Important note: Information in this article was accurate in August 2000. The state of the art may have changed since the publication date.I went into science a great deal myself at one time; but I saw it would not do. It leads to everything; you can let nothing alone. -- Mr. Brooke1
As the epidemic's second decade lurches toward its close, HIV headlines have been hogged by issues unanticipated even three years ago: adiposity with fat wasting, hyper lipids and hapless insulin, swashbuckling new strategies like structured treatment interruptions (STIs), and the high cost of importing anti-HIV meds into countries where oft-buried disbelievers in HIV are promenaded from sepulcher to soapbox.
Amid the tumult, one could almost be forgiven for assuming that the study of resistance had become frozen in time, like some malarial mosquito conveniently encased in amber, awaiting the inquisition of future scientists. But the wise men and wily women who ponder resistance have not been idle. Fourteen of them spent the past year meeting, virtually and verily, to craft advice on resistance testing in HIV-infected adults.2,3 And another 250--at bench and bedside--counted codons and rushed logistic regressions to meet the submission deadline for the 4th International Workshop on Drug Resistance & Treatment Strategies. (Meeting abstracts will be published as Supplement 3 to Volume 5 of Antiviral Therapy.)
Of those applicants, 197 (79 percent) made the cut and booked passage to Sitges, Spain, a coastal Catalonian enclave. But even in this idyllic setting, the deans and doyennes of resistance found it hard to forget that their discoveries have typically not cheered the antiretroviral experienced. Those who sought briefly to forget the burdens implicit in a fat and sharp-edged binder thick with poster and abstract reprints discovered no succor poolside, where the audio loop returned with dysphoric regularity to a rueful rock lyric, "Ain't nothin' but a heartache."
Indeed, many a study underscored the difficulties clinicians will face in interpreting genotypic tests, and many an attendee returned to the theme that "sensitive" and "resistant" cutoff values for phenotypic tests actually vary widely from drug to drug and still lack clinical correlates. Genotypic results, attendees relearned, don't cleanly mirror phenotypic results, and even different gene assays don't always yield matching data.
Although one study suggested some success with cyclic treatment interruptions, at least in half of a highly selected and small cohort, other work continued to uncover potential flaws in that strategy. In Switzerland transmission of drug-resistant virus apparently peaked three years ago, but studies in several US and UK cities tracked continuing slow rises in resistance transmission.
Heartache, though, was not the only sequela of news from Sitges. Abbott investigators continued to tout the cardiotonic4 attributes of lopinavir (ABT-378) for heavy-hearted folks with resistant virus. And sedulous work by the Virco team failed to link a distinct resistance mutation pattern to tipranavir, another investigational protease inhibitor (PI). Other research chipped away at mechanisms of resistance to zidovudine (AZT) and other nucleoside reverse transcriptase inhibitors (NRTIs), while one study suggested that these findings could lead to antiresistance agents.
And this Resistance Workshop did not lack for insightful forays into the clinical arena. One analysis of two trial cohorts upset conventional opinion in finding that blips from viral loads below 50 copies/mL do not inevitably herald full-fledged rebounds. And two studies of people who traded in their PI for nevirapine or abacavir recorded continued viral suppression more often among those who had not taken one or two NRTIs before starting their PI regimen.
Attempting, in four days, to absorb and digest the mounds of data served up in Sitges can leave one agreeing with Mr. Brooke, the doting but fatuous uncle in George Eliot's Middlemarch, that these damned scientists just keep uncovering more complications. And the science that cowed Mr. Brooke was "agricultural chemistry," not retrovirology. So anyone expecting more tidy answers than thorny questions at a resistance meeting will not be heartened by this report.
Still, dogged science has its rewards, workshop cochair John Mellors (University of Pittsburgh) reminded colleagues. Ignoring a problem may spare investigators the frustration of long and sometimes fruitless labor, and it will invariably keep tougher questions in the closet. But "when we don't understand a phenomenon like resistance to AZT," he said, "we should go back and do the fundamental work," because HIV's riddles are neither wholly discrete nor utterly inscrutable.
Can pipeline protease inhibitors do better than current PIs in outmuscling resistant virus? The folks at Abbott (lopinavir, ABT-378), Boehringer Ingelheim (tipranavir), and Bristol-Myers Squibb (BMS-232632) would like you to think so. And they showed up in Sitges to make their case.
