Something very much like that happened on May 16-17, 2002, in Chicago, when the International Association of Physicians in AIDS Care (IAPAC) gathered some of this era's top HIV thinkers and asked them to anatomize four specimens before an eager crowd. And the specimens were very much alive, and expertly vivisected though not human, mammalian, or even Drosophilian, but rather conceptual.
In anatomy lessons led by Christine Wanke (Tufts University, Boston), Diane Havlir (University of California, San Diego), Daniel Kuritzkes (Brigham and Women's Hospital, Boston), and Maurizio Bonacini (California Pacific Medical Center, San Francisco), a faculty of eight pared away superficial tissue and probed the tendons of four topics: antiretroviral tactics, their metabolic and morphologic side effects, resistance testing, and the confluence of HIV with hepatitis B and C viruses.
The IAPAC Sessions 2002 differed from classically limned anatomy lessons in another important way: Although the audience included a sprinkling of the merely curious, most attendees were anything but acolytes in HIV medicine. They were carefully selected clinicians seeing large quotas of people with HIV, and imbued with a matching commitment to the corpus of complication and contingency that defines the art in this science.
The reasons to treat later in the course of HIV progression, somewhere below 350 cells/mm3, but if possible above 199 cells/mm3, are familiar to all: The risk of progression is low even when treatment starts late, Havlir allowed, and later intervention does not greatly jeopardize the response to treatment. Antiretroviral regimens are tough to stomach and hard to stick with -- traits that favor missed doses and foster resistance. And then there is toxicity.
But Havlir challenged two key cohort studies that showed low risks of progression or death when therapy starts below 350 cells/mm3. These studies in Vancouver1 and at Johns Hopkins in Baltimore,2 which both tracked treated people beginning in 1996, appraised antiretroviral regimens more than a tad moldy by today's standards. So they fail to reflect the possible benefits of starting with a nonnucleoside (NNRTI) regimen, for example, or with a boosted protease inhibitor (PI). As a result, the rationale for later treatment, as framed now, overlooks the greater simplicity and tolerability of newer therapies, the profusion of salvage options, and a changing patient population: Those starting treatment in the past few years don't lug the baggage of sequential nucleoside (NRTI) monotherapies and duotherapies, or the class-sacrificing stratagem of adding a new inhibitor to a woebegone combo.
Follow-up in the Vancouver and Baltimore studies, 28 and 22 months, now represents a small fraction of the expected life span made possible by today's best combinations. If these cohorts could be tracked for even a few more years, Havlir proposed, differences in survival or AIDS diagnoses may emerge. Even if continued follow-up discerned no great morbid or mortal difference between early and late starters, she argued that an endpoint focus narrowed to AIDS or death blurs other potential benefits of early intervention -- lower incidence of HIV-related (though not AIDS-defining) diseases, increased energy and a better quality of life, enhanced cognition and performance, and greater prospects of benefiting from strategies that may work better in a better preserved immune system, such as therapeutic vaccines.
What about antiretroviral toxicity? That "major concern," Havlir conceded, lingers. The full spectrum of side effects exceeds our field of vision, and their mechanisms remain murky. They may not be treatable; they may not be reversible. And there's the rub for many clinicians seeing a person with newly diagnosed HIV infection, a CD4 count of 325 cells/mm3, and a viral load of 60,000 copies/mL. If she had the time, Havlir might have dwelled more on the better tolerability of new drugs -- the antiatherogenic potential of nevirapine,3 for example, or of the investigational PI atazanavir,4, 5 not to mention the durability and endurability that have made efavirenz a first-line winner in cohort after cohort.6-8
But even where she left it, Havlir's crusty little dialectic recalls a lesson HIV clinicians have learned many times over -- always question the standard of care, because in this field standard today may be time-worn tomorrow. And Havlir might have pointed to a list proposed by the opening session's chair Christine Wanke, counting at least three potential interventions (Table 1) for the most-concerning long-term toxicities -- glucose gridlock, lipid upsets, fat atrophy, and fat accumulation -- problems probed during the first half-day of the IAPAC Sessions 2002.
|Table 1: When Antitoxicity Measures May Be Warranted|
|Fat atrophy||Antiretroviral switches, rosiglitazone, cosmetic surgery|
|Fat accumulation||Diet, exercise, metformin, recombinant human growth hormone|
|Glucose abnormalities||Diet, exercise, glitazones, antiretroviral switches|
|Lipid abnomalities||Diet, exercise, anilipid agents, antiretroviral switches|
|Source: Christine Wanke|
Antiretrovirals, with PIs topping the list, deserve an ample share of blame for dysregulated lipids and glucose in people being treated for HIV. But Kathleen Mulligan (University of California, San Francisco) reminded attendees that the retrovirus itself starts stirring the metabolic pot well before HIV drugs spice the stew. It's been 10 years since Mulligan's colleague Carl Grunfeld charted higher triglycerides in people with HIV infection or AIDS than in uninfected people.9 In the same study total cholesterol -- including both the high- and low-density lipoprotein varieties (HDL and LDL) -- proved significantly lower in people with AIDS. When people started taking PIs, triglycerides climbed even more, total and LDL cholesterol shot up, and protective HDL cholesterol stayed low.
In healthy volunteers taking ritonavir, research showed marked jumps in triglycerides and cholesterol, striking in their rapidity and (for triglycerides) their magnitude.10 Indinavir given to healthy people for 4 weeks barely budged triglycerides, total cholesterol, or HDL cholesterol,11 but it did rile glucose, and quickly (see below). Other work documented significantly impaired endothelial function in people taking PIs.12 Research also showed thicker carotid artery walls in HIV-infected women with lipodystrophy13 and more plaques gumming up carotid and femoral arteries in people with HIV,14 but PIs didn't worsen either of these cardioconsequential omens.
Do these talismans -- favorable, unfavorable, and indeterminate -- portend a heart disease epidemic in people taking PIs? Mulligan summoned siftings from two massive cohorts in an attempt to answer that question, and so far (if one looks only at those studies) the answer is no. After 5.5 years and 14,823 person-years of followup in a California health cooperative, people with HIV had higher rates of coronary heart disease (P = 0.003) and myocardial infarction (P = 0.07) than did uninfected people.15 But neither antiretrovirals in general nor PIs in particular weighed in this equation. A Veterans Administration study embracing 121,936 person-years of follow-up logged decreased hospital admissions for heart disease and stroke since PIs took the stage.16 But this second survey did not compare HIV-infected and uninfected people, Mulligan observed. And the drops in admissions may reflect unmeasured factors, such as better controlled chronic infection in people taking potent antiretrovirals.
