Though the wisdom of nature can reason it thus and thus, yet nature finds itself scourged by the sequent effect.
-- William Shakespeare, King Lear (I, ii, 113-115)
And the point returned to may be one set 10 years ago, suggested antiretroviral maven John Mellors (University of Pittsburgh). While other CROIsters harked back to the discovery of AIDS a quarter century ago or the 10th birthday of triple therapy, Mellors pointed to a single study -- AIDS Clinical Trials Group (ACTG) protocol 320, the randomized trial that ratified the life-saving advantage of three antiretrovirals over two.1
Speaking after six STI researchers served up fresh results -- some of them mixed, some plainly awful -- Mellors suggested "it's ironic and fitting that we've learned what we already knew." ACTG 320 showed that triple therapy "was lifesaving and most important for people with an AIDS diagnosis and a low CD4 count," he observed. "Ten years later we've found that people with an AIDS diagnosis and a low CD4 count still need that therapy."
That discomfiting assessment may oversimplify what these latest STI trials mean, argued Staccato study chief Bernard Hirschel (University Hospital of Geneva). A person just starting antiretrovirals with a red-alert CD4 count, as in ACTG 320, is not the same as someone whose CD4 count drops toward the danger zone when that person suspends a regimen that worked for years.
But there was no escaping the humbling "rediscovery" that taking potent antiretrovirals day after day is good for people with HIV infection -- better, two trials showed, than taking antiretrovirals part time. Both the massive multinational SMART study [abstract 106LB] and the smaller Trivacan trial in Côte d'Ivoire [abstract 105LB] logged more than twice as many cases of HIV disease progression or death among people who timed drug breaks to CD4-cell swings rather than taking steady therapy.
Even more astonishing to some drug-holiday apologists, both randomized trials unhasped a pivotal linchpin of STI thinking -- that on-and-off therapy trims the risk of complications typically blamed on antiretrovirals, namely heart, liver, and kidney disease. Far from improving quality of life, in both SMART and Trivacan drug breaks put more people in the hospital -- or the morgue -- some with distinctly non-AIDS diagnoses. And it happened so fast that independent review panels had to shut SMART down early and close the CD4-guided arm in Trivacan.
Staccato, the third large randomized comparison of CD4-steered breaks versus unremitting therapy detailed at CROI, did not find a higher progression risk in the off-and-on arm [abstract 102]. Different CD4 stop and start signals in Staccato, and other variables considered below, may explain this happier outcome.
But when CROI roundtable chair Scott Hammer (Columbia University, New York) asked five STI experts if "CD4-guided STIs should be reserved for clinical research" until statisticians finish digesting these latest results, everyone agreed. And Staccato's Bernard Hirschel was one of those experts. Indeed, in a post-CROI e-mail to the IAPAC Monthly, he professed seeing no immediate need for further STI trials, at least until SMART's stat chefs hash through all the data.
Besides what these STI trials say about suspending antiretrovirals, the next most burning question is what they say about cardiovascular disease in people with HIV. This year's CROI also featured another iteration of the vast Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) study, confirming a higher risk of heart attacks with each extra year of therapy and now squarely placing the blame on protease inhibitors (PIs) [abstract 144]. But the relative rate (RR) of myocardial infarction (MI) has been dropping in this 11-cohort study, perhaps because people are switching from PIs, or starting antilipid therapy, or stopping smoking (see "MI Rates Drop in D:A:D," below).
Another variable almost certainly contributes to these seesaw MI rates: HIV itself. Uncontrolled or poorly controlled HIV infection sparks inflammation -- particularly in endothelial cells that line blood vessels. At the same time the virus apparently promotes thicker and tougher artery walls. And HIV probably contributes to pulmonary hypertension, as more CROI studies confirmed. All these things make the heart's job harder.
Do the HIV rebounds and immune upsets that invariably follow drug breaks threaten blood vessels enough to explain the higher heart disease risk in SMART's drug holiday arm? No one can say. But mounting evidence from that seminal trial, from other STI studies, and from research results stretching back several years, offer at least a loud hint that letting HIV off the antiretroviral hook poses clinical risks beyond progression to AIDS.
2-4 then in people with primary infection.5-7 And evidence-based speculation that drug breaks would render HIV wobbly-kneed when facing well-planned salvage regimens fell flat.8-11
Yet more than one cohort study hinted that drug breaks seemed safe in people whose CD4s climbed high with an antiretroviral boost or in those who started therapy with a T-cell count no longer considered a treatment start signal. Indeed, 48-week results of the randomized Staccato trial comparing steady therapy, drug holidays guided by CD4 ups and downs, and week-on-week-off (WOWO) therapy in 74 people showed comparable clinical outcomes in the continuous-therapy arm and CD4-guided group.12 WOWO proved a no-no, with a 31% virologic failure rate.
Curiously, people told to shelve their HIV meds in Staccato and other small studies rarely enjoyed the benisons of freshly regulated lipids, newly controlled glucose, or falling liver enzymes,12-15 though people taking their drugs staccato in Staccato did have self-reported lipodystrophy significantly less often than people in the control arm after 24, 48, and 72 weeks.
Then came SMART, surely the weightiest STI study ever. Its sheer size, its simple design, and its vow to pile up plentiful endpoints over eight years impressed many. But one morning a month before CROI, STI adherents and skeptics alike opened their e-mail and felt still-uncaffeinated eyes glaze further upon reading this headline: "International HIV/AIDS trial finds continuous [therapy] superior to episodic therapy."16
SMART was over. After only an average 14 months of follow-up, instead of the anticipated 96, SMART's data and safety monitoring board (DSMB) had seen enough.
At least the randomized trial is over. The data delving will surely continue for years. But already SMART's stat team has burrowed through mountains of data and handed the richest diggings to Wafaa El-Sadr (Harlem Hospital, New York), who came to CROI with 18 slides, none bearing good news on CD4-guided STIs [abstract 106LB]. (Slides from this and every talk at the 13th CROI appear on the conference Web site, www.retroconference.org/2006.)
SMART's straightforward plan called for signing up 6,000 people with a CD4 count above 350 cells/mm3, assigning half to continuous therapy and half to drug breaks beginning any time the CD4 count topped 350 cells/mm3 and ending when the count hit 250 cells/mm3. Trial designers anticipated 910 primary endpoints -- progression to AIDS or death -- over an average eight years of follow-up. But in January 2006, with 164 endpoints on the books, and an amazingly low 2% lost to follow-up, it became clear that drug breaks as defined in this trial are fraught with danger.
