-old US man with lipodystrophy, the absolute risk of heart disease in five years is 3.6, compared with 1.1 for a smoking 30-year-old US woman with lipodystrophy. Here, being a man instead of a woman ups the risk 3.3 times.
  
• For a smoking 50-year-old US man with lipodystrophy, the absolute risk of heart disease in five years is 14.0, compared with 8.3 for a smoking 50-year-old French man with lipodystrophy. For these men, living in the United States instead of France boosts the risk 1.7 times.
  

The age, smoking, gender, and nationality comparisons buttress what's already known about the risk of pinhole coronaries, but an absolute risk comparison of lipodystrophy or no lipodystrophy in every other category--men, women 30-year-olds,

50-year-olds, smokers, nonsmokers--invariably favors the no-lipo option (Table 3). Just one more example:

• A 30-year-old smoking US woman without lipodystrophy has an absolute five-year heart disease risk of 0.1, compared with a risk of 1.1 for the same woman with lipodystrophy--an order of magnitude increase.
  

Just as unsettling were Egger's NNT_H numbers (Table 2). Older age and cigarette smoking greatly tip the odds against antiretroviral safety for both men and women. Whereas 217 nonsmoking 30-year-old US women must be treated with antiretrovirals to harm one with drug side effects, only 15 nonsmoking 50-year-old US women or 10 smoking 50-year-old US women must be treated to harm one. The same brutal calculus applies to men.

Egger urged his listeners to assess cardiovascular risk routinely in HIV-infected people, taking into account age, sex, lipids, insulin resistance, and family history. Cardiologists, he noted, routinely carry heart risk calculators in their pockets, and "HIV doctors should be moving in the same direction." (See "Rating risk factors for heart disease.") If Egger's estimates are correct, or even close, erasing one risk factor can bump a person down into an entirely different risk bracket. For example, he observed, stopping smoking "can bring a patient [with lipodystrophy] all the way back to being a nonlipodystrophy patient."

Of course formulas of this ilk are less than Talmudic. "Obviously, what I showed you are estimates," Egger cautioned, "and they should be taken with a grain of salt." He assumed a severe form of lipodystrophy; he assumed continuous antiretroviral treatment; and he assumed that risk profiles in a seronegative population can apply to people with HIV infection. But on that last score, Egger suspects the bias may favor people without HIV. He believes, for example, that HIV-infected people smoke more than age-matched counterparts without HIV, so heart disease risk in seropositive people "may be worse than we think."

After hearing Egger's talk at the Lipodystrophy Workshop, Carl Grunfeld promptly pointed out that his own cardiovascular risk calculation showed that taking antiretrovirals, in itself, only modestly upped the odds of heart disease. But piling other risk factors on top of antiretroviral therapy raised the odds synergistically.

"We know that we will see a difference" in CHD risk with antiretroviral therapy, Grunfeld said. "We know that risk will not be outlandishly different from what is predicted" without therapy. But Egger's arithmetic--showing that lipodystrophy increases the five-year risk of heart disease anywhere from 2.5 to 20 times depending on age, gender, and smoking (Table 3)--may brook a more ominous interpretation. If his numbers are right, those odds may qualify as outlandish.

Return to Article Contents

The Cause

It's multifactorial. These days, multifactorial trips from the tongues of those yearning to explain lipodystrophy as nimbly as paradigm shift once tripped from the tongues of PI enthusiasts. It's surely the best explanation, but it's also another way of saying, "Lots of things are going on and no one's sorted them out yet."

Speaking at ICAAC, Harvard's Steven Grinspoon was as precise as anyone can be about the cause of lipodystrophy, explaining that "It's most likely a complex interaction between protease inhibitors, nucleosides, and nondrug factors." But which PIs and which NRTIs, or does it matter? And which nondrug factors are cause--and which consequences--of fat changes?

A multicenter study by German clinicians, for example, found in a multivariate analysis that hypertriglyceridemia more than doubled the risk of lipodystrophy [abstract I-1287, reviewed below]. But that correlation, they noted, "does not clarify whether hyperlipidemia is a cause or effect of the syndrome."

Lipodystrophy Workshop co-chair Morris Schambelan considered the flip side of the multifactorial doctrine in a recent online review.⁸ Perhaps it's not so much the factors that are multiple, he suggested, as the syndromes. "More than one syndrome of metabolic abnormalities and body habitus changes may be present in HIV-infected patients," Schambelan wrote, noting that "metabolic abnormalities and fat distribution changes are not always linked." And recent work, discussed below, suggests that distinct mechanisms may drive lipid and glucose metabolism in people taking PIs.

While experts like Schambelan struggle to bring the big picture into focus, others work to tease out tiny clues. And often this work doesn't require study participants to sign consent forms, because mice can't write. Marc Reitman (National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, Maryland, USA) studies lipoatrophy in A-ZIP/F-1 mice [abstract 1]. After some genetic rejiggering, A-ZIPs have practically no white fat, little brown fat, severe insulin resistance, and diabetes.

One big problem when adipose tissue disappears, Reitman learned, is that the body has to find new places to store fat--like the liver. In the mice, that caused hepatic steatosis. So what did Reitman do? He stuck globules of fat back into these fat-free mice and, lo, insulin sensitivity improved, insulin levels dropped, hyperglycemia disappeared, and complications like steatosis "were either partially or completely reversed."

In theory, Reitman said, fat transplants could ameliorate some problems in HIV-infected people with lipoatrophy and insulin upsets. "But you'd have to put a lot of fat in," he noted, because the dose-response effect in mice is marked. So retroliposuction does not appear to be on the horizon. And Reitman agreed with attendees who observed that abnormal distribution of fat--not fat loss alone--is often the problem in HIV lipodystrophy. Fat in different places does different things, he said.

Iichiro Shimomura (University of Texas, Dallas) prefers another murine model of lipodystrophy, the SREBP mouse [abstract 2]. These animals can't crank out full-length versions of a nuclear protein labeled nSREBP-1c. As a result their adipose tissue does a bad job producing fat-specific proteins like leptin, which regulates food intake and metabolism. SREBP mice grow up with hyperglycemia, hyperinsulinemia, insulin resistance, and lipodystrophy.

Shimomura decided to remedy the lack of leptin by affixing little leptin pumps to these mice and infusing 5 µg daily for 12 days. It worked. Plasma leptin levels returned to normal and reversed the hyperglycemia. Just like these mice, humans with generalized (non-HIV) lipodystrophy have little leptin to spare. So Shimomura has begun a trial of leptin supplements in them.

Return to Article Contents

PIs, fat, and when you eat

What works in SREBP mice may not work in humans with lipodystrophy, regardless of their HIV serostatus. Indeed, what works in SREBP mice didn't even work in Marc Reitman's mouse model of lipodystrophy. He found leptin infusions "only slightly effective"; lofty insulin and glucose levels persisted.

But SREBP does appear to drive insulin resistance in human proto-fat cells doused with indinavir. Jacqueline Capeau (St. Antoine Faculty of Medicine, Paris) found that high doses of this PI, 10 µg, impaired the differentiation of cells destined to become adipocytes at an early step in that process [abstract 3]. As in Shimomura's mice, the SREBP protein in these cells was crippled. Those findings led Capeau to propose that faulty SREBP function could explain the deficient formation of fat cells. So far, though, Capeau has worked only with a preadipocyte cell line, not with nascent adipocytes harvested from humans.

Work by Greg Stevens (Pfizer Global R&D, San Diego) supported Capeau's finding that PIs stunt the orderly growth of fat cells from precursors [abstract P3]. But he blamed a different culprit, 11ßHSD-1, an enzyme needed to maintain stores of intracellular cortisol. "Alterations in this enzyme," Stevens proposed, "could play a key role in alterations in adipocyte metabolism and function." Indinavir and ritonavir, it turned out, muffle expression of 11ßHSD-1 in adipocytes significantly more than saquinavir, nelfinavir, amprenavir, d4T, or efavirenz (P < 0.05). But his experiment could not nail down whether indinavir and ritonavir directly affect 11ßHSD-1, or whether scanty expression of the enzyme is simply a marker of stymied fat cell maturation.

Two studies suggested that fat may pile up in cells at different rates depending on whether someone takes a PI with a meal. The two research teams suggested that some PIs may best be taken with food, and others without. Oliver Distler (University of New South Wales, Sydney, Australia) spritzed human carcinoma (HepG2) cells and rat and human microsomes with ritonavir or saquinavir [abstract 19]. The PIs, he found, moderately stifle triglyceride synthesis. But ritonavir may unbalance lipids by blocking clearance of apolipoprotein B (apoB), the key component of insalubrious very low density lipoprotein (VLDL).

Distler proposed that, when a person taking PIs isn't eating, apoB gets stockpiled in cells. As soon as that person dines, the fat loading jacks up apoB levels so high that storage cells can no longer contain the protein. So it floods into circulation. Calling this process "a block followed by a burst," Distler suggested that taking PIs with meals may eliminate the effect.

Another study using HepG2 cells contradicted Distler's finding about triglyceride synthesis. James Lenhard (Glaxo Wellcome) found that ritonavir, saquinavir, nelfinavir, and lopinavir (ABT-378) fire up triglyceride forges significantly more than amprenavir or indinavir (P < 0.05) [abstract P9]. That difference from Distler's finding, Lenhard suggested, could be explained by different experiment designs. Distler measured the immediate effects of the PIs he studied, whereas Lenhard gauged triglyceride synthesis after soaking cells with PIs for 24 hours. On the other hand, in vivo work by Indiana University's Michael Dubé indicated that amprenavir does drive up triglycerides in PI-naive individuals, whereas indinavir does not [abstract P4, see "First-line amprenavir" below].

In food-deprived mice Lenhard found that ritonavir magnifies levels of glucose by 29 percent, cholesterol by 40 percent, and triglycerides by 99 percent. Nelfinavir, on the other hand, drives up none of these levels in hungry mice. But nelfinavir does boost serum triglycerides in fed mice. Fasting, Lenhard concluded, "reduces nelfinavir-associated hypertriglyceridemia."

Return to Article Contents

Fast effects of two PIs in people without HIV

No one proposed that the findings of Oliver Distler and James Lenhard should inform meal planning with your favorite PI. But the results do support Lenhard's long-standing claim that individual PIs vary in their metabolic effects. Two studies, one presented at the Lipodystrophy Workshop and another published earlier this year, bolstered that contention. And both studies showed that PIs themselves directly affect lipid, glucose, or insulin levels--and quickly--because the studies involved short courses in people without HIV infection.

In a double-blind study, investigators from the University of Washington, Seattle, gave ritonavir to 11 healthy volunteers and placebo to eight.⁹ It didn't take ritonavir long to send lipid levels skyward. Compared with the placebo group, people taking the PI had significantly higher quotients of plasma triglycerides, VLDL, intermediate-density lipoprotein (IDL) cholesterol, apoB, and lipoprotein when the two-week course ended. Their body composition didn't change. Apparently, ritonavir itself contributes at least to triglyceride and cholesterol abnormalities in people with HIV infection. Altered metabolism resulting from ritonavir-induced drops in HIV load obviously played no role in the seronegative people studied.

At the Lipodystrophy Workshop, Mustafa Noor (University of California, San Francisco) reported that indinavir also promptly riles metabolic markers in seronegative people, but not the same markers roused by ritonavir [abstract 10]. In this open-label study, 10 healthy volunteers served as their own controls. After four weeks of standard-dose indinavir, Noor charted no changes in standard fat values measured by DEXA or CT. And he found nothing askew in readings of triglycerides, free fatty acids, or total, LDL, or HDL cholesterol.

But several insulin and glucose measures vaulted significantly from baseline during the four-week indinavir course: fasting plasma glucose (P = 0.05) and insulin (P < 0.05); insulin/glucose ratio (P < 0.05); insulin resistance index (P < 0.05); and two-hour glucose (P < 0.05) and insulin (P = 0.05) by oral glucose tolerance test. Glucose and insulin sensitivity strayed so far in one person, in fact, that diabetes was diagnosed. Of course ritonavir may have similar effects in healthy volunteers, but the University of Washington team limited the study to lipid analysis.

Besides the eye-opening verification that PIs briskly bollix metabolics in people with no fat changes, these studies also hint that some in vitro findings are on track. For example, the different effects of ritonavir and indinavir on triglycerides reflect the HepG2 cell study by James Lenhard [abstract P9, above]. As in the human studies, ritonavir didn't take long to juice up triglyceride synthesis in Lenhard's cell study, whereas indinavir swayed triglycerides hardly at all. The human studies also make it clear that results of an experiment involving one or two PIs can't be blindly applied to all drugs in the class.

Another workshop study showed that PIs and NRTIs may latch onto different lipids in different ways. If verified, such differences could begin to explain the varied effects of these drug classes on fat. Lisa Ware (University of Southampton, UK) studied seven people with lipodystrophy who had taken only NRTIs (for an average 49 months) and six taking both PIs and NRTIs (for an average 12 months) [abstract 20]. Although nucleoside use by both groups somewhat clouds the differences Ware found, she claimed those differences are too great to be explained by the mere addition of another drug (the PI) exerting the same effect.