Those findings don't mean people taking protease drugs and doctors who prescribe them can forget about heart attacks, Mulligan cautioned. Rather she stressed one sure bet in warding off time in the cardio ward -- a renewed focus on reducing or eliminating known risk factors.
In bygone days before one ever worried about whether to say protease or proteinase, HIV had little measurable impact on glucose or insulin.17 If anything, Mulligan noted, insulin sensitivity seemed better in people infected with HIV than in seronegatives. But all that changed when PI prescribing began. She listed four studies that cataloged higher rates of impaired glucose tolerance or diabetes in PI takers or people with lipodystrophy18-21 (Table 2).
|Table 2: Diabetes and Impaired Glucose Tolerance (IGT) in the Age of Pls|
|Walli (1998)18||Taking Pls||24||37||17|
|Behrens (1999)19||Talking Pls||38||13||24|
|Carr (1999)20||Taking Pls||113||7||16|
|Hadigan (2000)21||With lipodystrophy||71||7||35|
|* 2-hour glucose >200 mg/dL|
** 2-hour glucose >140 mg/dL
Source: Kathleen Mulligan
If a heart disease epidemic does not yet plague people taking PIs, does a diabetes epidemic? Mulligan observed that the smaller studies listed in Table 2 18, 19 are not the only ones showing double-digit diabetes rates in HIV-infected people. Three large cohorts had rates ranging from 10 to 14 percent in HIV-infected people,22-24 and up to 19 percent in people coinfected with HIV and HCV.24 One of these surveys22 studied only people taking PIs, while the other two23, 24 traced no links between PIs and diabetes. But Mulligan said a rising incidence of diabetes in PI takers wouldn't surprise her because cohort studies show that these drugs bollix glucose metabolism (Table 2), and other research verifies that PIs destabilize glucose homeostasis.
Several studies in the past few years drive those points home, and the dysregulating derelicts include indinavir,11, 25-27 ritonavir,27 amprenavir,27 and nelfinavir28 -- so far. The only study to implicate amprenavir in insulin resistance did so in fat cells,27 whereas an 8-week amprenavir trial in whole humans found no insulin resistance.29 Mulligan added that atazanavir may also escape the taint of muddled glucose transport, but that finding remains to be verified in humans.
Although PIs that induce insulin resistance do so rapidly -- in people with or without HIV infection -- visceral fat gains that come with continued PI therapy only worsen the problem. At the same time numerous studies in mice, men, and women show that too little fat -- a frequent consequence of prolonged NRTI therapy -- also incite metabolic mayhem, including high triglycerides, hepatic steatosis, and severe insulin resistance. Yet switching away from PIs tends to improve insulin sensitivity, so protease drugs probably deserve most of the blame.
Tebas suggested the definition will find its greatest use as a benchmark for clinical investigation. Yet the 31-site, 1,000-person effort may prove clinically useful as well, because the research team will soon offer an easy-to-use online version of their equation. Clinicians will be able to plug in a person's data for all or some of the variables and get a lipodystrophy score that could sort out subtle cases.
One correlation that seems to hold true across studies is the one between more advanced HIV disease and fat abnormalities. HIV Outpatient Study (HOPS) investigators tightened this link in a survey that looked specifically at lipoatrophy incidence through the prism of nadir and maximum CD4+ counts.32 Their findings lend credence to Diane Havlir's contention that delaying therapy has risks not measured in studies weighing only AIDS diagnoses and death (see above). This 21-month analysis of 546 people with HIV found that atrophy appeared in only 3.3 percent of those whose CD4+ tally never fell below 350 cells/mm3, but atrophy plagued 12 percent with a nadir between 200 and 349 cells/mm3, and more still with nadirs below 200 cells/mm3, even if their CD4+ counts rose above 350 or 500 cells/mm3 with treatment.
The HOPS cohort consists largely of men, but women may be more prone to lipodystrophy -- or at least fat accumulation -- than men. Female gender was an independent risk factor in the EMEA case definition.30 And in an Italian cohort of 2,258 people with HIV, including 677 women, female gender proved the strongest predictor of fat abnormalities with an adjusted odds ratio of 2.019 (P = 0.001).33 Significantly more women (10.1 percent) than men (7.7 percent) in this cohort had fat accumulation (P = 0.0008), and significantly more women (22.4 percent) than men (9.7 percent) had both accumulation and atrophy (P = 0.0001). More men than women had atrophy alone, but this difference lacked significance.
If it could be bottled, everyone would prescribe it.
Metformin, the insulin-sensitizing agent licensed to treat diabetes, has shown merit in two randomized trials of HIV-infected people with hyperinsulinemia and central fat buildups.34-36 It improved insulin sensitivity, trimmed overall fat (including visceral fat), improved markers of cardiovascular disease,36 and proved tolerable. But because lactic acidosis is metformin's premier side effect, Mulligan cautioned against prescribing it for people with already high lactates. She noted that metformin won't reverse -- and may worsen -- peripheral fat wasting, and that its value in people with normal insulin sensitivity is unknown.
Glitazones (or, for unreformed polysyllabophiles, thiazolidinediones) also bear labels indicating their use for diabetes mellitus. The pace setter in this class, troglitazone, improved insulin sensitivity, pared visceral fat, and added some subcutaneous fat in obese, lipoatrophic, HIV-negative people with noninsulin-dependent diabetes.37-39 Troglitazone's ill effects on the liver led to its banishment, so attention turned to rosiglitazone. This drug improved insulin sensitivity in people with HIV lipodystrophy.40 But in this placebo-controlled trial involving 30 people with HIV lipodystrophy, rosiglitazone inflated triglycerides and had no effect on visceral fat or waist-to-hip ratio. Subcutaneous fat did rise a bit in the rosiglitazone group, but not significantly.
