Can Drugs Catch Up With HIV?
12th Conference on Retroviruses and Opportunistic Infections
February 22-25, 2005, Boston
Already a decade has passed since Louis Mansky and Howard Temin discovered that HIV-1 cranks out one mutation per genome per replication cycle.1 And antiretrovirals haven't caught up with the virus yet.
But it has not been for lack of effort.
Since early studies of nelfinavir (NFV) and bootless moil to prove adefovir's antiretroviral prowess, researchers have struggled to devise drugs that rein in resistant virus. Small battles have been won, but the big victories belong to HIV.
That tide may have turned (until the next new moon, anyway) at the 12th Conference on Retroviruses and Opportunistic Infections (CROI) held this year in Boston. The slide sessions on antiretroviral therapy (ART) showcased results on two protease inhibitors (PIs) -- TMC114 and tipranavir (TPV) -- that surpassed other PI options in slowing resistant virus. Both studies involved people with triple-class failure and plenty of protease mutations.
From the poster halls came word that triple-class failure itself may be on the wane, along with data confirming improved long-term responses to potent up-front regimens. But several groups agreed that year-in-year-out viral suppression does not repay the CD4 debt amassed by a prodigal HIV. And one team mapped a mortality plateau after the post-1995 freefall.
Part I of the IAPAC Monthly's 12th CROI review construes those trends, along with some good news -- and some worrying news -- on hepatitis coinfection. The May 2005 issue will report other clinically useful resistance data, news on antiretroviral side effects, and dark developments in the epidemic among gay men.
The international phase 3 trials pitting TPV against other PIs bear the memorable monikers RESIST 1 and 2. Researchers scouting out the best dose of another resistance-routing PI, TMC114, answered with POWER 1 and 2. The trial-naming contest may be a dead heat. But because TPV's RESIST results made their first splash in late 2004, TMC114 caused the bigger ripples at CROI.
Tipranavir and TMC114 were not the only ritonavir (RTV)-boosted PIs scrutinized at the Boston conference. Nor were PIs the only new resistance-ready antiretrovirals surveyed. The conference also told two more knells in obsequies for a once-favored resistance strategy -- presalvage treatment interruption.
TMC114 Clears First Hurdle Handily
Not until CROI's penultimate slide talk did attendees hear about what may be this year's ultimate PI answer to PI-resistant virus, Tibotec's RTV-boosted TMC114. People taking the highest dose of this resistance-confounding compound (600/100 mg twice daily) lopped 1.85 logs off their starting viral load in 24 weeks, while people taking a comparison single or double RTV-boosted PI whittled away only 0.27 log (P<0.0001) [abstract 164LB]. Richard Haubrich (University of California, San Diego) reported that nearly half of those taking 600/100 mg of TMC114/RTV ended 24 weeks with a viral load under 50 copies/mL, compared with 9 percent in the control group.
With colleagues at 90 sites in 14 countries, Haubrich randomized people with triple-class experience to one or two licensed boosted PIs or one of four TMC114/RTV doses -- 400/100 or 800/100 mg once daily, or 400/100 or 600/100 mg twice daily. The study group had a median of three primary protease mutations (everyone had at least one), and all enrollees started another set of antiretrovirals picked to subdue their mutant horde.
The 497 trial participants began salvage with a mean 79-fold resistance to lopinavir (LPV), 38-fold resistance to atazanavir (ATV), 29-fold resistance to saquinavir (SQV), but only 4-fold resistance to TMC114. Equivalent proportions in the TMC114 arms and the control arms (about 15 percent) had tried the fusion inhibitor enfuvirtide (ENF), and 8 percent once sampled TPV.
This heavily pretreated cohort had challenged HIV with an average of 11 antiretrovirals. They had fairly advanced disease, with a median CD4 count of 136 cells/mm3 in the TMC114 group and 163 cells/mm3 in the control arm. Baseline viral load averaged 4.61 logs in the TMC114 group and 4.47 logs in controls (about 40,700 and 29,500 copies/mL). (See note 2 for more baseline details.) The protocol blinded researchers to the TMC114 dose.
By week 24 the control arm had lost 51 people (51 percent), compared with 9 percent in the TMC114 groups. Toxicity rates looked similar in the TMC114 and comparison arms. While 4 percent stopped TMC114 because of side effects, 2 percent had treatment-limiting toxicity with comparison PIs. Notably, rash did not emerge as a TMC114 complication in this study, as it had in trials of unboosted TMC114.
The POWER team defined virologic failure as less than a half-log dip in viral load after week 8, less than a one-log drop after week 12, or consecutive loads a half-log above the lowest value reached. In a 24-week planned interim analysis using those criteria, Haubrich registered a 2 percent failure rate with TMC114 and a 43 percent rate with other PIs.
A noncompleter-equals-failure analysis found the steepest average 24-week viral load drop with 600/100 mg of TMC114/RTV twice daily:
While 47 percent taking the 600/100-mg dose reached a 24-week load below 50 copies/mL in a noncompleter analysis, only 9 percent taking a comparison PI regimen reached that mark (P<0.01). With the same type of analysis to figure sub-50-copy rates, the highest dose of TMC114 outdid the control PIs in subgroups of ENF-naive people starting that entry blocker (67 percent versus 16 percent), people not taking ENF (37 percent versus 8 percent), and people having three or more primary PI mutations (48 percent versus 5 percent), having more than 4-fold baseline susceptibility to TMC114 (45 percent versus 5 percent), or starting no other antiretrovirals rated active at baseline (31 percent versus 0 percent).
Tighter viral control with 600/100 mg of TMC114/RTV translated into the best CD4 gain in that group (75 cells/mm3), compared with 50- to 55-cell jumps with the other TMC114 doses and a 15-cell uptick with the other combinations.
As good as these results look, it bears remembering that they represent 24 weeks of data from a trial subset. POWER 1 and 2 will continue through 96 weeks while researchers plan phase 2 studies to test the 600/100-mg twice-a-day dose.
TPV Sooner -- or TMC114 Later?
Clinicians with candidates for TMC114 will probably have to make a tough choice by year's end -- whether to assail that PI-resistant virus with TPV/RTV or whether to wait for the TMC alternative. Reports at this year's CROI detailed resistance-specific responses to TPV/RTV [abstract 104] and dovetailed the two RESIST trials to compare TPV/RTV with LPV/RTV [abstract 560].
RESIST enrollees shared some baseline traits with the POWER cohorts assembled to test TMC114: RESIST participants randomized to TPV/RTV started with a median viral load of 4.82 logs (around 66,000 copies/mL) compared with 4.61 logs (about 40,700 copies/mL) among people starting TMC114/RTV in POWER. Respective median baseline CD4 counts measured 155 cells/mm3 in RESIST and 136 cells/mm3 in POWER.
While 60 percent assigned to TPV/RTV in RESIST claimed three or four primary PI mutations, three was the median primary mutation number in POWER. Median numbers of antiretrovirals on study entry charts were 12 in RESIST and 11 in POWER. Everyone in both trials had triple-class experience, and everyone got a resistance-tuned background regimen plus their new PIs. Instead of defining virologic failure benchmarks, as in POWER, the RESIST team defined response as at least a 1-log (10-fold) drop in viral load.
RESIST researchers randomized 582 people to start 500/200 mg of TPV/RTV twice daily and 577 to start LPV/RTV, SQV/RTV, indinavir (IDV)/RTV, or amprenavir (APV)/RTV. Six months later, David Cooper (University of New South Wales, Sydney) figured a 39.6 percent response rate in the TPV/RTV group versus 21.4 percent with LPV/RTV (P<0.0001) by a noncompleter-equals-failure analysis [abstract 560]. Among LPV-naive people, though, the better 24-week response with TPV did not significantly exceed that with LPV (45.3 percent versus 36.1 percent).
A 24-week noncompleter comparison between TPV/RTV and the combined comparison arms showed a consistently better response to TPV/RTV according to the number of other background antiretrovirals judged active at baseline:
In a more detailed resistance analysis, Jonathan Schapiro (Sheba Medical Center, Tel Aviv, and Stanford University) considered three sets of baseline mutations -- (1) all changes from the protease consensus sequence, (2) primary mutations at codons 30, 46, 48, 50, 82, 84, and 90, and (3) substitutions at codons 33, 82, 84, and 90, mutations originally rated particularly troublesome for TPV [abstract 104].
Tipranavir/RTV outperformed the other RTV-boosted PIs in comparisons based on all three mutation sets. Among people with one or more primary mutations (set 2), more people randomized to TPV/RTV lowered their viral load at least 1 log (10-fold) by week 24:
Despite the nonsignificant difference in the third comparison, the 1-log response rate to TPV/RTV stayed steady as primary mutations piled up, while the response to comparison PIs dropped after two appeared. How tightly a PI keeps HIV in its clutches as mutations stack up is literally a matter of hanging on, explained François Clavel (Bichat-Claude Bernard Hospital, Paris) in a symposium lecture [abstract 180].