After eating, people taking PI regimens had much higher levels of all major lipids than did the NRTI group (P = 0.05). While the PI takers had a tougher time clearing dietary lipids in the form of triacylglycerol (P < 0.05), people taking NRTIs struggled harder to clear dietary lipids in the form of nonesterified fatty acid (P < 0.05). Both drug classes, Ware concluded, may lead to hyperlipidemia, but by different routes.

Return to Article Contents

Mitochondria concertante (fuga)

Rex Parker (Bristol-Myers Squibb) challenged the hypothesis that NRTI-linked lipodystrophy involves mitochondrial toxicity [abstract 4]. He exposed murine fat cells to PIs and NRTIs to determine the drugs' effects on triglyceride accumulation and mitochondrial function. Nelfinavir, saquinavir, and ritonavir inhibited formation of triglycerides at concentrations relatively lower than those of indinavir, amprenavir, abacavir, d4T, or lamivudine (3TC). Similarly, nelfinavir, saquinavir, and ritonavir disrupted mitochondrial function (measured by ATP) at much lower concentrations than indinavir, amprenavir, or the NRTIs. Individual NRTIs did synergize with ritonavir to inhibit triglyceride formation.

In discussion, an audience member pointed out that Parker had not tested the potential synergism between two NRTIs and a PI, although that's how clinicians typically combine the drugs in practice. Parker noted that other study limitations are time dependence (cells were exposed to drug for four to six days) and dependence on the drug concentrations selected. Those limitations, of course, apply to other cell studies.

Some research in people with HIV infection, however, did tend to bolster the mitochondrial hypothesis. Perhaps the most compelling findings so far came from Cecilia Shikuma (University of Hawaii at Manoa, Honolulu), who correlated NRTI use and lipodystrophy with mitochondrial DNA (mtDNA) levels in adipose tissue biopsies from three sites--the back of the neck, the abdomen, and the upper thigh [abstract 7]. She reported results from four groups: eight people with physician-confirmed self-reported lipodystrophy (all had peripheral lipoatrophy) and at least a six-month history of NRTI-HAART; seven NRTI-HAART-treated patients without lipodystrophy; two NRTI- and HAART-naive patients; and seven people without HIV infection.

Biopsies of the NRTI-treated people with lipodystrophy had significantly less mtDNA than those of the seronegative controls (P < 0.001), the treatment-naive individuals (P = 0.001), or the NRTI-treated people without lipodystrophy (P = 0.018). Levels of mtDNA in peripheral blood mononuclear cells (PBMCs) didn't differ between any of the groups. The people in the NRTI group with lipodystrophy had a much longer treatment history than NRTI takers without lipodystrophy (101.1 versus 50.6 months, P < 0.05). But the groups did not differ in exposure to d4T.

This analysis does not clinch the case for mitochondrial toxicity, for a few reasons. The investigators might have used more objective measures of lipoatrophy, and pathologic levels of mtDNA in such tissues have never been established. Graeme Moyle (Chelsea and Westminster Hospital, London) observed that some of the people with lipodystrophy did not have depleted mtDNA. Shikuma suggested those may be the two people who traded their PI for an NNRTI, and that perhaps PIs but not NNRTIs exacerbate the mtDNA-draining effect of nucleosides. "But this is only a guess," she added. And if that guess is right it would warp the mitochondrial hypothesis with a whole new wrinkle.

A similar study by Ulrich Walker (Albert Ludwig University, Freiburg, Germany) offered a slightly cleaner drug-class distinction than Shikuma's analysis, and his results leaned in the same direction [abstract 6]. Along with 19 people taking NRTIs and PIs, Walker managed to recruit four who had tried only NNRTIs and PIs, one treatment-naive individual, and eight volunteers without HIV infection. Whereas Shikuma biopsied three sites, he biopsied only the buttocks. And the methods the two researchers used to tally mtDNA also differed: Walker used Southern blots, while Shikuma used PCR.

Walker found significantly less mtDNA in adipose tissue from the NRTI-treated people than from the five NRTI-naive individuals (P = 0.009). But mtDNA content did not differ significantly between the NRTI-naive people and the healthy controls. MtDNA was 38 percent lower among 11 treated individuals with clinical evidence of lipoatrophy than in 12 treated persons without lipoatrophy (P = 0.04). Most of the people with lipoatrophy were taking d4T, but they had been using AZT. There were not enough people taking their first NRTI regimen to permit valid comparisons between individual nucleosides.

The tough workshop audience challenged Walker on two points: First, the biopsies almost certainly included cells other than adipocytes, a criticism that holds for Shikuma's study. Second, the average mtDNA decrease Walker measured, 44 percent, is probably not pathologic. Still, the two studies do offer the first bits of evidence implicating mitochondrial depletion in fat atrophy among people taking NRTIs, though not particular NRTIs.

Returning to the rodent kingdom, I. Gaou and colleagues (Beaujon Hospital, Clichy, France) took a closer look at whether d4T saps mtDNA in mice genetically programmed to be skinny and mice programmed to be fat [abstract I-1628]. Gaou used dot blot hybridization to tote mtDNA in liver, brain, heart, skeletal muscle, and white adipose tissue of lean mice given d4T at a dose of 100 or 500 mg/kg daily for one to six weeks and in obese mice given 100 mg/kg daily for six weeks.

In lean mice 500 mg/kg daily--roughly 400 times the d4T dose downed by humans--scrubbed mtDNA from liver and skeletal muscles, but not from brain or heart. The 100-mg dose--about 80 times what humans take--left mtDNA stores unchanged in skeletal muscle and white fat. In obese mice 100 mg/kg daily for six weeks sliced mtDNA levels by 45 percent in white fat but didn't deplete mtDNA in liver or skeletal muscles.

Gaou concluded that the doses of d4T needed to drain mtDNA from mice far exceed human doses. "This suggests," the poster read, "that concomitant factors including peculiar genetic/metabolic background, might be necessary to trigger major mtDNA depletion." But the French team allowed that "further investigations will be needed to determine whether specific mtDNA depletion in [white adipose tissue] may trigger lipoatrophy in predisposed individuals."

Return to Article Contents

The Drugs

At least a dozen studies at ICAAC and the Lipodystrophy Workshop addressed what may be the most clinically pressing question about lipodystrophy: Is it worse, or better, in people taking certain antiretrovirals? Glaxo researchers have long contended that amprenavir will foster fewer fat changes than other PIs, and James Lenhard's workshop poster offered in vitro evidence that amprenavir--and indinavir--forge fewer triglycerides than the other PIs [abstract P9, above]. But three other reports had in vivo results, which weren't always consistent. And one of them raised questions about Lenhard's triglyceride finding.

More contentious is the ongoing fray over d4T. Is this nuke the fat wasting kingpin, or an innocent fall guy caught in some Hitchcockian nightmare of mistaken identity? If one reads online discussions among people with HIV, like Lipidlist or the PI-Treatment List, it's obvious that lots of folks have heard this story and worry about their d4T when a newly hollowed face stares back at them from the morning mirror. And some of their physicians, they report, are trading their d4T for AZT or abacavir, and hoping for the best. Are those docs jumping the gun, or is it past time to act? Together the Toronto meetings yielded a trove of new evidence on d4T, some incriminating, some exculpatory.

Return to Article Contents

First-line amprenavir churns lipids

Recent studies, described in the preceding section of this article, make it clear that ritonavir and indinavir have rapid--but different--metabolic consequences when taken by healthy volunteers. Ritonavir quickly drives up lipids of every ilk, while indinavir leaves lipids alone (at least in the first few weeks) and mucks up measures of insulin and glucose. What about amprenavir, whose developers worked hard to prove it's a cleaner PI? Two studies looked at people who switched from another PI to amprenavir, and a third charted lipid and glucose changes in people taking amprenavir as their first protease drug.

A 48-week open-label study tracked 84 men and seven women with hyperlipidemia, with or without body fat changes, who swapped their PI for amprenavir [abstract P88]. About two-thirds of them were taking 3TC, and about half d4T. Louise Pedneault (Glaxo Wellcome) reported prompt swoons in fasting triglycerides, cholesterol, and glucose among 48 people still taking amprenavir after 12 weeks and among 15 people after 48 weeks (Table 4).

Table 4. Median changes in fasting metabolic measures after switching
to amprenavir from another PI* (compare with Table 5)

	Week 12	Week 48
Triglycerides (mg/dL)	188 (n = 48)	389 (n = 11)
Cholesterol (mg/dL)	21 (n = 47)	32 (n = 12)
Glucose (mg/dL)	6 (n = 48)	7 (n = 15)
*All study participants had grade 1 to 4 hypertriglyceridemia or hypercholesterolemia before switching to amprenavir. Baseline values were not reported. Source: Louise Pedneault, abstract P88.

But Pedneault saw "little or no improvement" in fat redistribution calculated by anthropometric measures and a standard form used by physicians in this multisite study. As PI-to-NNRTI switch studies show, objectively measured fat abnormalities only rarely resolve within a year of the switch and may never resolve. That appears to be true for amprenavir as well.

A retrospective French study of 30 men and one woman with lipodystrophy who switched from a PI to amprenavir because of virologic failure did show limited body shape improvements but failed to confirm the postswitch tempering of metabolic measures that Pedneault reported (Table 5) [abstract P62].

Average cholesterol and glucose levels were just below the upper limit of normal when the French patients started amprenavir. But average baseline triglycerides stood at a high 257 mg/dL and changed not at all during six months of amprenavir. The high-normal cholesterol and glucose measures didn't change much either. (Pedneault didn't report baseline lipid and glucose numbers.) The French study chief, Willy Rozenbaum (Rothschild Hospital, Paris), noted that background antiretrovirals rarely changed with the switch to amprenavir, so other drugs probably didn't affect whatever metabolic impact the switch may have had.

Among the 31 people studied, seven (23 percent) had some regression in clinical signs of lipodystrophy at six months, determined by physician evaluation and verified by two other physicians. Nineteen (61 percent) had no physical changes, and one (3 percent) had new signs of lipodystrophy.

A small but carefully designed study of nine men and one woman taking amprenavir as their first PI suggested again that different protease drugs have different early metabolic effects [abstract P4]. Indiana University's Michael Dubé reported that eight weeks of amprenavir (plus abacavir and 3TC) drove up total cholesterol and LDL cholesterol but did not upset glucose, insulin, insulin resistance, or insulin sensitivity.

Those results contrast diametrically with what Dubé found in a similar study of people starting indinavir, in whom lipids stayed steady but fasting glucose rose and insulin sensitivity fell in the first eight weeks of treatment.¹⁰ And those earlier indinavir findings held true in people without HIV infection who took that PI for four weeks, as Mustafa Noor reported at this year's workshop [abstract 10, above].

Dubé found that eight weeks of the amprenavir regimen significantly boosted total cholesterol (from 163 to 213 mg/dL, P = 0.002), directly measured LDL (from 77 to 100 mg/dL, P = 0.006), LDL calculated by an equation (from 113 to 152 mg/dL, P = 0.002), and LDL estimated by NMR Lipoprofile (from 105 to 139 mg/dL, P = 0.002). Those LDL gains appeared to drive the jump in total cholesterol. Triglycerides also rose, though not quite significantly, from 103 to 144 mg/dL (P = 0.08). This early in vivo waxing of triglycerides casts some doubt on Glaxo's in vitro finding that amprenavir, like indinavir, doesn't crank up triglyceride synthesis. If amprenavir doesn't speed synthesis, triglycerides must pile up for some other reason.

Food diaries kept by study participants showed that they didn't suddenly load up on calories or fat during the study, so Dubé suggested that the cholesterol surge he measured reflects "a direct drug effect, possibly with a component of improved virologic control." Everyone responded well virologically to amprenavir.

Dubé's nice study shows yet again that individual PIs do different things when they broach the body's metabolic factories. "The divergent results with indinavir and amprenavir regimens suggest that PI-associated glucose and lipid dysregulation may not always co-exist," Dubé proposed "Thus, there may be different pathogenic mechanisms responsible for glucose and lipid derangements."

If longer studies confirm Dubé's findings, will they dilute the allure of amprenavir as a first-line PI? Probably not, an ICAAC audience survey suggested, because most HIV docs have apparently made up their mind about first-line prescribing of a PI with such a high pill burden and no virologic advantage over other drugs in the class. In an interactive session attended by several hundred AIDS physicians, Judith Currier described the case of a 48-year-old man who smokes, has a family history of myocardial infarction, and wants to start antiretroviral therapy.

What would you add to a two-nucleoside backbone, Currier asked, efavirenz, nevirapine, indinavir/ritonavir, nelfinavir, amprenavir, or something else? Amprenavir proved the least popular option, at 2 percent, while efavirenz led the pack at 45 percent. Other attendees split their votes among nevirapine (18 percent), nelfinavir (16 percent), and indinavir/ritonavir (11 percent).