The DSMB pulled the plug after counting 117 primary endpoints in the CD4-guided arm versus 47 in the steady-therapy group. Trial planners projected an overall progression or death rate of 1.3% per year in the first 24 months, and that's about the rate they saw in the continuous-therapy arm -- 1.5 per 100 person-years. But the treatment-break arm chalked up an event rate of 3.7 per 100 person-years. Statisticians figured a 2.5 times higher risk of progression or death in the CD4-guided arm (95% confidence interval [CI] 1.8 to 3.6), a highly significant difference (P < 0.0001).
People taking drug breaks had a 1.9 times higher risk of death alone, a 6.1 times higher risk of serious progression, a 3.3 times higher risk of nonserious progression, and a 2.2 times higher risk of serious progression or death.
Bad enough, but then El-Sadr delivered what may be the coup de grâce for this kind of STI: compared with the continuous-therapy arm, the STI group suffered more "severe complications," which meant cardiovascular-, hepatic-, or renal-related deaths, or nonfatal heart, liver, or kidney complications. A composite endpoint conflating all these endpoints -- the very setbacks drug holidays look to avert -- proved 50% more likely in the holiday group.
SMART also shredded another shibboleth of STI hermeneutic -- that people with higher pretreatment CD4 nadirs can take longer drug breaks safely. Splitting CD4 nadirs into six echelons, SMART's number nudgers saw almost no difference in progression rates. For example, people with a CD4 nadir under 50 cells/mm3 had a 2.9 times higher risk of progression or death than the control arm, precisely the same risk as people with a nadir between 200 and 299 cells/mm3.
Nor did CD4 count at the beginning of SMART affect progression rates. Indeed, the risk of progression proved lower (but still bad) in the lowest CD4 bracket, 350 to 459 cells/mm3 (odds ratio [OR] 1.5), than in the higher CD4 brackets (OR 4.3 for 450 to 549 cells/mm3, OR 3.1 for 550 to 649 cells/mm3, and OR 2.9 for 650 or more cells/mm3). The CD4 count measured on the study visit just before an endpoint "event" did not predict those endpoints. Finally, having AIDS at study entry -- as 24% of these people did -- had no effect on progression risk.
Strictly speaking, all these results pertain only to the population studied, fewer than 5% of whom began SMART with no antiretroviral experience, 24% of whom already had an AIDS diagnosis, and 30% of whom enrolled with a viral load above 400 copies/mL. Yet a study of the same CD4-guided tactic in younger West African adults, all with viral loads under 300 copies/mL, yielded almost precisely the same risk of progression (see "Trivacan," below). Whereas 74% of SMART enrollees were men, 77% in the African Trivacan study were women. Trivacan participants had a median age of 34 years, compared with 46 years in SMART. But statistical analysis showed that neither age nor gender swayed progression risk in SMART.
Any way you look at it, SMART upends several tenets of STI thinking. Writing for the Web site of the National AIDS Treatment Advocacy Project (NATAP), David Margolis (University of North Carolina, Chapel Hill) calls the higher non-AIDS complication rate in the drug-break arm a "particularly powerful" finding because "it is directly in opposition to the hypotheses that the SMART study investigators sought to prove."17
And that's not the only theoretical turnaround SMART turned up. In CROI notes shared with the IAPAC Monthly (and reviewed by SMART's top statistician), STI expert Bernard Hirschel spells out several more (Table 1).
At least when governed by a 350-cells/mm3 stop sign and a 250-cells/mm3 green light in Côte d'Ivoire -- the same stop and go signs SMART used -- drug breaks proved as dangerous in Africa as elsewhere [abstract 105LB]. The incidence rate ratio for serious morbidity measured 2.6 (95% CI 1.3 to 5.6) in the CD4-guided arm compared with continuous therapy, nearly mirroring SMART's 2.5 OR for progression with CD4-regulated breaks.
Christine Danel (University of Treichville, Abidjan) reported 17.6 cases of serious morbidity per 100 patient-years in the STI arm versus 6.7 in the steady-therapy group. People stopping and restarting antiretrovirals had a 1.5 times higher incidence of tuberculosis, a 2.6 times higher incidence of oropharyngeal candidiasis, and a 15.9 times higher incidence of invasive bacterial infection than people who never took a drug break.
Most of the bacterial infections -- which drove the higher STI-group morbidity rate -- happened at CD4 counts above 250 cells/mm3. And everyone continued cotrimoxazole prophylaxis during the study; 85% of those with bacterial infections had cotrimoxazole-resistant bacteria.
After an average 19.4 months of followup, Danel counted 0.6 deaths per 100 patient-years in the ongoing-therapy group and 1.2 deaths per 100 patient-years in the CD4-guided group. With these results in hand, the Trivacan safety panel shut down the CD4 STI arm but allowed researchers to continue comparing steady therapy with a two-month-off-four-month-on strategy.
Trivacan enrolled 651 adults with a median age of 34 years, a CD4 count above 350 cells/mm3, and a viral load below 300 copies/mL and assigned them in a 3-2-1 ratio to two-month-off-four-month-on STIs, CD4-guided breaks, or continuous therapy. Median nadir CD4s measured 272 cells/mm3 in the 216 people assigned to CD4-guided STIs and 274 cells/mm3 in the 110 who never stopped therapy. Respective median baseline counts stood at 457 and 461 cells/mm3.
Higher morbidity rates in the CD4-managed group meant those people spent significantly more time in the hospital than people who continued treatment (68 versus 36 days, P < 0.001) and had more clinic visits (233 versus 159, P < 0.001). Researchers often proffer cost savings as one rationale for STIs in poor countries, but in this trial extra care for the STI group probably ate up any money saved on drugs. (Danel did not have a cost analysis.) Genotyping spotted twice as many resistant samples in the drug-break group (11% versus 5%), though the difference fell short of statistical significance in this small study group (P = 0.13).
Did drug breaks have a big impact on toxicity? No. After 12 months of study, rates of low-density lipoprotein (LDL) cholesterol, high glucose, self-reported lipodystrophy, and arterial hypertension were only marginally lower in the drug-break group. The break takers had modestly higher rates of anemia (low red cells) and thrombocytopenia (low platelets) than the always-on group, and similar rates of neutropenia (low white cells) and elevated liver enzymes.
If Staccato's holiday takers had the same progression rate as break takers in SMART, Staccato chief Hirschel observes in notes shared with the IAPAC Monthly, they would have endured 15 to 20 such "events." But they suffered only one -- a death from colon cancer.
