Keeping an Eye on HIV Treatment: From Substance Abuse to Side Effects
Poor William Stewart, Surgeon General of the United States in Lyndon Johnson's White House. Dr. Stewart had the misfortune to make this brave pronouncement in 1967:
The time has come to close the books on infectious diseases. We have basically wiped out infection in the United States.
Other equally luminous visionaries made similar predictions at the time, but Stewart's words were the ones remembered by Charles Hicks (Duke University, Durham, North Carolina) when he opened the IAPAC Sessions 2003, a two-day by-invitation meeting of veteran HIV clinicians -- on both sides of the speaker's lectern.
Within a year of Stewart's ill-timed avowal, researchers now calculate,1 HIV-1 had already crossed the US border and -- almost immediately -- began its exponential spread. Even when clinicians in Los Angeles and elsewhere started noticing symptoms of this new infection, its gravity escaped many. Hicks noted, for example, that the 1981 MMWR issue announcing "Kaposi's sarcoma and Pneumocystis pneumonia among homosexual men" ran on an inside page.2 The lead article explored the menace of "Dengue type 4 infection in US travelers to the Caribbean."
These days, on the contrary, a worried citizenry can track new infectious diseases on the first page of the local newspaper -- even those that pose a slimmer threat than the plucky retrovirus. IAPAC Sessions delegates, most of whom have treated HIV-infected people since pre-zidovudine days, well understand the vicissitudes of public attention. They know that sometimes too much attention can distract as much as too little attention can dismay. All recognized the legendary Newsweek cover that Hicks resurrected via PowerPoint. As the protease inhibitor (PI) era dawned, Newsweek asked if we could anticipate "The End of AIDS?" Those three-drug combos certainly changed the face of the epidemic, but at a price. The Newsweek cover suggests part of that price -- side effects, intolerance, poor adherence -- in the appealing regimen depicted. Hicks observed that it included two big favorites, full-dose ritonavir and chewable didanosine (ddI).
Happily, highly active antiretroviral therapy (HAART) has changed for the better. Hicks traced the accelerated rate of approval -- one drug in the 1980s, five from 1991 to 1995, nine from 1996 to 2000, three already since 2001, and more nearing the pipeline's end. As a result, intent-to-treat analyses of some recent trials show 48-week sub-50-copy rates exceeding 80 percent.
Of course many people with HIV, even in wealthy countries, don't share in that success story. As one delegate noted, while drawing knowing nods from colleagues, "I have patients whose list of drugs taken is the list of drugs available." But even people starting antiretroviral therapy now, and starting before getting an AIDS-defining disease, face more than one obstacle with these multidrug regimens. IAPAC Sessions planners framed the meeting around several of these obstacles -- coexisting disease and substance abuse, risks and benefits of simpler regimens, poor adherence, variable pharmacokinetics, resistance, and metabolic side effects.
While antiretroviral therapy continues to evolve, some things don't change. Working in Newark with poor and often indigent people, Patricia Kloser has seen more than one "triple-diagnosed" person -- with hepatitis, mental illness, and substance abuse on top of HIV. So she often diagnoses HIV infection the same way clinicians did in the early 1980s -- by diagnosing an AIDS-defining disease first. Just in the past week, she noted, one man sought care with Pneumocystis carinii pneumonia (PCP) and a CD4 count of 50 cells/mm3, and another came in with central nervous system (CNS) lymphoma.
That happens everywhere, not just in Newark. But it happens more in poor inner cities rife with unemployment, homelessness, crime, substance abuse, high resident turnover, and lack of social contact. All of those problems may contribute to a disorder that routinely complicates management of HIV infection -- mental illness.
The major mental illness Kloser faces in her HIV clinic is depression, which has its seeds not only in HIV infection, but also possibly in cocaine use or alcoholism. Depression remains highly underdiagnosed in people with and without HIV. Kloser proposed the symptom clues appearing in Table 1, adding that a blinkered focus on CD4 counts and viral loads can make clinicians overlook depression.
The clinician should also observe whether the patient has psychomotor retardation.
Although depression is the most common mental health problem in people with HIV, patients may also have underlying psychoses such as schizophrenia or bipolar disorder. Kloser mentioned one person with apparent schizophrenia who claimed a voice was telling her not to take her antiretrovirals. People with bipolar disorder typically fail to keep appointments when at either extreme of their mood swings.
Organic diseases that may cause or contribute to mental problems in HIV-infected people include CNS atrophy, progressive multifocal leukoencephalopathy, toxoplasmosis, cryptococcosis, HIV encephalopathy, and dementia. Kloser stressed the importance of screening for such diseases as well as for metabolic abnormalities.
HIV dementia manifests itself in three spheres. Clues pointing to these problems become easier to spot when a clinician gets to know a patient better:
HAART itself may rouse neuropsychiatric nettles, including headache, insomnia, neuropathy, nightmares, and -- again -- depression. In turn, depression and other mental illnesses can promote poor adherence.
Kloser proposed a checklist for managing mental health problems in people with HIV infection:
Kloser emphasized that treatment of mental illness should be planned in consultation with a specialist. But primary clinicians may have to write some prescriptions for psychotropic agents to manage acute problems before a psychiatric consult can be arranged.
The "big three" abused drugs are cocaine, heroin (often both), and benzodiazepines. A related problem is alcoholism, which Kloser notes especially among the women for whom she cares.
In the United States, cocaine caused 175,000 emergency room visits in 2000, 57 percent of which led to admissions. Kloser spelled out these symptoms of cocaine use:
Cocaine has dangerous cardiovascular effects -- myocarditis, atherosclerosis, cardiomyopathy, ischemia, and arrhythmia. In Kloser's cachement, cocaine abuse is most common among black males, but she notes that female abusers have gravitated toward the drug because it doesn't have to be injected. Most cocaine users also smoke tobacco.
In the Newark area heroin preys on the growing population of immigrants, who often share needles. Heroin has infectious complications -- cellulitis, sepsis, endocarditis, and osteomyelitis -- and may cause nephropathy, which often comes on suddenly in people with HIV. Dialysis-dependent nephropathy may lead to sepsis resulting from poor shunt care and even illicit use of shunts. Hypertension frequently accompanies nephropathy.
Severe depression, delusions, and paranoia may signal heroin withdrawal. Although propranolol and the antidepressants desipramine and bupropion may help in the acute phases of withdrawal, there is no good treatment. Buprenorphine or methadone may also ease heroin withdrawal, and clonidine may diminish the severity of symptoms. Rapid and ultrarapid detoxification programs use a variety of medications and naloxone-induced withdrawal under anesthesia or heavy sedation. Inpatient detox programs have proved more successful than outpatient programs.
No one argues that the key to survival with HIV infection is HAART, Kloser noted. But advanced liver disease makes HAART intolerable. Especially in the form of hepatitis virus infection, liver disease has become a subepidemic among people with HIV infection, and especially among substance abusers who share needles. Kloser considered four types of hepatitis:
Toxic, drug-induced hepatitis is often idiosyncratic and unpredictable but not dose dependent. Often caused by drug metabolites, toxic hepatitis manifests itself differently from person to person and can be confused with infectious hepatitis. Kloser ticked off a laundry list of hepatotoxic agents taken by HIV-infected people (Table 23, 4). One could add the antiretrovirals zidovudine (AZT), ddI (especially when combined with stavudine [d4T]), nevirapine, indinavir, and atazanavir. Because isoniazid and rifampin assault the liver, Kloser added, liver function test variables are almost always elevated in people taking those agents for tuberculosis. But the clinical side effects of those drugs vary.
Kloser noted that 15 to 20 percent of people in general primary care practice report alcohol-related health problems. The CAGE questions, she said, can help identify alcoholics:
C: Can you cut down on your drinking?