Return to Article Contents

A d4T requiem by CHORUS?

Sitting through several slide reports that tie d4T to lipoatrophy, say, or reading encyclopedic posters that impugn this NRTI (but no others) in multivariate analyses, can easily encourage notions that d4T's days are numbered. Then one watches the next slide presentation, or drifts down the poster aisle, and there are studies that nose out not one link between d4T and cratered faces or high lactates. As damning and redeeming data on this drug rain down from cell studies, mouse studies, cohort studies, prospective and retrospective inquisitions, deciding what to make of it still stumps the experts.

"It's really hard," sighed workshop co-chair David Cooper when asked to make sense of the d4T free-for-all. "Whether there's a [lipodystrophy] hierarchy among nucleosides--I don't think the jury's in," he said. "There have got to be prospective, randomized studies." Studies that split treatment-naive people between d4T and non-d4T regimens have begun. But while results accrue, the cells, mice, and cohorts will have to do.

A 4025-person analysis of the Glaxo-supported CHORUS Observational Cohort found marginally but consistently higher rates of lipodystrophy among people taking d4T regimens than among those taking similar AZT-based combos [abstract I-1286]. A proportional hazards analysis showed significantly higher risks of lipodystrophy with several d4T-containing combinations compared with AZT/3TC or AZT-based multidrug regimens. A.J. Scarsella (Pacific Oaks Medical Group, Los Angeles, California) presented this data-drenched poster and described how the study worked.

Clinicians at four US clinics (in Nashville, New York, San Francisco, and Los Angeles) tracked cohort members from when they signed up for the study to when those clinicians detected fat atrophy or hypertrophy. Then researchers counted the number of cases that had evolved in groups of at least 40 individuals taking specific regimens when they entered the study. Finally they calculated hazard ratios for lipodystrophy, using AZT/3TC or AZT-based combos as the reference. Median follow-up stretched to 26 months.

Signs of lipodystrophy never appeared among 57 people taking AZT/3TC/nevirapine, one of the 11 regimens taken by 40 or more people. In ascending order, lipodystrophy rates for the other 10 regimens (with the number taking that regimen in parentheses) were

• 2 percent for AZT/3TC (214)
  
• 3 percent for d4T/3TC (198)
  
• 4 percent for AZT/3TC/nelfinavir (186)
  
• 5 percent for didanosine (ddI)/d4T/nelfinavir (75)
  
• 6 percent for d4T/3TC/nelfinavir (267)
  
• 6 percent for d4T/3TC/nevirapine (95)
  
• 10 percent for AZT/3TC/indinavir (333)
  
• 14 percent for d4T/3TC/indinavir (312)
  
• 14 percent for AZT/3TC/ritonavir/saquinavir (44), and
  
• 25 percent for d4T/3TC/ritonavir/saquinavir (51).
  

A few things strike one right away. New diagnoses of lipodystrophy seem low in this cohort. Even people taking reportedly dangerous double-PI regimens have an incidence of only 14 percent with two PIs plus AZT/3TC and 25 percent with PIs plus d4T/3TC in medians of 20 and 18 months from study entry. As Morris Schambelan noted when opening the Lipodystrophy Workshop, lipodystrophy prevalence hovered between 40 and 50 percent in three other large, recent cohort analyses. Measuring incidence in CHORUS may explain the difference.

Still, overall incidence in this study--that is, including people taking regimens other than the 11 listed above--came to only 7 percent over a median 26 months. And at three of the CHORUS centers, the 26-month incidence was 2.5 percent or lower, while at the fourth center the incidence was 16.1 percent. By comparison, the 36-month rate of new lipodystrophy diagnoses in a 221-person German cohort measured 34 percent [abstracts P68 and I-1287, below]. Maybe the CHORUS clinicians are particularly conservative in diagnosing lipodystrophy. But if clinicians in the CHORUS center that diagnosed lipodystrophy in 16.1 percent also favored certain drug combinations significantly more than the other centers, the regimen comparison could be skewed. As Scarsella and colleagues pointed out, "caution must be used in interpreting these results as they may be explained by residual confounding."

If one puts the possibility of confounding aside for the moment, another neon lesson from the above ranking of regimens is that d4T plus a certain PI or NNRTI always yielded a higher rate of lipodystrophy than AZT and that same PI or NNRTI. For example, d4T/3TC/nelfinavir gave a 6 percent incidence, compared with 4 percent for AZT/3TC/nelfinavir; and d4T/3TC/ritonavir/saquinavir yielded a 25 percent incidence, compared with 14 percent for AZT/3TC/ritonavir/saquinavir. But only one such head-to-head comparison, d4T/3TC/indinavir versus AZT/3TC/indinavir, yielded a statistically significant difference (favoring the AZT regimen) in a hazards ratio analysis looking for lipodystrophy predictors (Table 6).

It's important to know that the regimens listed by Scarsella and colleagues are those people were taking when they agreed to enter the study; they may have changed to something else during follow-up. But statisticians tried to account for that possibility by figuring hazards ratios for lipodystrophy in two ways: A baseline (intent-to-treat) model locked the analysis to the first regimen. Then a time-dependent (as-treated) model used CD4⁺ count and viral load as covariates and censored people from analysis if they switched to some regimen other than those in the list above. The resulting risk analyses indicting d4T regimens remained statistically significant in both models (Table 6). Four separate regimen comparisons found d4T on the short end of the statistical stick, although two of them stacked the deck against d4T by comparing a d4T-plus-PI combo with AZT/3TC.

Still, this study lines up with other cohort analyses tying d4T to lipodystrophy. Also as other studies show, making a PI part of the regimen substantially escalates that risk. But as the authors observed, this cohort study, like all cohort studies, is open to confounders. There is no accounting for regimens people may have taken before a d4T combo. There is no accounting for potentially different prescribing patterns in centers with statistically significant differences in diagnosing lipodystrophy. And of course cohort studies only suggest associations; they don't prove causes. All those caveats would stand if lipodystrophy consistently arose more in people taking AZT rather than d4T. But that didn't happen. So readers will have to toss this one into the hopper and see how the data distill. But don't take too much time with that, because there are lots more data to harvest.

Return to Article Contents

More witnesses for the d4T prosecution

Let's move on to that German cohort study in which clinicians diagnosed lipodystrophy in 34 percent of 221 people, all of whom began antiretroviral therapy between July and September 1996. Stefan Mauss (Dusseldorf, Germany) presented the results at both Toronto meetings [abstracts P68 and I-1287]. His multivariate analysis linked four factors to lipodystrophy, and PI treatment wasn't one of them. Taking d4T for more than 12 months was.

Investigators determined lipodystrophy rates by having clinicians and study participants answer standardized questionnaires. They also used anthropometric measures, but anthropometrics proved relatively insensitive and nonspecific. Most people, 69 percent, began treatment with two NRTIs; 90 percent or more took AZT and 3TC at some point, compared with 62 percent who ever took d4T. And they generally took AZT and 3TC longer than d4T, with 86 percent on 3TC more than 12 months, 63 percent on AZT more than 12 months, and 44 percent on d4T more than 12 months. Nearly three quarters took a PI at some point, and 62 percent took a PI longer than 12 months. Statisticians sifted the data three years after people began antiretrovirals.

Taking d4T for more than 12 months increased the odds of a lipodystrophy diagnosis 2.1 times (P = 0.02); a CD4⁺ nadir below 200 cells/mm³ upped the odds 2.2 times (P = 0.03); and hypertriglyceridemia inflated odds 2.3 times (P = 0.001). NNRTI therapy for more than 12 months proved protective, slicing the lipodystrophy risk 80 percent (P = 0.03).

This study is one of very few that do not brand PI use or older age (over 36 years here) as lipodystrophy risk factors, not even in univariate analyses. The authors do not speculate on these departures from other studies. Although 62 percent took a PI longer than 12 months, only 28 percent began with triple therapy. Perhaps the cumulative PI experience simply wasn't long enough for PIs to prove themselves risky. But the even briefer NNRTI experience (8 percent took NNRTIs for more than 12 months) readily landed that drug class in the statistically significant protective column. Only 8 percent of people with lipodystrophy had isolated lipohypertrophy (usually associated with PIs), but 74 percent had mixed fat atrophy and hypertrophy. So a preponderance of isolated lipoatrophy diagnoses (usually associated with NRTIs) cannot explain the absent link between PIs and overall lipodystrophy.

If one cared to dig for extenuating evidence favoring d4T, one might note that 3TC and AZT were both the most-used and the longest-used drugs in the study. But that argument doesn't quite work here, because lipodystrophy diagnoses in this analysis were not keyed to the regimen being taken at that moment. To put it another way, the analysis didn't yoke lipodystrophy to current drugs, but rather to how long people took any drug. As Mauss and colleagues argued, "confounding by difference in duration of antiretroviral treatment is excluded by the nature of the cohort." In other words everyone started antiretrovirals at the same time, and taking 3TC or AZT for a long time didn't correlate with lipodystrophy. More than 12 months of d4T did. Add this one to the hopper.

Estaban Martínez and colleagues (Hospital Clinic Universitari, Barcelona, Spain) counted lipodystrophy diagnoses in 494 consecutive individuals beginning treatment with a PI and two nucleosides between October 1996 and September 1999 [abstract 14]. This study broke down diagnoses into subcutaneous lipoatrophy, central obesity, or any lipodystrophy. In a multivariate analysis of the "any" cadre, female gender increased the risk of lipodystrophy 1.9 times (P = 0.03), and every additional six months of antiretrovirals raised the risk 1.4 times (P = 0.04). But no specific drug made "any" lipodystrophy more likely.

When Martínez turned to lipoatrophy, duration of d4T treatment nudged the risk up 1.2 times, but that increase proved significant in a multivariate analysis (P = 0.02). About half of his clinic's patients began treatment with a d4T regimen. So this study reflects the d4T duration finding of the German study just reviewed. But the Spanish analysis suffers somewhat, a workshop attendee argued, because the diagnoses depended entirely on the clinical impressions of the physicians and not on a standardized survey or any objective measures. And the timing of these diagnoses seemed vague to some audience members. Although the study's start date was October 1996, presumably Martínez and colleagues didn't diagnose lipodystrophy before early to mid-1998, that is, before Andrew Carr and his Sydney colleagues first drew wide attention to what had been an underappreciated syndrome.

Matthew Law (University of New South Wales, Sydney, Australia) looked at lipodystrophy in people who had taken part in two randomized controlled trials of nucleoside combinations plus indinavir or nevirapine, Ozcombo 1 and Ozcombo 2 [abstract 28]. The NRTI duos were AZT/3TC, d4T/3TC, and ddI/d4T. The analysis is somewhat unbalanced, Law noted, because only 47 percent of the original Ozcombians volunteered for the lipodystrophy study, and 75 percent of them came from one of the d4T arms. Mean follow-up ranged from 26 to 27 months for Ozcombo 1 treatment groups and from 12 to 14 months for groups enrolled in the more recent Ozcombo 2.

People from the d4T/3TC groups (P = 0.001) and the ddI/d4T groups (P = 0.007) had significantly more peripheral lipoatrophy than people in the AZT/3TC groups. Total DEXA-measured body fat was significantly lower in the ddI/d4T arms than in the AZT/3TC arms (P = 0.05). Serum lactate levels were also highest among people taking ddI/d4T, but not significantly so. Nor did the NRTI groups differ in overall visceral adipose tissue, subcutaneous adipose tissue, skinfold thickness, or glycemic index.

Law noted in his abstract that the results "need careful interpretation" because of the disparate retention of patients originally randomized to the three NRTI combos and because of the low total carryover of study participants from the Ozcombo trials. But Law's caveats didn't stop attendees from trying to poke more holes in the study.

Chelsea and Westminster's Graeme Moyle observed that lactate levels, though higher among d4T takers, did not exceed the normal range. Moyle also worried that some people assigned to AZT/3TC in Ozcombo 1 or 2 may have declined to volunteer for this study because they didn't feel well, perhaps because of lipodystrophy. Ian Weller (University College London Medical School) wondered about the timing of Law's analysis. Because the well-versed Australian clinicians who made the diagnoses were probably aware that d4T had been linked to fat atrophy, their unblinded evaluation of patients could be biased.

But the prosecution did not rest its circumstantial case against d4T there. A small, single-center study by R. Polo (Hospital Carlos III, Madrid, Spain) found significantly higher rates of lipoatrophy among people taking a d4T regimen (11 of 21, 52.4 percent) than among those taking AZT (1 of 10, P = 0.046) [abstract I-1282]. Polo calculated the relative risk of lipodystrophy as 5.24 with d4T compared with AZT (P = 0.046). Triglycerides among d4T takers topped those in the AZT group (P = 0.026), but total cholesterol and glucose levels were equivalent.