Compared with unceasing therapy, CD4-timed breaks yielded higher rates of oral and vaginal candidiasis (P = 0.03) and thrombocytopenia (P = 0.06). People who never interrupted therapy had more neuropathy (P = 0.03) and diarrhea (P = 0.04). Of course CD4 counts dipped during the drug breaks (see "Trying to Explain the Differences," below). These ups and downs led Margolis to observe that "one seems to be trading one set of adverse events for another, at the cost of lower CD4 counts."17
As in earlier reports on Staccato, setting off on drug holidays didn't do much to rein in jumpy lipids. The steady and STI arms did not differ in triglyceride changes. Although the STI group ended up with a lower median cholesterol reading (4.6 versus 5.0 mmol/L, or 161 versus 193 mg/dL), both group levels lie in the optimal range. Resistance mutations emerged no more often with drug breaks than with steady therapy.
But Hirschel doesn't think CD4 thresholds hold the whole answer. Other differences he and Jintanat noted (Table 2) were the longer median prestudy treatment durations in SMART (72 months versus 15 months in Staccato) and the older median age in SMART (46 years versus 34 years in Staccato). But how these differences explain the divergent results remains unclear, Hirschel adds. Indeed, younger age didn't help Trivacan's African break takers. And Trivacan enrollees had even less prestudy antiretroviral experience than Staccato participants.
So one comes back to the CD4 cutoffs. How many rich-country clinicians these days would let a treatment-naive person drift under 250 cells/mm3 -- the SMART and Trivacan restart flags -- without getting edgy? Does it make more sense to let CD4s drop that low in people with a seven- or 72-month treatment history who may have a CD4 nadir under 200 cells/mm3, an AIDS diagnosis, or resistance mutations?
But baseline AIDS, baseline CD4s, and CD4 nadir had no apparent tie to progression risk in SMART. And only a handful of SMART break takers ever registered a count below 200 cells/mm3 during treatment suspensions (Table 3). SMART's STI assignees spent considerably more patient-years below 250 cells/mm3, and more still below 350 cells/mm3, especially when compared with the no-break group. More than one third of break takers in Staccato saw their T-cell counts waft under 350 cells/mm3, compared with 4% in Staccato's always-on group. And SMART break takers spent almost one third of their patient-years under 350 cells/mm3, compared with 8% of patient-years in SMART's steady arm.
These rates suggest at least two interpretations:
Another study detailed at CROI sheds light on this question. Charting opportunistic infection (OI) incidence in people with CD4 quotients assumed to protect against certain OIs, EuroSIDA researchers confirmed low rates of such rude surprises in 11,229 cohort members [abstract 783]. But the six OIs studied proved much more likely at these presumed "high" CD4 counts among people not taking a potent regimen.
Amanda Mocroft (Royal Free Hospital, London) and the EuroSIDA team looked for cytomegalovirus (CMV), Mycobacterium avium complex (MAC), or toxoplasmosis in people with more than 100 cells/mm3, Pneumocystis pneumonia or esophageal candidiasis in people with more than 200 cells/mm3, and tuberculosis (TB) in people with more than 300 cells/mm3. They tallied only 212 such OIs, but incidence per 1,000 patient-years proved consistently higher in people not taking a stout regimen:
Mocroft reckoned that starting potent therapy lowered the risk of the first five OIs by 54%. An even more telling observation, vis-à-vis STIs, was that lower CD4 counts -- even those still above 200 or 300 cells/mm3 -- meant a bigger OI risk. Opportunistic infection incidence per 1,000 patient-years measured 0.1 with a count at or above 500 cells/mm3, 0.4 at 400 to 499 cells/mm3, 1 at 300 to 399 cells/mm3, 2 at 200 to 299 cells/mm3, and 8 at 100 to 199 cells/mm3.
Viral load also made a difference. People with loads topping 1,000 copies/mL ran a 4.4 times higher risk of CMV, MAC, or toxo and a 2.63 times higher risk of Pneumocystis pneumonia or candidiasis (the most common OI in SMART). Perhaps progression risk during STIs reflects not only falling CD4 counts during the drug break, but also rebounding viral loads. Finally, just as in SMART, neither nadir nor baseline CD4 count affected OI risk in this EuroSIDA analysis.
The EuroSIDA study did not address non-AIDS progression -- the heart, liver, and kidney complications that bedeviled treatment interrupters in SMART. But several other CROI studies -- and a rich literature of earlier reports -- do suggest some answers about that SMART shocker. This article will consider such studies below (see "HAART, Kidneys, Liver, Heart," below), after reviewing CROI's other STI trials.
One thing SMART does not say, El-Sadr insisted, is that the relatively low 3% progression rate in the CD4-guided group means this kind of STI is worth the risk. Progression risk rose consistently in the STI group as follow-up continued and would have been worse without a sharp-eyed safety panel. And people inclined to take a drug break can't hire SMART's safety watchdogs to monitor their course.
What can one say about Trivacan? Although antiretrovirals came late to Africa -- too late for millions -- people now beginning therapy have one vast advantage over counterparts in Europe, the Americas, and even Thailand: They can learn from mistakes made in richer countries. Sadly, market forces stuck many Africans (and Thais) with a fixed-dose combination that includes stavudine (d4T), a toxic drug no longer recommended in lands of plenty. One hopes they learn faster about STI failings.
Bruno Marchou (Purpan Hospital, Toulouse) and ANRS confrères signed up 403 people with a CD4 count at or above 450 cells/mm3 and a viral load below 200 copies/mL for at least six months while taking the same regimen. The protocol excluded people with a CD4 nadir below 100 cells/mm3, anyone taking nevirapine (NVP) or abacavir (ABC), and anyone coinfected with hepatitis B virus (HBV) and taking tenofovir (TDF) or lamivudine (3TC). So Window's results don't apply to anyone with those traits.
The study group was a stable contingent of steady responders. Median baseline CD4 count measured 739 cells/mm3 in the STI group and 748 cells/mm3 in the control arm after an average five years of therapy. Only about 20% of enrollees were taking their first regimen when Window opened.
An intent-to-treat analysis at week 96 determined that seven people (3.6%) in the on/off group versus three (1.5%) in the steady-therapy group met the primary endpoint of a confirmed CD4 count under 300 cells/mm3. Despite this more than doubled rate of immunologic failure in the STI arm, the difference met the predetermined criterion establishing intermittent therapy as noninferior to continuous treatment. An on-treatment analysis counted six immunologic failures in the STI arm and only one in the control arm. Even this difference fell within statistical bounds of noninferiority for treatment interruptions.