Care for an alcoholic patient must be supportive and nonjudgmental, yet assertive. Kloser finds that self-help groups alone are not enough to help people stop drinking. Clinicians caring for an alcoholic patient should be prepared to treat withdrawal symptoms (delirium, tremors, seizures) and to order inpatient treatment for severe depression, suicidal ideation, or psychotic symptoms. Benzodiazepines, carbamazepine, and valproate can promote seizure-free withdrawal.
Hepatitis A affects 40 to 70 percent of US residents with HIV infection by age 30. Markers of hepatitis B virus turn up in 60 to 80 percent of injecting drug users. Kloser stressed the urgency of preventing both by vaccination. Between 150,000 and 300,000 US residents with HIV infection also have hepatitis C virus (HCV). HCV may be a marker for:
More than three alcoholic drinks daily speeds progression of liver disease. Yet in one Veterans Administration study, 30 percent of people with HCV infection also drank alcohol.5 Every effort should be made to help HCV-infected drinkers stop.
Recent outbreaks of hepatitis D infection have been noted among drug abusers, with a fatality rate of 5 percent. When people infected with hepatitis B also become infected with hepatitis D, the fatality rate climbs to 20 percent. Hepatitis E is less common in the United States but is a concern in developing countries.
Liver transplants have proved successful in carefully selected people with HIV. Selection criteria include stable antiretroviral therapy, a CD4 count above 200 cells/mm3, and a low or undetectable viral load. "Ten years ago," Kloser confessed, "I wouldn't have dreamed this would be possible."
What antiretroviral regimens does Kloser prefer for people burdened by mental illness, substance abuse, and/or hepatitis: a simpler, more tolerable nonnucleoside (NNRTI) combination, or a PI regimen with a higher barrier to resistance? She often opts for a simpler regimen, but first, Kloser added, "you have to teach that person how to be a patient." Because this could well be a patient's first contact with healthcare, she may try to establish adherence by prescribing a vitamin or PCP prophylaxis and monitoring carefully. Ironically, she noted, it may be easier to explain the importance of adherence to a drug user who understands the grave consequences of missing a day's dose.
If antiretroviral therapy must begin before adherence can be established, Kloser leans toward Trizivir, the three-in-one pill combining AZT, lamivudine (3TC), and abacavir. Besides being the simplest of regimens, its failure -- which usually starts with resistance to 3TC -- leaves an array of backup options.
Trizivir suffered a recent setback, though, in a randomized comparison with efavirenz regimens. And preliminary analysis of that study left much grist to mill during Charles Steinberg's presentation on simplified antiretroviral regimens.
Is simpler better? Charles Steinberg had a simple answer: Yes.
At least we humans intuitively sense that simpler almost always works better in the long run. Just look at antiretroviral therapy. The granddaddy of all HAARTs -- AZT, 3TC, and indinavir -- raised eyebrows and dropped jaws in the Merck 035 study, convincing all that these new triple therapies handily surpassed single and dual nucleosides in potency. But the 035 regimen wasn't exactly easy to take. The ideal patient, Steinberg reminded colleagues, took the combo in nine discrete steps:
Who could have guessed that less than a decade later easy once- and twice-daily regimens -- with no dietary or aqueous codicils -- would abound? It's true that planning therapy remains tough for the prescriber, but fewer pills, scarcer side effects, and convenient coformulations can make treatment much simpler for the patient (Table 3).
Those considerations made Steinberg modify his "yes" to a "yes, but."
While simpler may always be easier, it's not always safer. Steinberg referred colleagues to results of the 2NN study, which compared once-daily nevirapine, twice-daily nevirapine, once-daily efavirenz, and efavirenz/nevirapine.6 After 48 weeks of treatment, liver function tests were elevated in 13.2 percent taking once-daily nevirapine versus 7.8 percent taking the standard twice-a-day dose. The higher rate with once-daily dosing could reflect the higher peak concentrations when a person swallows 400 mg of nevirapine all at once.
But once-a-day strategies with twice-a-day drugs don't always turn out bad. Take, for example, once-daily extended-release d4T (d4T-XR) compared with the twice-daily immediate-release formulation. The peak concentration with twice-daily dosing (694 ng/mL) more than doubles that with the XR version (339 ng/mL). In the clinic that difference has translated into lower reported rates of peripheral neuropathy, hyperlactatemia, pancreatitis, and lipodystrophy.
Most antiretrovirals favored in simpler regimens, however, have familiar and -- for some -- inescapable side effects:
A nagging concern with once-a-day dosing is its possibly unforgiving nature. Forgiveness has become the term favored to explain a drug's ability to hang around long enough, at levels high enough, to prevent emergence of resistant virus even if a dose is missed. When people miss a once-daily dose, they could wind up going 48 hours without sending any drug cellward.
But Steinberg stressed that the half-lives of many once-daily drugs are probably long enough to "forgive" a missed dose: Efavirenz, nevirapine, tenofovir, and ddI-EC fall into that group. Other drugs that already -- or may later -- do once-daily duty have shorter half-lives: 3TC and abacavir, for example. With nucleosides and the first nucleotide, Steinberg reminded delegates, what matters is not the plasma half-life but the intracellular half-life. By that measure ddI-EC, tenofovir, and the investigational nucleoside emtricitabine (FTC) all look like friendly forgivers (Figure 1). Again, 3TC and abacavir do not. A US Food and Drug Administration (FDA) sanctioned once-daily boosted PI, amprenavir/ritonavir (1,200/200 mg) also sustains high levels through 48 hours.
This five-country study also found that the number of pills in a regimen determines whether people will take them all in one sitting: Whereas 92 percent said they would take three pills once a day and 84 percent said they would take four pills, only 59 percent would agree to swallow six pills all at once and 38 percent eight pills. An unpublished 265-person US survey confirmed those trends: 73 percent liked the idea of downing four pills once a day. In comparison, only 18 percent favored an easy-sounding one pill in the morning and two at night.
Cutting down pill counts or simplifying schedules would address several problems that people cite when asked why they miss doses. Steinberg showed the results of one survey of 133 people who gave 16 reasons for missing a dose.8 Taking fewer pills less often might solve seven of these problems, including the three most frequent (Table 4).
How simple is it to build a simple regimen? Pretty simple, nowadays, even when thinking about second- or third-line regimens.
The nucleoside/nucleotide wedge of a once-daily pie can now draw on four drugs:
FTC may soon be another nucleoside ingredient. The NNRTI-PI piece already includes:
Long-range planning with a treatment-naive person could even yield a scheme for a not-too-daunting three-regimen sequence, Steinberg suggested, referring to a presentation by Graeme Moyle (Chelsea and Westminster Hospital, London) at the 2002 Glasgow meeting:
But there's a problem with this scenario, one delegate noticed. While all these regimens and many other possibilities look good on paper, few have ventured through the clinical trial gauntlet. This is no quibble, as an interim analysis of ACTG 5095 shows. The trial randomized treatment-naive people to one of three arms: Trizivir, Trizivir plus efavirenz, or Combivir (AZT/3TC) plus efavirenz. The study's independent review board stopped enrollment into the Trizivir-only arm when a higher proportion of people in that group (21 percent) than in the combined efavirenz arms (10 percent) reached a protocol-defined virologic failure after an average 32 weeks.13 Time to virologic failure proved significantly shorter in the Trizivir arm (p < 0.001). And among 33 percent of people who reached study week 48, 74 percent taking Trizivir alone versus 89 percent in the efavirenz arms had a viral load below 200 copies/mL. Trizivir failed in equivalent proportions with baseline loads above and below 100,000 copies/mL.
Is simpler really worse when it comes to this one-pill-twice-daily regimen? The British HIV Association thinks so, suggesting in its 2003 draft guidelines that clinicians avoid first-line Trizivir.14 But several clinicians at the IAPAC Sessions think it's too early to tell. Steinberg observed that blinding the ACTG 5095 trial stripped Trizivir of its simplicity because people in the Trizivir-only arm had to take five dummy pills along with their two Trizivir tabs.