Pablo Tebas noted that this small 24-week Finnish study rocked researchers planning to study rosiglitazone for lipodystrophy. The AIDS Clinical Trials Group (ACTG) canned one rosiglitazone trial but will continue ACTG 5082, which compares rosiglitazone, metformin, both, or neither in people with unhealthy waist-to-hip ratios and insulin resistance. Australian researchers pursuing a large study of rosiglitazone for lipoatrophy issued something of a manifesto critiquing the 30-person placebo-controlled trial. They raised four points:
Tebas added that the 8-mg rosiglitazone dose used in the Finnish study seems adequate, but it was given once daily instead of in twice-daily 4-mg doses, which may work better.
In the end clinicians may not need glitazones to remedy lipoatrophy, not if results of three drug-switch studies hold true. In two randomized trials41, 42 and one nonrandomized study,43 three groups charted small but significant limb fat gains after switching from stavudine (d4T)41, 43 to another NRTI or from two NRTIs -- d4T and lamivudine (3TC) or zidovudine (AZT) and 3TC -- and a PI to abacavir plus Combivir (AZT/ 3TC)42 (Table 3).
|Table 3: Peripheral Fat Gains in Three Drug-Switch Studies|
|Author, Design||DEXA Results||CT Results||Comment|
d4T or AZTABV
(n = 111x24 w)
|0.39 kg limb fat|
|Improvement||No participant or clinician noticed any fat gain|
2 NRTIs + PIABV/CBV
(n = 39x48 w)
|0.018 kg/month fat gain in arms; 0.013 kg/month fat gain in legs||Not done||No lipid improvements|
d4TABV or AZT
(n = 118x24 W)
|25% fat fain in arms, 6% in legs, and 9% in trunk||Improvement||Participants self-reported fat gains|
|ABV = abacavir; CBV = Combivir (AZT/3TC)|
Tebas tempered any enthusiasm these results may inspire by noting that both the studies and the fat improvements were small -- so small that people in the 24-week Australian trial and their clinicians didn't notice the fat gain.41 Perhaps the most hopeful finding came from the other Australian study,42 which showed that fat gains in arms and legs continued from week 24 through 48. ACTG 5110 will pursue this strategy, randomizing people to trade in d4T or AZT for either an NNRTI/PI regimen or for abacavir plus the baseline PI or NNRTI.
Swapping d4T or a PI for different drugs may slowly promote subcutaneous fat gains in arms and legs but appears to do nothing to replace lost facial fat. Tebas briefly mentioned early studies of polylactic acid (New-Fill) facial implants, saying "the aesthetic result is remarkable." This promising technique remains unavailable in the US.
Kathleen Mulligan and others continue their struggle to find a tolerable yet still effective dose of recombinant human growth hormone (Serostim) for lipodystrophy. The drug cut visceral fat and shrank buffalo humps in open-label studies, Christine Wanke noted. But the dose used for wasting, 6 mg daily, far exceeds what's needed for lipodystrophy. Mulligan and colleagues found that even 3 mg daily upsets glucose homeostasis,44 and they are now evaluating 1 mg daily. The other common side effect of growth hormone, swelling, can be painful and can be mistaken for reversal of atrophy. Finally, the drug's effect is transient, so it would have to be started again when the effect wears off.
Smith made three arguments for non-PI regimens: (1) Their efficacy equals or exceeds that of PI combos. (2) They cause fewer metabolic side effects. (3) They improve adherence. DuPont's seminal 006 study showed that efavirenz outmatched standard-dose indinavir in treatment-naive people,45 and it maintained that edge through 96 weeks.46 Nevirapine plus Combivir did as well as twice-daily nelfinavir plus Combivir in keeping viral loads below 20 copies/mL through 12 months in a naive population.47 And abacavir plus Combivir more than held its own against indinavir plus Combivir for 48 weeks in another naive cohort.48
This last study underlined an earlier concern about triple-nuke regimens, their suspect performance in people starting therapy with a high viral load. In an as-treated analysis 59 percent in the abacavir arm versus 73 percent in the indinavir arm with a baseline viral load over 100,000 copies/mL had fewer than 50 copies/mL at week 48. But an intent-to-treat analysis discerned no RNA difference between these two groups. Another trial comparing abacavir/Combivir with nelfinavir/Combivir found equivalent proportions with sub-50-copy viral loads in the two arms in a 48-week missing-data-equal-failure analysis.49 But Smith observed that researchers designed this trial to compare lipid changes in treatment-naive people (half of them women) and not to compare virologic outcomes. The 48-week viral load results involved only 15 people taking abacavir and 19 taking nelfinavir.
Smith did not broach the growing hoard of cohort data supporting first-line efavirenz regimens. If she had time, she might have mentioned that efavirenz outpaced PIs6-8 (including ritonavir/ saquinavir)6 and nevirapine7 in suppressing viral load, at least over the short term. The potency and tolerability of NNRTIs swayed the British HIV Association to list them as "recommended" for first-line therapy, while advising that PIs may be "considered" up front.50 The latest US Department of Health and Human Services (DHHS) Guidelines for the Use of Antiretrovial Agents in HIV-Infected Adults and Adolescents give efavirenz the nod as a first-line option, but not nevirapine.51 They also put indinavir, nelfinavir, and three dual PIs in the first-line column. Renslow Sherer, assigned to defend PI therapy, nonetheless proposed abacavir and nevirapine as first-line cornerstones, a concession demonstrating the allure of starting with a non-PI combo.
But Sherer also reviewed trial data suggesting that two PIs may constrain the virus longer than efavirenz. Although efavirenz bettered ritonavir/saquinavir virologically in one eight-month cohort study,6 a four-year intent-to-treat follow-up of people taking those two PIs counted 70 percent with a viral load under 200 copies/mL.52 Lopinavir/ritonavir kept 76 percent of one study group under 50 copies/mL for three years in an intent-to-treat analysis,53 while 52 percent in the landmark efavirenz study had a three-year sub-50 viral load by intent-to-treat reckoning.54 Cross-study comparisons like this hardly prove that double or ritonavir-boosted PIs suppress HIV longer than efavirenz. But this exercise shows how hard it is to compare PIs and NNRTIs on an ever-evolving playing field.