Protease inhibitors work by catching hold at the very center of the butterfly-shaped protease protein. A few protease mutations disrupt drug binding in this pocket, Clavel noted, but most do not. These other mutations pop up all over the protein's "wings" and stretch the binding pocket ever wider, loosening the PI's grasp. Think of climbing a tree, Clavel proposed. You do best when you use both hands and both feet. If you hang from the tree with only one hand -- as a PI does when the binding pocket yawns -- shaking the tree just a little makes you fall: When enough mutations emerge, just one more makes the virus resistant. Although Clavel did not extend his analogy of TPV, it seems to take more mutational "shaking" to knock TPV out of the protease tree.
Among people with 12 or fewer protease mutations, the median viral load dropped 1.85 logs with TPV/RTV (versus 0.44 log with control PIs, P<0.0001). In this analysis TPV maintained a significant edge over the other boosted PIs as the baseline mutation number climbed: -0.63 versus -0.42 log for 13 to 15 mutations (P = 0.0105) and -1.08 versus -0.16 log for 16 to 18 mutations (P<0.0001). When faced with 19 or more mutations, TPV/RTV finally ran out of steam but still nipped replication significantly better than comparison PIs, by -0.36 versus -0.2 (P<0.0001).
The four mutations once judged prime TPV nemeses -- at positions 33, 82, 84, and 90 -- did not slow this new PI more than others in Schapiro's resistance shakedown. But a deeper analysis, he reported, turned up 21 mutations that may conspire to curtail responses.3
A mutation score based on that arm-long list may emerge with further study. If it does, clinicians not already privy to expert resistance counsel will have to find a friendly expert fast.
Are there hints on how TPV and TMC114 compare in people with PI-resistant virus? Cross-study comparisons, all readers know, are statistically unsupportable, probatively problematic, and sometimes downright dangerous. But HIV clinicians and the people they treat live dangerously, and they'll have to weigh some data or other if deciding whether to try TPV (maybe late this year) or to hold out for TMC114. As already noted, baseline traits looked similar in the RESIST and POWER studies, and work so far shows no toxic surprises with either PI. So, if you are bold, consider Table 1.
ATV as a Salvage Drug
Atazanavir has earned favor as a page-one PI because of its once-daily dosing and steadfast disinterest in stirring lipids or glucose. But for people with a handful of PIs in their treatment history, can ATV help bring HIV to heel? One CROI study found synergy between ATV and SQV against ATV-resistant but SQV-susceptible virus. Other work uncovered one salvage combo that bears close watching -- ATV plus nevirapine (NVP) -- but not for good reasons.
The ATV/SQV study involved three site-directed mutants with two to four mutations and low-level resistance to ATV and/or SQV and 11 clinical isolates with four to nine mutations and higher levels of resistance to the PIs [abstract 715]. Sophie Lebel-Binay (VIRalliance, Paris) gauged inhibition of these viruses at four ATV/SQV ratios -- 1:1, 2:1, 4:1, and 8:1. She found no hint of synergy against wild-type (nonmutant) virus but significant synergy at all four PI ratios for a virus in which the I50L and A71V mutations yielded low-level resistance to ATV (5.35-fold change in 50 percent inhibitory concentration [IC50]) and susceptibility to SQV (0.66 fold change in IC50). The PIs also proved synergistic against two other SQV-susceptible viruses.
What do these findings mean? Lebel-Binay noted that synergy rose with each higher ATV/SQV ratio -- a finding strongly suggesting that ATV somehow enhances SQV activity. One possibility, she proposed, involves higher intracellular SQV levels, perhaps because of saturation of intracellular proteins or inhibition of drug transporters by ATV. The same thing happened in an earlier study of LPV/SQV that measured synergy against virus with high resistance to LPV and less to SQV.4 And other research clocked long intracellular SQV half-lives compared with other PIs.5
Together with work charting higher ATV and SQV exposure when RTV boosts the paired PIs rather than each PI singly,6,7 Lebel-Binay's findings position ATV/SQV/RTV as a rescue regimen worth study -- at least in people with resistance to ATV but not SQV.
On the other hand, ATV probably should not replace RTV as an SQV booster in people just starting PIs. That conclusion emerged from a study of eight healthy men and seven hardy women who volunteered to take SQV/ATV at doses of 1,600/400 mg and 2,000/400 mg once daily for 10 days [abstract 655]. Stephen Becker (Pacific Horizon Medical Group, San Francisco) charted significantly higher SQV exposure with an RTV boost at a once-daily dose of 1,600/100 mg. Notably, levels of all three PIs climbed higher in the women than in the men even after statistical adjustment for weight (P<0.05).
The potential dangers of ad hoc salvage concoctions came to light in a study of 100 people taking ATV [abstract 656]. Measuring ATV troughs in this group, Alan Winston (University of New South Wales, Sydney) recorded levels below 50 µg/L in some of them, all of whom had an undetectable viral load. Only one variable predicted low ATV levels in a multivariate analysis -- cotherapy with nevirapine (NVP) (mean ATV trough 350 µg/L with NVP and 726 µg/L without, P = 0.011). In the same analysis only RTV boosting independently predicted a higher ATV trough. Factors such as fat content of meals, drinking buffered beverages, and adherence did not affect ATV troughs.
Clinicians questioned study participants about other prescription, nonprescription, and herbal medicines they took with ATV. Although an earlier study found 76 percent lower ATV exposure among people taking 40 mg of omeprazole with 300/100 mg of ATV/RTV,8 ATV levels proved no lower in 12 people taking that stomach acid quencher in Winston's cohort. Two factors may explain the lack of an omeprazole-ATV interaction, Winston suggested: 10 of the 12 people taking omeprazole took it at least 10 hours earlier or later than ATV, and two took only 20 mg daily.
Only further study can determine whether people need a higher ATV/RTV dose with NVP, Winston cautioned. But the findings offer an admonitory notice on possible risks of untested antiretroviral combinations.
One other notable CROI report on ATV -- to be detailed in part 2 of this article -- came from Eoin Coakley (ViroLogic), who documented what seems to be the first case of ATV/RTV failure with primary resistance to ATV mediated by the N88S mutation, a change not seen earlier in the clinic without ATV's signature mutation, I50L [abstract 716].
Resistance to the Sole Entry Inhibitor
Envelope (env) is HIV's most volatile gene -- not a surprise when one considers the deft gymnastics env evolved to escape neutralizing antibodies. What do these skittish gyrations mean for resistance to ENF, a drug often mixed into incompletely suppressive salvage regimens?
To find out, a team from François Clavel's lab, headed by Beatrice Labrosse (Bichat-Claude Bernard Hospital, Paris), tracked env evolution in six people starting ENF as part of a salvage mélange and followed for 12 to 50 weeks [abstract 97]. She found highly variable baseline susceptibility to the fusion inhibitor. But regardless of that baseline value, the virus stubbornly hatched new resistance mutations in the heptad repeat (HR1) region of viral gp41 until high-level resistance emerged.
In every case these HR1 mutations arose from an env quasispecies different from the one dominant before ENF therapy. (A quasispecies consists of all viral variants that make up a population in one person.) In four of the six people the post-treatment env sequences proved more fit than those cataloged at baseline.
These findings appear to mean that envelope's whole genetic repertoire plays "a critical role in the expression and selection of HR1 mutations," Labrosse suggested. In other words, Clavel proposed in his review lecture [abstract 180], involvement of the whole env quasispecies could facilitate emergence of resistance to ENF and preserve replication capacity.
And virus resistant to ENF appears to pop up posthaste when an ENF regimen flounders. Sequencing virus from 30 people when they started ENF salvage and repeatedly over the next few weeks, George Beatty (University of California, San Francisco) spotted ENF resistance mutations immediately upon viral rebound, and viral loads typically rushed right back toward baseline [abstract 581]. Although no commercial assay tracks ENF mutations, clinicians should suspect failure of an ENF regimen when it does not trim the viral load, or when it does and the load rebounds.
A (Minor) Role for Replication Capacity?
Fervid research in the past few years has sought to divine the impact of resistance on viral replication capacity, and the effect of replication capacity on disease progression. So far, though, this work has not driven clinicians to add replication capacity -- or RC -- to their prognostic tool kit. New work by Andrea De Luca (Catholic University, Rome) nominated high RC as a useful guide to rescue therapy [abstract 692]. But resistance remained a better predictor of long-term response.
With ViroLogic colleagues (who developed an RC assay), De Luca looked at viral susceptibility and RC in people from the ARGENTA trial,9 which tested the merits of genotyping to elect a new regimen after virologic failure. In the 139 people De Luca analyzed, a higher (better) baseline phenotypic susceptibility score correlated significantly with a sharper drop in viral load through 34 months of follow-up (P = 0.011) and with a bigger CD4 boost through 18 months (P = 0.039).