One delegate argued that clinicians shouldn't "junk the whole concept" of first-line Trizivir until the ACTG presents a fuller 5095 analysis. Will we learn something by finding out who endured the Trizivir failures? Do they represent some subgroup in whom first-line Trizivir may indeed pose a higher risk of failure? IAPAC Sessions Co-Chair Renslow Sherer (University of Chicago Hospitals) added that the preliminary response rate in the solo Trizivir arm wasn't that awful -- 74 percent under 200 copies at 48 weeks in an intent-to-treat analysis. For now, Sherer said, he will continue to prescribe up-front Trizivir for selected individuals, for example, someone who asks for a simple regimen, has a viral load under 100,000 copies/mL, and has some reason not to start with an NNRTI.
To balance his summary of why simpler drug medleys may not be better, Steinberg summed up several reasons why they may:
Successful adherence, some say, means taking at least 80 percent of doses. But Judith Feinberg observed that HIV medicine inherited this benchmark from the antihypertensive literature and that "it may have nothing to do with anti-infective prescribing." To complicate things, this notion of successful adherence paid no special attention to intervals between doses or dietary restrictions. "Eighty percent adherence," Feinberg said, "is not going to get you below 50 copies."
So what is successful adherence for HAART? Two studies suggest that it lies somewhere above 90 percent of all doses, but accurate estimates probably vary with the regimen. The classic study of antiretroviral adherence by David Paterson and colleagues found that 80 percent of people who were more than 95 percent adherent to a PI regimen had a viral load under 400 copies/mL after three months of therapy.15 Only about 60 percent of people who had a 90 to 95 percent adherence score reached that level of suppression. Some have observed that Paterson's study, conducted from 1997 to 1999, evaluated unboosted PIs and not the more user-friendly -- and often more forgiving -- boosted PI, nonnucleoside, and Trizivir regimens of today.
A later study that compared Combivir/abacavir with Combivir/indinavir found that almost everyone with 90 to 95 percent adherence to the triple-nuke regimen had a 48-week viral load below 400 copies/mL.16 But only about 65 percent of people with that level of adherence to Combivir/indinavir notched a sub-400 viral load. When adherence exceeded 95 percent, the two groups achieved equivalent levels of viral suppression. A second recent study also found that only near-perfect adherence -- better than 95 percent -- correlates with durable suppression of viremia below 50 copies/mL.17 Worse adherence raised the risk of virologic breakthrough.
An ACTG survey confirmed what seasoned HIV clinicians know about why people skip antiretroviral doses:
A survey of 196 women charted four key reasons for poor adherence: 57 percent forgot, 39 percent complained of side effects, 22 percent "felt well," and 20 percent suspected their antiretrovirals weren't working.18 Although the just-noted study of Combivir/abacavir versus Combivir/indinavir found better adherence to the all-nucleoside regimen,16 the reasons for poor adherence were similar with both regimens: forgetting, being out in public, and side effects.
No one can predict with certainty who will adhere well to antiretroviral combos and who will not. During a discussion, Richard Elion (George Washington University, Washington, DC) noted that four of eight gay white men enrolled in a treatment support group attended sessions for a full year before confessing their slipshod adherence. On the other hand Feinberg cited the case of a homeless woman who wanted to enroll in an antiretroviral trial. When Feinberg asked how she'd manage to take her drugs on time, she answered, "I could if I had a watch." Feinberg bought her a watch and she had the best drug levels in the group. Still, Feinberg proposed that certain variables often work as positive or negative predictors of adherence, while others are not predictive (Table 5).
Sometimes bad adherence, Feinberg suspects, reflects ineradicable guilt "about whatever they did to get HIV." They cope with that guilt by seeking punishment -- and a certain way to do that is to shun their medications. She offered an anecdote of a 27-year-old man who rarely took his antiretrovirals and died with a highly treatable 3TC-resistant virus. At the other end of the spectrum sit people who are perfectly content with their every-eight-hour indinavir and wouldn't dream of tinkering that into a twice-daily ritonavir boost. Despite such intangibles, Feinberg urged colleagues to flip through their mental checklist of factors that may affect adherence whenever planning therapy or addressing potential poor adherence (Table 6).
Beyond the burden of side effects, Feinberg noted, a certain percentage of people invariably misunderstands dosing directions. And that percentage grows as treatment continues. In an observational cohort of 289 women, researchers found that 80 percent started off understanding food restrictions, 75 percent understood dosing frequency, and only 63 percent understood both.19 The proportion of women who understood dosing frequency fell with time. Women who correctly understood dosing and were taking once- or twice-daily regimens with no food restrictions had a 60 percent lower chance of skipping doses in the preceding three days than did women taking more complex regimens.
Feinberg offered two bedrock dictates of adherence-promoting prescribing:
How can a clinician put these dictates into action? Before starting antiretroviral therapy, Feinberg suggests four steps:
Adherence tools -- pill boxes, medication charts, calendar stickers, timers -- can help some people. Of these, Feinberg finds that pill boxes work best for her patients. Finally, clinicians must remember that adherence counseling only begins before the first dose. It must continue throughout therapy, as part of regular monitoring.
When Feinberg feels adherence may be difficult for someone, she addresses potential problems up front. Besides recognizing and addressing mental illness and drug use (see the presentation by Patricia Kloser), Feinberg tries to help people overcome fundamental obstacles like transportation, housing, and psychosocial support. IAPAC Sessions Co-Chair Renslow Sherer believes his team increases retention in the clinic about 20 percent by providing transportation. If that's not possible, he suggests trying the tactic Paul Farmer (Harvard University) uses in Haiti -- recruiting a family member or neighbor to visit people taking antiretrovirals and encourage full adherence.
Like Farmer, Feinberg enlists family or friends in the effort. She asks people starting antiretrovirals to bring a friend or family member to pretreatment discussions about therapy. Under less stress than the patient, that person often remembers more when treatment begins. If some type of time-limited directly observed therapy (DOT) is feasible, Feinberg thinks it can get some people off on the right foot.
Each patient's regimen should be selected for optimal convenience, simplicity, tolerability, and potency. For people starting therapy, Feinberg sticks to once- or twice-daily combinations. She believes, though, that an advantage for once-daily dosing over twice-daily dosing remains uncertain. Data comparing the efficacy of once- versus twice-daily regimens are sparse, while limited data link once-a-day regimens with better adherence. The "forgiveness" of once-daily combinations is not well understood. (See the presentation by Charles Steinberg for more on adherence and forgiveness with once-daily therapy.) Pharmacokinetic considerations such as interpatient variability, intracellular concentrations, and drug-drug interactions may have a greater impact on once-a-day combos.
Feinberg suggests a two-part approach when adherence goes awry despite the best planning:
When resources allow, other stratagems can help resolve adherence problems:
Feinberg summed up her insights on adherence with this advice:
For further adherence advice, Feinberg listed some online sources.20
Adherence matters, said Richard Elion, because today's antiretrovirals can achieve close to 100 percent virologic success in treatment-naive people. Evidence to support this claim comes from a study comparing HIV-infected prisoners taking directly observed HAART and people taking HAART on their own in ACTG trials.21 Everyone in both groups was treatment naive. After 88 weeks all of the prisoners had a viral load below 400 copies/mL, compared with 80 percent in the ACTG trial group.
If the drugs work so well, why are virologic success rates among first-time HAART takers so much lower in clinical practice? Elion gave two reasons: antiretroviral regimens are demanding, and adherence is poor. Citing Paterson's study linking increasing levels of adherence to higher rates of virologic success,15 he stressed that "midrange adherence" -- taking about 60 to 90 percent of doses on time -- is the most fertile breeding ground for drug-resistant virus.
A study comparing two adherence tactics with no intervention supports this contention. Researchers based the comparison on a simulation model using published parameter estimates; they evaluated the model monthly over a one-year simulation in 50,000 people taking antiretrovirals.22 In the no-intervention group adherence averaged 58.9 percent. But in an adherence case management (ACM) group and a modified directly observed therapy (MDOT) group, adherence rates improved to only 67.5 percent and 84.1 percent respectively. Although these higher adherence rates correlated with higher proportions reaching viral loads below 50 copies/mL (16.2 percent for no intervention, 22.4 percent for ACM, and 38.5 percent for MDOT), evolution of drug-resistant mutants proved equivalent across the three groups.