The ANRS fixed the noninferiority threshold at less than or equal to a 7% difference in the upper bound of the 95% confidence interval when figuring the primary endpoint. The upper bound difference came to 5.6% in the intent-to-treat analysis. Analyzing these results on the NATAP Web site, Margolis suggests that although the number of CD4 drops below 300 cells/mm3 remained small, "an extra handful of events in the [treatment interruption] arm would have made it inferior."17 In the on-treatment analysis, the difference in this metric, 6.5%, edged right up to the preset threshold.
In notes shared with the IAPAC Monthly, Hirschel also voices some hesitation about embracing this kind of STI. Although the trial arms did not differ significantly in the proportion with a count below 300 cells/mm3 after 96 weeks of follow-up, he observes, they did differ in the proportion under 400 cells/mm3. And since CD4s drift inexorably downward during drug breaks, "there is a risk of low CD4 counts, were this regimen continued past 96 weeks."
Although no one in either arm ended up with AIDS before the trial ended, the STI group endured more cases of mucosal candidiasis (10 versus six), thrombocytopenia with bleeding (two versus none), and herpes simplex (four versus two), but not herpes zoster (two versus three). Eleven break takers had grade 3 or 4 thrombocytopenia during the study, compared with two in the steady-therapy group. Three people endured the acute retroviral syndrome when they suspended therapy.
At week 96, after being back on treatment for eight weeks, 81% in the drug-holiday group had a viral load under 400 copies/mL versus 90% in the always-on group, a significant difference (P Margolis makes another piquant point about these high breakthrough rates in a group doing well on therapy for so long. "One wonders if the study selected patients hoping for the [STI] arm," he muses, "driving up failure rates in the disappointed ones assigned to the [continuous therapy] arm." If that hunch were true, it would mean STI trials hold dangers for people in steady-therapy arms as well.
One should also note that fixed-interval drug breaks sometimes fail outright, as week-on-week-off interruptions did in Staccato.12
If STIs are supposed to make life easier for people with HIV, they failed miserably in this study group. Two thirds of those in the STI arm (66.5%) dropped out during the 24-month trial, compared with 19.4% in the steady-treatment group. People in the STI group had a 4.6 times higher risk of dropping out (95% CI 3.0 to 7.3) than people in the control arm.
Statistical analysis keyed to the primary endpoint -- a CD4 count above 500 cells/mm3 at month 24 -- determined that these STIs were "not noninferior" to uninterrupted therapy. Translated from StatSpeak into English, that means these STIs were worse than steady therapy. But some would question a CD4 threshold of 500 cells/mm3 as a primary endpoint in an STI trial because CD4s invariably ebb when people with HIV stop taking antiretrovirals.
The treatment arms did not differ when rated by the main virologic endpoint: just over 90% of both groups had a 24-month viral load below 400 copies/mL. And 24-month cumulative risk of virologic failure -- defined as a viral load above 400 copies/mL at any time in the steady-therapy group and at any time on treatment in the STI group -- stood at 24% in the control group and 26% in the STI group.
Resistance-related mutations piled up during each of the first four drug breaks. At the study's end 38 of 136 people (27.9%) randomized to try STIs had at least one mutation. Nearly three quarters of the people randomized to an STI (72.8%) took a nonnucleoside reverse transcriptase inhibitor (NNRTI), and 20% of them came away with a mutation. But half of the interrupters taking a PI -- and most took unboosted PIs -- ended up with PI mutations. Picking up mutations raised the risk of virologic failure 2.6 times in this trial.
These resistance findings suggest two lessons: Don't interrupt (or don't prescribe) a nonboosted PI regimen, and don't interrupt any NNRTI regimen. Palmisano and coworkers took the precaution of stopping the NNRTI a set number of days before stopping nucleosides in an attempt to avoid de facto NNRTI monotherapy as efavirenz (EFV) or NVP slowly clears from the body. The Window trial protocol (see preceding section) called for the same safeguard. But research convincingly shows high patient-to-patient variation in NNRTI clearance rates,18,19 as well as persistence of EFV levels high enough to provoke resistance up to 21 days after some people stop that drug.18 So stopping EFV seven days before stopping nucleosides may avert resistance in certain people, but not in others.
Multivariate analysis singled out two independent predictors of virologic failure in the Italian study -- unboosted PI therapy and archived mutations in viral DNA at study entry. Palmisano suggested checking DNA for occult mutations before letting a person essay STIs, but she could not explain how those mutants got archived in people supposedly taking their first antiretrovirals. One must pick from three possibilities: Some people had unreported failures before signing up for the study; they got infected with resistant virus; or the mutations evolved by chance.
Palmisano reported no new AIDS diagnoses or deaths in this two-year study, though she did not detail the fates of the many study dropouts. She counted 14 "serious adverse events" in each treatment arm but attributed only one of them to antiretrovirals or the trial protocol -- acute retroviral syndrome in one break taker.
Daniel Skiest (Baystate Medical Center, Springfield, Massachusetts) and ACTG coworkers enrolled 167 people with a pretreatment nadir above 350 cells/mm3 (median 436 cells/mm3), current CD4 count above 350 cells/mm3 (median 833 cells/mm3, with 92% above 500 cells/mm3), and a viral load below 55,000 copies/mL (median pretreatment load 26,611 copies/mL, with 71% below 50 copies/mL). They had taken antiretrovirals for a median of 4.5 years, and all said they wanted a drug holiday.
At the end of follow-up only 17 of 144 people who stayed in the study (11.8%) had a confirmed CD4 count at or below 250 cells/mm3, which was part of a composite primary endpoint that also included a new AIDS "event," death, or the need to resume therapy.
In this study group a nadir count below 400 cells/mm3 raised the risk of reaching the primary endpoint 1.95 times (P = 0.02), and a viral load above 400 copies/mL at study entry raised the risk 2.75 times (P = 0.02). In a multivariate analysis a lower CD4 nadir and a viral load above 50 copies/mL at entry independently raised the risk of reaching the primary endpoint. A hefty majority of people starting antiretrovirals today -- anywhere in the world -- probably have sub-400-cell nadirs.
Fourteen people restarted the regimen they stopped, and five of them (36%) endured virologic failure, a sad result. Thirty-two people resumed treatment with a new regimen, four (12.5%) of which failed.
Of the three people who died of MI and the one who died with arrhythmias, three had a history of coronary artery disease. Only one of the five deaths came at a low CD4 count (137 cells/mm3), while the other four deaths came at counts of 378, 523, 626, and 845 cells/mm3. As Margolis, cochair of this study, writes in his CROI review, "the possibility that morbid events not classically associated with HIV infection occur at an increased rate in patients off [ART] is of potential concern, and worthy of further study."17
Borkowsky and the PACTG 1015 team enrolled 14 children whose viral loads lay below 400 copies/mL for more than one year and below 50 copies/mL when the study began. They started with a three-day STI, then added two days to each successive drug break. Three children dropped out of the trial for personal reasons, and one can imagine many "personal reasons" interfering with a complicated on-and-off protocol like this.