Replication capacity correlated positively with the number of baseline drugs to which HIV proved susceptible on phenotyping (R = 0.034, P<0.001). For every PI tested, mean RC proved significantly higher in PI-susceptible virus than in PI-resistant virus (P<0.01 for APV, ATV, IDV, LPV, NFV, RTV, and SQV). For each PI, a lower RC meant a significantly decreased susceptibility to that drug.
People with an RC above 65 percent gained significantly more CD4 cells through 24 months of follow-up than did people with a lower RC (approximately +40 versus -80 cells/mm3, P = 0.007). A higher RC also correlated with a worse three-month virologic response in people who did not control HIV and after stratification for the number of drugs to which HIV proved susceptible at baseline.
But in a univariate Cox regression model, RC did not predict a higher risk of clinical progression, whereas US Centers for Disease Control and Prevention (CDC) class C disease, older age, and CD4 and RNA changes did. The protease mutations K20M/R and I84V also ratcheted up the risk of progression in this analysis (K20M/R hazard ratio [HR] 5.41, P = 0.0003; I84V HR 2.74, P = 0.03).
Resistance, De Luca concluded, remains a stronger driver of RNA and CD4 response to rescue therapy. He proposed that susceptibility testing -- phenotyping -- should be "the primary tool" to guide treatment decisions after ART failure. But in people who cannot stop replication with salvage therapy, De Luca suggested, "replication capacity can be a secondary tool to drive the treatment decision."
The Novel and the Not-so-New
The parade of antiretrovirals termed "novel" has not slowed since the first CROI. To the dismay of many but the surprise of none, most of these elixirs never get named on a prescription pad. Predicting eventual winners remains a chore that shames both drug development doyens and Wall Street panjandrums.
To stand a chance in today's florabundant antiretroviral market, any new drug must do something special -- stop virus via some unique modus operandi, strangle off resistant strains, or go down like honey. A viral maturation inhibitor from Panacos Pharma met the first criterion. Aside from the new PIs discussed above, a fresh nonnucleoside reverse transcriptase inhibitor (NNRTI) from Tibotec met the second. Alas, dogged research has yet to integrate an integrase inhibitor into antiretroviral plans, as attendees learned in hearing the first clinical trial results of such a drug.
PA-457, the Panacos aspirant, stymies maturation of the viral capsid protein p24 (yes, the protein once used to gauge antiviral response) and so renders new virions impotent [abstract 159]. A double-blind study tested single doses of PA-457 in 24 HIV-infected people who had not taken antiretrovirals for at least four weeks, reported Panacos researcher David Martin.
Twenty days of follow-up showed a significant dose-response relationship (P<0.05) with median viral load dips of 0.17 log with placebo, 0.27 log with 75 mg of PA-457, 0.45 log with 150 mg, and 0.5 log with 250 mg. Viral suppression with the two highest doses, though not huge, matched results in single-dose studies of other antiretrovirals, Martin noted. And the solitary dose kept viral loads down for 10 days.
Two people started the study with multi-resistant virus. One person assigned to 250 mg had the L210W, K103N, and Y181C reverse transcriptase mutations plus the V77I and L90M protease mutations. Nonetheless, that person's viral load fell 0.73 log. Despite starting treatment with the M184V and K103N reverse transcriptase substitutions and L10I and V77I in protease, a person taking 150 mg of PA-457 notched a 0.53-log RNA drop. No notable side effects marred this single-dose study. PA-457 efficacy trials will start later this year.
Tibotec is putting money on two NNRTI ponies with activity against virus resistant to NVP and efavirenz (EFV). The first, TMC125, looked good in test tube studies aired at last year's CROI.10 The second, TMC278, started its clinical run for the roses this year in a three-country sprint presented by Frank Goebel (Ludwig Maximilians University, Munich) [abstract 160]. With colleagues in Russia and Britain, he tested five doses -- 25, 50, 100, and 150 mg -- given once daily for seven days to 44 people with viral loads ranging from 3.5 to 5.9 logs (about 3,200 to 794,000 copies/mL) and no resistant virus.
The median viral load dropped 1.2 logs with no apparent difference by dose. The median CD4 count shot up 55 cells/mm3 in this week-long study. No one had to stop TMC278 because of side effects, and no resistant virus emerged. A multinational phase 2 study has begun.
Viral load drops averaged 1.77 logs with 400 mg of a Merck integrase inhibitor given as twice-daily monotherapy for 10 days to treatment-naive and -experienced people, and 1.73 logs with 200 mg [abstract 161]. Susan Little (University of California, San Diego) reported respective CD4 vaults of 89 and 73 cells/mm3 in those 10 days. Six of 16 people taking the higher dose reached a viral load below 400 copies/mL. But the development gauntlet for integrase inhibitors still has no end in sight: Merck had to shelve this compound when toxicities arose in animal studies. The company will now try a related agent.
Two other drugs with dog-eared development dossiers -- the nonnucleoside capravirine and the nucleoside amdoxovir (DAPD) -- did not fare terribly well in regimens tooled for people with treatment experience. The nonnucleoside trial teamed 700 or 1,400 mg of capravirine twice daily (or placebo) with NFV and hand-picked nucleosides in 179 people with NNRTI-resistant virus but no PI experience [abstract 555]. Baseline loads averaged 4.4 logs (about 25,000 copies/mL) and average starting CD4 counts were in the low 200s. Earlier work indicates that capravirine has some activity against NNRTI-resistant virus.
After 48 weeks viral loads dropped 2.1 logs in the placebo group, 2.3 logs in the 700-mg group, and 2.4 logs in the 1,400-mg group. Defining failure as less than a half-log RNA decline by week 4 or a half-log drop followed by a rebound, Pfizer's Rick Pesano chalked up a 48-week failure rate of 24 percent with placebo, 15 percent with 700 mg of capravirine, and 13 percent with 1,400 mg -- nonsignificant differences. Dropout rates were high: 44 percent with placebo, 42 percent with 700 mg, and 30 percent with 1,400 mg. The overall poor response seems odd since everyone started NFV without PI experience.
Because the nucleoside reverse transcriptase inhibitor (NRTI) amdoxovir thwarts virus resistant to zidovudine (AZT) and lamivudine (3TC), Barbara Gripshover (University Hospital of Cleveland) and US AIDS Clinical Trials Group (ACTG) mates added it or placebo to a salvage regimen including ENF and three to five other drugs for 18 people with medians of six NRTI mutations, one NNRTI mutation, and seven PI mutations [abstract 553]. This group began salvage with an abysmal CD4 count of 36 cells/mm3 (range 11 to 537 cells/mm3) and an average starting viral load of 100,000 copies/mL.
The runaway baseline resistance proved too tough even for a regimen boasting two drugs with novel resistance profiles. In 24 weeks the mean viral load slipped 1.11 logs among the nine people randomized to amdoxovir and 0.8 log in the placebo group, a nonsignificant disparity. Three of nine people taking amdoxovir versus one of nine taking placebo reached a viral load below 200 copies/mL. And the average CD4 count climbed 70 cells/mm3 with amdoxovir versus 54 cells/mm3 with placebo (also nonsignificant). Gripshover recorded six new AIDS diagnoses and two deaths, with no advantage for the amdoxovir arm.
But because the drug showed some activity in this advanced group, the ACTG urged further study. Whether that will happen remains unclear because Gilead dropped development of amdoxovir last year and turned the license back to Emory University and the University of Georgia Research Foundation.
Two More Dead-End STI Salvage Studies
One day clinicians will stop studying structured treatment interruptions (STIs) as a way to prime virus for salvage therapy. The odds on success with that tactic got longer yet at CROI, as two more groups found no advantage with a presalvage break versus immediate therapy. For those still keeping score, that makes it five11-13 to one14 for the "no benefit" team.
People in Canada have apparently kept pace with STI salvage news, because a randomized trial of a 12-week STI versus immediate treatment had to close accrual early with a dearth of enrollees [abstract 580]. Still, Sharon Walmsley (University of Toronto) and coworkers randomized 67 people to each study arm. Study participants had to have at least two active antiretrovirals they could put in a new regimen. All started three to five drugs with plentiful treatment experience but with relatively well-controlled HIV infection (median viral load 3.9 logs in both groups; median CD4 count 320 cells/mm3 in the STI group and 360 cells/mm3 in the control group).
After 60 weeks of follow-up, Walmsley saw no virologic differences between the groups but a significant CD4 deficit in the STI arm. While 43 people (64 percent) in the control arm reached a sub-50-copy load at least once, 53 people (79 percent) in the immediate-treatment group reached that mark. The median viral load drop at week 60 measured 1.7 logs in both arms.
The STI covey lost an average 80 cells/mm3 during their drug break and never caught up with the control arm. After 60 weeks of follow-up median CD4 gains measured 25 cells/mm3 in the STI takers and 95 cells/mm3 in the immediately treated (P = 0.04). Although that difference may seem clinically marginal to groups starting the trial with more than 300 cells/mm3, the early ebb in the STI arm caused trouble for at least one and maybe four people. Despite a protocol stipulation to start Pneumocystis prophylaxis if the CD4 count fell under 200 cells/mm3, one person ended up with Pneumocystis pneumonia at week 7. Two others endured Candida esophagitis (at weeks 28 and 43), and one had lymphoma by week 57. No one in the immediate-treatment group added an AIDS diagnosis during the study.