Besides raising the risk of resistance, imperfect adherence poses a public health threat. People who adhere poorly to antiretrovirals tend to adhere poorly to safer-sex guidelines, Elion said. As the prevalence of antiretroviral resistance in the United States grows, so do rates of unsafe sex and sexually transmitted diseases. Several lines of evidence suggest that better adherence may lower the risk of HIV transmission:23-27
For most people, adherence gets worse as treatment continues. Among the many reasons for this withering resolve as the months roll by, nagging side effects weigh heavily. And the most unsightly yet hardest to hide side effect is lipodystrophy. Elion made the point bluntly: "People don't like to look like they're taking antiretrovirals."
In a study of 74 people infected with HIV for more than five years, 78 percent reported some body shape change.28 While 30 percent had switched their antiretrovirals and 7 percent had stopped them because of lipodystrophy, only 7 percent vowed they would definitely not revamp their regimen because of fat changes. Among people who hadn't switched to different drugs, 57 percent said they were thinking about it, and 46 percent said they would switch if their lipodystrophy worsened. A survey of 75 people attending an HIV clinic found that 20 percent would give up four years of life to avoid lipodystrophy.29 More than 10 percent would give up five or six years.
To counteract waning adherence in people starting HAART, Elion suggested that an updated version of induction-maintenance therapy may work for some. Offer people the option of starting treatment with a PI-based regimen, explaining the advantage of a high barrier to resistance. But hold out the possibility of switching to a simpler yet still suppressive regimen when the viral load becomes undetectable.
A more structured approach to adherence education can pay virologic dividends, according to results of one randomized study involving 116 people starting HAART.30 An "experimental group" received counseling (including information about therapy and the hazards of poor adherence), an adapted medication schedule, and a training session to solve common problems. The control group got standard care. Researchers measured adherence by self-report and plasma drug levels, which mirrored patient reports in 93 percent.
Adherence averaged more than 90 percent throughout the study in the experimental group, while dwindling from 90 percent to about 70 percent in the control group. After 48 weeks 88.9 percent in the experimental group and 65.8 percent in the control group had a viral load below 400 copies/mL (P = 0.026). Among people with better than 95 percent adherence, 85 percent had a viral load below 400 copies/mL.
Another adherence tactic that worked for one group is a pre-HAART practice trial.31 The trial may also give clinicians a better fix on who will adhere well once actual therapy begins. The study involved 179 active or current drug users who took a two-week practice run with dummy antiretrovirals. Electronic bottle caps measured adherence during that time. Sixty-five of them later began antiretroviral therapy, and electronic caps again recorded pill taking for two weeks. Adherence during the practice run correlated with adherence during actual treatment (r = 0.50), though adherence itself was less than perfect -- 70 percent during practice and 75 percent during therapy.
No single adherence strategy will work for everyone, Elion cautioned, but he proposed a general approach that should always guide clinicians -- fit the antiretrovirals to the patient's lifestyle. If a regimen forces people to make many changes in their lives, he maintained, that regimen is not going to work. Factors to consider in making this fit include:
A study of 1,910 people found that their perception of how well a regimen fit their lifestyle directly affects adherence (Table 7).32 A 299-person survey in six cities determined what lifestyle "fitness factors" matter most.33 The variables, ranked by order of importance, are:
The just-cited study33 adds to the recent HIV literature indicating an adherence advantage for once-daily versus twice-daily regimens (see Steinberg and Feinberg presentations). When Elion asked meeting delegates how many think taking pills once a day improves adherence, only about 10 percent raised their hands. He argued that a much bigger percentage of patients think so.
A December 2001-January 2002 survey of 536 people with HIV found that 80 percent thought they were likely to remember all their pills if taking them once daily, while 63 percent claimed they would miss no doses in a twice-daily regimen (p < 0.001).34 When asked to rate their preferences of different dosing schedules, 68 percent opted for four pills once a day, 24 percent voted for one pill in the morning and two at night, and 5 percent preferred one pill in the morning and four in the evening (p < 0.001). A large majority of people, 73 percent, felt a once-a-day regimen would fit their daily lives better than a twice-a-day regimen (p < 0.001).
Elion also cited the 504-person study reviewed by Charles Steinberg showing that interest in once-daily regimens drops off as the pill count climbs past four.7 One convenient once-daily regimen -- ddI, 3TC, and efavirenz -- matched one twice-daily combination (AZT/3TC/efavirenz) and beat another (AZT/3TC/nelfinavir) in 12-month virologic efficacy.35 This randomized study of treatment-naive people found that about 80 percent in the once- and twice-daily efavirenz groups had an undetectable 12-month viral load versus 50 percent in the nelfinavir group (p < 0.02).
But not all once-daily drugs are created equal, as revealed in a randomized comparison of saquinavir/ritonavir (1,600/100 mg daily) and efavirenz (both with twice-daily nucleosides).36 The study involved 152 treatment-naive people with an average CD4 count around 325 cells/mm3. An intent-to-treat analysis determined that 51 percent in the saquinavir arm and 71 percent in the efavirenz arm had a 48-week viral load below 50 copies/mL. A substantially higher rate of side effects -- mostly nausea and vomiting -- explained the poorer response in the saquinavir group.
Although staunch adherence is critical to successful therapy, there are some things adherence simply cannot do. One of those things came to light in a study of 1,420 treatment-naive people beginning HAART in British Columbia from 1996 through 2000.37 Defining good adherence as taking medications more than 95 percent of the time, researchers tracked people through March 2002 to find predictors of viral load suppression and death.
In a multivariate model adjusted for adherence, a pretreatment viral load above 100,000 copies/mL remained an independent predictor of death, with an adjusted hazards ratio of 1.81. To learn why, the researchers looked for differences between adherent study participants. They found that people with a baseline viral load topping 100,000 copies/mL suppressed viral replication more slowly than those with lower loads and were significantly less likely ever to reach an undetectable level. Compared with good adherers who started therapy with fewer than 50,000 copies/mL, good adherers starting with more than 100,000 copies/mL were 73 percent less likely to see viremia vanish.
Therapeutic Drug Monitoring, or Integrating Antiretroviral Pharmacokinetics, HIV Resistance, and GeneticsPresented by Eugene D. Morse, State University of New York at Buffalo
Although many European HIV clinicians consider therapeutic drug monitoring (TDM) standard of care for certain patients, antiretroviral TDM has not caught on in the United States. The reason may be low demand. When Eugene Morse asked the clinicians at the IAPAC Sessions how many had patients who asked for TDM, only a handful raised their hands. But the clinicians themselves proved anxious to learn about measuring antiretroviral levels, grilling Morse after his talk more than any other IAPAC Sessions presenter.
The potential for unwanted drug interactions is the main reason for interest in TDM, Morse said. He counted 21 antiretroviral formulations in the United States, and the recently approved 625-mg nelfinavir tablet makes it 22. A salvage regimen might include two ritonavir-boosted PIs, an NNRTI, and a few nucleosides, or tenofovir, or even enfuvirtide (T-20). Add to that mix the medications taken for comorbidities, not to mention herbal remedies the clinician may or may not know about, and potential interactions escalate.
Most TDM research has focused on protease inhibitors because (1) measuring active triphosphates of nucleosides inside cells remains difficult, and (2) nonnucleosides typically reach high concentrations and have long half-lives. But whether it's better to gauge one drug level or another -- trough, peak, or area under the curve -- has not been determined (Figure 2).