Polo and colleagues verified lipodystrophy via physical examination, anthropometric measures, bioelectric impedance analysis, CT, photos, and patient questionnaire. One weakness of the analysis is that it compares two groups on the basis of their current regimen. As it turns out, people in the d4T group had taken their current regimen for an average nine months, compared with 26 months in the AZT group (P = 0.002). Since the duration of total antiretroviral treatment in the two groups was equivalent (43 months for the d4T group and 48.5 months for the AZT group), it is plausible that lipodystrophy evolved in the d4T group while they were taking other nucleosides.

A 14-center cross-sectional study, also presented by Polo, compared 279 patients with lipodystrophy and 293 without to ferret out possible risk factors [abstract I-1281]. This time Polo reported significantly higher rates of overall lipodystrophy in people now taking d4T versus AZT (54 percent versus 39.9 percent, P = 0.004). But fat hypertrophy proved significantly more common in the AZT group than in the d4T group (27 percent versus 11.8 percent, P = 0.002). The overall lipodystrophy rate was higher among people taking NRTIs plus a PI than among those taking NRTIs alone (58.8 percent versus 37.5 percent, P < 0.001). On the other hand, lipoatrophy was more common in those taking NRTIs alone than in those taking NRTIs plus PIs (45.6 percent versus 24.3 percent, P < 0.001). Those last findings confirm the impression that NRTIs spur atrophy and PIs hypertrophy.

A.J. White (Glaxo Wellcome) and researchers from the Centre for Clinical Immunology and Biomedical Statistics in Perth, Australia, offered a meta-analysis of 15 studies rating the relative risk of lipodystrophy with d4T versus AZT and another meta-analysis of four studies estimating the risk of elevated blood lactates with d4T versus AZT [abstract P82]. Fourteen of the 15 lipodystrophy studies found an increased risk with d4T, and all four lactate studies chalked up higher risks with d4T.

White and colleagues concluded that "d4T therapy is a common risk factor for both lipodystrophy and hyperlactatemia." And, they suggested, "raised plasma lactate levels may be useful for predicting the development of subcutaneous fat wasting." Of course the meta-analyzed studies are all cohort studies, with their already noted limitations. And White's survey leaves out at least one cohort study that found no link between lipodystrophy and d4T.¹¹

Findings on d4T and lipoatrophy presented before the Toronto meetings convinced one physician to substitute another NRTI for d4T in 11 people with facial fat wasting [abstract P85]. Andrew Clark, a clinician in Cape Town, South Africa, recorded an overall average gain of 5.4 kg after the switch and improved facial fat in 10 of 11 people. Facial wasting was assessed by the clinician, a second observer, and the patient and was recorded by photography. No one had a viral load rebound after abandoning d4T.

Of course this is a tiny and uncontrolled study by a clinician who didn't have money to spend on fancy fat tests. But the results lend weight to earlier nonrandomized studies by Polo¹² and Thierry Saint-Marc,¹³ who reported reversal of lipoatrophy after people traded d4T for another nucleoside.

Evidence from switch studies like these has swayed William Powderly (Washington University, St. Louis) to conclude that d4T, and maybe ddI, figure tangibly in fat loss. Summarizing switch studies at the Lipodystrophy Workshop, he proposed that the "data are consistent with a role of d4T and possibly ddI in the evolution of lipodystrophy syndrome and lactic acidosis" [abstract 27]. "Switch studies have the potential to provide important insights into the pathogenesis of the metabolic syndromes," Powderly wrote in his presentation abstract. "However, there is a need for larger, controlled studies and a more standardized approach to definition of metabolic abnormalities."

Return to Article Contents

Indinavir, d4T, and nondrug factors

That's the evidence against d4T, some of it strong, none of it conclusive. A case can also be made in this nucleoside's defense, starting with HIV Outpatient Study (HOPS) data presented earlier this year and reviewed again by Kenneth Lichtenstein (University of Colorado, Denver) at the Lipodystrophy Workshop.

This survey of 1077 people with HIV infection at eight US centers looked for links between lipodystrophy and NRTIs (d4T or 3TC), the PI indinavir, and a host of nondrug factors: age older than 40 years, HIV infection for seven or more years, AIDS for two or more years, nadir CD4⁺ count below 100 cells/mm³ or CD4⁺ percentage below 15 percent, three or more years since nadir CD4⁺ cell count, body mass index loss of 1.0 kg/m² or more, body mass index change of 2.0 mg/m² or more, and hemophilia. No matter how long people took d4T, or indinavir for that matter, they never had signs of fat maldistribution if they had no nondrug risk factors. But percentages of cohort members with misplaced fat climbed inexorably as the number of nondrug risk factors rose.

The HOPS analysis made it clear that cumulative drug experience--when compounded by nondrug risk factors--drives up rates of lipodystrophy. So the drugs themselves play some part in this many-layered drama. But the HOPS study distances individual drugs from the word cause and should give pause to anyone bent on ready answers. "Studies until now furiously try to answer questions" about lipodystrophy, Lichtenstein said. "We furiously try to ask them."

Lichtenstein devoted the main part of his workshop talk to a June-to-September 2000 update of the lipodystrophy survey involving 307 of the first 1077 people studied [abstract 13]. Fat atrophy, the HOPS team learned, has become more common in this population, its prevalence bounding from 17 to 33 percent in only 20 months. The prevalence of fat accumulation, on the other hand, stayed nearly flat, inching from 10 to 13 percent over the same span. Fat accumulation improved in 40 percent of those affected, whereas atrophy improved in only 17 percent.

A workshop attendee observed that these trends dovetail with a model proposed by French investigators, who describe lipodystrophy as fat accumulation, fat atrophy, or a mixed syndrome. Observation of the French cohort suggests to the investigators that the mixed syndrome represents a transition from accumulation to atrophy.

At this point such concepts, and even the HOPS trends, must be seen as broad strokes in the overall lipodystrophy picture. HOPS does not factor in whether people are switching from PIs to NNRTIs, for example, from d4T to abacavir, or from AZT to d4T. And HOPS does not account for worsening adherence or structured (or unstructured) treatment interruptions in attempts to counter ugly fat changes. But one constant in this evolving syndrome seems certain: Almost everyone taking antiretrovirals takes nucleoside analogs and continues to take them as other parts of the regimen change. That could explain the steady trend to more lipoatrophy. It's hard to single out one of six nucleosides as the prime mover in this multiyear fat loss epidemic, although the cohort studies reviewed in the preceding section point in that direction.

To complicate things a little more, two studies of Asian populations found high rates of lipoatrophy and lipohypertrophy among people with HIV infection taking no antiretrovirals at all. Nick Paton and co-workers (Tan Tock Seng Hospital, Singapore) used structured questionnaires, body and skinfold measures, and bioelectric impedance analysis to measure fat atrophy and accumulation in 319 people, 85 percent of them men and 82 percent of Chinese ancestry [abstract 15].

The largest group, accounting for 36 percent in the study, was taking a PI regimen; but nearly as many, 34 percent, were taking only NRTIs with or without an NNRTI; and 30 percent had never tried an antiretroviral. Rates of lipoatrophy proved almost identical across the three groups: 43 percent for those on PIs, 39 percent for those on NRTIs, and 39 percent for the naive group. But rates of fat accumulation were significantly higher in the PI group, 53 percent, compared with 34 percent in the NRTI group and 25 percent among treatment-naive people (P < 0.001).

Another cross-sectional survey, this one in Japan, rated body shape changes by physical exam in 192 people, most of them men and nearly all Japanese [abstract 17]. The group included 17 percent who had never taken an antiretroviral and 20 percent naive to PIs. Helen Fraser (International Medical Center of Japan, Tokyo) reported a high prevalence of fat accumulation or atrophy in this population, 64 percent. Of those with lipodystrophy, 75 percent were taking PIs, 14 percent were taking non-PI combos, and 11 percent were taking no antiretrovirals. In a multivariate analysis, PI treatment for 26 months or more raised the risk of lipodystrophy 4.73 times. NRTI therapy for 54 months or more hiked the risk only 1.19 times, a nonsignificant increase.

Neither Paton nor Fraser had a ready explanation for lipodystrophy in the antiretroviral-naive people. One might speculate that Fraser's diagnostic criteria were loose, but Paton's were tighter and he found even more lipodystrophy among the naive. Paton also excluded people with wasting or opportunistic infections in the past six months. Perhaps the typically slender Asian body type favors a diagnosis of atrophy, especially in people with an untreated retroviral infection. Whatever the explanation, Fraser's failure to correlate more than 4.5 years of NRTI treatment with lipodystrophy suggests that nucleosides sometimes contribute little to the syndrome, at least in certain populations.

Return to Article Contents

Lipo toss-ups with first-line d4T versus AZT

One way to pin down how much a particular NRTI drives fat or lipid changes is to restrict the analysis to people taking their first antiretroviral regimen. That's the tack taken by Graeme Moyle in a recently published study and by Johannes Bogner in a report at the Lipodystrophy Workshop.

Moyle studied 135 people seen at London's Chelsea and Westminster Hospital who began treatment with two NRTIs and an NNRTI.¹⁴ Most of them, 97, started with a d4T-based combo, and 58 used ddI as their second nucleoside. After an average 132 days on their first-line regimens, Moyle found no differences in cholesterol or triglyceride levels between the d4T group and the AZT group in univariate or multivariate analyses. In the multivariate model only older age (P = 0.004) and triglyceride level (P = 0.015) predicted hypercholesterolemia. And high cholesterol was the only variable that predicted high triglycerides (P = 0.003). Neither nevirapine nor efavirenz made high cholesterol or triglycerides more likely.

This study was too brief to detect any body shape changes between the d4T group and the AZT group. But Moyle argued that his findings cast a shadow over NRTI-lipodystrophy links in cohort studies. "Given the apparent association between lipid handling abnormalities and fat redistribution syndrome," he wrote, the failure to tie d4T, AZT, or either NNRTI to lipid abnormalities "suggests that differences between drugs reported in cohort studies may be best explained by confounding biases."¹⁴ Some would contest that argument by noting that lipid and body fat abnormalities may arise separately in people with HIV infection.

The study by Bogner (University of Munich, Germany) confirmed Moyle's finding that d4T and AZT have a similar impact on cholesterol and triglycerides [abstract P54]. Unlike Moyle, Bogner tracked substantial increases in those measures, and in glucose as well, because the 39 people taking d4T and the 76 taking AZT were also taking a PI, and because the follow-up stretched to an average 23.8 months.

That longer follow-up also allowed Bogner to see whether d4T or AZT contributed more often to lipodystrophy, because none of the 115 study participants changed the NRTI backbone of their regimen. Neither drug could be linked to more lipoatrophy (33.3 percent for d4T, 39.5 percent for AZT), lipohypertrophy (30.8 percent for d4T, 27.6 percent for AZT), or any lipodystrophy (46.2 percent for d4T, 50 percent for AZT). All percentage differences were nonsignificant.

In a multivariate analysis Bogner correlated five factors with lipoatrophy: Caucasian race (odds ratio [OR] 5.5, P = 0.41), baseline viral load above 100,000 copies/mL (OR 5.2, P = 0.007), HAART for more than 104 weeks (OR 4.4, P = 0.002), age older than 40 years when starting HAART (OR 3.2, P = 0.016), and baseline cholesterol above 200 mg/dL (OR 0.36, P = 0.047). Two variables correlated with lipohypertrophy, PI versus non-PI therapy (OR = 3.83, P = 0.014) and HAART for more than 104 weeks (OR 2.69, P = 0.028).

The final bits of data to mull come from two cohort studies--one with over 2000 people--that departed from the cohort surveys reviewed above in their conclusions on d4T's role in lipodystrophy. Massimo Galli (University of Milan, Italy) spelled out the most recent findings in the 2258-person Lipodystrophy Italian Multicenter Study [abstract I-1284]. A multivariate analysis including d4T and factors other than antiretroviral drug classes did not link this nucleoside to body shape changes (OR 1.073, P = 0.5). When statisticians put antiretroviral drug classes into the equation, d4T remained a nonsignificant risk factor (OR 1.026, P = 0.82).

On the other hand Galli did tie indinavir to fat changes in both multivariate models (OR 1.382, P = 0.002 without accounting for drug class; OR 1.316, P = 0.01 with drug classes factored in). Variables fettered to any fat change were longer duration of antiretroviral treatment (OR 1.007), older age (OR 1.023), female gender (OR 2.125), four versus three versus two versus no antiretrovirals (OR 1.579), undetectable versus detectable viral load (OR 0.815), and treatment with two NRTIs plus one PI (OR 2.1) (P = 0.001 for all). Being antiretroviral naive slashed the risk of body fat changes 80 percent (P = 0.002).

Finally, Pere Domingo (Hospital de la Santa Creu i Sant Pau, Barcelona, Spain) sorted out the impact of AZT (n = 23) and d4T (n = 75) on fat distribution and metabolic numbers in people also taking 3TC as part of HAART [abstract I-1289]. More people in the d4T group (53 percent) than in the AZT group (30 percent) had AIDS, and more taking d4T (29 percent) than taking AZT (0) were naive when they started their antiretrovirals. The AZT group had taken ddI significantly longer than those in the d4T group (11.6 versus 4.9 months, P = 0.002).