The already mentioned study by George Beatty (University of California, San Francisco) enrolled 30 people with much more advanced HIV infection (median baseline CD4 count 39 cells/mm3, range 12 to 135 cells/mm3) and with resistance to at least two NRTIs, at least one NNRTI, and at least two PIs [abstract 581]. With Steven Deeks (University of California, San Francisco), Beatty cited earlier work15 in hypothesizing that treatment breaks may work for people with multidrug-resistant virus who can build a new combo around a drug from a novel class, such as ENF. But that didn't happen in this study of 15 people who took a 16-week STI before starting an ENF regimen and 15 who started salvage pronto.
Twenty-four weeks after randomization (six weeks after treatment began in the STI arm), both groups sawed about 1.5 logs off their viral load, and that lack of difference persisted through week 48. At 24 weeks eight of 15 people (53 percent) in the immediate arm and five of 14 (36 percent) in the STI group had a viral load under 75 copies/mL, a nonsignificant difference. CD4 counts rose about 100 cells/mm3, on average, through week 48. During the drug break median CD4s swooned 27 cells/mm3 from an already treacherous baseline of 47 cells/mm3.
Baseline susceptibility to ENF did not predict virologic outcome, but overall susceptibility to the salvage regimen did. Everyone with a phenotypic susceptibility score at or below 1 had a virologic failure during salvage, whereas 63 percent with a score above 1 had an undetectable load at 24 weeks.
Final results of the already published11 CPCRA 064 study confirmed a persistent CD4-cell tumble after 24 weeks of treatment in the STI arm (-3.2 cells/mm3 versus +39.6 cells/mm3 in the immediate-treatment group, P = 0.07) [abstract 579]. Jodi Lawrence (University of California, San Francisco) reported 44 cases of disease progression in the group that took a four-month drug holiday before salvage versus 29 in the control group (1.66 hazard ratio adjusted for baseline CD4, viral load, AIDS diagnoses, and study site, P = 0.04).
Values of Avoiding Drug Breaks
Compared with proving the merits of stopping therapy for people with resistant virus and risky CD4 ratios (see preceding section), proving the virtues of continuing therapy for such people seems almost simple. Daniel Kaufmann did it in the Swiss HIV Cohort Study,16 Veronica Miller did it in the Frankfurt17 and EuroSIDA18 cohorts, and now Isabelle Kousignian (Pierre and Marie Curie University, Paris) has done it in the French Hospital Database on HIV [abstract 592]. Even with the lowest CD4 counts and detectable viral loads, she reported, maintaining ART wards off AIDS diagnoses.
Kousignian and colleagues charted new AIDS-defining diseases in five groups:
The median treatment hiatus measured 4.6 months (interquartile [IQR] range 2.7 to 8.9 months).
As one might expect, the overall rate of new AIDS diagnoses proved significantly higher in the untreated and monotherapy groups than in the other three groups. And new AIDS rates proved lowest among people who kept their viral load under 500 copies/mL (group 5). The most interesting comparison involves groups 3 and 4. Group 4 (people with virologic failure but no break from potent therapy) did significantly better in averting AIDS than group 3 (people who interrupted potent therapy) regardless of CD4 count (Table 2).
Other interesting disease-specific findings emerged: The treatment break cohort had a higher rate of fungal infections than all other groups -- even the untreated and monotherapy contingents. And compared with the untreated or monotherapy groups, the treatment interrupters had the same rates of Pneumocystis pneumonia, toxoplasmosis, and pulmonary or esophageal candidiasis. For people in the French Hospital Database, holidays from potent therapy meant sailing back to the monotherapy era.
Antiretroviral history is replete with stirring advances, rueful setbacks, and the odd thunderclap. But more than anything that history is replete with repeats, building a record of almost soporific sameness as the best regimens and soundest strategies rise like fine cream in study after study.
The just-reviewed virtues of continuing -- rather than suspending -- treatment for people with advanced disease and scant options have resounded for a half decade. The same span has elapsed since first reports from Dupont's 006 study installed EFV and two NRTIs in the firmament of simple, durable first-line regimens. The 12th CROI showed, again and again and again, that this frill-free trio is hard to beat.
Mounting long-term, massive, multinational trials -- difficult by definition -- can have its rewards when the P power grows big enough to show that one regimen, or one strategy, clearly outperforms rivals. But these megatrials can also have a hollow ring when their length leaves an anachronistic echo.
INITIO, begun in the last century, now seems a victim of remorseless time. As in ACTG 384, unveiled 2.5 years ago at the XIV International AIDS Conference19 in Barcelona, Australian and French INITIO collaborators found that EFV with two nucleosides works better than NFV plus two nucleosides as first-line therapy [abstract 165LB]. And as in the ACTG trial,20 a four-drug regimen piggybacking EFV and NFV with two NRTIs lagged the simpler EFV combo.
All of INITIO's initial regimens rested on didanosine/stavudine (ddI/d4T), reported Patrick Yeni (Bichat Hospital, Paris). As a result, lots of people switched to new nucleosides over the median 3.7 years of follow-up. But significantly fewer people had to stop or switch from EFV/ddI/d4T (44 percent) than from those NRTIs plus NFV (63 percent) or EFV/NFV (76 percent) over the study's course. An intent-to-treat analysis after three years showed significantly more sub-50-copy responses in the EFV/ddI/d4T arm:
Efavirenz plus two NRTIs also did better than the NRTI-sparing regimen EFV/LPV/RTV in people with advanced disease who had responded to a three- or four-drug first-line PI or NNRTI [abstract 162]. The higher LPV/RTV dose needed with EFV (533/133 mg twice daily) caused more toxicity in that arm and probably contributed to its poorer performance.
Reporting for ACTG A5116 colleagues, Margaret Fischl (University of Miami) explained that everyone in the study had started a three- or four-drug medley with a CD4 count at or below 200 cells/mm3 and a viral load atop 80,000 copies/mL. They had no resistance to AZT, ddI, d4T, PIs, or NNRTIs, never had two consecutive loads above 400 copies/mL, and had an RNA tally under 200 copies/mL when randomized to EFV/LPV/RTV or EFV plus two nucleosides (usually AZT/3TC) in ACTG A5116.
After a median 110 weeks of follow-up, Fischl counted seven virologic failures among 118 people (6 percent) in the EFV/NRTI group versus 14 of 118 (12 percent) assigned to EFV/LPV/RTV. In the EFV/NRTI arm 74 percent kept their viral load under 50 copies/mL, compared with 66 percent in the NNRTI/PI group. While six people quit the EFV/NRTI combo because of side effects, the comparison arm had 20 toxicity dropouts. In a 160-week intent-to-treat analysis, time to virologic or toxicity failure proved significantly shorter in the EFV/LPV/RTV group (P = 0.0015).
In a separate analysis of this trial, Pablo Tebas (Washington University, St. Louis) reported progressive limb fat loss in the EFV/NRTI group and a progressive gain in the NRTI-sparing arm [abstract 40]. But at what a price. Besides the higher risk of failure with EFV/LPV/RTV, people taking those drugs endured elevations of total cholesterol, low-density lipoprotein (LDL) cholesterol, and triglycerides. Glucose and insulin rose in the EFV/NRTI group, but so moderately that Fischl discounted the clinical import of those gains.
Together these studies confirm the worth of plain old triple EFV therapy for people starting therapy or shifting from another suppressive regimen. Confirming these confirmations, John A. Bartlett (Duke University, Durham, North Carolina) decocted data from 49 clinical trials of triple regimens (counting a boosted PI as one drug) in people with little or no treatment experience [abstract 586].
Rating antiviral vim in 48-week intent-to-treat analyses, he found that 17 of the 21 combinations yielding the highest sub-50-copy rates consisted of one NNRTI and two NRTIs. Fourteen of those 17 winners included EFV -- along with 3TC/abacavir (ABC) in four, d4T/3TC in three, AZT/3TC in two, ddI/3TC in two, ddI/emtricitabine (FTC) in two, and 3TC/tenofovir (TDF) in one. Ritonavir-boosted PI regimens did win this 49-study scrum by one criterion -- best 48-week CD4 gains.
What Goes on Under the 50-Copy Cutoff?
When treatment pushes your viral load under 50 copies/mL, does it matter if you have 30 copies or 3? It's hard to say, says Robert Siliciano (Johns Hopkins University, Baltimore), who has spent years finessing sub-50 findings.
Speaking in one of CROI's last sessions, he defined two viral load strata under the 50-copy limit -- one representing ongoing low-level replication (though this remains unproved) and one representing release of virus from a stable reservoir of about 1 million resting CD4 cells -- a mere dewdrop in the body's ocean of T lymphocytes [abstract 179]. When something pricks these cells into predatory mode (it happens all the time), they spill virus back into circulation, where hawkeyed handlers of supersensitive assays can count it.