The good news is that median off-treatment viremia peaked at 26,667 copies/mL during cycle seven, then never rose above 10,000 copies/mL, even though subsequent STIs grew longer and longer. Among eight children who reached STI cycle 13, seven had at least a 10-fold rise in HIV-specific CD4 cells, and HIV-specific lymphoproliferative responses also improved.
Median CD8 percentage climbed from 24% to 29.5%, but median CD4 percentage dipped from 40.5% to 36.5%. Resistance mutations emerged in three children -- the 3TC-provoked M184V in one, the PI mutation V82A in one, and M184V plus assorted PI mutations in the third.
These mixed early results with a complex strategy suggest much remains to be learned about crafting safe STIs for children.
But this empyrean logic rarely intersected a rife quadrangle of research showing that life-sustaining end organs -- the heart, liver, and kidneys -- seemed to end up better after people started antiretrovirals. So should we be surprised that people who interrupted therapy in the SMART trial endured more cardiovascular, hepatic, and renal setbacks? (See "Are There Any Smart STIs?," above.)
Kidney disease did not burst onto the HIV scene when people started taking indinavir (IDV) and TDF. Paul Klotman (Mount Sinai Medical Center, New York), the renal disease dean among HIV experts, noted in a recent study of acute renal failure that HIV-infected people risk renal disease not only from antiretroviral toxicities, but also from HIV-associated nephropathy or immune complex disease,20-23 thrombotic microangiopathy, 24-26 and HBV- or hepatitis C virus (HCV)-related kidney disease.23,27-29 People who need hospital care for HIV complications also risk acute renal failure from volume depletion, hemodynamic stress, and urgently administered kidney cripplers like radiocontrast. Not a short inventory.
A prospective cohort study of 754 people with HIV found acute renal failure more likely in people with a CD4 count under 200 cells/mm3, a viral load above 10,000 copies/mL, AIDS, or HCV infection.30 Possibly because people with these markers of advanced disease were obvious candidates for therapy, antiretroviral treatment also correlated with acute renal failure in this analysis.
In Klotman's study of New York state residents released from the hospital, diagnoses of acute renal failure proved more frequent in people with HIV than in people without HIV.31 That heightened risk of renal failure dropped from a 4.62-fold elevation in the days before potent antiretroviral combos to a 2.82-fold elevation in 2003.
Klotman did find a 2-fold higher incidence of acute renal failure in 2003 than in 1995. Yet the incidence rose in both the HIV and non-HIV cohorts. And potent antiretrovirals promote longer survival of more people who also have diabetes and liver disease -- both of which raised the risk of acute renal failure in Klotman's 2003 cohort.
Despite fears that antiretroviral-riled liver enzymes would fuel an epidemic of liver disease in people with HIV, much research now shows that people with hepatitis virus infection do better with than without antiretrovirals. Not long after PIs reversed the course of HIV disease, French researchers found that PIs also slowed the pace of liver fibrosis in people coinfected with HCV.32Lack of PI therapy -- as well as a lower CD4 count -- independently paved a faster track to cirrhosis.
A retrospective German study of HIV/HCV-infected people seen from 1990 through 2002 chalked up lower liver death rates in people taking potent antiretroviral combinations (0.45 per 100 person-years) than in people treated only with nucleosides (0.69 per 100 person-years) or people who never took antiretrovirals (1.70 per 100 person-years).33 Potent regimens lowered the risk of liver-related death almost 90% in this analysis.
More recently, an 800-person cohort study with three years of follow-up found that starting antiretrovirals earlier in the course of HIV infection lowered the risk of non-AIDS deaths including liver disease.34 And the multicohort D:A:D study determined that up to seven years of antiretroviral therapy do not raise the risk of liver-related death.35
36 buttressed widely held opinion that HIV drugs hammer the heart. But which antiretrovirals? And does the trend persist?
The D:A:D team came to CROI with evidence indicting PIs -- and perhaps exculpating NNRTIs -- for boosting the MI risk, apparently by jacking up lipids [abstract 144]. But this work also charted a dwindling MI rate in the 23,437-person D:A:D cohort. And other research detailed at the conference failed to link PIs to a heightened risk of heart disease.
D:A:D embraces 11 cohorts from Australia, Europe, and the United States with follow-up through February 2005. All cohort members took a three-drug regimen including either a PI or an NNRTI, but median duration of PI therapy far exceeded that of NNRTI treatment (3.0 years versus 0.9 years). The median age of D:A:D members when they entered the cohort in 2000-2001 was 39 years (interquartile range [IQR] 34 to 45 years).
Through 94,469 person-years of follow-up, D:A:D tabulators counted 345 MIs for an incidence of 3.7 per 1,000 person-years. Stated another way, a researcher would have to monitor 1,000 antiretroviral-treated people for a year to count 3.7 heart attacks. Nina Friis-Møller (Copenhagen HIV Program) reported that MI risk in the cohorts dropped by half from 1997 through 2003-2004 (RR 0.50) -- after statistical adjustment for demographic and cardiovascular risk factors as well as antiretroviral exposure.
When statisticians adjusted that analysis for the latest total cholesterol, "good" high-density lipoprotein (HDL) cholesterol, and triglycerides, the improvement in MI incidence vanished. That finding suggests that improving lipid profiles -- perhaps because of switching from PIs to NNRTIs, taking antilipid drugs, or getting serious about diet and exercise -- apparently circumvented the potential for an increased MI risk as the cohort aged. An earlier reckoning of D:A:D data showed a switch away from PIs to NNRTIs, a doubling in use of lipid lowerers (from 3.5% to 7.6%), a small drop in proportion of current smokers (from 47.4% to 45.0%), and a big jump in proportion of ex-smokers (from 16.0% to 24.7%) from 2000 to 2003.37
The latest D:A:D analysis figured a 16% higher MI risk with each year of antiretroviral therapy. And when D:A:D statisticians broke that down by drug class, they found a 16% higher MI risk with every year of PI therapy (P = 0.0001). Factoring lipid changes into that equation whittled the MI risk with each extra year of PIs down to 10%. Risk of MI changed not at all with added years of NNRTI therapy, but Friis-Møller reminded attendees that the analysis rests on relatively fewer years of NNRTI therapy.
D:A:D's math mavens also ran sensitivity analyses considering people who took PIs but never took an NNRTI, and people who tried an NNRTI but never tasted a PI. Once more, every 12 months of PI therapy inflated the MI risk 16%, while MI odds seemed to dwindle for each year of NNRTI treatment. Despite the cohort's huge size, statisticians still can't weigh the impact of individual antiretrovirals.