NIQ = patient IQ/reference IQ
Protein binding is only one variable that frustrates the simple reckoning of drug-drug interactions involving antiretrovirals. Pharmacologists must also consider sped-up or slowed-down drug metabolism by the liver. But before drug even gets to the liver, it may be thwarted by efflux pumps such as P-glycoprotein (P-gp), which affects antiretroviral absorption via the gut and penetration of cerebrospinal fluid (CSF) and T cells. Intracellular interactions also influence drug levels in CSF, semen, and other sites.
These variables have different effects on different antiretrovirals. Indinavir, for example, attains good semen-to-blood plasma and lymph node-to-plasma ratios, while nelfinavir and lopinavir do not.38 Indinavir also penetrates CSF better than nelfinavir or lopinavir.
Genes regulate all of these mechanisms. Polymorphisms in genes controlling cytochrome P450 isoforms can influence drug metabolism. Polymorphisms in the MDR-1 gene, which controls P-gp, can change drug absorption and distribution.
Boosting PIs with low-dose ritonavir has smoothed out spikey drug level curves. But the lofty and durable curves seen in PK study slides don't tell the whole story, Morse warned. If you add confidence intervals around each data point, it becomes clear that not all patients attain the average high concentration. And ritonavir-boost studies rarely show the effects of other drug interactions.
Confounders like those reviewed in the preceding section inspire caution about the wholesale use of TDM in clinical practice. But the fundamental rationale for checking drug levels is sturdy:
The principal trial underpinning the value of TDM, though convincing, has its limits. The TDM study of the ATHENA cohort randomized people starting nelfinavir or indinavir to have drug level results, and advice, reported to their clinicians, or to have TDM but not have results reported.39 The findings may have diminished applicability because everyone was treatment-naive when the study began, and because the unboosted PIs studied are used less often now than when the trial began. But 12 months of follow-up showed that TDM did what it's supposed to do.
The study involved 92 people taking nelfinavir and 55 taking indinavir. Whereas 17.4 percent randomized to the TDM group discontinued one of the PIs by month 12, 39.7 percent in the control group did so. Among people taking nelfinavir, 2.4 percent in the TDM group versus 17.6 percent in the control group had virologic failure. In the indinavir arm, toxicity affected 14.3 percent in the TDM group and 29.6 percent of controls. According to a noncompleter-equals-failure analysis, significantly more people in the TDM group (78.2 percent) than in the control group (55.1 percent) had a 12-month viral load below 500 copies/mL.
Where can a US clinician send a sample for TDM? A few commercial labs and medical centers offer the service, Morse said, but many have no proficiency testing. A TDM lab must also be CLIA certified and should be able to report results promptly. (CLIA stands for Clinical Laboratory Improvement Amendments, administered by the Centers for Disease Control.)
Six ACTG sites perform TDM and have proficiency testing -- Johns Hopkins, Stanford, the University of Alabama at Birmingham, the University of California, San Francisco, the University of Colorado, and Morse's team at the State University of New York at Buffalo. Although the Buffalo group will accept samples from approved sites as part of a research protocol, Morse could not speak for the other sites.
The Buffalo Antiretroviral TDM Registry has three goals:
Figure 4 outlines how the registry works. TDM will be done on samples from anyone taking a PI or NNRTI who meets one of the following criteria:
Morse is gradually making TDM a routine part of HIV disease care in his clinic by linking it with resistance testing. A panel of clinicians had already been meeting regularly to discuss resistance test results from people cared for at the clinic. Morse encouraged his colleagues to order a drug level every time they order a resistance test, and now the panel discusses both results at the same time. Most of the people who have had TDM at Buffalo are taking complex rescue regimens.
Session chair Mark Dybul (National Institute of Allergy and Infectious Diseases, Bethesda) suggested the Buffalo model may be a way to integrate TDM into routine care, at least at medical centers or large practices. Dybul himself has ordered TDM for some patients and typically calls a few HIV pharmacologists to help him interpret the results.
A hurdle to TDM interpretation is the lack of formally established high or low concentration cutoffs for each agent, a task complicated by the proliferation of ritonavir-boost doses. At least two cutoff schemes have been proposed, one by the editorial board of HIVPharmacology.com41 and the other by University of Alabama pharmacologists Edward Acosta and Jennifer King.42 Morse cautioned that cutoff determinations often rely on data derived from people taking their first antiretrovirals and so may not be valid for people with heavy treatment experience.
Which patients are good candidates for TDM? Several groups have proposed lists, including an international panel of HIV pharmacologists and clinicians headed by the University of Liverpool's David Back.43 Mark Dybul proposed the following list for Morse's review:
If using TDM to check adherence, which may be the problem in Dybul's fifth scenario, Morse suggested getting a morning trough level then comparing it with a trough after an observed dose. He also proposed three more TDM candidates:
The international panel's long list43 included most of these, as well as:
Yet Morse reminded delegates that evidence supporting the clinical value of resistance testing remains thin. Because of intraindividual variation in drug levels, you can't be certain that a level will be the same the next time you measure it, even if you don't change the dose. And if you do change the dose, you can't be certain the drug level will also change.
The panel that writes antiretroviral guidelines for the Department of Health and Human Services (HHS) is pondering a change in advice on resistance testing. At the time of the IAPAC Sessions, the HHS guidelines recommended resistance testing only in people with virologic failure or suboptimal suppression of viral load after starting antiretrovirals.44 Those guidelines say testing should be "considered" before starting therapy in someone with acute HIV infection, but that scenario may be bumped up into the "recommended" category. At the time of the IAPAC Sessions, the HHS labeled resistance testing "not generally recommended" before starting therapy in someone with chronic HIV infection. That scenario may be promoted to the "considered" category.
But most clinicians attending the IAPAC Sessions seem to be taking their cue not from the HHS but from the British HIV Association, which thinks people with chronic infection should have a pretreatment resistance test.45 The HHS panel justified its reluctance to endorse pretreatment testing during chronic infection by noting that resistance mutations tend to fade to undetectable minorities in someone not taking antiretrovirals. That wasn't the rationale offered by the single IAPAC Sessions delegate who spoke against testing in such patients, arguing that it's "crazy" in the current economic climate. The county where he works has already twice run out of resistance test vouchers, so he believes the tests should be reserved for those who stand the best chance of benefiting.
Delegate opinion was more mixed on the value of ordering a resistance assay for an untreated person with acute infection. David Katzenstein maintained that the clinical benefit of testing in acute infection remains unproved, noting that reports of virologic failure resulting from primary infection with resistant virus have been rare. IAPAC Sessions Co-Chair Diane Havlir, on the other hand, worried that failure to test acutely infected people would make it impossible to spot the potential emergence of multidrug-resistant virus.
Other clinicians in the room proposed that decisions on testing people with acute infection should be based on the prevalence of resistance transmission in that region. A physician from North Carolina, for example, found almost no cases of resistance in 40 patients with acute infection, most of whom lived in rural areas. But another clinician from Orlando, Florida, said transmission of resistant HIV is not uncommon there.
The virtue of genotyping is that it gives a straightforward result, detecting at least majority populations of mutants that confer resistance to specific drugs. Then the difficulties begin, Katzenstein lamented, because interpreting genotypes can be tough. Certain mutants have complex effects on some drugs -- but not others. Expert algorithms often disagree on what a complicated mutation set means.
In a Utopian world stripped of all confounders, interpreting a genotype would yield a clear prediction of viral susceptibility to antiretrovirals of interest. Knowing precisely which drugs or classes would stifle a patient's virus, the clinician could then decide on a course of action:
For some drugs -- such as abacavir, tenofovir, and lopinavir -- knowing which mutants dominate a person's viral population tells the clinician a lot. A study of people beginning abacavir found that 85 percent of those with wild-type (nonmutant) virus had more than a half-log drop in viral load or an undetectable load after 12 weeks of treatment.46 With one or two mutations conferring resistance to AZT, the response rate fell to 77 percent. With one or two AZT mutations plus the 3TC-inspired 184V change, the response rate dipped to 60 percent. Three or more AZT mutations, and three or more AZT mutations plus 184V, trimmed the response rate even more. But other work suggests that 184V alone increases susceptibility to the thymidine analogs, AZT and d4T.47, 48
A study of people beginning tenofovir found the best 24-week response in those with no AZT/d4T mutations, a lessened response in those with one or two, and little response in those with three AZT/d4T mutations including 41L or 210W.49 Abbott researchers demonstrated different 24-week response rates to lopinavir/ritonavir depending on whether people began treatment with zero to five PI mutations, six or seven PI mutations, or eight to 10 PI mutations.50
Studies like these may clarify the effect of discrete mutations or mutation sets on the probability that a person will respond to certain drugs. But, Katzenstein asked, can such information be readily adapted to clinical practice? Genotypes can be translated into treatment advice in three ways -- by a rules-based algorithm, by expert opinion, or by a database comparison (as with the VirtualPhenotype, described below).