The groups did not differ significantly in CD4⁺ count, viral load, percent with undetectable viral load, caloric intake, total time on NRTIs (94.5 months for d4T, 107.3 months for AZT), NRTI use in HAART (61.8 months for d4T, 62.1 months for AZT), duration of HIV infection, physical activity, total cholesterol, triglycerides, LDL, HDL, insulin, testosterone, glucose/insulin ratio, or rates of hypercholesterolemia, hypertriglyceridemia, or diabetes mellitus.

In a bivariate analysis, the sole factor predicting lipodystrophy was a sedentary job (in 80.3 percent with lipodystrophy versus 54 percent without, P = 0.01). Two factors correlated with lipoatrophy--less physical activity (P = 0.01) and packing in more calories per day (P = 0.04). Time taking any NRTI did not correlate with lipoatrophy. One factor predicted lipohypertrophy, older age (P = 0.01), while the correlation with less physical activity was nearly significant (P = 0.06).

The d4T jury may now retire to deliberate.

Return to Article Contents

The Fix

How do you fix lipodystrophy? And how do you fix those metabolic perturbations? So far the metabolic part of the equation has been easier to solve, a trend supporting the proposed disjunction between metabolic and body shape changes. The cleanest fixes--diet and exercise--remain the least studied. The easiest fixes--changing or stopping certain antiretrovirals--continue to be tested in cohorts large and small, in studies formal and perfunctory. A third option--adding more drugs on top of the antiretrovirals--has gained currency. Early results of small drugs-for-drug-side-effects studies have begun to appear, and bigger trials are signing people up.

One recent exercise-only study came from Simon Jones and colleagues (Liverpool University, UK) [abstract P97]. They compared six people with lipodystrophy who underwent a 10-week course of aerobic and resistance exercise and six people with lipodystrophy who didn't exercise. The exercise began with 20 minutes of cycling at 70 percent of heart rate, followed by an hour of upper and lower body resistance training. No one took lipid lowerers during the study.

Exercise did lots of good things. Total cholesterol fell an average 19.8 percent in the exercise group (P < 0.001 versus baseline), and triglycerides slid 25 percent

(P < 0.05). Body mass rose an average 3.6 kg among the exercisers (P = 0.03), and percentage of body fat fell from 21.9 to 18.25 percent (P = 0.01). Waist-to-hip ratio tightened from 1.07 to 1.02 (P = 0.045), arm circumference rose from 23.6 to 25.8 cm (P = 0.038), and leg circumference grew from 44.05 to 47.6 cm (P = 0.0024). Sedentary controls enjoyed no such gains. HDL, LDL, and body mass index did not change substantially in either group. This small study adds to the still tiny but consistent heap of data showing that structured exercise can rapidly reverses metabolic and fat abnormalities at virtually no risk to the exerciser.¹⁵

Return to Article Contents

Small studies of facial implants and L-carnitine

Patrick Amard (ARME, Paris) and colleagues erased some signs of lipoatrophy in the faces of 33 men with an implant they called "biocompatible, bioabsorbable, and immunologically inactive" [abstract P94]. Polylactic acid (New-Fill) is unlike biodegradable implants such as collagen and hyaluronic acid, or nonbiodegradable implants such as silicone. It works, Amard reported, by spurring growth of a fibrous dermal layer in response to the injected implant.

The 33 study participants had taken antiretrovirals for a median of 64.6 months. They received an average of four injections of polylactic acid. Mean facial fat thickness, measured by ultrasonography, increased 0.014 mm (3.11 percent) after 24 weeks of treatment, while mean dermal thickness increased 5.3 mm (153.9 percent). "All patients showed major improvement in their condition," Amard reported. At least the three before-and-after photos on the poster looked convincing.

A request to the manufacturer for further information about potential side effects, cost, and availability of the implant has gone unanswered. Biotech Industry SA in Luxembourg makes polylactic acid and offers some details about the product at www.new-fill.com. It has been used for years in reconstructive surgery and passed muster in toxicology trials, according to Biotech Industry. "There is absolutely no risk of allergic reaction," the company Web site says, and "no preliminary skin test is required since the polymer is not of animal origin." Two larger trials are under way in Europe, and the product will soon be introduced in the UK.

Dusseldorf clinician Stefan Mauss gave 1000 mg of L-carnitine twice daily to 12 men with lipodystrophy to see if this agent lives up to its reputation as a fat-change remedy [abstract I-1291]. Twelve weeks of treatment did significantly lower total cholesterol (from a median of 250 to 219 mg/dL, P < 0.05) and LDL cholesterol (from 177 to 132 mg/dL, P < 0.05). But L-carnitine did nothing for fat abnormalities when measured separately by clinicians and study participants on a visual analog scale, by various anthropometrics, or by bioelectric impedance analysis. HDL and triglyceride levels stayed flat.

Return to Article Contents

Who will benefit from metformin?

Recent reports on metformin, an insulin-sensitizing agent, have been more encouraging than the L-carnitine study. Speaking at ICAAC [abstract I-1373], Harvard's Steven Grinspoon noted that "in the endocrine world, [metformin] is emerging as the drug of choice in insulin-resistant states." His randomized, double-blind, placebo-controlled trial of the drug in 26 people with HIV lipodystrophy documented significant drops in mean insulin area under the curve (P = 0.01), weight (P = 0.005), and diastolic blood pressure (P = 0.009) with metformin versus placebo.¹⁶ People taking metformin also shed visceral fat (P = 0.08).

Grinspoon used 500 mg of metformin twice daily, a dose substantially lower than the 850 mg three times daily used in an open-label French study.¹⁷ The smaller dose could account for the low rate of side effects and absence of elevated lactates, which may be more common at higher doses and could compound NRTI-induced lactate jumps.

But metformin is not for everyone with lipodystrophy. At the Lipodystrophy Workshop, an attendee observed that metformin has been linked to lipoatrophy, and Grinspoon concurred that the drug may not be appropriate for people primarily with lipoatrophy. Grinspoon and colleagues restricted the published study to people with impaired glucose tolerance on oral glucose tolerance testing and/or hyperinsulinemia.¹⁶ They barred people with high serum creatinine, low hemoglobin, or a history of liver dysfunction, renal failure, diabetes mellitus, or substance abuse. No one could be taking insulin, antidiabetics, glucocorticoids, testosterone, megestrol acetate, growth hormone, estrogen, or anabolic steroids. These researchers stressed that their results "cannot be extrapolated" to everyone with HIV infection and lipodystrophy.

Metformin also trims tallies of tissue-type plasminogen activator (tPA) antigen and PAI-1, which forebode heart disease, reported Grinspoon's colleague Colleen Hadigan (Massachusetts General Hospital, Boston) [abstract 22]. Her three-month placebo-controlled trial enrolled 25 people with lipodystrophy and insulin resistance or impaired glucose tolerance. Levels of insulin and the two heart disease markers were equivalent at baseline in the 11 people who took placebo and the 15 who took 500 mg of metformin twice daily. PAI-1 dropped 16 percent in the metformin group (P = 0.03), and tPA antigen fell 11 percent (P = 0.02). Levels of both markers rose slightly in the placebo group. Increased tPA antigen correlated highly with fasting insulin (r = 0.3, P = 0.004) and 120-minute insulin (r = 0.35, P = 0.001) on the oral glucose tolerance test.

Hadigan's case-control study involving 86 people with lipodystrophy and 258 age- and weight-matched controls from the Framingham Offspring Cohort confirmed significantly higher levels of insulin, PAI-1, and tPA antigen in the lipodystrophy group (P < 0.0001). Glucose levels and waist-to-hip ratios were equivalent in the two groups.

In his ICAAC lecture Grinspoon again warned clinicians against prescribing metformin for anyone with lipodystrophy. Results so far, even though from placebo-controlled trials, are preliminary. But, he said, if further studies confirm metformin's effects on insulin resistance, markers of cardiovascular disease, and body fat changes, measuring fasting insulin would make sense to determine when to give metformin.

Return to Article Contents

Growth hormone doses go lower and lower

Viscerally obese people have paltry levels of growth hormone, but no one knew whether the same applies to people with HIV lipodystrophy and central adiposity. To remedy that oversight, Harvard's Steven Grinspoon cajoled 61 people into staying up all night so he could measure their growth hormone every 20 minutes from 8 PM to 8 AM [abstract 11]. This nocturnal cohort included 21 men with HIV lipodystrophy, 20 HIV-infected men without lipodystrophy, and 20 healthy volunteers--all matched for age and body mass index.

Growth hormone concentrations did prove to be significantly lower in the lipodystrophy group when compared with either control group (P < 0.05). And in a multivariate analysis that controlled for age, body mass index, body fat, and visceral fat, only visceral fat predicted growth hormone concentrations (r² = 0.4, P = 0.0036).

Grinspoon noted that his study says nothing about the physiologic relevance of these findings. "This is just an interesting bit of physiology that needs to be followed up on," he cautioned. His results mean only that "some percentage" of people with lipodystrophy may be "relatively or completely growth hormone deficient."

The Harvard endocrinologist maintained this prudent tone in his ICAAC lecture on insulin resistance and lipodystrophy [abstract I-1373]. Growth hormone can be a "double-edged sword" in people with lipodystrophy, he maintained. It can at first worsen insulin sensitivity and glucose tolerance and so cause hyperglycemia. "Using the same dose [of growth hormone] used for HIV wasting is wrong," he added.

Joan Lo (University of California, San Francisco) agrees. When she gave recombinant growth hormone to seven men with HIV infection and fat accumulation, she used only 3 mg daily for six months [abstract 12]. Even at that dose, half the dose typically given for wasting, one person had to drop out after three weeks because of hyperglycemia. An oral glucose tolerance test confirmed diabetes despite normal fasting glucose. Lesson--screen people with the oral test before starting growth hormone. Another person moved out of town and left the study, so only five completed the six-month course. One of them had to cut the 3-mg dose in half because of arthralgias.

Buffalo humps shrank, as did abdominal girth. Average lean body mass increased by 5.4 kg, total fat fell by 4.4 kg, and trunk fat dropped by 3.7 kg. Average visceral adipose tissue decreased from 17,166 mm² to 10,863 mm² (P = 0.08). Subcutaneous adipose tissue did not wane. Insulin sensitivity fell in the first month, and fasting glucose rose. But both came back toward baseline levels as the study continued. Lo proposed that even lower doses of growth hormone should be tested for HIV-related fat accumulation.

Ellen Engelson (St. Luke's-Roosevelt Hospital, New York) is doing just that in people with visceral adiposity and HIV infection [abstract P95]. Her work confirms Lo's finding that visceral fat drops in 24 weeks while subcutaneous fat does not. Engelson gave a slightly lower dose than Lo, 4 mg every other day, to 14 people who had participated in an earlier six-month trial of 6 mg daily. Average trunk fat dwindled significantly among eight people who completed 24 weeks of treatment (P < 0.01), while arm and leg fat remained stable. The drop in visceral adipose tissue was statistically significant (P < 0.02).

This study also highlighted the risks inherent in growth hormone therapy. Mean fasting glucose rose significantly from 86 to 103 mg/dL (P < 0.02), but average HDL rose from 38.4 to 46.2 mg/dL (P < 0.002). Two people dropped out because of "self-perceived lack of efficacy." Two others stopped taking growth hormone because of abnormal liver function tests, and one because of severe pain. But all three resumed treatment.

A large chart review followed by prospective follow-up suggested that taking anabolic agents for wasting or hypogonadism will not ward off fat redistribution calculated as a DEXA-derived ratio of trunk fat to total extremity fat [abstract P69]. G. McComsey (Case Western Reserve University, Cleveland) and colleagues studied 28 people taking oxandrolone, 58 taking testosterone, 23 taking oxandrolone plus testosterone, and 37 taking growth hormone. In all four groups the trunk-to-extremity-fat ratio indicated fat redistribution in 30 to 40 percent, rates similar to those in larger cohorts not taking anabolic agents.

"While this study was not designed to compare effectiveness of [anabolic] treatments for" fat redistribution syndrome, the authors concluded, "one would expect to see at least some difference if one agent clearly played a role in prevention of this syndrome."

Return to Article Contents

Lessons from PI switch studies

While metformin and growth hormone findings are too preliminary to guide clinical decisions, results of antiretroviral switch studies already looked conclusive when this millennium was only weeks old. After the Jan. 30-Feb. 2 Conference on Retroviruses, this journal reviewed 24 studies in which clinicians replaced a protease inhibitor with a nucleoside or nonnucleoside reverse transcriptase (RT) inhibitor.¹⁸ The results were consistent:

• RT inhibitor regimens usually preserved virologic suppression secured by PIs, especially if the PI regimen was first-line therapy.
  