If PIs swell the risk of heart attacks, why would men taking them more than six years have fewer MIs and a lower relative risk than men taking them two to six years? The unsurprising answer mirrors D:A:D disclosures -- people are smoking less, taking antilipid pills, and reining in high blood pressure. Also, people with bad PI-induced lipid profiles would probably not keep taking PIs year after year if they could switch to a non-PI combo or to the more lipid-friendly atazanavir (ATV).
From 2002 through 2005 the percentage of smokers among 5,430 HIV-infected men in Klein's cohort dropped from 21.1% to 17.9%, and from 2000 to 2005 the percentage with systolic blood pressure above 140 mm Hg fell from 22.0% to 11.6%. The proportion of PI-treated men taking lipid-lowering agents rocketed from 1% in 1997 to 27% in 2005 (P < 0.001). And from 2003 through 2005 the percentage of PI takers using ATV climbed from 6% to 35% (P < 0.001). (But ritonavir [RTV]-boosted ATV does jack up cholesterol quotients in some children. See "Arteries and Lipids in Adults and Kids," below.)
Probably because of the moves toward ATV and lipid lowerers, 59.3% of the HIV cohort had total cholesterol below 200 mg/dL in 2004 to 2005 compared with 43.6% in 2000 to 2001. In that same period the percentage with HDL cholesterol above 39 mg/dL edged up from 47.5% to 52.2%.
Overall age-adjusted coronary heart disease (CHD) and MI rates remained significantly higher in men with HIV than in men without HIV (6.0 versus 2.9 CHD events per 1,000 person-years, P < 0.001; MI 3.6 versus 2.2 events per 1,000 person-years, P = 0.002). But the Framingham Risk Score for CHD among HIV-infected men actually eased from 2000-2001 to 2004-2005 (8.6% to 8.3%), even though the cohort grew significantly older (P < 0.001).
A second US cohort inquest, the HIV Outpatient Study (HOPS), confirmed the importance of traditional cardiovascular risk factors in people with HIV [abstract 735]. But this analysis by Kenneth Lichtenstein (University of Colorado, Denver) winkled out no links between specific antiretrovirals or classes and CHD.
From 1993 through 2005 the HOPS team tallied 57 MIs, 22 strokes, 86 cases of coronary artery disease, and 22 cases of peripheral vascular disease in 8,024 people with HIV, most of them white men. Diagnoses of MI peaked at the turn of the millennium, then fell as cohort members started taking more antihypertensives and antilipid drugs. Incidence of MI among cohort members approached 3.5 per 1,000 person-years in 2000, remarkably close to the overall 3.7 incidence in D:A:D, then plunged to about 1.1 in 2005.
Multivariate analysis involving 1,807 people who sustained 57 cases of cardiovascular disease flushed out five independent predictors of heart disease:
Individual antiretrovirals, antiretroviral classes, and switches from older PIs to ATV or any NNRTI had no impact on the risk of cardiovascular disease. But among 363 people with hyperlipidemia, taking antilipid agents cut the risk of heart disease 66% (AOR 0.34, 95% CI 0.14 to 0.85, P = 0.021). In the high-lipid subset smoking upped the odds of heart disease 2.22 times (P = 0.057), age over 40 years made heart disease 2.38 times more likely (P = 0.087), and diabetes raised the risk 2.45 times (P = 0.052).
For now the bottom line from these three studies appears to be that PIs make MIs more likely. But D:A:D, the California study, and HOPS all strongly suggest that easing lipids and other traditional risk factors may wipe out the MI risk pinned on PIs.
38 But the conviction remains in question, as other CROI studies (including the just-reviewed HOPS data) failed to turn up PI-incriminating evidence.
A three-year study of matched threesomes found no significant differences in carotid artery intima-media thickness (IMT) -- an accepted marker of coronary artery atherosclerosis and cardiovascular disease -- in HIV-infected people taking a PI, HIV-infected people who never took a PI, and people without HIV, reported Judith Currier (University of California, Los Angeles) [abstract 145]. But the PI takers did have moderately faster carotid wall thickening, and 61% of the PI group did not get a boost from RTV.
Earlier work by a different group found a 10 times higher IMT gain in 121 people with HIV after one year of follow-up than in 27 age- and gender-matched uninfected controls, and lower CD4 nadirs inflated the risk of thickening carotid walls.39 Another one-year study of 346 HIV-infected people graphed a slow but significant IMT increase but did not tie it to type or duration of ART.40 A third study of 233 people in the same HIV cohort traced a rise in median IMT during the first 12 months of ART, then a drop through 36 months of follow-up.41 In that time more people started taking lipid lowerers, stopped taking PIs, and stopped smoking. These Aquitaine cohort researchers posited a succinct conclusion: "The progression of atherosclerosis in HIV-infected patients can be controlled."
Currier and ACTG 5078 colleagues matched 44 sets of three people for age, race, gender, blood pressure, smoking, and menopausal status. They excluded people with a history of MI, diabetes, or uncontrolled hypertension. Baseline IMT proved similar in the PI-treated, PI-naive, and HIV-uninfected sets and barely changed through 144 weeks of follow-up (Table 5).
Higher fasting blood sugar and fasting insulin did correlate with IMT in this analysis. Currier noted that the small size of the cohort and normal lipid levels among some PI-treated people may have diluted the study's power to spot a difference between groups. So even if this trial does not completely vindicate PIs in promoting cardiovascular disease, it hints that PI therapy without lipid abnormalities may not pose a grievous threat to the heart.
A second and much larger case-control study did chart significantly thicker carotid walls in an HIV cohort than in a large cohort of healthy people not tested for HIV [abstract 738]. But this age and gender matching of 292 HIV-infected "cases" and 1,168 "controls" could not tie carotid IMT changes to any antiretroviral class regardless of carotid segment scanned, antiretroviral treatment duration, or statistical model.
The HIV cohort and the matched, prospectively tracked Carotid Atherosclerosis Progression Study group had grayed to a median age of 53.2 years, reported Christoph Stephan (JW Goethe University Hospital, Frankfurt). But the control group had a worse cardiovascular risk profile, with significantly higher body mass index, systolic and diastolic blood pressure, need for blood pressure medication, total cholesterol, and LDL cholesterol. More people with HIV took antilipid drugs (18.5% versus 4.1%, P < 0.0001), and the HIV group averaged 27.0 cigarette pack-years compared with 13.4 in the control group (P < 0.0001).