Right now one part of that field tipping decidedly in PIs' favor is resistance, especially when one considers ritonavir boosting in treatment-naive people. Sherer borrowed the step diagram fashioned by Abbott's Dale Kempf to make this point. HIV has to reshape itself with a handful of mutations before it attains a high level of resistance to PIs. But this advantage dissolves in the face of low trough concentrations typical of single PIs (Figure 1, Single PI). Those low troughs, perhaps deepened by missed doses, allow resistance mutations to pop up one after another. NNRTIs, in contrast, have comfortably safe, high troughs. But the comfort zone can shrink fast because only one or two mutations will overwhelm efavirenz or nevirapine (Figure 1, NNRTI). Several mutations must evolve to sink boosted PI combos (Figure 1, Enhanced PI), which notch troughs much higher than single PIs. This picture could change again, though, if second-generation nonnukes fulfill the promise to erect a higher barrier to resistance than their first-generation forebears.
Pressing the resistance argument further, Sherer cited a revealing but small study comparing the impact of up-front regimens on later resistance profiles. This analysis of 47 treatment-naive people starting triple therapy with a PI and 17 starting with an NNRTI found that 41 percent taking an NNRTI versus 6 percent taking a PI ended up with dual-class resistance if their viral load rebounded above 1,000 copies/mL.55 Those results suggest that resistance to NRTIs may evolve more quickly when an NNRTI fails than when a PI does.
When Kimberly Smith outlined the traits of complex regimens -- thrice-daily dosing, a high pill burden, food and fluid requirements, and special storage needs -- she could have made a good case against solo indinavir, amprenavir, or ritonavir. But how often are those PIs prescribed alone nowadays? She also cited week 44 to 48 self-reported adherence data from the trial comparing abacavir/Combivir with indinavir/Combivir.56 Significantly more taking abacavir (72 percent) than indinavir (45 percent) claimed better than 95 percent adherence (P Renslow Sherer brought this contest up to date when he compared a few PI combos now in vogue with standard NNRTI dosing. Lopinavir/ritonavir (three pills twice daily) and indinavir/ritonavir (three pills twice daily) don't suffer much from comparison with nevirapine (one pill twice daily) or efavirenz (one or three pills once daily). He observed that research has yet to nail down an adherence advantage for once-daily versus twice-daily dosing. So the question becomes how much sheer pill number affects adherence. Analysis of 23 clinical trials tied higher pill burden to worse 48-week virologic control.57 But this analysis found no virologic difference between regimens centered on a PI, an NNRTI, or an NRTI, and it evaluated no double or boosted PIs.
Smith made a strong case for greater safety with abacavir, efavirenz, and nevirapine than with PIs, especially when it comes to lipids and glucose. Indeed, her toughest task in this part of the debate must have been picking from the profusion of studies that back her points. She cited the 48-week comparison of abacavir/Combivir and nelfinavir/Combivir to demonstrate significantly lower total and LDL cholesterol in the abacavir group (P 50 Yet lactates were no higher in the triple-NRTI group than in the nelfinavir/Combivir group.
Smith also invoked results of two randomized trials in which people traded a PI for abacavir, nevirapine, or efavirenz.58, 59 Triglycerides fell most in the nevirapine groups, which also enjoyed the sharpest jumps in HDL cholesterol. The biggest drops in total cholesterol came in the abacavir groups. The studies also mapped significant declines in insulin resistance. The abacavir groups suffered the most virologic breakthroughs, which researchers tied to baseline reverse transcriptase mutations.
Sherer could only concede that faulty metabolic markers start falling back in line when people stop PIs. He cited an outline of switch studies by William Powderly (Washington University, St. Louis) summarizing lipid and insulin improvements with nevirapine, efavirenz, or abacavir (Table 4). But Sherer worried that "we're too obsessed with lipids, out of proportion to other issues affecting people in our care." Besides a welter of deranged lipids and glucose, he reminded attendees, PIs can also induce nausea, diarrhea, kidney stones, and liver toxicity. On the nonnuke side of the ledger, he listed rash and liver toxicity for nevirapine and efavirenz, and central nervous system side effects, teratogenicity, and hypercholesterolemia for efavirenz. "I'm not sure there is a winner in toxicity," he concluded.
|Table 4: Metabolic Changes in PI-to-Non-PI Switch Studies|
|Virologic suppression||Maintained||Maintained||Avoid abacavir in people with NRTI experience|
|Insulin resistance and hyperglycemia||Improved||Improved||Improved|
|Triglycerides||Lower or no change||Lower or no change||Lower or no change|
|Cholesterol||Lower or no change||No change||Lower|
|Source: William Powderly|
The lipid obsession, Sherer added, will prove unfounded if a theory advanced by Werner Richter (Ludwig-Maximilians-University of Munich) is right. Richter holds that the very low-density lipoprotein (VLDL) and apolipoprotein B patterns conjured by the liver in people with PI-induced hyperlipidemia resemble those of people with familial hypertriglyceridemia, not those of people with familial combined hyperlipidemia. What's the difference? Familial combined hyperlipidemia confers a high risk of heart disease, while familial hypertriglyceridemia confers no risk or a low risk of heart disease. If true, that hypothesis could partly explain why two big cohort studies failed to forge a link between PIs and heart disease.15, 16
But behind this endorsement of resistance assays lie persistent uncertainties about when to use these tests and what they mean. Those uncertainties bedevil not only front-line HIV clinicians, but also the experts, as David Katzenstein (Stanford University), Richard Haubrich (University of California, San Diego), and session chair Daniel Kuritzkes (University of Colorado Health Science Center) made clear. Assigned to probe the limits of resistance testing, Katzenstein averred that, despite such limits, he sees these tests as a crucial component of antiretroviral planning. Assigned to stress these assays' merits, Haubrich spent much of his time explaining why they fail.