• The new RT-only regimens generally improved lipid measures, though switching to efavirenz often did not.
  
• Stopping PIs rarely improved fat atrophy or hypertrophy within six to 12 months of the switch.
  

In an update on switch studies at the Lipodystrophy Workshop and at ICAAC, William Powderly (Washington University, St. Louis) added these insightful refinements [abstracts 27 and I-1375]:

• Most switch studies involving fat redistribution are poorly performed or inadequately controlled.
  
• PIs may have a causal role in insulin resistance, which can almost always be reversed by trading a PI for a different antiretroviral.
  
• PIs probably affect triglyceride metabolism in some patients, but their effect on cholesterol metabolism is less clear. They may contribute to decreased HDL in some individuals.
  
• Current switch studies do not support a role for PIs in maintaining fat gains or loss.
  
• Because the randomized multisite Australian study [abstract P84, Table 8] found that lipoatrophy worsened in both the continued PI group and the non-PI switch group, Powderly wondered about the wisdom of switching people from PIs if they primarily have lipoatrophy.
  
• The NRTI d4T may contribute to lipoatrophy, because the condition has resolved when another nucleoside replaces d4T (see The Drugs above).
  

Powderly prefaced those conclusions with cautious words about interpreting what happens--or what doesn't happen--when an RT inhibitor replaces a PI (Table 7). And although some workshop attendees disagreed, he argued that the world does not need more cohort studies looking to assign blame to this drug or that. Instead researchers should plan more randomized controlled trials with well-characterized baseline abnormalities, including studies of NRTI-sparing regimens.

If one can judge from the number of cohort studies already reviewed in this article, Powderly's voice may be one crying in the wilderness. And the two Toronto meetings certainly did not lack for switch studies. The Lipodystrophy Workshop and ICAAC featured at least 14 studies in which nevirapine, efavirenz, or abacavir ousted a PI (Tables 8 and 9). Do they say anything new?

No epiphany-caliber news, that's for sure. But the relentless accretion of data confirms some trends and infers others. The Australian randomized switch trial [abstract P84, Table 8] turns out not to be the only study demonstrating worsening lipoatrophy with a new non-PI regimen. Three small studies also cataloged progressive fat atrophy, usually peripheral, after trading a PI for efavirenz plus abacavir [abstract I-1531, Table 8], for efavirenz alone [abstract I-1532, Table 8], or for nevirapine [abstracts P92 and P93, Table 8]. One of these teams, headed by Heribert Knechten (Aachen, Germany) showed worsening lipoatrophy in nine of 45 people who switched to efavirenz but kept taking d4T and 3TC. But among 19 people who started efavirenz while keeping AZT and 3TC, Knechten perceived no worsening lipoatrophy.

A biopsy study by Pere Domingo (Hospital de la Santa Creu i Sant Pau, Barcelona, Spain) tracked rates of fat cell apoptosis in three people who had biopsies while taking a PI and again after swapping the PI for nevirapine [abstract I-1584]. After an average postswitch follow-up of 8.3 months, cholesterol and triglyceride readings dropped, and waist-to-hip ratios improved in all three study participants. But the follow-up biopsies of lipoatrophic sites showed that adipocytes kept dying. All three of these people were taking d4T, two of them with ddI. The findings, according to Domingo, suggest that a simple switch to nevirapine "will be useless for reversal of lipoatrophy."

The PI-NNRTI switch studies confirmed that nevirapine reliably rights topsy-turvy lipid measures, whereas efavirenz often fails in this chore. A randomized Spanish study in which 26 people continued their PI regimen, 26 turned to nevirapine, and 25 started efavirenz made this point most clearly [abstract I-473, Table 8]. After six months of follow-up, cholesterol plummeted in the nevirapine arm (P = 0.023 compared with the other groups), but dropped a mere 1 mg/dL with efavirenz. Cholesterol, triglycerides, and insulin sensitivity (when measured) improved in all four studies that dumped a PI in favor of abacavir (Table 9). But when German clinicians exchanged a PI for abacavir and efavirenz, metabolic abnormalities remained stable or worsened [abstract I-1531, Table 8].

Most studies, regardless of the switch regimen, found that disfiguring emblems of lipodystrophy lingered for six months to more than a year after PI treatment stopped. One notable exception was a 48-week substudy of an international trial that randomized people to continue their PI or trade it for abacavir [abstract P89, Table 9]. Willy Rozenbaum (Rothschild Hospital, Paris) reported that 20 of 26 clinical signs recorded at the switch resolved with abacavir, a surprising feat given results of other studies. This happy reversal of fortune may at least partly reflect the clinical scoring system these clinicians used; among people who continued their PI, 12 of 29 signs of lipodystrophy also resolved.

A more important question emerged from abacavir switch studies: Just how well does this NRTI keep HIV in check? A 163-person trial that randomized people to continue a PI or switch to abacavir plus Combivir (ZDV/3TC) charted 13 rebounds in the abacavir arm (15 percent) versus five in the PI arm (6 percent) after a median of 68 weeks [abstract I-476, Table 9]. No one had the AZT-associated codon 215 mutation at baseline, at least not in their viral RNA. But it's hard to spot mutations in people, like these, with viral loads below 50 copies/mL. Later analysis of baseline viral DNA in PBMCs did turn up stored AZT and 3TC mutations, which may have re-emerged during treatment with abacavir and Combivir.

Milos Opravil (University Hospital, Zurich, Switzerland) reported that a history of AZT therapy inflated the odds of virologic failure five times in the abacavir arm of the trial. He suggested that abacavir maintenance regimens may be vulnerable in some people with AZT experience and archived AZT mutations. But Jonathan Schapiro (National Hemophilia Center, Tel Hashomer, Israel) wondered whether that caveat should cover only people with AZT experience. D4T, Schapiro reminded colleagues, can evoke these same "AZT mutations" in people naive to AZT. On reflection, Opravil proposed that abacavir maintenance may be best for people who begin treatment with a PI regimen and remain aviremic for at least six months.

That describes a 205-person cohort in another PI-to-abacavir switch study [abstract I-477, Table 9]. These people had viral loads below 50 copies/mL and had never suffered a virologic breakthrough. After 48 week investigators logged only four rebounds above 400 copies in the abacavir arm (4 percent) and two in the continued-PI group (2 percent).

Earlier in the same session, Schapiro raised an even more interesting point about the lessons of NNRTI switch studies. Dozens of such inquests have redundantly drummed home the message that switching to a nonnuke is virologically safe when a PI combo has kept HIV under wraps for six months or more. Maybe, Schapiro suggested, the PI-to-NNRTI handoff should come sooner rather than later, before PI side effects mature toward potential irreversibility.

Perhaps an idea worth pursuing, as long as one remembers that NRTIs alone have atrophic side effects and that NNRTIs tote their own bag of toxic tricks. Phillipe Clevenbergh (Nice University Hospital, France) taught that lesson in a poster summarizing his group's experience with 181 people who started taking nevirapine or efavirenz as first-line therapy, or after intolerance to a PI, or as part of a salvage combo [abstract I-1549]. He reported that 81 members of this observational cohort (45 percent overall) had to stop their NNRTI. Reasons for stopping included virologic failure (48 percent of failures), neurologic side effects (18 percent), GI toxicity (11 percent), rash (8 percent), or other side effects (15 percent). These so-called adverse events outpaced those in the INCAS study of nevirapine or the DMP-006 study of efavirenz, and the neurologic snarls with efavirenz continued to take a toll longer than others have reported.

Those comparisons with randomized trials are rough because, with 56 percent in the Nice cohort starting an NNRTI in salvage, these people had more advanced HIV disease than those enrolled in INCAS or DMP-006. Still, the high failure rate for toxicities attributed to NNRTIs suggests that switching to nevirapine or efavirenz will not be a joy ride for people already burdened by PI side effects.

There's one sure way to shirk drug toxicities: stop the drugs. But even that solution--with all its attendant risks--offers less than a perfect solution to side effects. As noted time and again, some side effects may prove a life-long albatross; others may take more than a year to disappear, and then may do so only among avid exercisers, strict dieters, or people who accept the added risk of other pharmaceuticals.

Researchers at the National Institute of Allergy and Infectious Diseases (NIAID) found that five- to 10-week antiretroviral interruptions by 26 aviremic men turned around some treatment-related aberrations in short order.¹⁹ They chronicled significant drops in total cholesterol (from 194 to 159 mg/dL, P < 0.001), LDL cholesterol (from 114 to 96 mg/dL, P = 0.0013), triglycerides (from 261 to 185 mg/dL, P = 0.008), and 24-hour urinary 17-hydroxycorticosteroids (from 15 to 5 mg/24 hours, P < 0.001). Glucose and insulin readings stayed flat, and fat abnormalities (in 17 of the 26 men) didn't get better.

A study of 19 men who stopped treatment for 12 weeks had similar results, although these men stopped with viral loads above 2500 copies/mL [abstract P86]. They had taken NRTIs for an average of seven years and PIs for an average of three. Rebecca Hoh (University of California, San Francisco) reported that total cholesterol fell 20 percent 12 weeks after the interruption (from 194 to 106 mg/dL, P = 0.022), and the holiday halved triglyceride readings (from 343 to 161 mg/dL, P < 0.05). LDL cholesterol fell 19 percent over 12 weeks (from 106 to 89 mg/dL, P = 0.22).

As in the NIAID study, fasting insulin did not change. But the San Francisco study participants put on an average 1.3 kg in 12 weeks (P = 0.04), and bioelectric impedance analysis attributed 77 percent of this gain to lean body mass (P = 0.02). Hoh noted, thought, that research has not validated this test in people with lipodystrophy, which about half of these men reported.

This study also underlined the dangers of stopping antiretrovirals in the face of uncontrolled viral replication. The average CD4⁺ count ebbed from 256 to 180 cells/mm³ (P < 0.0001), while the viral load jumped from 4.4 to 5.1 logs (P = 0.004). These changes had clinical consequences: Two new opportunistic infections, including Pneumocystis carinii pneumonia (PCP), cropped up during the 12-week treatment hiatus. And one person suffered "rapidly progressive peripheral neuropathy" despite stopping his drugs.

Return to Article Contents

The Bones

This summary of Toronto meeting news on bone problems in people with HIV infection follows the section on potential remedies for a good reason. Unless you count hip replacement or high-dose hydrochlorothiazide as a fix, no one has figured out what to do about the bone maladies now being cataloged in people with HIV disease. And at this stage of the game no one is too sure whether certain antiretrovirals--protease inhibitors, for example--may be causing osteopenia, osteoporosis, or osteonecrosis, or making them worse. Indeed, some work suggests antiretrovirals are making them better.

It's still a complicated picture and, yes, mechanisms behind weakened bones are probably multifactorial. The origins might be traced back to childhood nutrition and to days spent basking beneath the vitamin-D-breeding sun. Corticosteroids and alcohol abuse have been tied to osteonecrosis, which often turns up in tandem with lipid disorders, diabetes, hepatic steatosis, or pancreatitis--familiar faces all. After completing an osteonecrosis survey of nearly 400 people with HIV infection and finding 4.4 percent with osteonecrosis, one surprised researcher told The New York Times, "We think its prevalence will increase."²⁰

The prevalence of reports on osteopenia and osteonecrosis has certainly increased. Two posters addressed these problems at the 1999 ICAAC. This year's ICAAC and Lipodystrophy Workshop featured at least 10 studies of note, most of them focused on osteopenia.

Return to Article Contents

Do PIs cause--or quell--osteopenia?

Probably recalling some precipitant theories on the cause of lipodystrophy, researchers are circling warily round the question of what drives bone mineral loss in HIV-infected people. Suggestions include:

• Protease inhibitors (which some say are protective)
  
• Fat atrophy (often attributed to NRTIs)
  
• Lactic acidosis (more often attributed to NRTIs)
  
• HIV itself

Pablo Tebas (Washington University, St. Louis) assumed the daunting task of probing for links between PIs and debarking bone minerals. Earlier he used DEXA scans to chart significantly lower bone density scores in people taking PIs than in those taking only non-PI regimens or in healthy volunteers.²¹ Bone density scores did not correlate with fat changes in this 112-man study. At the Lipodystrophy Workshop, Tebas stepped to the plate with two studies, one in vivo and one in vitro, one clocking fast bone turnover in people taking PIs and one suggesting PIs squelch activation of vitamin D, critical to bone health.

Tebas began with a microprimer on weak bones. At varying rates we all make new bone to replace what's lost. About 90 percent of the skeleton that shapes us today will still be around a year from now. If all goes well, 10 percent of that 2001 skeleton will be new. Things haven't been going well for some people with HIV infection, but researchers don't know whether that reflects impaired formation of new bone, swifter loss (resorption) of old bone, or both. Together, formation and loss add up to remodeling.