Algorithms are fallible because the experts who build them cannot be infallible. A study comparing four often-used algorithms made that abundantly clear.51 Feeding 2,045 viral sequences into the four decision trees, Katzenstein's colleague Robert Shafer charted which sequences the algorithms rate sensitive, intermediate, or resistant to given drugs. He found complete concordance for only 66.4 percent of interpretations, and most of those concordant calls involved sensitive virus. While 15.4 percent of interpretations disagreed on sensitive-versus-intermediate calls, 13.8 percent involved intermediate-versus-resistant discordance.
On another 4.4 percent of interpretations, algorithms disagreed on whether a sequence was sensitive or resistant to some drugs. For example, two algorithms called virus with reverse transcriptase mutations 74V, 184V, and 215F or Y sensitive to d4T, one called it resistant, and one called it intermediate. Two algorithms predicted that virus with the protease changes 84V plus 90M would be resistant to amprenavir, one called it sensitive, and another called it intermediate.
Virco's VirtualPhenotype estimates the likely susceptibility of a submitted genotype by searching its database of phenotyped viral samples with the same genotype. A comparison of the VirtualPhenotype and real phenotyping in 201 heavily pretreated people found that the two tests performed similarly in picking new suppressive regimens.52 After 48 weeks of follow-up, people randomized to real phenotyping had virtually the same virologic responses as people randomized to the VirtualPhenotype (Table 8). One way of interpreting these results, Katzenstein offered, is that a VirtualPhenotype will add no valuable new information to a real phenotype.
Randomized, controlled trials comparing genotyping with standard of care show a consistent, if often small and short-term virologic benefit with testing. The genotyped groups in four trials53-56 all had better average 12-week responses than the standard-of-care groups, ranging from a 0.48-log improvement in VIRADAPT53 to a 0.18-log improvement in ARGENTA.54
Results in phenotyping trials have been mixed. VIRA 3001 found an average 0.37-log 12-week advantage for phenotyping over standard of care,57 but NARVAL58 and CCTG 57559 found no benefit with phenotyping. (Two other randomized trials not reviewed by Katzenstein also differed on whether phenotyping can help pick a new regimen.60, 61) The varying results of phenotyping trials can be partly explained by different study designs, different populations, and recent approval of more drugs that counter some resistant virus.
As with genotyping, though, phenotyping cannot yield certain conclusions because the assays -- and their interpretation -- are imperfect. A phenotypic assay measures how much drug one needs to inhibit replication of a virus by 50 percent, the IC50. That value is then compared with the drug's IC50 against wild-type virus to yield a "fold change" in viral susceptibility to the drug. But it's not so easy to set IC50 cutoffs that signal when a drug will be active, inactive, or partially active against a virus.
Katzenstein and colleagues tried to improve the reliability of phenotyping by toting a continuous -- rather than dichotomous -- phenotypic sensitivity score (PSS). Resistance, he explained, is usually not a dichotomous, black-or-white phenomenon. It is a continuous variable that changes as the viral population evolves, embracing greater or lesser proportions of sensitive and resistant virus.
A dichotomous PSS can be figured by assigning a value of 1 to each drug in a regimen against which the virus has a 2.5-fold or less change in susceptibility compared with wild-type virus (or 1.5-fold or less for ddI and d4T). Drugs score 0 if the fold change exceeds 2.5. The scores for each drug in the regimen get added up to yield the dichotomous PSS.
A continuous PSS attempts to account for gray areas in viral susceptibility by adding some flourish to the homely dichotomous equation. Again, each drug scores 1 if the fold change is 2.5 or less (or 1.5 or less for ddI and d4T). If the fold change tops 10, the drug scores 0. For fold changes between 2.5 and 10 (or 1.5 and 10 for ddI and d4T), a value between 0 and 1 is calculated thus:
1 - (fold change - 2.5)/fold change
Katzenstein retrospectively tested the continuous PSS in viral samples from ACTG 364, which enrolled 195 people with only nucleoside experience and randomized them to nelfinavir, efavirenz, or both plus two nucleosides.62 Everyone began the new regimen with a viral load above 2,000 copies/mL.
Defining virologic failure as a confirmed viral load of 2,000 copies/mL or more at week 16 or later, the ACTG team found that a PSS of 3 or more correlated with durable suppression of viremia through 144 weeks. In the group with a PSS at or above 3, the proportion remaining free of virologic failure stayed steady after week 24 at about 80 percent. In the group whose PSS lay between 2 and 3, the proportion free of virologic failure fell steadily over the study period, ending up near 40 percent. And in the group with a PSS of 2 or less, only 20 percent had escaped virologic failure by week 144.
Katzenstein concluded that an effective rescue regimen must contain at least three "fully active" drugs or the continuous PSS equivalent of 3. He told IAPAC Sessions delegates that he believes the future of clinical resistance testing will involve a measure like the continuous PSS, but he cautioned that this formula -- and any resistance formula humans devise -- has its limits. Although it is more elastic than a dichotomous PSS, the continuous version must still rely on cutoffs that cannot accurately reflect viral susceptibility for every individual. The ACTG 364 PSS analysis does improve on earlier resistance assay trials in length of follow-up, but the analysis was retrospective.
Finally, Katzenstein urged his clinician colleagues to remember that "resistance is a relatively small part of a large picture -- maintaining a clinically effective treatment strategy." With that perspective in mind, he suggested the following approach to interpreting resistance tests.
Introducing the two talks on metabolic abnormalities in people with HIV, Kathleen Mulligan (University of California, San Francisco) posed a question that captured some key conflicts in this field:
Can we study it and manage it even if we can't agree on what to call it, how to define it, or what causes it?
The answer appears to be a qualified yes, if one can judge from the volume of work done on the metabolic complications of HIV and antiretrovirals, and from some progress in controlling them. And a definition of lipodystrophy has been proposed, although results of a large case-control study of fat abnormalities in people with and without HIV infection63, 64 don't agree with the first rigorous case definition based on a case-control study of HIV-infected adults with or without apparent symptoms.65 "We now have a case definition," Mulligan noted. "I'm not going to say we have consensus."
The case definition uses a scoring system incorporating 10 variables. Clinicians can access an interactive online program of that definition or of two simpler versions, plug in numbers from their patients, and see if they meet the case definition.66 But IAPAC Sessions delegates apparently saw little value in that exercise, since none had done so. One veteran HIV clinician explained his lack of enthusiasm by pointing to the case definition study's design. "A sophisticated regression model like that won't help me decide on the borderline cases," he said. By eliminating people with iffy lipodystrophy, the study eliminated the very people who interest him most. "I already know which people have severe lipodystrophy."
Mulligan listed HIV-related metabolic and morphologic abnormalities in four separate boxes:
Can the boxes be connected, she asked. If the boxes overlap, can distinct phenotypes still be defined? Are antiretrovirals or host factors more to blame? What is their long-term clinical significance? And what's the best way to manage them? The two speakers at this session -- Colleen Hadigan and Morris Schambelan -- didn't have all the answers. But they suggested more than a few.