Siliciano's lab showed that this replication-ready HIV can drip back into circulation for years. But if a person is taking a potent regimen and keeps the viral load under 50 copies/mL, that virus will not mutate into resistant species.21 The baseline viral population will not evolve at all. Further work by his group published just before CROI confirmed that blips in viremia do not portend sustained rebounds or open the door to resistant virus.22 Instead, he concluded, blips probably reflect variation in hair-trigger RNA assays.
Scouring data from the Dutch ATHENA cohort, Irene van Valkengoed (HIV Monitoring Foundation, Amsterdam) found equivalent CD4, RNA, and progression trends in 87 blippers and 134 nonblippers tracked during three years of ART [abstract 602]. Defining a blip as an RNA reading between 50 and 1,000 copies/mL after at least two loads below 50 copies/mL and followed by another under that mark, she estimated an overall blip rate of 6.3 per 100 person-years in 1,730 ATHENA members with more than one year of on-treatment followup. The blip rate rose from 2.9 per 100 person-years in the first year of treatment to 7.6 in the second, then settled back to 3.5 in the third.
The 87 blippers matched 134 no-blip controls in age, gender, HIV transmission group, time since HIV diagnosis, year of first treatment, drug levels, baseline CD4 and RNA levels, and baseline AIDS status. Through three years of follow-up, the median CD4 count rose from 140 to 430 cells/mm3 in the blip group and from 180 to 410 cells/mm3 in nonblippers. Three blippers (4 percent) and four controls (3 percent) had rebounds above 1,000 copies/mL. Eight blippers (9 percent) and 15 controls (11 percent) had a new AIDS diagnosis.
What separates people with loads bubbling under the 50-copy mark from those with the very lowest loads? Studying 145 responders in a trial comparing LPV/RTV with NFV,23 Sarah Palmer (National Cancer Institute, Frederick, Maryland) found only one discriminating factor: Lower pretreatment viral load correlated significantly (P<0.001) with lower on-treatment viremia -- gauged by an assay with a 1-copy limit -- after 60 weeks of treatment [abstract 163]. She reckoned a viral load of about 3 copies/mL in both treatment arms.
Just after CROI a study using a 2.5-copy cutoff assay confirmed Palmer's correlation between lower pretreatment viral load and lower sub-50 viremia.24 This analysis of 100 people in the trial comparing TDF with d4T (plus 3TC and EFV) also linked loads under 2.5 copies/mL with lower HIV DNA quotients in peripheral blood mononuclear cells and with TDF rather than d4T therapy. But how far below 50 the viral load went did not affect CD4 counts or the viral rebound risk through 72 weeks of follow-up.
Another Try With Triple Nukes
Last year's CROI confirmed the scalding failure of triple-NRTI amalgams lacking a thymidine analog (AZT or d4T). And AZT-containing Trizivir didn't measure up to EFV combos in a big ACTG trial.25 So you would think the itch to try triple nukes had been well scratched.
Not at all. David Rey (Strasbourg University Hospital, Strasbourg, France) and confreres still have the yen to test an AZT/3TC combination -- but this time with TDF instead of ABC -- at least in a nonrandomized trial [abstract 599]. The rationale for this tactic rests on competing resistance pathways of AZT and TDF -- mutations conferring resistance to these drugs almost never share a genome. Alone among licensed nucleosides, AZT retains activity against the TDF-provoked K65R mutant.26 And 24-week results from a trial of first-line once-daily Trizivir plus TDF showed a reasonable virologic response, though more people with a baseline load below 100,000 copies/mL (85 percent) than above that brink (61 percent) claimed fewer than 50 copies/mL at six months.27
The Strasbourg study involved 42 treatment-naive people with a median CD4 count of 233 cells/mm3 (range 23 to 425 cells/mm3) and a median viral load of 4.88 logs (about 75,000 copies/mL). Over a median eight months of follow-up, nine people (21 percent) stopped the three NRTIs, four because of virologic failure and five because of side effects (including two cases of anemia). On-treatment genotyping of the four people with virologic failure showed only the K65R mutation in one, the AZT-evoked M41L and T215N mutations (detected before treatment) in another, and mixes of AZT mutations and the 3TC-disabling M184V in the others. The median CD4 count climbed 82 cells/mm3.
Rey did not break out results of the 19 people who started treatment with more than 100,000 copies/mL. The nonrandomized design of this pilot study makes it impossible to rate Combivir/TDF against sanctioned first-line remedies.
A nonrandomized trial in Uganda gave the same regimen to 200 adults with a median CD4 count of 100 cells/mm3 and a median viral load of 333,000 copies/mL [abstract 22]. After 24 weeks only 54 percent had a viral load under 50 copies/mL in a missing-data-equal-failure analysis.
Induction-maintenance -- another once-touted first-line tactic that fell by the wayside -- also got a fresh look at CROI. Three early studies of this stratagem flopped, partly because both the induction and the maintenance regimens look substandard from today's vantage.28-30
But Diane Havlir (University of California, San Francisco), who headed one of those studies,28 argued at CROI that induction-maintenance deserves another go [abstract 181]. Eighteen-month follow-up of people in the French trial29 who stuck with their maintenance drugs despite their doctors' advice, Havlir noted, showed that 14 of 17 taking only AZT/3TC and 14 of 15 taking AZT/IDV kept replication under wraps.31
British researchers obliged Havlir's zeal in a randomized trial of standard therapy (one NNRTI plus two NRTIs) versus 24 to 32 weeks of induction with one NNRTI, two NRTIs, and one PI [abstract 575]. People taking the quadruple induction combo could drop the PI if they had two consecutive RNA readings below 50 copies/mL after week 24 to 32. (True enough, this "maintenance regimen" is authorized as full-time therapy in current guidelines, but the value of four-drug induction still needs proving.) The induction-maintenance group ended up with a lower rate of virologic failure, but that result defies easy explanation.
Clive Loveday (International Clinical Virology Center, Buckinghamshire, UK) and colleagues randomized 122 people with an AIDS-defining median CD4 count of 160 cells/mm3 (IQR 90 to 260 cells/mm3) and a mean starting viral load of 4.94 logs (about 87,000 copies/mL). The study group was largely male (88 percent) and gay (71 percent). More people started NVP (n = 77) than EFV (n = 44), and because the trial began in 1999 the most popular PI was NFV (in 44) followed distantly by LPV/RTV (in 17).
After a median 79.6 weeks of follow-up in the induction-maintenance group and 82 weeks in the control arm, rates of protocol-defined virologic failure proved significantly higher in the control group (Table 3).
Aside from the better outcome in the experimental group, the result that leaps out is the high virologic failure rate with standard NNRTI therapy. Yet although 43 percent looks like a high failure rate, it reflects 80 weeks of followup, not the 48 weeks reported for most studies. The under-50-copy rate for 48 weeks, 65 percent, is more in line with recent studies. As Loveday observed in a note to the IAPAC Monthly, the median 48-week NNRTI failure rate in a multi-study review presented at the conference measured 64 percent [abstract 586].
The data presented suggest no baseline imbalance between the induction-maintenance and standard-of-care arms: Rates of baseline resistance were similar in the two study groups, at about 15 percent. And early control of viremia does not explain the difference in virologic outcome: Intense viral load monitoring of 34 people in the first 15 days of therapy drew nearly superimposable RNA slopes in the two study groups. Further analysis of HIV-1 subtype and proviral DNA offered no hints toward explaining the difference in virologic failure.
The result may look surprising in that most trials comparing three-drug and four-drug first-line therapy saw no advantage with the extra drug through 48 weeks. In a report at the 5th Annual Workshop on Clinical Pharmacology of HIV Therapy, Andrew Hill (then at Roche) found only one in nine randomized trials that saw a virologic edge with quadruple therapy.32 Most recently, QUAD study researchers discovered that AZT/3TC/EFV did as well as AZT/3TC/ABC/EFV in people starting treatment with a six-figure viral load.33 But Loveday pointed out that QUAD and most of the studies reviewed by Hill combined four drugs from two classes, while his trial combined four drugs from three classes. And one of the four was a boosted PI -- an ingredient not used in ACTG 384 or INITIO (see "Something Not New ...," above). If others can confirm Loveday's result, taking a boosted PI for the first eight months of treatment does not seem so onerous a burden when the reward is better long-term control.
At last -- a Possible Advantage of Aging
The graying of the HIV population in lands of antiretroviral plenty raises questions about how older age may affect responses to first-line therapy. The answer from two US cohorts seems to be that people over 50 reap greater virologic rewards than younger folk [abstract 596]. And despite worries about waning T-cell output with age, the senior contingent gained CD4 cells as avidly as their juniors.