Because of the inability to correlate thicker carotid arteries -- measured at six sites -- to antiretroviral therapy, Stephan and colleagues blamed the worse IMT on HIV itself. Statisticians figured that the heightened IMT risk ratio in people with HIV made them four to five years older in "vascular age" than the control group and gave them a 4% to 14% higher risk of one "vascular event" in five years.
A single-center cross-sectional study of 132 people with HIV found two independent predictors of subclinical atherosclerosis defined by a carotid IMT above 0.8 mm or plaque detection -- combination antiretroviral experience (OR 10.5, 95% CI 2.8 to 39) and a 10-year coronary risk at or above 10% (OR 4.2, 95% CI 1.5 to 12).42
The study included 93 people taking antiretrovirals and only 39 treatment-naive people, 32 of whom (82%) earned a very low-risk rating on the Framingham scale (10-year coronary risk below 5%). Besides consisting predominantly of antiretroviral-treated people, the group with subclinical carotid atherosclerosis was the oldest, had the lowest CD4 counts, had the highest HIV duration and the highest Framingham risk scores, and had abnormal fat distribution. So independent scrutiny of the statistical methods might prove worthwhile.
Children with HIV also have thicker carotid walls than uninfected children, according to results of a small, single-center, case-control study by Grace McComsey (Rainbow Babies and Children's Hospital, Cleveland). Besides having higher carotid IMTs, children with HIV had higher levels of myeloperoxidase (a cardiovascular risk surrogate) and higher lipids than did controls without HIV [abstract 691].
The analysis involved 27 children with HIV and 17 age-, gender-, race-, and weight-matched controls. The HIV-infected children were all taking a stable regimen, 52% with an NNRTI, 41% with a boosted PI, and 7% with both an NNRTI and a boosted PI. Eighteen children in the HIV group were girls, and 18 were African American. The study excluded children with a family history of diabetes or premature cardiovascular disease.
Median myeloperoxidase proved significantly higher in the HIV group (1,299 versus 746 pmol/L, P = 0.01), and median C-reactive protein, another heart risk factor, was nonsignificantly higher in the children with HIV (0.67 versus 0.43 mg/L, P = 0.15). The HIV group also had significantly higher fasting total cholesterol, non-HDL cholesterol, and triglycerides than did controls.
Common carotid IMT was comparable in the two groups, but median left internal carotid artery IMT was significantly higher in children with HIV (0.50 versus 0.45 mm, P = 0.038). The HIV group also had a higher median right internal carotid artery IMT, but that difference fell shy of statistical significance (0.50 versus 0.45 mm, P = 0.08).
An earlier, similarly small case-control study in France found no greater carotid IMT in children with HIV than in those without HIV.43 None of the vascular variables measured -- IMT, endothelium-dependent dilation, and endothelium-independent dilation -- differed between 34 children taking antiretrovirals and 15 treatment-naive children with HIV. So HIV's impact on children's arteries remains an open question. Answers may come from McComsey's planned 144-week follow-up of her study group.
José Ramos (Hospital 12 de Octubre, Madrid) and colleagues at other Spanish hospitals reported a significant jump in high cholesterol rates (more than 200 mg/dL) in 91 children starting a PI -- from 9% before treatment to 48% after 24 months (P Despite generally good lipid reports in adults starting or switching to ATV, some US and South African children taking RTV-boosted ATV saw cholesterol readings climb toward the danger zone, reported Grace Aldrovandi (Children's Hospital of Los Angeles, California) and PACTG 1020A colleagues [abstract 689]. Total cholesterol rose from below to above 180 mg/dL in eight of 81 children (10%) after 24 weeks of ATV/r and in eight of 48 (17%) after 48 weeks. Three of 82 children (4%) wound up with cholesterol readings above 200 mg/dL at week 24, as did four of 49 (8%) at week 48. Aldrovandi traced a significant correlation between RTV boosting and higher cholesterol at weeks 24 (P = 0.01) and 48 (P = 0.03). Unfasted triglyceride levels rose nonsignificantly.
Those findings emerged from a 3,456-person analysis of SMART enrollees with an electrocardiogram (ECG) and without evidence of earlier heart disease. Their age averaged 46 years, one third were black, and they had taken antiretrovirals for an average six years. Only 5.3% had never tried antiretrovirals, 32.2% never sampled a PI, and 36.4% claimed NNRTI naiveté. Everyone had a CD4 count above 350 cells/mm3, and 66.7% had a viral load under 400 copies/mL. Average total cholesterol stood below the danger zone at 193 mg/dL (4.99 mmol/L), but the group's average HDL cholesterol was low (43 mg/dL, 1.12 mmol/L), and their triglycerides were high (231 mg/dL, 2.59 mmol/L).
Andrew Carr (St. Vincent's Hospital, Sydney) reported that 10.1% had ECG evidence of asymptomatic ischemic heart disease -- Q waves and/or ST depression. In contrast, only 1.5% had symptomatic ischemic heart disease. Carr tallied 2.1% with arrhythmias and 1.4% with left ventricular hypertrophy.
Neither gender, race, smoking, nor antilipid therapy swayed the risk of asymptomatic ischemic heart disease, but four factors did -- older age, current antihypertensive therapy, diabetes, and one to three years of NNRTI experience, the last of which lowered the risk (Table 6). Carr and SMART colleagues figure that current antihypertensive therapy reflects longstanding hypertension. Current lipid levels may not have nudged the risk equation, they speculate, because lifelong lipid levels may "have changed recently with [ART] (95% of patients) and/or lipid-lowering therapy (15%), so altering any association between lipid levels and ischemic heart disease."
These are interim data, Carr noted. The SMART team plans to analyze asymptomatic ischemic heart disease risk in another 1,000 trial participants.
But why would ischemia threaten people with HIV infection if antiretrovirals can't be blamed or -- in the case of NNRTIs -- may actually lower the risk? The HIV literature is pregnant with one fecund suggestion: infection-induced inflammation and cell damage. And a CROI study may help bring this pregnancy to term.