A cross-sectional study involving 73 people taking PIs for more than six months suggested an overall remodeling speed-up [abstract 29]. Tebas gauged spine and hip bone mineral density by DEXA and measured a dozen markers of bone metabolism in serum and urine. Vitamin D metabolism looked normal, as did testosterone. While serum calcium levels were on target, half of the study participants excreted more than 200 mg of calcium in the urine every day. That and other evidence suggested to Tebas that their bone-making machines had shifted into high gear.

Whether this increased bone turnover "reflects a direct effect [of PIs] on bone metabolism or an indirect effect on bone metabolism or an indirect effect mediated through vitamin D metabolism or renal handling of calcium," Tebas speculated quadrifactorially, "requires further study."

Moments later Tebas was back at the lectern with a further study. This one looked at how ritonavir, indinavir, and nelfinavir affected activity of an enzyme called 1-hydroxylase in a human monocyte-macrophage cell line [abstract 30]. The kidney uses an identical enzyme to manufacture a form of vitamin D essential for calcium regulation. All three PIs handcuffed the relevant enzyme and snuffed production of calcium-regulating vitamin D, but to different degrees--ritonavir by 80 percent, indinavir by 66 percent, and nelfinavir by 32 percent.

Tebas cautioned against besieging the vitamin counter in quest of vitamin D supplements. High-dose vitamin D is toxic. And he added that vitamin D disruption probably is not the sole mechanism explaining osteopenia and osteoporosis in people taking PIs, but it may be one of "multiple steps in a very delicate balance." Tebas suggested just how tenuous even that conclusion is by pointing back to his in vivo study, where vitamin D metabolism was normal.

Another stab at pinning down the cause of bone thinning came from David Nolan (Royal Perth Hospital, Australia), who also saw a link to protease inhibitors [abstract 31]. But he proposed that indinavir increases bone mineral density. At the same time Nolan divined a correlation between bone mineral density and subcutaneous fat--the less subcutaneous fat in the thigh, the less dense were bone minerals in the spine. Tebas's published study found no such correlation.²¹

Nolan's analysis depended on a longitudinal survey of 52 men with HIV infection who had DEXA scans two times at intervals of more than 100 days. Everyone was taking two nucleosides plus nelfinavir or indinavir. Among people taking indinavir, Nolan figured that the average z score, a measure of bone mineral density, increased by 0.37 per year (P < 0.001), whereas z scores didn't change appreciably among those taking nelfinavir. But in a cross-sectional analysis of 171 men, he found high rates of osteopenia (49 percent) and osteoporosis (17 percent) among people taking PIs and much lower z scores in PI-experienced men than in PI-naive men (P = 0.055).

Putting the longitudinal and cross-sectional studies together does not bring the PI-osteopenia picture into sharp focus. Nolan suggested that bone mineral density may begin to fall with HIV infection, before people begin treatment. By the time they start a PI regimen, many already have low z scores and weakened bones. Certain PIs--nelfinavir in this cohort--do nothing to repair the damage. But indinavir may turn things around. The association Nolan found between indinavir and increasing bone mineral density is "in keeping with in vitro experimental data indicating that indinavir favors osteogenic differentiation of mesenchymal stem cells via retinoid signaling mechanisms."

Perhaps, but Tebas wasn't convinced. He suggested that the z score comparisons in the longitudinal study may be unreliable because Nolan based them on only two DEXAs for each study participant. A few scoring mistakes in one direction or the other could distort the results. But Chelsea and Westminster's Graeme Moyle turned up data that supported Nolan's analysis, or part of it.

Moyle rated bone mineral density by whole-body DEXA scans in three groups with HIV infection: 52 antiretroviral-naive individuals (51 of them men), 22 PI-treated men, and 10 men treated only with NRTIs. Average total bone mineral densities in the three groups were equivalent, while t scores, another measure of bone mineral density, were lower in the untreated individuals than in the PI group or the NRTI group (P = 0.322). The lower t score in antiretroviral-naive people, though not statistically significant, is intriguing because they were significantly younger than the PI and NRTI groups, and bone mineral density dwindles with age (38.8 years for the naive group versus 46.5 for the PI group and 41.4 for the NRTI group, P = 0.003).

In a multivariate Cox proportional hazards model with the untreated individuals as the reference group, PI treatment sliced the risk of osteopenia 83 percent (P = 0.003), and NRTI-only treatment cut the risk 68 percent (P = 0.018). A viral load below 62,347 copies/mL (the median value) halved the risk of osteopenia (P = 0.052).

"Osteopenia in persons with HIV infection may be a consequence of HIV infection per se," Moyle concluded, "rather than antiretroviral therapies. Virological control with therapy may diminish the risk of osteopenia." But NRTIs, anyway, won't get off the hook that easily, according to results of a study by Andrew Carr. And scrutiny of osteopenic trends in the Australian PI switch study further clouded the role of both NRTIs and PIs.

Return to Article Contents

Little change in osteopenia during 48 weeks without PIs

Andrew Carr's study involved 221 HIV-infected men, 32 of them naive to antiretrovirals, 42 with only NRTI experience, and 147 with NRTI and PI experience [abstract 32]. According to t scores, 44 (20 percent) had osteopenia and seven (3 percent) osteoporosis. Pretreatment weight proved significantly higher in men without osteopenia (76 kg) than in those with bone disease (71 kg, P = 0.012). The men with osteopenia had been infected with HIV for 9.2 years, compared with 7.7 years among those without osteopenia (P = 0.04). Lactic acidemia (>2 mmol/L) affected 27 men with osteopenia and 11 men without it (P = 0.033). Lactates in those with osteopenia averaged 3.9 mmol/L, compared with 2.1 mmol/L in men without osteopenia (P = 0.002).

Thirty men currently taking AZT had normal bone mineral density, while 12 had osteopenia (P = 0.017). The odds ran in the opposite direction for current d4T: 71 taking that NRTI had osteopenia and 49 had healthy bones (P = 0.013). Duration of d4T treatment also stretched longer among men with osteopenia (18.0 versus 12.9 months for men with normal bones), but this difference was not statistically significant (P = 0.15).

In a multivariate analysis, two factors correlated with osteopenia or osteoporosis: Higher weight lowered the risk of bone disease (odds ratio [OR] 0.94 for every 1 kg higher, P = 0.006). And higher lactates raised the risk (OR 2.39 for every 1 mmol/L higher, P = 0.002). Three factors favored higher lactates: current ddI therapy (OR 6.1, P < 0.0001), current d4T therapy (OR 2.9, P = 0.013), and a higher CD4⁺ count (OR 1.02 per every 10 cells/mm³ higher, P = 0.005).

Higher body weight has traditionally proved a protective variable, presumably because the bones get stronger from lugging more weight. The higher lactate correlation could be a marker for longer NRTI treatment, especially in light of Carr's finding that osteopenia is more common in people infected with HIV longer. The once-removed correlation between the two d-nucleosides and bone disease--that is, they predict lactic acidemia, and lactic acidemia predicts osteopenia--appears to be the first such link between specific drugs and weaker bones. Of course these links between current d4T, current ddI, and osteopenia do not factor in the possible impact of earlier NRTIs. And given Moyle's opposite conclusion that NRTIs, in general, protect against osteopenia, where these chips may fall is anyone's guess.

That question was not answered by analyzing possible ties between antiretrovirals and bone mineral density in the Australian PI switch study [abstract P32]. Jennifer Hoy (The Alfred Hospital, Melbourne, Australia) reported that baseline bone mineral density, t score, or z score did not correlate with baseline HIV disease stage, CD4⁺ count, viral load, duration of antiretroviral therapy, current d4T versus current AZT, or exposure to AZT, ddI, d4T, 3TC, saquinavir, ritonavir, indinavir, or nelfinavir.

But rates of osteopenia (28.8 percent) and osteoporosis (8.8 percent) were high in this 81-person cohort, and Hoy did link total bone mineral density to five variables: older age (P = 0.031), higher weight (P < 0.001), higher body fat (P = 0.021), higher lean body mass (P < 0.001), and lower osteocalcin, a bone formation marker (P = 0.009).

This is the cohort that Carr and colleagues randomized to continue PI therapy or to switch to nevirapine, abacavir, adefovir, and hydroxyurea [abstract P84, Table 8]. DEXA scans taken at screening, then 12, 24, and 48 weeks after randomization, showed precious little bone mineral difference between the continued-PI group and the non-PI switch group.

At the trial's switch point, rates of osteopenia and osteoporosis were, respectively, 25 and 9 percent for the PI group versus 31 and 8 percent for the non-PI group. When study participants could change their assigned regimen 24 weeks after randomization, rates of osteopenia and osteoporosis in the PI group came in at 23 and 10 percent respectively versus 36 and 9 percent in the non-PI group. After 48 weeks, when the non-PI group grew because people crossed over from the PI group, rates of osteopenia (29 percent) and osteoporosis (7 percent) still hadn't changed much. The two treatment groups did not differ significantly in DEXA-measured bone mineral density at any point in the study.

One could put a good spin on Hoy's results: Neither PI treatment nor non-PI combos increased rates of osteopenia over 48 weeks. The study also shows that switching from a PI does nothing to spur new bone formation over that time. Because the Australians don't have bone mineral records from the antiretroviral-naive days of this group, therapy's overall impact on osteopenia cannot be judged.

To complete the circle, a DEXA study of 74 HIV-infected men led a French team to the same conclusion Moyle reached--that HIV infection itself is the prime mover in osteopenia and osteoporosis [abstract I-1304]. But the French based that proposition on results that differ a little from Moyle's. And they made a startling finding about calcium.

Eric Billaud and colleagues (Hospital Hôtel-Dieu, Nantes, France) reported high overall rates of osteopenia (38 percent) and osteoporosis (7 percent), in line with findings of other investigators. Their univariate analysis sniffed out no trail between low bone mineral density and duration of antiretroviral therapy in general or PIs in particular. But a longer time since HIV infection did correlate with osteopenia, determined by lumbar spine t score (P = 0.046). Those findings led Billaud to conclude that "osteoporosis or osteopenia in HIV patients could more likely to be due to the duration of HIV disease than to the use of [antiretroviral therapy]. Chronic stimulation of the immune system has been described as a risk factor for bone mineral loss in other disease."

But, unlike Moyle, Billaud and co-workers did turn up some evidence implicating antiretrovirals. Among 60 people taking antiretrovirals, 42 percent had osteopenia, compared with 18 percent of treatment-naive individuals. And 8 percent of those with antiretroviral experience had osteoporosis compared with none in the naive group. Those differences between the naive and experienced groups were not statistically significant, but they do run counter to Moyle's finding of worse t scores in untreated people than in those taking antiretrovirals.

The French study produced another odd finding--a counterintuitive positive correlation between high daily calcium intake and osteopenia in a univariate analysis (P = 0.03). The researchers could not explain this surprise, except to venture that some people with bone disease may have had early symptoms and tried to compensate by upping their Camembert rations. Another possibility is that the tool used to measure daily calcium intake--a diet questionnaire--may be too crude a gauge. Whatever the explanation, Billaud proposed that people with HIV infection should give serious thought to a calcium-rich diet and exercise.

Return to Article Contents

No sure necrosis role for PIs

Ebbing bone minerals are frightening enough; when bone cells start dying, the urgency to find out why quickens apace. As with osteopenia, one question seems central with osteonecrosis: Does blame lie with protease inhibitors or any other antiretrovirals? PIs certainly figure in hyperlipidemia, a risk factor for necrosis.

Just before the Toronto meetings, an FDA file search turned up 25 cases of avascular necrosis in people with HIV infection, three of them men taking the appetite stimulate megestrol acetate for wasting.²² Megestrol acetate contains corticosteroids, a certain risk factor.

Even more ominously, a pre-Toronto study used magnetic resonance imaging (MRI) to spot 15 cases of asymptomatic osteonecrosis in 399 HIV-infected people (4.4 percent) being cared for at the National Institutes of Health Clinical Center.²³ Henry Masur and colleagues (NIAID, Bethesda, Maryland, USA) started the survey after they saw four cases of painful osteonecrosis that required hip replacements in people with HIV infection. Neither physical exam nor routine x-rays could detect the asymptomatic necrosis.

The NIAID team couldn't tie osteonecrosis to any particular pattern of HIV treatment or to more advanced HIV disease. Factors they did link to osteonecrosis were body building (P = 0.03, a marker for self-administered steroids?) and use of testosterone (P = 0.01), systemic corticosteroids (P = 0.02), or lipid-lowering agents (P = 0.004).

At ICAAC a retrospective review of 11 people with HIV infection and avascular necrosis established no sure link between antiretrovirals and this bone disease [abstract I-1302]. Although nine of these individuals had taken a PI regimen for an average 19.4 months, two had never tried any antiretrovirals. CD4⁺ counts at diagnosis of necrosis ranged from 73 to 900 cells/mm³, but 10 of these 11 people had CD4⁺ nadirs below 50 cells/mm³. Nine had an AIDS-defining illness and 10 had hepatitis C virus (HCV) infection.