Katzenstein and Haubrich agreed wholeheartedly on one point: as HIV therapies improve, it's getting harder and harder for randomized trials to show a significant virologic benefit for resistance testing. Haubrich told this tale of diminishing returns in a concise table 60-67 (Table 5). The difference in RNA drop between the genotyping arms and standard-of-care arms in the first two randomized resistance studies, VIRADAPT and GART, came close to 0.6 log. But the difference between the assay arm and the control arm in all the following studies averaged 0.06 log.
|Table 5: Is the Resistance Assay Advantage Waning With Time?|
|Study||Design||First PI Failing at Study Entry (%)||Low Change in HIV RNA (w 12-16)||<400 copies/mL (%)|
|VIRADAPT60||G vs SOC||40||-1.04 vs -0.46*||29 vs 14*|
|GART61||G vs SOC||50||-1.19 vs -0.61*||34 vs 22*|
|Kaiser62||P vs SOC||25||-0.25 vs -0.4(NS)||NA|
|VIRA 300163||P vs SOC||100||-1.23 vs -0.87*||46 vs 34 (NS)|
|NARVAL64||P vs G vs SOC||<30||-0.7 vs -1.1 vs -1.0||34 vs 41 vs 34|
|ARGENTA65||G vs SOC||47||-0.6 vs -0.4 (NS)||26 vs 12*|
|HAVANA66||G vs SOC||44||-1.5 vs -1.2*||66 vs 53*|
|CCTG 57567||P vs SOC||80||-0.7 vs -0.7** (NS)||45 vs 46 (NS)|
|G = genotyping; NA = applicable, NS = not significant; P = phenotyping; SOC = standard of care.|
* P< 0.05
Source: William Powderly
Haubrich saw two reasons for this tailspin: Salvage options today are more plentiful and more potent than they were in the infancy of resistance testing. As one attendee aptly noted, genotyping in VIRADAPT and GART might have proved less advantageous if clinicians in both arms could have prescribed lopinavir/ritonavir along with, say, tenofovir. Haubrich's second point was that clinicians picking rescue regimens in the standard-of-care arms are doing a much better job these days. Analyzing his own phenotyping study, CCTG 575,67 he noted that the lack of difference between study arms (Table 5) shows "not that phenotyping failed, but that standard-of-care did so well." But if clinicians have become so savvy in picking rescue drugs, Kuritzkes wondered, what does phenotyping add?
There are other ways to analyze the results in Haubrich's resistance assay timeline (Table 5). One is that phenotyping has failed in three62, 64, 67 of four studies. The only study in which it passed muster, VIRA 3001,63 had the least treatment-experienced population. Those results run counter to the idea that genotyping works best in people with a short history of treatment failure, while phenotyping works best in those with a long history. As Kuritzkes explained this concept, after a first or second PI failure, genotyping will pick out the sentinel mutations and so suggest salvage options. But after several PI failures, the hash of primary and secondary protease mutations becomes too hard to decipher. At that point a phenotype may tell you more. But how much more? Katzenstein proposed that, if (as trials show) phenotyping gives you at best a marginal edge, it may not be worth paying up to three or four times more than you would pay for a genotype.
Another complication in analyzing the randomized trials on record, Katzenstein observed, is that most of them have short follow-ups, generally 12 to 16 weeks. The longest study, CCTG 575, showed a virologic dead heat between phenotyping and no phenotyping after 12 months. Would genotyping's advantage vanish if follow-up continued so long? Perhaps not, if 48-week results68 of the 1997-1998 VIRADAPT study hold true in 2002.
Short-term endpoints, Katzenstein said, can be "contaminated" by expert opinion. And that's what some people say happened in GART.61 Prescribers in that study didn't get raw genotypes; they got genotypes interpreted by virologists who spend most waking hours thinking about resistance. So, this argument goes, taking away the experts may take away genotyping's advantage.
The more recent Havana study bolsters that argument. This trial randomized people taking a failing regimen to get new drugs based on genotyping, expert advice, both, or neither.66 Genotyping raised the odds of a good response 1.7 times (P = 0.016), but failure of a second regimen sliced the odds of success by 60 percent (P = 0.057), and a third failure cut the odds of success by 70 percent (P = 0.0001). For people coming off their second failed regimen, though, expert advice tripled the chance of virologic success (P = 0.016). In other words, when things get tricky, call an expert.
But if you can tap expert opinion, do you even need a genotype? Or, to put it another way, if one adds up all these trends, do they mean resistance testing is a waste of time? No one ventured that opinion, but Haubrich and Katzenstein took a hard look at when testing does make sense (Table 6). Both were less gung-ho than the IAS-USA panel that issued guidelines in 200069 and a bit less enthusiastic than DHHS experts who offered advice earlier this year.51 Though Haubrich didn't endorse testing for treatment-naive people, he said that could change in the future if transmission of resistant virus continues to rise (as it is already in some parts of North America and Europe). But in untreated people diagnosed simultaneously with HIV and their first AIDS-defining illness, he saw little use for resistance testing. At that point, Haubrich explained, any resistant virus transmitted during primary infection will have reverted back to wild type.
|Table 6: Evolving Advice on When to Test for Resistance|
|Disease Stage||IAS-USA Panel69 (5/2000)||HHS Panel51 (2/2002)||Katzenstein (5/2002)||Haubrich (5/2002)|
|Primary infection||Consider||Consider||Uncertain||Perhaps in future|
|Chronic infection||Consider||Not generally recommended||Probably not||Perhaps in future, but only if infection is recent|
|First regimen failure||Recommended||Recommended||Yes, though rationale diminishing||Yes|
|Multiple regimen faliure||Recommended||Recommended||Yes||Yes|
|Source: HHS panel,51 Hirsch et al,69 David Katzenstein, Richard Haubrich|
Kuritzkes cast the question of testing treatment-naive people in monetary terms. If paying is no problem, testing all naive people seems reasonable because you will pick up a few transmitted mutations. But early research suggests that testing a naive population becomes cost effective only when at least 5 percent of that population (excluding those with primary infection) got infected with resistant virus.
But because Katzenstein's role at the IAPAC Sessions 2002 was to dissect the limits of resistance testing, he went ahead and did so -- even though he called some of his own criticisms "carping." He began with perhaps the most vexing resistance riddle:
Susceptibility and resistance evolve along a continuum for any given drug or virus. Researchers, labs, and assay makers propose cutoffs, typically expressed as a fold change in inhibitory concentration. But HIV doesn't know about cutoffs. So for any resistance readout, the clinician must ask two questions:
That those questions must be answered suggests science has not yet replaced art in interpreting resistance.
Richard Haubrich called Virco's biologic cutoffs a big step in improving clinicians' ability to use resistance testing. Ultimately, though, Haubrich said clinical cutoffs will be needed, that is, cutoffs defined from data that relate resistance assay results to clinical response. Two cutoffs may be needed for each drug -- the first defining the inhibitory concentration at which a drug's antiviral activity begins to wane, and the second marking the point where the drug has no activity. Researchers at GlaxoSmithKline and Gilead proposed clinical cutoffs for abacavir and tenofovir. But both efforts depended on studies that added each drug to a failing regimen. And such studies become less likely as this so-called intensification strategy falls from favor.