Kristine Patterson (University of North Carolina at Chapel Hill, USA) and colleagues at Johns Hopkins University in Baltimore ran a case-control comparison of people over or under 50 years old who started antiretrovirals with a potent regimen since January 1998. The 63 cases and 183 controls included 70 women and 176 men, two thirds of them black. About one in five picked up HIV by injecting drugs. The median age of cases stood at 54 years (IQR 51 to 60), compared with 38 years (IQR 32 to 48) in controls.
About half in each age group started a PI regimen and half started an NNRTI. Pretreatment CD4 counts were substantially higher in older women than in the other groups (Table 4). But equivalent proportions of the two age strata had a CD4 gain topping 25 percent through six months of follow-up, and equivalent proportions reached a viral load below 400 copies/mL (Table 4).
Patterson and colleagues don't speculate on why the over-50 set harnessed viral replication better than the younger crowd. One might speculate that better adherence -- begot by a keener appreciation of mortality among older people -- could be a contributor. The higher CD4 count at diagnosis among older women may also reflect a greater concern for their own health. These researchers did not try to rate adherence. But a California group did test adherence in a randomized trial:
Five Lessons in Better Adherence
With clinicians at five southern California clinics, Glenn Wagner (Rand Corporation, Santa Monica, USA) randomized people starting, restarting, or switching antiretrovirals to one of two adherence plans or to "usual care" [abstract 614]. The adherence programs significantly improved pill taking compared with the control group, although that difference has not paid virologic dividends through 24 weeks of follow-up. The trial, which also includes a drug level monitoring component, will last 48 weeks.
Wagner offered a 24-week analysis of 199 people starting their first antiretrovirals and randomized to (1) pretreatment practice plus cognitive/behavioral intervention, (2) cognitive/behavioral intervention alone, or (3) usual care. (What is "cognitive/behavioral intervention"? Note 34 explains.) The adherence sessions came in five servings -- three before therapy started and two afterwards. Pretreatment pill-taking practice added nothing to cognitive/behavioral training, so Wagner combined group one and two results in the comparison with group three.
The cohort was 50 percent Latino, 29 percent white, 13 percent black, and 20 percent female. Just over one third were employed, and 42 percent had no insurance. The mean starting CD4 count measured 188 cells/mm3 and the mean viral load 5.2 logs. An intention-to-treat analysis 24 weeks after treatment began documented much better adherence in the intervention groups, as measured by tattletale electronic bottle caps:
Despite better pill taking in the intervention group, viral load responses did not differ significantly from the control group at the study's 24-week halfway point. For example, the average viral load dropped 2.15 logs in the intervention group versus 1.97 logs among controls. But adherence reckoned as percent of doses taken did correlate significantly with the 24-week viral load drop (r = 0.2, P = 0.03).
Dutch insurers made headlines just after CROI when they decided to write life policies for people with HIV -- but only for those who pulled out of their CD4 nosedive with the help of antiretrovirals, who have no other life-threatening maladies, and who don't inject drugs.35 Still, the shrewd underwriters' estimate that they can make money betting on the lives of people with HIV as readily as those of people with heart disease or diabetes says something about the boons of ART.
Work detailed at the 12th CROI explained why life expectancy keeps climbing in the West, even as it crashes across Africa: Antiretroviral response rates mount with each passing year, according to a multi-cohort EuroCanadian study [abstract 593], while rates of triple-class failure fall [abstract 594].
But treatment has its limits. Three big studies agreed that scudding CD4 gains flatten out after three or four years of treatment -- usually before scrabbling into normal territory [abstracts 609, 611, 612]. HIV death rates also stumbled into flatland after the big post-1995 plummet, at least at London's Royal Free Hospital [abstract 957]. The overall risk of death among Norwegians with HIV quadruples the death risk of all Norwegians, but the risk of a non-HIV death is now lower in HIV-infected people [abstract 959].
Perhaps the most useful survival news came from the Vancouver group, who showed that pretreatment CD4 percent predicts survival even among British Columbians who start therapy with 200 to 350 cells/mm3 [abstract 589].
With confederates in Calgary, Barcelona, Nice, and Frankfurt, Fiona Lampe (Royal Free Hospital, London) charted at least a halving of the virologic failure rate among 4,143 adults who started antiretrovirals with a potent regimen from 1996 through 2002 (Table 5) [abstract 593]. The risk of gaining fewer than 50 CD4 cells/mm3 in the first year of treatment also dropped significantly.
But Lampe showed that this shift did not fully account for improving virologic trends, because the second adjusted risk ratio in Table 5 considers first-line drugs and still shows a year-after-year drop in risk of failure. So, Lampe ventured, "factors such as increases in adherence and improvements in clinical management ... may have made an additional contribution."
Nicholai Lohse (Odense University Hospital, Odense, Denmark) and colleagues across Denmark agreed that "better patient coaching" in recent years could explain better treatment responses, measured in this study as rates of triple-class failure [abstract 594]. Although this inquest involved only Danish people taking a potent regimen, it included 2,722 individuals -- everyone cared for at a public HIV clinic.
Defining failure as a viral load above 1,000 copies/mL for 120 (not necessarily consecutive) days on a regimen, Lohse counted 177 three-class failures over a median 3.7 years of follow-up. The incidence of triple failure swelled in the first three years of therapy, then swooned significantly. From the third to sixth year of treatment, treble-class failures dropped 16 percent yearly (P = 0.04), including a 20 percent per annum plop among treatment-experienced people (P = 0.022).
Incidence of triple burnout peaked at 3.7 failures per 100 person-years in 2000, then fell to 1.6 per 100 person-years in 2001, 0.7 per 100 person-years in 2002, and 0.4 per 100 person-years in 2003. From 1997 through 2003 the incidence of triple failure skidded 12 percent yearly (P = 0.002).
A Lifelong CD4 Deficit?
The Swiss, the Spanish, and the French may not agree on cheese, chocolate, or the toughest climb in the Tour de France, but they agree that CD4 counts stop climbing after three or four years of potent therapy. Nationwide studies in all three countries logged muscle-popping T-cell ascents among virologic responders in the early years of treatment -- followed by tabletop plateaus linked to starting CD4 counts.
The biggest and longest analysis came from Gilbert Kaufmann (University Hospital Basel, Switzerland) and the Swiss HIV Cohort Study, who tracked CD4 changes in 6,497 people for seven years or more [abstract 612]. The Swiss team divided their cohort into 2,449 people who never interrupted treatment (642 took only one regimen, 1,425 took several, and 400 took several highly active antiretroviral therapy (HAART) plus non-HAART regimens) and 4,048 who had breaks in their treatment (2,936 took only HAART and 1,112 also took a non-HAART regimen).
Across all these groups, the biggest CD4 gains came in the first three to four years of therapy. The overall median CD4 count rose from 210 cells/mm3 (IQR 91 to 355 cells/mm3) to 443 cells/mm3 (IQR 295 to 625 cells/mm3). So at the end of follow-up more than half of the cohort still had a count below 500 cells/mm3. After seven years of treatment, 41 percent inched across the 500-cell line.
Absolute CD4 gains looked similar in people who stayed with the same regimen and people who tried multiple combinations. But T-cell upticks proved significantly higher in people who never stopped their antiretrovirals than in those who took holidays. Kaufmann plotted a significant inverse correlation between the number and duration of antiretroviral breaks and CD4 jumps at seven years (r = -0.32, P<0.001 for both correlations).
Angels Jaén (Center for Epidemiological Studies on HIV/AIDS, Badalona, Spain) and colleagues in 11 Spanish hospitals found that rising CD4 tallies pancaked after four years of potent therapy in everyone except those starting with fewer than 100 cells/mm3 [abstract 611]. This analysis of 1,452 people beginning treatment after 1997 also traced a significant correlation between starting CD4 tally and the last measured count (P<0.001):
The AIDS rate during follow-up proved significantly higher in the lower baseline strata -- 19.9 percent, 7.3 percent, 2.3 percent, and 2.5 percent (P<0.001). But mortality did not differ significantly by starting CD4 count. People who entered the PISCIS cohort when older than 40 years and responded to treatment gained fewer CD4 cells than younger responders.
French APROCO cohort researchers sighted a T-cell tableland after three years of suppressive therapy in 1,281 people who started a PI regimen in 1997 or later [abstract 609]. After a median follow-up of 57 months, Vincent Le Moing (University Hospital of Montpellier, France) confirmed the Swiss and Spanish finding that people starting therapy at lower CD4 counts reached lower plateaus than those starting with more CD4 cells.
The overall upward CD4 slope measured 29.9 cells/mm3 monthly before month four, 6.4 cells/mm3 between months four and 36, and 0.6 cells/mm3 monthly after month 36. An unadjusted analysis picked out three factors that predicted a slow but steady CD4 gain after three years of treatment: male gender (+1 cell/mm3 monthly, P = 0.04), treatment naive at baseline (+2 cells/mm3 monthly, P = 0.02), and baseline CD4 count under 100 cells/mm3 (+2.6 cells/mm3 monthly, P<0.01). Variables that did not predict a growing CD4 quotient after 36 months were HIV transmission group, hepatitis C virus status, PI prescribed at baseline, and switch from a PI to a non-PI regimen.