Sizing up endothelial function is one way to get a fix on cardiovascular health. As observed by David Nolan and Simon Mallal (Royal Perth Hospital, Perth, Australia),44 "endothelial dysfunction is an early phase of atherogenesis in which the vasodilatory properties of the endothelium are diminished, preceding the development of atherosclerotic plaque in the arterial wall."45 Cell studies link HIV's gp120 protein46 and Tat protein47,48 to endothelial damage, while gp120 incites killing of neonatal rat ventricular myocytes and human coronary artery endothelial cells.49 Tat and the inflammatory cytokine TNF-alpha apparently team up to help leukocytes stick to endothelial cells.50 Mice genetically jogged so their cells express HIV-1 nef get heart disease.51
What does HIV do to human endothelium still inside humans? You can get a good noninvasive guess by mapping flow-mediated dilation of the brachial artery. A study comparing 22 PI-treated people and 15 PI-naive people with HIV linked PI therapy to high triglycerides and triglyceride-rich lipoproteins, then tied those changes to impaired endothelial function.52 Nolan's study suggested preserved endothelial function in people with advanced infection.44
But a more recent prospective case-control study in Israel charted an inverse correlation between viral load and endothelial dysfunction -- the higher the load, the worse the dysfunction (P 53 That finding bolsters the idea that infection -- especially an infection left uncontrolled for years, or intermittently controlled as treatment stops and starts -- is bad for the heart. As the Israeli researchers note, endothelial dysfunction, hypercoagulability, hypertriglyceridemia, and abnormal coronary artery pathology plagued people with HIV infection long before they started taking PIs. But as results of these three studies suggest, no one can say for sure whether PIs or HIV is tougher on endothelial cells.
At the 13th CROI, Priscilla Hsue (San Francisco General Hospital) offered evidence that atherosclerotic progression to heart disease in people with HIV may involve the infected person's ongoing struggle not only with HIV, but also with CMV [abstract 741]. Almost everyone who reaches the age of sexual consent -- and then consents -- has CMV. People with a working immune system, including most people taking ART, keep CMV under control by directing CMV-specific T cells against the virus. But what does that ever-smoldering inflammatory battle against CMV do to the circulatory system?
Because CMV apparently hastens atherosclerosis in heart transplant patients,54-56 and because HIV infection upsets CMV-specific immunity, Hsue hypothesized that immune responses to this herpesvirus may explain faster atherosclerosis in people with HIV. To explore this possibility, she mounted a case-control study of 93 people with HIV and 37 uninfected controls, then gauged carotid IMT in 12 predefined segments. People with HIV had to be on a stable antiretroviral regimen for at least a year (and 92% were) or taking no antiretrovirals. Median CD4 count in the HIV group measured 354 cells/mm3, and 57% had an HIV load under 75 copies/mL
Compared with controls, the HIV-infected group had higher proportions of smokers (42% versus 19%, P = 0.03) and people with hypertension (30% versus 11%, P = 0.03). Controls had significantly higher HDL and LDL cholesterol, while people with HIV had significantly higher triglycerides.
Median carotid IMT proved significantly greater in the people with HIV (0.95 mm) than in controls (0.68 mm) (P < 0.001), a finding confirming results of Hsue's earlier IMT study.39 After statistical adjustment for other cardiovascular risk factors, the HIV group averaged 27% thicker carotid walls than the non-HIV controls (P < 0.001). People with HIV also had significantly higher median levels of C-reactive protein (CRP), a heart disease marker, than did controls (1.1 versus 0.8 mg/L, P = 0.05). But neither traditional nor HIV-specific cardiovascular risk factors correlated with CRP readings.
CD4- and CD8-cell activation proved significantly higher in the HIV group (P "Taken together," Hsue concluded, "these findings suggest that the accelerated atherosclerosis in HIV disease may be mediated by an increased inflammatory response that is directed against CMV." But does that finding contradict the assumption that controlling HIV with antiretrovirals is enough to stop progression to atherosclerosis? Hsue did not distinguish between virologic responders and nonresponders in her analysis. And she told the IAPAC Monthly this line of research has a way to go before answering that question.
Starkly refutatory results like SMART's invariably inspire the unhelpful bromide that new findings raise more questions than they answer. Even when that cliché proves true, ducking behind it does not excuse clinicians from answering whatever questions they can.
SMART, for example, is hardly the first study to find that lack of antiretroviral therapy may pose a bigger threat of organ damage than taking these imperfect drugs day after day. In all the talk on this seminal study -- at least all the talk this reporter heard -- no one mentioned the 9th CROI study of Samuel Bozzette (University of California, San Diego) showing that heart disease and stroke admissions to Veterans Administration (VA) hospitals fell by half as HIV-infected vets took ever-more-potent antiretroviral combinations.57
And swallowing PI or NNRTI regimens for four or more years sent no more people to the hospital for heart trouble or stroke than taking those combinations for four, three, or two years. The risk of admission for cardiovascular or cerebrovascular disease slipped 21% for people taking PIs more than 24 months versus not at all, though that improvement lacked statistical significance. From 1995 though 2001, as PI and NNRTI regimens became routine, all-cause mortality plunged from 21.3 to 5.0 deaths per 100 patient-years.
Bozzette's conclusion back in 2002 sounds sounder than ever today: "Fear of accelerated vascular disease should not deter patients and providers from using the highest-quality care for HIV, as defined by the use of combination [ART] that is compatible with current guidelines."
Not long after Bozzette published the VA study, other US researchers reported that severe clinical complications -- "grade 4 events" like liver and heart disease -- had become more common than new AIDS diagnoses among 2,947 people taking potent regimens in clinical trials (11.4 versus 5.6 per 100 person-years).58 That finding fanned fears that triple therapy with PIs or NNRTIs caused these new complications.
Now it appears that fear got overblown. As Bozzette explained in a recent e-mail, "many of the newer complications that came with HAART were attributed to PIs or NNRTIs because they came together." But people forgot to factor in the "huge increase" in NRTI use around the same time and the much longer survival of people with HIV infection.
The latest big cohort studies confirm the end-organ-protecting prowess of strong antiretrovirals. In a just-published 22-cohort Concerted Action on SeroConversion to AIDS and Death in Europe (CASCADE) analysis of 7,680 people with HIV, for example, mortality from organ failure dropped by more than half after potent combinations arrived, from 1.0% to 0.4%.59 Meanwhile, the death rate from non-AIDS infections fell from 8% in the 15 years after HIV seroconversion in pre-1997 cohort members to 3% in more recent years. And what about classic AIDS-defining infections? Of course they kill many fewer people today than they did before 1997. But CASCADE researchers found that classic OIs remain the leading cause of death of people with HIV -- and a real good reason to take ART.
When viewed from this perspective, SMART's finding that people who interrupt antiretrovirals run a higher risk of disease progression and death -- from AIDS and "non-AIDS" causes -- than people who keep taking their drugs sounds as intuitive as can be.
And how does Bozzette put these pieces together?
"I know this sounds moronic," he told the IAPAC Monthly, "but my reaction is that HIV is a bad virus that should not go unchecked. I cannot imagine that we would not suppress from day one of care if we could."
Mark Mascolini writes about HIV infection (firstname.lastname@example.org).