Besides defining clinical cutoffs, Katzenstein noted, research must grapple with the still blunted sensitivity of these assays. "We're in a technologic bind," he said, in that assays detect resistance only if it affects an ample slice of the viral population. Resistance-conferring mutations must sully more than 20 percent of a viral population before a genotypic assay spots them as a mixture, he estimated. More than 50 percent must have ebbing sensitivity to a drug before phenotyping can figure an accurate 50 percent inhibitory concentration (IC50). With the Virtual Phenotype, he said, the 20 percent rule applies because the test actually matches genotypes.
One advantage of the Virtual Phenotype, session chair Daniel Kuritzkes offered, is that Virco constantly updates the genotype database with which a submitted sample is compared. Reports that come with standard genotyping, on the other hand, use rules set by experts who meet periodically, so their advice may lag new findings. But Kuritzkes cautioned that Virtual Phenotypes need careful interpretation. The report will tell you, for example, that 85 percent of the database isolates matching the submitted sample's genotype are susceptible to a certain drug. One way to read this result is that the sample has a 15 percent chance of being resistant to that drug. And when the percent of matching isolates in the database susceptible to a drug becomes 65 percent or 60 percent, he added, "you're much closer to a coin toss."
Even when a resistant population is big enough to set off a resistance assay alarm, Katzenstein said, other factors may muddy application of the results. Yes, your patient's virus may be susceptible to drug Z, but can your patient keep drug Z concentrations at the level needed to stymie replication? Or will shaky adherence, sluggish absorption, or drug interactions yield dangerously low troughs?
Then there are those inky questions of mutational interactions. One could write a fat tract on this issue alone -- and solve nothing doing so. Katzenstein offered one example, the 184I or V mutation that makes virus resistant to 3TC. Work by his colleague Nancy Shulman showed that these mutations lowered the IC50 of AZT 9-fold and of d4T 2-fold, while boosting the IC50 of ddI 2-fold and of abacavir 7-fold.70
Haubrich tossed in another caveat: Different labs interpret different resistance results differently. He referred to the ENVA-3 study, which sent 132 labs five coded viral samples, asked them to rate drugs as active, partially active, or inactive against the isolates, then asked for an interpretation.71 For a virus with the 90M protease mutation, for example, 70 percent of the labs advised that nelfinavir would prove inactive, while 22 percent predicted it would be partially active. For someone with a history of a few PI failures, that call could be important.
And that honest difference in interpretation does not even touch the issue of quality assurance. Plenty of labs, ENVA-3 showed, just make the wrong call, especially when sizing up an isolate that's part mutant and part wild type. How can you tell whether your lab is up to snuff? Kuritzkes had some practical suggestions: Ask what quality assurance panels the lab participates in, and ask for their last quality assurance scores. Ask what they do to prevent cross-contamination of isolates.
But, for all its complexity, who would do without resistance testing? Haubrich predicted that the utility of both genotyping and phenotyping will improve as interpretation algorithms gain precision. And, despite difficulties in defining clinical cutoffs, he believes large, ongoing efforts to tie genotype and phenotype to virologic outcome will sharpen interpretation of these tests.
Barbara McGovern (New England Medical Center, Boston) quoted estimates that one third of people with HIV infection also carry HCV. The two viruses share a raft of similarities:
But there's one big difference between HCV infection and HIV -- the first can be cured, and the second can't. Among people in whom treatment renders HCV undetectable for six months, few suffer relapses. The same clearly cannot be said for HIV. Why? McGovern proposed three answers:
HIV complicates the natural history of HCV, and vice versa. For starters, McGovern said, advanced HIV infection can make HCV harder to diagnose. False-positive ELISAs for HCV become much more common at CD4+ counts below 100 cells/mm3. About 91 percent of coinfected people have HCV detectable in blood, compared with 76 percent infected only with the hepatitis virus. In people with HCV infection who then pick up HIV, the HCV load tends to keep climbing. But people carrying only HCV typically reach a viral set point.72
HIV treatment also unleashes HCV in coinfected people. Sandro Vento proposed three possible explanations.73
The dangerous liaison between these two viruses suggests worse morbidity and mortality in people who harbor both. Some studies endorse that suggestion; others don't. McGovern and colleagues reviewed causes of 84 deaths in HIV-infected cohorts in 1991, 1996, and 1998-1999.74 Most people in those groups -- 55, 75, and 77 percent, respectively -- injected illicit drugs, and most had HCV infection. Alcohol abuse rose from 35 percent and 36 percent in 1991 and 1996 to 73 percent in 1998-1999. So several factors weighed on the morbidity and mortality trends McGovern found.
Deaths from end-stage liver disease accounted for 11 percent of all deaths in 1991. That rate inched up to 14 percent in 1996, then exploded to 50 percent in 1998-1999. More than half who died with end-stage liver disease had an undetectable HIV load or a CD4+ count above 200 cells/mm3 within six months of death. But those hopeful harbingers changed in a good portion of the 1998-1999 cohort, one third of whom had to suspend antiretroviral therapy because of liver toxicity.
A chart review of 263 people with HIV alone, 60 with HCV alone, and 166 with both viruses confirmed the burden of coinfection.75 Fourteen in the coinfected group (8 percent) had decompensated liver disease, but nobody in the other two groups did. Researchers counted 19 deaths (11 percent) among coinfected people, 18 (7 percent) in the HIV-only group, and none in the HCV-only group. Of the 19 deaths among people with HIV plus HCV, nine (47 percent) involved liver disease, compared with none of the deaths in the HIV-only group.
But a large retrospective study found low and stable rates of end-stage liver disease in coinfected populations surveyed in 1995 and 1997.76 And these were huge cohorts -- 17,000 people in 1995 and 26,000 people in 1997. So McGovern concluded that questions of morbidity and mortality in coinfected people remain open.
Liver fibrosis rates are higher in men, people who drink alcohol, and people who are older when infected with HCV. In coinfected people, fibrosis progression correlates with a CD4+ count below 200 cells/mm3.77 In that chart review of 63 coinfected people taking a PI and 119 not taking a PI, liver fibrosis scores were significantly higher in the non-PI group (P = 0.03). But McGovern wondered whether PI therapy, averaging only 14 months, had lasted long enough to account for that difference.