An analysis adjusted for age, gender, baseline naive or AIDS status, and baseline CD4 and RNA found only two factors that foretold a positive CD4 slope from month four to month 36 -- age under 50 years (+2.84 cells/mm3 monthly, P<0.0001) and baseline viral load above 100,000 copies/mL versus under 10,000 copies/mL (+ 2.77 cells/mm3 monthly, P = 0.001).
What Is an "HIV-Related Death?"
Three findings stand out in a mortality study by Caroline Sabin (Royal Free Hospital, London) [abstract 957]. First, the death rate at the Royal Free has been low but fairly flat since 1998. Second, deaths among just-diagnosed people do not explain this persisting mortality. And third, unlike other researchers Sabin did not chart a significant drop in the HIV-related death rate over time.
In a cohort that includes nearly 3,000 people since potent therapies arrived, Sabin plotted an 8.1 percent death rate in 1996, 4.0 percent in 1997, 1.8 percent in 1998, 1.8 percent in 1999, 2.1 percent in 2000, 1.0 percent in 2001, 0.9 percent in 2002, and 0.8 percent in 2003. So mortality measured about two deaths per 100 person-years in 1998 to 2000 and about one in 2001 to 2003. She attributed 52 percent of all deaths directly to HIV, and that quotient did not fall significantly after 1996 (P = 0.18).
Of the 231 deaths chronicled, only 31 (13 percent) involved people diagnosed with HIV in the preceding six months, with no clear trend in that statistic over time. Almost everyone who died had low hemoglobin, and half had abnormal triglycerides or albumin.
The steadfast HIV-linked death rate in the Royal Free cohort prompted Sabin to call for closer scrutiny of other big data sets "to clarify how the classification of 'HIV-related' should be made in the HAART era."
With Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) study colleagues, Sabin tried to do just that by analyzing 1,248 deaths over 76,893 person-years of follow-up in this collaboration of European, US, and Australian researchers [abstract 595]. Reporting for the group, Rainer Weber (University Hospital, Zurich, Switzerland) made plain the difficulties in isolating HIV as the factor or a factor that leads to death.
Weber and colleagues grouped deaths into four clusters:
Even these broad groups have blurry lines between them, since liver failure can be an antiretroviral side effect, some "non-AIDS malignancies" may well belong on the AIDS list, and the D:A:D team itself demonstrated that more years of ART raise the myocardial infarction risk [abstract 42]. Weber's findings tended to broaden the blur.
AIDS led this field of grim reapers, accounting for 30 percent of deaths, followed by liver failure (14 percent, nearly all due to hepatitis), heart disease (9 percent), and "non-AIDS malignancies" (8 percent). Minor contributors to the death rolls (at rates between 2 percent and 5 percent) included other problems familiar to HIV clinicians -- suicide, pancreatitis, lactic acidosis, and renal failure.
As D:A:D workers anticipated, HIV/AIDS deaths correlated strongly with CD4 count, at an adjusted relative rate of 96.4 with a count below 50 cells/mm3 versus above 500 cells/mm3 (P<0.0001). But the risk of death from liver failure and "non-AIDS malignancies" also rose significantly with a sub-50 CD4 count -- at a relative rate of 26.6 for liver deaths and 23.5 for malignancies (P<0.0001 for both).
Weber and coworkers proposed that deaths at very low CD4 counts "can only be categorized as non-HIV-related if there is clear evidence that the patient's immunodeficiency did not contribute to death." Indeed, these findings led D:A:D to toss out its mortality code and build a new one with a sharper focus on immunodeficiency.
In Norway the overall five-year risk of death among people with HIV compared with the general population fell by almost 85 percent after the dawn of brawny therapies [abstract 959]. And the risk of a non-HIV death proved lower after 1995 in the HIV cohort than among all Norwegians.
Vidar Ormaasen (Ullevål University Hospital, Oslo, Norway) and colleagues at Akershus University Hospital in Lørenskog, Norway reckoned death rates in 782 people seen from January 1982 through December 1994 and in 398 seen from January 1997 through March 2004. Compared with the general population, the death risk ratio for the pre-1995 group measured 22.6 (95 percent CI 19.5 to 26.4) and for the post-1995 group 3.96 (95 percent CI 2.25 to 6.67).
The risk ratio for non-HIV-related deaths in the pre-1995 group measured 3.70 (95 percent CI 2.60 to 5.25) compared with Norwegians at large, and 0.61 (95 percent CI 0.15 to 2.44) in the post-1995 group. In other words, Norwegians treated for HIV infection since 1997 had about a 40 percent lower risk of a non-HIV death than did the general population. And when Ormaasen eliminated injecting drug users from the analysis, the pre-1995 group no longer had a higher risk of a non-HIV death than the overall population.
Forget Baseline Count, Try CD4 Percent?
Work by the Vancouver group found CD4 percent a strong predictor of survival among people starting a robust regimen after mid-1996 [abstract 589]. David Moore (British Columbia Centre for Excellence in HIV/AIDS, Vancouver, Canada) also showed that CD4 percent remains a reliable harbinger of death among people who fall into the CD4 count gray zone of 200 to 350 cells/mm3 before starting therapy.
The larger analysis involved 1,623 people starting their first antiretrovirals in August 1996 or later and followed until death or June 30, 2003. In a multivariate analysis adjusted for age, adherence, initial viral load, and pretreatment AIDS diagnosis, Moore isolated six factors that foretold a higher risk of nonaccidental death:
Then Moore focused on 417 people who began treatment without AIDS and with a CD4 count between 200 and 350 cells/mm3. In this subgroup three independent predictors of nonaccidental death emerged:
Moore and coworkers suggested that CD4 percent "should be considered for inclusion in therapeutic guidelines to determine when to start therapy." It may also add more evidence when making tough individual calls about when to start.
Missed Opportunities for Diagnosing HIV
One doesn't need an advanced degree in statistics, or even a fancy calculator, to cite one immutable death predictor in people with HIV -- getting treated late, or not at all. Yet US clinicians routinely miss HIV diagnoses, according to one CROI study. Edward Gardner (Denver Public Health Department, Denver, USA) figured that one third of people with HIV newly diagnosed in the Denver Health system received care there within the preceding three years, and more than a few had problems hinting at HIV [abstract 966].
The Denver Health population reflects poor and underserved groups in many US cities. Area residents without insurance typically use Denver Health as their sole source of care, often showing up at the emergency room or urgent care clinic. Gardner underlined the key limitation of this retrospective analysis -- the inability to establish when people newly diagnosed with HIV actually became infected. So there may have been few clues of HIV when they made earlier visits. But in some cases the clues seemed clear.
Gardner ran a case-control comparison of 120 just-diagnosed people who came to Denver Health in the preceding three years and 228 just-diagnosed controls without an earlier visit. The cases included more females (22 percent versus 10 percent, P = 0.001) and had a lower median CD4 count at diagnosis (370 versus 458 cells/mm3, P = 0.01) and a higher viral load (4.7 versus 4.5 logs, P = 0.06). The case group also included more Hispanics (37 percent versus 19 percent) and more primary Spanish speakers (17 percent versus 11 percent), but those differences lacked statistical significance.
Among people who sought care at Denver Health in the preceding three years, most visits involved the urgent care clinic (58 percent) or the emergency room (50 percent), though 29 percent of visits were in primary care offices and 16 percent after hospital admission. Thirty-four of the 120 cases (26 percent) had made five earlier visits.
At least 18 cases (15 percent) had a clinical indicator of HIV at their earlier visit, though six of those 18 refused an HIV test at the time. Twenty-eight cases (23 percent) had an earlier respiratory infection, 19 (16 percent) had an earlier sexually transmitted disease (STD), and 13 (11 percent) had an earlier skin infection. Workers in the STD clinic eventually made most of the HIV diagnoses. Diagnosis with a CD4 count below 200 cells/mm3 proved more common among older people and among Spanish speakers. Gardner also stressed the missed opportunities for spotting HIV among women and people with multiple infections.
The growing number of reports on hepatitis B virus (HBV) and hepatitis C virus (HCV) coinfection at CROI mirror growing concern over liver-related morbidity as people taking antiretrovirals skirt the dangers of more exotic opportunists. Attendees at this year's conference saw disturbing signals that HCV-provoked liver fibrosis may get worse faster in HIV-infected people than most anticipate. Research on entecavir -- a novel anti-HBV drug just recommended for approval in the United States -- showed that the drug does well in people already taking 3TC. And TDF looked good in a comparison with adefovir among HBV/HIV-coinfected people.
Risk of Fast Progression From Mild Fibrosis
When liver biopsy shows little or no fibrosis, physicians tell people with HCV they can delay treating this slowly progressive disease -- perhaps until more potent and tolerable drugs arrive. But that advice may not work for some people coinfected with HIV, reported Mark Sulkowski (Johns Hopkins University, Baltimore) [abstract 121]. He found that more than one quarter of coinfected people with little or no fibrosis on their first biopsy had a two-stage jump in fibrosis on a biopsy three years later.