Hadigan started by asking two questions of her own:
Most delegates picked the right answers before Hadigan reviewed studies that yielded remarkably consistent numbers. Her own work found that 7 percent of 71 people with changes in body fat met World Health Organization criteria for diabetes, and 35 percent had impaired glucose tolerance.67 An Australian study charted the same diabetes prevalence in 113 people taking a protease inhibitor.68 And Spanish clinicians found that 5.8 percent of PI-treated men with lipoatrophy had diabetes.69
In Hadigan's study 57 percent had fasting triglycerides above 200 mg/dL; the same proportion had cholesterol readings above 200 mg/dL.67 This case-control analysis found significantly higher proportions with hyperglycemia, hypercholesterolemia, hypertriglyceridemia, and decreased high-density lipoprotein cholesterol (HDL-C) in the 71 people with HIV and lipodystrophy than in the 213 HIV-uninfected controls (p < 0.03). Thirty HIV-infected people without lipodystrophy had a metabolic profile similar to that of 90 controls, except that about 40 percent in the HIV group had an HDL-C below 35 mg/dL compared with about 5 percent of controls (p < 0.01). A review involving 159 Australian men with HIV found triglycerides above 177 mg/dL in 52 percent and high cholesterol in 44 percent, regardless of fat distribution.70
A study of 1,927 children with HIV infection counted 22 percent with cholesterol levels above 200 mg/dL compared with 9 percent of HIV-uninfected but perinatally exposed controls.71 In the HIV group 12 percent had a cholesterol tally above the 95th percentile for healthy children. The researchers linked these high cholesterols to PI use, adherence to therapy, and younger age. Nonnucleosides had a protective effect. With effective antiviral therapy, Hadigan observed, these children will have lived with high lipids for 20 years by the time they're young adults.
Comparing 91 HIV-infected men and women with 273 age-, sex-, and body mass index-matched controls without HIV, Hadigan learned that 46 percent in the HIV group met criteria for the metabolic syndrome (see note 72) versus 15 percent of controls (P = 0.001).73 Nearly 30 percent of the people with HIV had more than a 10 percent risk of cardiovascular disease in 10 years versus fewer than 15 percent of controls (P = 0.001).
An array of research in the past few years implicates certain antiretrovirals in insulin resistance, dyslipidemia, and adipocyte abnormalities. Hadigan offered this outline:
Researchers at Washington University in St. Louis determined that indinavir, ritonavir, and amprenavir squelch insulin-stimulated glucose uptake in adipocytes by inhibiting the glucose transporter GLUT-4.74 Morris Schambelan's group showed that a single dose of indinavir decreases insulin sensitivity in healthy men without HIV infection.75 Last year a study of ritonavir-fed mice traced a 30 percent jump in very low-density lipoprotein (VLDL) cholesterol.76 That increase doubled when the mice also ate a high-fat diet. HIV alone boosted hepatic production of VLDL apolipoprotein B, and either PI therapy or NNRTI therapy reduced its clearance.77
Nelfinavir increases lipolysis -- the release of fatty acids from adipocytes into the circulation -- in a dose-dependent manner.78 Hadigan found higher rates of lipolysis in 19 people with HIV than in eight uninfected controls; and d4T upped that rate regardless of treatment with a PI.79 Levels of adipogenic transcription factors SREBP-1 and PPAR-gamma proved significantly higher in 26 people taking antiretrovirals than in 18 uninfected controls.80 Everyone in this study was taking a PI, and most were taking d4T.
ACTG 384 charted declines in limb fat after 16 weeks of therapy with regimens containing either AZT/3TC or ddI/d4T, although the drops were significantly greater in the ddI/d4T group (p < 0.05).81 Some recent work attributes peripheral lipoatrophy to nucleoside-induced mitochondrial toxicity, but Hadigan doesn't think the case has been clinched.
Bravely, she attempted to summarize much of this mechanistic research in a single slide outlining the effects of HIV and antiretrovirals on the liver, of PIs and nucleosides on fat cells, and of PIs on muscle (Figure 5). Perhaps her most important point was that the effects on fat and metabolic variables cannot be separated.
Hadigan reviewed work suggesting two approaches to managing metabolic complications:
Several studies show that replacing d4T (and sometimes a PI) with another nucleoside (and the PI with abacavir) can slowly but measurably reverse peripheral lipoatrophy.82-86 Most recently, a study of 13 people who swapped d4T for abacavir or AZT graphed gains in arm, leg, and trunk fat of 25 percent, 15 percent, and 23 percent.87 Adipocyte apoptosis, perhaps driven by mitochondrial toxicity, waned after people stopped d4T.
The protease inhibitor atazanavir may have an advantage over other PIs even greater than its once-daily dosing. Several trials convincingly show that atazanavir barely affects -- and sometimes improves -- lipid parameters. A published study that compared atazanavir with nelfinavir for 48 weeks in treatment-naive people found 20 percent to 40 percent gains in total cholesterol, low-density lipoprotein cholesterol (LDL-C), and triglycerides with nelfinavir, but little change with different doses of atazanavir.88 Portentous LDL-C dropped in the group taking 400 mg of atazanavir.
The safest lipid-lowering therapy, diet, works in people taking PIs. A Spanish study of people with elevated lipids found that a low saturated fat diet dropped triglycerides by 50 percent and cholesterol by more than 20 percent in PI takers who stuck with the diet.89 Lipid levels fell only modestly in people not taking PIs.
A 12-week trial of diet with or without gemfibrozil in PI-treated people with lipodystrophy and high triglycerides found a nonsignificant improvement in triglycerides in the gemfibrozil group versus the diet-only group (P = 0.06).90 Only one person reached a normal triglyceride level, and gemfibrozil did not change cholesterol, HDL-C, glucose, or insulin.
Niacin extended-release tablets significantly lowered but did not normalize triglycerides and cholesterol in 14 HIV-infected men with high triglycerides or LDL-C.91 Although no one suffered a grade 3 or 4 liver function test elevation in the 14-week study, indicators of insulin sensitivity worsened.
Atorvastatin and pravastatin are the preferred anticholesterol agents for people taking PIs because protease drugs drive simvastatin and lovastatin levels to dangerously high reaches. Early studies of atorvastatin and pravastatin in antiretroviral-treated people with high lipids show some drops in lipids, but usually not into normal ranges.92-94 When statins fall short, a fibrate may be added, but at the risk of liver toxicity and rhabdomyolysis.
Hadigan studied metformin at a dose of 500 mg twice daily in 25 HIV-infected people with a waist-to-hip ratio above 0.9 and insulin levels exceeding 15 µU/mL.95 After 12 weeks insulin area under the curve had dropped significantly in the metformin group compared with the control group (P = 0.01). Mean weight and waist circumference also improved significantly with metformin while rising with placebo.
Metformin decreased visceral adipose tissue in another study of people taking PIs,96 but it also cut subcutaneous fat in both studies, so it may worsen peripheral lipoatrophy. Other potential side effects of metformin are diarrhea and hyperlactatemia. Hadigan added that a just-completed study by her group charted an additive effect of metformin and exercise on strength, body composition, and insulin sensitivity.
A study of the thiazolidinedione troglitazone in people with non-HIV congenital lipodystrophy showed that it increases thigh and abdominal subcutaneous tissue while lowering visceral fat.97 When troglitazone got pulled from the market because of liver toxicity, research attention shifted to rosiglitazone and pioglitazone. The first placebo-controlled trial of rosiglitazone found improved liver function tests and decreased liver fat after 24 weeks in the active treatment group, along with improved insulin.98 But subcutaneous fat did not increase significantly, and triglycerides did.
A smaller, nonrandomized study found that six to 12 weeks of rosiglitazone increased subcutaneous fat by 23 percent (P = 0.05) and cut visceral fat by 21 percent (P = 0.04) in HIV-infected people with insulin resistance.99 Glucose disposal also improved, but lipids rose.
Beyond considering switching antiretrovirals or adding drugs to counter their side effects, Hadigan said, getting people to stop smoking will probably do more to lower their cardiovascular risk.