Data on how HCV affects HIV disease are also conflicting, McGovern reported. But at least one study suggests that HIV progression is worse in people coinfected with HCV. In a Swiss HIV Cohort Study involving 3,111 people with HIV, including 1,157 coinfected with HCV, researchers independently linked coinfection and active injection drug use with a higher probability of a new AIDS diagnosis or death.78 Because HCV coinfection also correlated with a lower CD4+ recovery after starting antiretroviral therapy, the Swiss team concluded that "HCV and active intravenous drug use could be important factors in the morbidity and mortality among HIV-1-infected patients, possibly through impaired CD4-cell recovery."
Although some aspects of HIV/HCV coinfection remain debatable, the US Public Health Service spelled out a few clear management guidelines, outlined by Stuart Ray (Johns Hopkins University School of Medicine, Baltimore):
Ray agreed with Renslow Sherer's observation that getting people to give up alcohol "could be the greatest contribution" a clinician makes to HIV/HCV management. Session chair Maurizio Bonacini went so far as to muse that complete abstention may help the liver as much as pegylated interferon (peg-IFN). He asks people beginning HCV therapy to disavow alcohol for six months. If they can't make that commitment, he said, they are unlikely to follow other advice.
McGovern advised delaying HBV vaccination in HBV-negative people with a CD4+ count below 200 cells/mm3 until antiretrovirals boost the count. But Ray urged attendees to set a vaccination deadline with such people -- perhaps six or 12 months after starting antiretrovirals -- and then to vaccinate regardless of CD4+ count. Otherwise, he maintained, the temptation to procrastinate could rob these people of whatever benefit they may derive from the HBV vaccine.
Clinicians familiar with the link between high HIV loads and progression to AIDS must realize, Ray stressed, that HCV RNA levels do not predict prognosis of liver disease. Nor do alanine aminotransferase (ALT) elevations. Thus he emphasized the value of liver biopsy in HCV-infected people. By determining the level of fibrosis, biopsy offers the same prognostic vantage that CD4+ counts do for HIV infection. Besides guiding decisions on HCV therapy, a biopsy may also detect other diseases. He cautioned, though, that the liver is not a "well-mixed" organ, so a biopsy can miss fibrosis. To increase the odds that the sample was good, make sure the pathologist could see at least five portal tracts. Ray added that he does not biopsy people with HCV type 2 or 3; he just begins peg-IFN therapy because they respond at much higher rates than people with type 1 HCV. If they don't respond, he then gets a biopsy to guide further management.
When should you treat HCV infection in people with HIV? Ray proposed these guidelines:
This advice may well change, he added, if more tolerable drugs such as HCV protease inhibitors prove effective. If they do, it may make sense to treat everyone infected with HCV.
Maurizio Bonacini disagreed only with Ray's last guideline, arguing that some people with decompensated liver disease can tolerate interferon if treatment begins with a low dose -- 0.5 µg/kg. If that proves tolerable, he doubles the dose. Bonacini believes a trial should be mounted to test interferon in this population.
Ray advised checking the HCV load 12 weeks after treatment starts. A good response is unlikely if HCV RNA has fallen less than 1 log. In that case continued treatment should be weighed against the burden of toxicity and biopsy-predicted disease progression. After 24 weeks of treatment, fewer than 5 percent of people with a still-detectable HCV load will respond. Maintenance therapy might be considered for some people whose HCV RNA remains detectable.
If people responded to standard interferon then suffer a relapse, Ray doesn't hesitate to try peg-IFN plus ribavirin. The response rate may be higher in relapsers than in all peg-IFN candidates, he said, because they've already demonstrated a response to interferon. Bonacini estimated a 30 percent response rate to peg-IFN after relapse following a course of standard interferon.
Twenty-four-week results of an ongoing 48-week study appear to confirm the superiority of peg-IFN plus ribavirin over standard interferon in people coinfected with HCV and HIV.79 Ray cautioned, though, that ACTG 5071 has another 24-weeks to go and that people are on treatment during the trial. When they stop therapy, some will have relapses.
This multicenter study randomized 134 people to take 180 µg of peg-IFN weekly or standard-dose interferon. Everyone also took ribavirin, starting with 600 mg daily then escalating to 1g daily. All study participants had abnormal liver histology, including compensated cirrhosis. They could be antiretroviral naive with a CD4+ count above 300 cells/mm3 or could be taking a stable antiretroviral regimen with a CD4+ count above 100 cells/mm3 and an HIV load below 10,000 copies/mL.
At study entry the groups did not differ in age (44 years for peg-IFN versus 45 years for interferon), median CD4+ count (452 versus 500 cells/mm3), percent with HIV RNA below 500 copies/mL (60 versus 58 percent), or percent taking antiretrovirals (90 versus 87 percent). Nor did the groups differ in HCV load, percent with HCV genotype 1, fibrosis score, or ALT elevations. At treatment week 24, peg-IFN proved virologically superior regardless of HCV genotype (Table 7). Median CD4+ counts fell in both treatment groups, by 83 cells/mm3 in the interferon group and by 137 cells/mm3 in the peg-IFN group. But CD4+ percents rose in both groups. Neither treatment affected control of HIV.
|Table 7: Peg-IFN Versus Standard Interferon for HCV in People With HIV|
|Interferon + Ribavirin||Peg-IFN + Ribavirin||P|
|Overall HCV RNA <60 IU*||10 (15%)||29 (44%)||0.0003|
|Genotype 1 HCV RNA <60 IU**||4 of 52 (8%)||17 of 51 (33%)||0.0014|
|Non-1 genotype HCV RNA <60 IU**||6 of 15 (40%)||12 of 15 (80%)||0.06|
** Response with genotype 1 versus non-1, P <0.0001
Source: ACTG 507179
Four factors predicted an HCV load below 60 IU in a multivariate analysis: treatment with peg-IFN (P P = 0.016), a Karnofsky performance score of 100 (P = 0.046), and a fibrosis score of 0 to 2 on a scale of 0 to 6 (P = 0.021). People taking peg-IFN had significantly more grade 4 lab abnormalities (17 versus 4, P = 0.0012), but people tolerated both regimens well. Eight in each group dropped out before week 24, a low rate for an interferon/ribavirin trial that may reflect the dose escalation of ribavirin.