Sulkowski's group pinched liver samples from 61 HCV/HIV-coinfected people twice over a median of 2.84 years (IQR 2.05 to 3.41 years). The same pathologist blinded to the biopsy sequence rated them on the Ishak scale, which scores no fibrosis 0, some portal area fibrosis 1, most portal areas with fibrosis 2, occasional portal-to-portal bridging fibrosis 3, and marked bridging fibrosis 4. The analysis eliminated people with a 5 or 6 score indicating cirrhosis on their first biopsy.
The 61 study participants had a median 23.8-year duration of HCV infection and a median age of 44 years at their first biopsy. While 21 percent had fewer than 200 CD4 cells/mm3 at that time, 57 percent had a viral load under 400 copies/mL and 82 percent had taken ART.
From the first to the second biopsy, 13 percent had a one-stage jump in fibrosis grade, 13 percent a two-stage jump, and 14 percent a three-stage jump. Among those who got a 0 or 1 score on their first biopsy, 28 percent had at least a two-stage vault on their second. Sulkowski found only one significant difference between people with at least a two-stage change and people with no change: 43 percent of fast progressors versus 17 percent of nonprogressors had an alanine aminotransferase (ALT) above 100 U/L on at least one third of their assays (P = 0.05). But four factors suggested tighter HIV control among nonprogressors:
Sulkowski called for more research to pinpoint predictors of faster fibrosis progression in coinfected people. Until such predictors emerge, he proposed repeating a biopsy every three years in coinfected people with no or mild fibrosis.
A noninvasive gauge of liver fibrosis emerged from analysis of 832 coinfected people enrolled in the AIDS PEGASYS Ribavirin International Coinfection (APRICOT) trial of pegylated interferon. If validated, the FIB-4 score could allow coinfected people to make informed decisions on ART without facing biopsy's big needle, proposed Richard Sterling (Virginia Commonwealth University, Richmond, USA) [abstract 120].
The APRICOT team split study participants into 555 for a "training set" to figure out the best noninvasive score and 277 for a "validation set" to test the candidate score. Using the Ishak fibrosis system they made three fibrosis groups: no or mild fibrosis (grade 0 to 1), moderate fibrosis (grade 2 to 3), or advanced fibrosis (grade 4 to 6). Picking out salient variables in univariate then multivariate analyses, Sterling settled on the following fibrosis formula:
Nimble arithmeticians figured that a FIB-4 score of 1.45 or less had a 70 percent sensitivity, 74 percent specificity, 42 percent positive predictive value, and 90 percent negative predictive value for distinguishing between advanced (grade 4 to 6) and mild to moderate (grade 0 to 3) fibrosis. An upper cutoff of 3.25 had a 22 percent sensitivity, 97 percent specificity, 65 percent positive predictive value, and 82 percent negative predictive value for making the same split.
Applying these cutoffs to the 277-person validation set, Sterling reckoned the score would correctly rate fibrosis in 172 of 198 people (87 percent) whose FIB-4 fell outside the cutoffs. For those comfortable with 87 percent accuracy, that means 198 of 277 people (71 percent) could avoid liver biopsy.
Sherri Stuver (Boston University) and colleagues at other Boston sites relied on more familiar methods to predict liver disease progression in an urban coinfected cohort of injecting drug users -- weighing viral and nonviral variables in people who do or do not have progressive disease [abstract 947]. The cohort included 231 coinfected people, 70 percent of them men, 51 percent black, 27 percent white, and 21 percent Hispanic. Most cohort members, 88 percent, had injected drugs at some point, 29 percent were injecting during this prospective study, and 21 percent drank alcohol to excess.
Twenty-two coinfected people endured a new liver-related setback or died to yield a progression rate of 5.1 per 100 person-years. In a univariate analysis, being Hispanic raised the progression risk 5.2 times (95 percent CI 1.1 to 24.3) and having a CD4 nadir below 100 cells/mm3 raised the risk 15.8 times (95 percent CI 2.0 to 122). On the other hand, reaching a viral load below 75 copies/mL with antiretrovirals clipped the progression risk 71 percent (HR 0.29, 95 percent CI 0.08 to 1.00). Being black, having a CD4 nadir between 100 and 199 cells/mm3, and currently injecting drugs also raised the risk of progression, but not significantly. Neither dangerous drinking nor age at HCV or HBV infection boosted progression risk.
In a multivariate analysis, nadir CD4 count remained an independent predictor of liver disease progression:
Taking antiretrovirals and being black or Hispanic were marginal progression predictors in the multivariate analysis.
Entecavir or TDF for HBV
Not long after CROI the US Food and Drug Administration's (FDA) advisory panel gave its blessing to entecavir for treating HBV infection.36 Although animal studies raised the specter of a cancer risk with entecavir, the panel decided that the drug's potential benefit outweighs this possible risk and did not propose trumpeting the cancer findings in a "black box" warning. But the FDA, burned more than once in recent months by surprise side effects of approved blockbuster drugs, asked for further study of the cancer risk in people taking entecavir.
The new antiviral won the FDA advisors' sanction by bettering 3TC in a randomized trial. With HBV resistance to 3TC rife among HBV/HIV-coinfected people, how does entecavir do when the retrovirus and the hepatitis virus team up? The answer means a lot to coinfected people because entecavir has no anti-HIV activity and so does not select HIV-resistant virus. The drug's potential for interacting with antiretrovirals also appears to be low. Speaking for colleagues running a phase 2 international trial of entecavir in coinfected people, Mario Pessoa (Infectious Disease Institute, São Paulo, Brazil) reported that adding the drug to a regimen swiftly lowers HBV load and sometimes brings ALTs back to normal [abstract 123].
The ongoing trial randomized 51 people to add 1 mg of entecavir once daily to a regimen containing 150 or 300 mg of 3TC and 17 people to add placebo. Everyone had compensated liver disease and an ALT at or below 10 times the upper limit of normal at screening. The average baseline HBV DNA load stood at 9.1 log copies/mL, and the average HIV RNA load under 200 copies/mL.
At week 24 the mean HBV load fell 3.66 logs with entecavir versus 0.11 log with placebo. After 24 weeks of treatment, 84 percent taking entecavir and no one taking placebo had an HBV load below 400 copies/mL or at least a 2-log drop. At the same point 46 percent in the entecavir group and 17 percent on placebo reached an ALT under 1.25 times the upper limit of normal. Alanine aminotransferase levels fell to the upper limit of normal in 34 percent taking entecavir and 8 percent taking placebo (P = 0.08).
Nobody quit entecavir because of side effects, and only one person had a serious complication -- hepatic encephalopathy and esophageal varices judged unrelated to therapy. Five people (10 percent) in the entecavir group had grade 3 or 4 treatment-related problems. The drug's most common side effects in general population trials are headache, fatigue, diarrhea, and upset stomach.
The antiretroviral TDF also controls HBV replication, though it lacks an indication for anti-HBV therapy. Adefovir, TDF's nucleotide cousin, flunked its antiretroviral exam but won an anti-HBV license at a dose of 10 mg daily. Plenty of clinicians already giving TDF to HBV/HIV-coinfected people found reassurance in the CROI report that the drug does at least as well as adefovir in coinfected people -- maybe better.
A trial enrolling coinfected people randomized 25 to take adefovir (plus TDF placebo) and 27 to take 300 mg of TDF daily (plus adefovir placebo) [abstract 124]. About three quarters of both groups had an HIV load under 400 copies/mL, and median CD4 quotients stood in the 400s. Everyone continued their antiretroviral regimen during the study. Marion Peters (University of California, San Francisco) reported starting mean HBV loads of 9.5 logs in the TDF group and 8.8 logs with adefovir. Most people in both groups had 3TC experience -- an important point since both drugs control 3TC-resistant virus.
Peters and ACTG colleagues planned the trial to demonstrate the "noninferiority" of TDF to adefovir. After 52 people had taken the drugs for a median of 75 weeks, the trial's independent panel decided it had seen enough to rank TDF as a noninferior agent, so it stopped the study. Because of that early closure, though, statisticians could not establish TDF's superiority to adefovir, even though it bettered adefovir by about 1 log in lowering HBV load in two 48-week intention-to-treat analyses and the on-treatment analysis.
Hepatitis B virus loads dropped smartly with both drugs through the first 12 weeks of treatment. But HBV copies kept dwindling in the TDF arm through 60 weeks of follow-up, while declines leveled out in the adefovir arm. Yet the confidence intervals surrounding the means for each drug in that analysis overlapped, so Peters could not call TDF the better drug. The crucial analysis of people who began TDF or adefovir with 3TC-resistant virus was not finished when Peters spoke. Side effect rates did not differ much between treatment arms.
Mark Mascolini writes about HIV infection (email@example.com).
This article was provided by International Association of Physicians in AIDS Care. It is a part of the publication IAPAC Monthly.