Fat abnormalities in people with HIV began appearing before they started taking protease inhibitors, as Morris Schambelan learned in the mid-1990s. That's when his colleagues started referring HIV-infected patients to see if their buffalo humps might mean they had Cushing's disease. In eight people referred between June 1995 and October 1997, Schambelan ruled out Cushing's, discovering instead a fat change specific to people with HIV.100
With body mass indices (BMIs) averaging 24.7 kg/m2, these eight people were not obese. And their CD4 counts were not remarkably low. But they had had a buffalo hump for one to 26 months. Compared with 15 age-, CD4 count-, and BMI-matched controls, they had significantly more trunk fat as a percentage of total fat (65.3 versus 56.8 percent, P = 0.03). Three of them were taking indinavir and one nelfinavir, but the other four were taking only one or two nucleosides. This early study of fat accumulation in antiretroviral-treated people made it plain that PIs were at most a contributing factor.
The arrival of protease inhibitors did have an impact on body fat. Comparing body measures in people who switched from AZT/3TC to ddI, d4T, and nelfinavir, Schambelan saw almost no change in overall weight (Table 9). But DEXA-measured total fat fell after nine months of the new regimen, trunk fat increased, and peripheral fat fell. Visceral adipose tissue rose, while subcutaneous adipose tissue dropped. What this kind of analysis cannot sort out is the cumulative effect of nucleosides or the independent effect of changing from AZT/3TC to ddI/d4T.
Explaining central girth gains in people taking antiretrovirals remains difficult, as the FRAM study shows.63, 64 The loss of peripheral fat, on the other hand, has been linked convincingly to antiretroviral therapy. Schambelan and Kathleen Mulligan offered one of the clearest demonstrations of antiretroviral effects on limb fat in a cross-sectional comparison of 44 people without HIV infection, 23 HIV-infected but untreated people, 30 people taking only NRTIs, and 26 taking NRTIs plus a PI.102
Mulligan and Schambelan found that the treated groups did not differ significantly from the untreated and uninfected groups in trunk fat divided by height. But the people taking antiretrovirals had significantly less limb fat per height than the two control groups and a significantly higher trunk-to-limb fat ratio. The NRTI-only group did not differ from the NRTI/PI group in either of these two measures.
Some evidence suggests, though, that nucleosides and protease inhibitors have an additive effect on limb fat loss. Most recently, Schambelan noted, limb fat proved significantly lower in treatment-naive people randomized to take nelfinavir than in those randomized to efavirenz in ACTG 384.81
Mitochondrial DNA depletion in fat cells and elsewhere among people taking nucleosides has buttressed the theory that lipoatrophy is a nucleoside-induced mitochondrial toxicity. But some researchers, including Colleen Hadigan, are not yet swayed by these arguments. The tissue of interest is not fat, she observed in an open discussion, but muscle, where mitochondrial function is critical.
A study by Schambelan's group addressed the role of mitochondria in muscle of eight people with severe peripheral lipoatrophy and eight antiretroviral-treated controls without atrophy.103 Although mitochondrial damage has been noted in muscle of people with lipoatrophy, Schambelan's group wanted to know what the functional consequences may be. The atrophy group was significantly older than controls, but the groups didn't differ in body mass index, CD4 count, duration of HIV infection, or current PI or NNRTI use.
Measuring subcutaneous fat in the right calf, thigh, and gluteal region by MRI, Schambelan's team found significantly less in all three areas among the people with clinically apparent atrophy (p < 0.001 for all comparisons). Although the atrophic patients had less DEXA- measured total body fat and limb fat than controls, their total and appendicular lean body mass proved equivalent with controls. In other words the people with atrophy did not have muscle wasting.
Next Schambelan and colleagues compared measures of the tibialis anterior muscle (which makes the foot tap) in cases and controls. They found no significant differences in cross-sectional area, maximal voluntary contraction, or specific strength. Phosphocreatine recovery after exercise proved virtually identical in the two groups, another finding indicating no difference in muscle function. Finally, with study participants at rest, the researchers looked at muscle metabolites that might signal worse muscle function; they found no differences between the two groups.
Although the study was small, Schambelan concluded that people with severe peripheral lipoatrophy do not differ from antiretroviral-treated people without atrophy in muscle-specific strength, intramuscular energy metabolism, or indices of mitochondrial function.
Schambelan suggested three lines of evidence suggesting that fat loss in people taking antiretrovirals contributes to insulin resistance in some of these people:
A study comparing 15 HIV-infected people with lipoatrophy, 14 HIV-infected people without atrophy, and 12 healthy controls established the correlation between low limb fat and insulin resistance among people with HIV.104 The gold- standard clamp test for insulin resistance showed that people with atrophy had about half the glucose uptake as healthy controls, while the HIV-infected control group had an intermediate level. More limb fat correlated positively with higher insulin sensitivity (r = 0.60, P = 0.001).
Several research teams strengthened this association in transgenic mice with virtually no subcutaneous fat. Besides severe insulin resistance, the mice have lofty triglycerides and hepatic steatosis. Transplanting fat from normal mice eases such metabolic abnormalities in one of these strains.
Another murine model -- obese, hyperlipidemic, diabetes-prone Zucker rats -- led Roger Unger (University of Texas Southwestern Medical Center, Dallas) to formulate the lipotoxicity hypothesis.105 As circulating fatty acids begin roosting in nonadipose tissues, beta cells falter in secreting insulin, skeletal muscle become insulin resistant, and heart muscle weakens. Surveying six mouse or rat models, Schambelan listed disorders of fat regulation resulting from no adipocytes -- or from too many adipocytes. "Adipocyte mass ranges from absent to massive," he noted, "but the common feature of these disorders is lack of leptin action."
A fat transplant from normal mice eased metabolic abnormalities in one of two mouse strains lacking subcutaneous fat. But leptin reversed metabolic abnormalities in both models. That finding has been extended to HIV-uninfected people with severe acquired lipodystrophy.106 In one 17-year-old woman with no subcutaneous fat, leptin replacement brought levels back to the lower limit of normal, slashed stratospheric triglycerides, and controlled diabetes well enough to make antidiabetic drugs unnecessary. In eight people with severe non-HIV lipodystrophy, four months of leptin sharply reduced glucose and triglyceride levels.
Will leptin work in people with antiretroviral-induced lipodystrophy? Schambelan asked himself that question while evaluating 39 men referred primarily for fat accumulation, though most also had fat atrophy. He found that 16 (41 percent) had leptin levels considered low in a National Institutes of Health trial -- below 3 ng/mL. Levels in that nether range have proved more common in people referred primarily for atrophy.
Schambelan and colleagues have begun a trial in HIV-infected men with leptins below 3 ng/mL and women with levels below 4 ng/mL. Study participants must have fasting triglycerides between 299 and 1,000 mg/dL and clinical evidence of fat abnormalities. The protocol calls for 0.01 mg/kg of leptin for men and 0.02 mg/kg for women for three months. In the next three months the doses climb to 0.03 mg/kg for men and 0.06 mg/kg for women.
The IAPAC Sessions 2003 spanned a panoply of problems facing HIV clinicians. Even this two-day look at the challenges of HIV medicine suggests that clinicians must have more than a little facility in:
No one can hope to excel in all those areas, but everyone must know what questions to ask and where answers might be found. As IAPAC Sessions Co-Chair Diane Havlir noted before the first talk, the best HIV clinicians are "the kind of people who know that the more you know, the more you don't know." And as each succeeding presentation made clear, anyone seeking a unifying theory of HIV disease had best seek retraining in a simpler discipline. Physics, maybe. These days an HIV paradigm, Havlir proposed, "is kind of an oxymoron."
Mark Mascolini writes about HIV infection (email@example.com).
Early Virologic Non-Response in Patients With HIV Infection Treated With Lamivudine, Abacavir and Tenofovir
This article was provided by International Association of Physicians in AIDS Care. It is a part of the publication IAPAC Monthly.