Ray advised trying nonsteroidal anti-inflammatories and fluids for the fevers and aches that accompany interferon therapy, and selective serotonin reuptake inhibitors (SSRIs) for the depression. He starts SSRI therapy along with interferon in people with a history of alcoholism or a personal or family history of depression. In others he starts an SSRI only if they become depressed while taking interferon. Early results of studies of G-CSF for interferon-associated neutropenia and of erythropoietin for ribavirin-induced anemia look promising.
A study of 5,292 gay men grouped by HIV and HBV serostatus charted a climbing rate of liver-related mortality dating from the dawn of PI therapy in 1996.83 In the entire cohort liver disease accounted for 2.5 deaths per 1,000 person-years before 1996 and 4.0 deaths per 1,000 person-years afterwards (P = 0.16). The rate of liver-related mortality per 1,000 person-years proved highest in men infected with both viruses (14.2, P <0.0001 compared with men carrying neither virus). That rate measured 1.7 in men infected only with HIV (P P = 0.04). Coinfected men with a CD4+ count below 100 cells/mm3 had an 11.6 times higher risk of liver-related death than men with higher T-cell counts.
For people with chronic HBV infection, the treatment options are interferon and antivirals. Stuart Ray recommended a dose of 5 million units daily or 10 million units three times weekly with interferon-alpha 2a or 2b. Therapy should last four to six months in people positive for hepatitis B early antigen (HBeAg) and more than 12 months in HBeAg-negative people. A survey of 15 controlled trials of interferon given for four to six months in HBeAg-positive people calculated a sustained HBV DNA response in 37 percent versus 17 percent in control arms.84 Respective sustained responses measured by HBeAg were 33 versus 12 percent, and by hepatitis B surface antigen (HBsAg) 8 versus 1 percent. In four controlled trials of interferon for six to 12 months in HBeAg-negative people, interferon-treated people consistently had higher sustained response rates than comparison groups when measured by HBV DNA, ALT, both, or HBsAg.85
Lamivudine at a dose of 100 mg daily in people without HIV infection has yielded 4-log drops in HBV DNA. (Of course the standard 150-mg twice-daily dose should be used in coinfected people.) DNA responses are similar in HBeAg-positive and -negative people. HIV clinicians don't have to be told the problem with lamivudine monotherapy. In one study of 58 people without HIV infection, 39 had resistance-driven virologic breakthroughs after one year of lamivudine at the 100-mg HBV dose.86
Although adefovir did not make the grade as an antiretroviral, it has done well in HBV trials, including studies of HIV-coinfected people with lamivudine-resistant virus. In one study of 35 coinfected people with lamivudine-resistant HBV while taking the nucleoside as an antiretroviral, adding adefovir at 10 mg daily chopped 4.8 log copies of HBV DNA off the viral load after 72 weeks of treatment.87 Fibrosis scores also fell, and no nephrotoxicity emerged. Tenofovir, adefovir's nucleotide cousin, took 4.6 logs off the HBV DNA load in 12 HBeAg-positive people with HIV infection.88 The drug worked equally well in people with and without lamivudine-resistant HBV when starting 300 mg of tenofovir daily, and no new mutants emerged.
Ray concluded that the goal of HBV therapy should be DNA suppression, and the means should be peg-IFN or combination antiviral therapy with lamivudine plus adefovir or tenofovir. He recommended delaying therapy, if possible, in people with an ALT less than two times the upper limit of normal, with or without active histology. HBeAg clearance, he explained, is low in people with low ALTs.89
Diane Havlir, referee of the antiretroviral strategy session, put TIs on the table in a brief traversal of controversial issues. She proposed six risks that accompany treatment breaks:
But Havlir did not propose closing the book on TI trials. Though some question whether the every-other-week tactic studied by Mark Dybul90 can work in clinics with typical three-month follow-up visits, Havlir argued that it's "premature to say we could never get patients" to stick with an every-other-week schedule.
Pablo Tebas noted, however, that even in Dybul's tightly controlled study of 10 people, two dropped out and another had a virologic breakthrough (later controlled) when he skipped treatment for 10 days instead of seven. Without much mental strain one can imagine how much faster lost follow-ups and breakthroughs might accrue if trying this strategy in the clinic.
Of course most people who take TIs -- with or without their physician's consent -- stop drugs for much longer than seven days. So they risk not only a nearly certain viral rebound, but also a steep slide in CD4+ cells. Tebas recalled his own retrospective study of 72 people with undetectable viremia who interrupted treatment for at least 12 weeks.91 They lost an average 16 CD4+ cells/mm3 monthly, and the only factor that predicted their CD4+ loss was how many T cells they gained with treatment. In other words, Tebas explained, most people tended to sink back to their pretreatment CD4+ nadir. So a CD4+ count of 600 cells/mm3 after successful therapy does not equal a 600-cell count in someone who never took antiretrovirals.
Franco Lori agreed that there's something "magic and tragic" about CD4 and RNA set points. In his yet-to-be-presented study of 123 people taking continuous therapy or switching off and on monthly, most virologic failures (<200 RNA copies/mL) in the switch group came in those with a CD4+ nadir below 50 cells/mm3. People with low nadirs, Lori elaborated, returned to them when they interrupted treatment. And people with high baseline viral loads quickly rebounded to those pretreatment levels. He read this seeking of the set point as an argument to begin antiretrovirals sooner.
Lori's study, like Dybul's, reckoned a significant improvement in total cholesterol in the TI group (P = 0.01). But he did not repeat Dybul's finding of significantly improved triglycerides.
Kimberly Smith echoed many clinical researchers' concerns about drug holidays, arguing that, "We're tinkering around with these people without really knowing what we're doing." But more than one clinician delegate maintained that many people now take drug breaks whether their doctors agree or not. Ignoring that reality instead of engaging patients on the question, they said, helps no one.
And covert break-taking tactics can be surprisingly wily. A clinician practicing in a Brooklyn public hospital discovered some people simultaneously testing their virologic control, their clinician's savvy, and their own luck. They started with a two-day TI before their next checkup. If the RNA assay counted no virions, they pushed the TI to seven days before the next checkup, then to one month, then to two -- always with a certain outcome.
Mark Mascolini writes about HIV infection (firstname.lastname@example.org).
Back to the July 2002 issue of IAPAC Monthly.