Guillem Sirera (Germans Trias í Pujol University Hospital, Barcelona) reported several other risk factors for avascular necrosis in this case series: steroid use by seven, alcohol abuse by four, hyperlipidemia in two, and septic infection (with Salmonella typhimurium) in two. Ages ranged from only 23 to 32 years in these patients, nine of whom had been infected by sharing needles. Three of the 11 were women.

A French case series of seven men with avascular necrosis had a few things in common with the Barcelona group. Nadir CD4⁺ counts in the French cohort were also strikingly low when antiretrovirals began, with a median of 18 cells/mm³ and a range from 5 to 150 cells/mm³ [abstract P34]. As in the Spanish study, most of the French patients--six of seven--had an AIDS-defining condition. But unlike the Spanish series, everyone in the French group was taking a PI when their physicians diagnosed avascular necrosis.

All the French patients had either elevated cholesterol or high triglycerides, they were older than the Spanish contingent (median 38.3 years), and they had taken PIs longer (median 32 months). Three in the French group had diabetes, and one had taken a corticosteroid, prednisone, for 18 months.

Laurent Roudière and Jean-Paul Viard (Hôpital Necker, Paris) confirmed the diagnosis in these individuals with scintiscans or MRIs. Everyone in this group had symptomatic osteonecrosis, and four of them needed new hips. The prevalence of necrosis in the French series, 1.4 percent, was substantially lower than the 4.4 percent in the NIAID cohort. But Roudière and Viard reported only symptomatic cases, whereas the NIAID team screened all their patients with MRIs and found only asymptomatic cases.

To begin segregating real risk factors from mere coincidence in the evolution of avascular necrosis, Marshall Glesby (Cornell University Medical College, New York) did a case-control study involving 14 HIV-infected people with necrosis diagnosed between 1992 and 2000 and 28 controls matched for date of first HIV clinic visit and baseline CD4⁺ stratum [abstract 33]. Twelve of the 14 necrosis diagnoses dated from after 1996--in the PI era--but PI use didn't correlate with the diagnosis in a multivariate analysis. The diagnosis increase since 1996 could simply reflect higher clinician awareness of the disease, Glesby suggested.

Cases and controls did not differ in many potential risk factors, some of them noted in the preceding case series: HCV infection, diabetes, alcoholism, nadir CD4⁺ count, or triglycerides. Other variables that made no difference in Glesby's analysis were year of HIV diagnosis, race, gender, HIV risk factor, hepatitis B virus serostatus, random glucose levels, viral load, or a history of cancer or pancreatitis. But 11 of the 14 people with avascular necrosis had one or more of the following risk factors, some of which overlap with variables in the two case series: alcoholism, corticosteroid use, radiation therapy, triglycerides above 400 mg/dL, and cholesterol above 240 mg/dL.

The controls had taken antiretrovirals for a shorter time (average 14 months) than the case patients (average 32 months) when necrosis was diagnosed (P = 0.06). A matched univariate analysis plucked out five factors that correlated with necrosis: PI or d4T use at diagnosis, a greater than 50-cell rise from the CD4⁺ nadir, PCP, and corticosteroid use. In the multivariate analysis, though, only PCP remained marginally predictive, increasing the risk of necrosis 5.8 times (P = 0.056). Glesby noted that PCP may be a marker for steroid use. He also mentioned that no link between osteopenia and osteonecrosis has been defined.

Return to Article Contents

The End

No one can say where the lipodystrophy story will end, much less the tales of osteopenia and osteonecrosis. At an ICAAC plenary talk on antiretroviral history, Martin Hirsch (Massachusetts General Hospital, Boston) chronicled four ages of anti-HIV therapy through the year 2000--conception, birth, childhood, and early adolescence [presentation 611]. By Hirsch's reckoning, antiretrovirals passed two main milestones in the early adolescence years of 1996 to 2000: proof that regimens combining three or more drugs are superior to one or two antiretrovirals, and confirmation that drug toxicities and resistance are entrenched foes of treatment success.

A similar chronology for treatment toxicity might put lipodystrophy in the toddler stage, while leaving bone afflictions in the incubator. Two years of intense study made it clear that lipodystrophy is much more than a protease inhibitor side effect. Yes, those ritonavir studies in healthy volunteers found a very short fuse on the triglyceride rocket.⁹ Yes, similar studies of indinavir showed glucose significantly sweetening serum in short order [abstract 10]. And yes, those metabolic markers often ease back toward the comfort zone as soon as PIs stop (Tables 8 and 9).

But the dogged wanderlust of body fat after a year or more without PIs brand other antiretrovirals as co-conspirators. Then there are all those pesky nondrug factors. Yet even as you read, big prospective trials, bigger cohort studies, and case definition struggles (variously named FRAM, DAD, and LARD) are sorting plump from lean, drug from drug, risk from root, and cause from coincidence to pin down some answers and squelch some speculation. Meanwhile, as the University of Bristol's Matthias Egger amply demonstrated [abstracts 23 and I-1374], nothing is more urgent than modifying those modifiable risk factors.

There has been progress; no one sends people with lipodystrophy to be worked up for Cushing's syndrome any more. Answers may still seem elusive, but many are asking the right questions. The inquest into osteopenia and avascular necrosis, with luck, will mature as quickly. But these newest companions of HIV may throw down an even sterner challenge. Some of the earliest evidence on the role of antiretrovirals seems contradictory, or at least deeply blurred.

When potent therapies for HIV dawned in that retina-scorching 1996 fireball, certain keen observers shielded their eyes and wondered about the toxic fallout. A prime worry, voiced by pediatric neurologist Leon Epstein (University of Rochester, Rochester, New York, USA), was that HIV seeded in the brain would come back to burn long-term responders, just as some children who survived leukemia when treatment improved later fell to CNS leukemia.²⁴ That worry may still prove true. But who could predict veins popping from skinny legs, fat humps on the neck, or hip transplants?

So no one predicts this will be the last of it.

"We've discovered some of the toxicities," Martin Hirsch said at ICAAC. "There will be more."

Return to Article Contents

References and Notes

*This article covers reports made at the 2nd International Workshop on Adverse Drug Reactions and Lipodystrophy in HIV (Sept. 13-15, 2000) and the 40th Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC, Sept. 17-20, 2000), both held in Toronto. Abstracts preceded by an I are ICAAC presentations. Abstracts of slide presentations at the Lipodystrophy Workshop are not preceded by a letter, while Workshop posters begin with a P. ICAAC abstracts are online at http://www.asmusa.org/mtgscr/40icaac.htm. Antiviral Therapy will publish the Lipodystrophy Workshop abstracts as supplement 5 to volume 5.

1. The Australian cohort, the HIV Outpatient Study (HOPS) cohort, and the Swiss HIV Cohort Study.

2. Agouron Pharmaceuticals, Inc. Lipodystrophy study: physicians' beliefs about its contributing factors. (A telephone survey conducted by Roper Starch Worldwide for Agouron in August 2000.) September 2000.

3. Grunfeld C, Pang M, Doerrler, et al. Lipids, lipoproteins, triglycerides clearance and cytokines in human immunodeficiency virus infection and the acquired immunodeficiency syndrome. J Clin Endocrinol Metab 1992 May;74(5):1045-52. Grunfeld's gloss on this study and related work appears online in the following reference.

4. Grunfeld C. Hyperlipidemia and insulin resistance due to HIV infection and its therapies. Medscape/HIV. 2000. Available at: http://hiv.medscape.com/Medscape/HIV/AnnualUpdate/2000/mha.update05.12.grun/mha05.grunfeld-01.html. Accessed October 6, 2000.

5. Rickerts V, Brodt H, Staszewski S, Stille W. Incidence of myocardial infarction in HIV-infected patients between 1983 and 1998: the Frankfurt HIV-cohort study. Eur J Med Res 2000 Aug 18;5(8):329-33.

6. Mellors JW, Munoz A, Giorgi JV, et al. Plasma viral load and CD4⁺ lymphocytes as prognostic markers of HIV-1 infection. Ann Intern Med 1997 Jun 15;126(12):946-54.

7. Carr A, Samaras K, Thorisdottir A, et al. Diagnosis, prediction, and natural course of HIV-1 protease-inhibitor-associated lipodystrophy, hyperlipidemia, and diabetes mellitus: a cohort study. Lancet 1999 Jun 19;353(9170):2093-9.

8. Schambelan M. Metabolic and morphologic complications of HIV. Available at: http://hivinsite.ucsf.edu/medical/iasusa/3098.0095.html. Accessed October 6, 2000.

9. Purnell JQ, Zambon A, Knopp RH, et al. Effect of ritonavir on lipids and post-heparin activities in normal subjects. AIDS 2000 Jan 7;14(1):51-7.

10. Dubé M, Aqeel R, Edmondson-Melançon H, et al. Effect of initiating indinavir therapy on glucose metabolism in HIV-infected patients: results of minimal model analysis. Antiviral Ther 1999;4(suppl 2):34.

11. Schwenk A, Breuer JP, Kremer G, et al. Risk factors for the fat redistribution syndrome in HIV infected patients--a cross-sectional single-center study. Presented at: XIII International AIDS Conference. July 9-14, 2000. Durban. Abstract WePeB4236.

12. Polo R, Verdejo J, Martínez S, et al. Contribution of NRTIs combinations on lipodystrophy and impact of therapy switching. Presented at: 1st International Workshop on Adverse Drug Reactions and Lipodystrophy in HIV. June 26-28, 1999. San Diego, California, USA. Abstract 67.

13. Saint-Marc T, Touraine JL. Reversibility of peripheral fat wasting (lipoatrophy) on stopping stavudine therapy. Presented at: 1st International Workshop on Adverse Drug Reactions and Lipodystrophy in HIV. June 26-28, 1999. San Diego, California, USA. Abstract 24.

14. Matthews GV, Moyle GJ, Mandalia S, et al. Absence of association between individual thymidine analogues or nonnucleoside analogues and lipid abnormalities in HIV-1-infected persons on initial therapy. J Acquir Immune Defic Syndr 2000 Aug 1;24(4):310-5.

15. Clinicians and people with HIV infection may want to have a look at www.hivfitness.org, a well-organized and thoughtful Web site that suggests numerous aerobic and resistance training programs, always planned with the advice of a physician.

16. Hadigan C, Corcoran C, Basgoz N, et al. Metformin in the treatment of HIV lipodystrophy syndrome: a randomized controlled trial. JAMA 2000 Jul 26;284(4):472-7.

17. Saint-Marc T, Touraine JL. Effects of metformin on insulin resistance and central adiposity in patients receiving effective protease inhibitor therapy. AIDS 1999 May 28;13(8):1000-2.

18. Mascolini M. Can HIV docs become compassionate conservatives? J Int Assoc Physicians AIDS Care 2000;6:78-83.

19. Hatano H, Miller KD, Yoder CP, et al. Metabolic and anthropometric consequences of interruption of highly active antiretroviral therapy. AIDS 2000 Sep 8;14(13):1935-42.

20. Altman LK. Rare bone disorder found in H.I.V. patients. New York Times. September 9, 2000. The quotation is by NIAID clinician Joseph Kovacs.

21. Tebas P, Powderly WG, Claxton S, et al. Accelerated bone mineral loss in HIV-infected patients receiving potent antiretroviral therapy. AIDS 2000 Mar 10;14(4):F63-7.

22. Reuters. Megestrol acetate linked to aseptic bone necrosis in HIV-infected men. Available at: http://hiv.medscape.com/26839.rhtml. Accessed October 6, 2000.

23. Masur H, Miller KD, Jones EC, et al. High prevalence of avascular necrosis (AVN) of the hip in HIV infection: magnetic resonance imaging of 339 asymptomatic patients. Presented at: IDSA 2000 (annual meeting of the Infectious Diseases Society of America). September 7-10, 2000. New Orleans. Abstract 15.

24. Mascolini M. Outing HIV (or, the tailor retailored). J Int Assoc Physicians AIDS Care 1996;2(No. 9):56.

Return to Article Contents

001110
IA001102

Copyright © 2000 - International Association of Physicians in AIDS Care. All rights reserved. http://www.iapac.org

ÆGiS is made possible through unrestricted grants from Roxane Laboratories, Inc., iMetrikus, Inc., the National Library of Medicine, and donations from users like you. Always watch for outdated information. This article first appeared in 2000. This material is designed to support, not replace, the relationship that exists between you and your doctor.

ÆGiS presents published material, reprinted with permission and neither endorses nor opposes any material. All information contained on this website, including information relating to health conditions, products, and treatments, is for informational purposes only. It is often presented in summary or aggregate form. It is not meant to be a substitute for the advice provided by your own physician or other medical professionals. Always discuss treatment options with a doctor who specializes in treating HIV.

Copyright ©1990, 2000. ÆGiS & the Sisters of Saint Elizabeth of Hungary. All materials appearing on ÆGiS are protected by copyright as a collective work or compilation under U.S. copyright and other laws and are the property of ÆGIS and the Sisters of Saint. Elizabeth of Hungary, or the party credited as the provider of the content.