The study reviewed at the Sitges meeting recruited people who had taken at least two earlier PIs, simultaneously or sequentially, but again they had never tasted an NNRTI [abstract 89]. The 57 people who enrolled in this study had sampled a median of three PIs and four NRTIs. With an average baseline viral load of about 32,000 copies/mL, they simultaneously began lopinavir/ritonavir (400/100 mg or 533/133 mg twice daily), efavirenz, and nucleosides chosen by their clinicians.
Most baseline isolates (79 percent) had one or more primary PI mutations (at positions 82, 84, or 90) plus secondary mutations at positions 10 and/or 71. Abbott's Eugene Sun reported that 55 percent of those isolates had more than a 4-fold drop in susceptibility to lopinavir, 43 percent had more than a 10-fold drop, and 14 percent had more than a 40-fold drop.
After 24 weeks of treatment, 42 of 56 people (75 percent) who had baseline phenotyping and genotyping notched a viral load below 400 copies/mL. Ten people had a viral load above 400 copies/mL at week 24 or dropped out earlier because of a 400-plus viral load. Four people quit before week eight.
The virologic response correlated closely with the number of baseline resistance mutations (P= 0.019) and with baseline susceptibility to lopinavir (P= 0.01). Almost everyone (96 percent) with 0 to 5 primary PI baseline mutations had a 24-week viral load below 400 copies/mL, as did 76 percent with 6 to 7 baseline mutations, and 33 percent with 8 to 10 baseline mutations. Among people with less than 10-fold resistance to lopinavir at base line, 93 percent had a 24-week viral load below 400 copies/mL, compared with 78 percent with 10- to 20-fold baseline resistance, 67 percent with 20- to 40-fold baseline resistance, and 50 percent with more than 40-fold baseline resistance.
These findings led Sun to propose that people with up to a 20-fold swoon in susceptibility to lopinavir will enjoy a "robust response" with lopinavir plus ritonavir. People starting the two PIs with 20- to 40-fold resistance, he conjectured, fall into the viral load response "gray zone." But a 67 percent 24-week response after earlier PI experience is a bright shade of gray.
Still, some workshop attendees remained unconvinced by these findings because many studies show good responses among PI-experienced people beginning their first NNRTI. "Efavirenz is likely doing a huge part of the work" in this study, argued Steven Deeks (University of California, San Francisco). "In terms of clinically meaningful [susceptibility] cutoffs," he saw Sun's analysis as "a bit of a stretch."
The poster for Sun's presentation offered Abbott's retort to that reasoning: "While the use of efavirenz in the study regimen undoubtedly affected the overall response rate, the correlation of response with baseline susceptibility to [lopinavir] demonstrates the antiviral activity of [lopinavir/ritonavir] in this population." This debate could be settled in an ongoing study of lopinavir plus ritonavir in people with PI and NNRTI experience.
Meanwhile, Abbott's Dale Kempf is hard at work tracking down genotypic talismans of lopinavir/ritonavir failure [abstract 38]. To do so, he has had to parse results from the two studies of lopinavir in PI-experienced persons, because mutations have not been documented after 72 weeks of treatment with lopinavir/ritonavir, stavudine (d4T), and lamivudine (3TC) among treatment-naive individuals.
Kempf's analysis involved 112 isolates from people enrolled in Abbott's two PI-experienced studies. Matching mutations with decreased susceptibility scores, he linked changes at 11 positions with resistance to lopinavir--the familiar changes at codons 10, 20, 24, 46, 54, 63, 71, 82, 84, and 90, and a "newcomer" at position 53. Sixteen isolates with more than a 20-fold decrease in susceptibility to lopinavir all had mutations at positions 10, 54, 63, plus 82, and/or 84. They also contained a median of three mutations at positions 20, 24, 46, 53, 71, and 90. Continuing study, Kempf said, could single out the mutations most responsible for resistance.
Most interestingly, Kempf showed that an isolate could harbor up to five PI resistance mutations and still have only a median 4.2-fold drop in 50 percent effective concentration (EC50) to lopinavir (Figure 1). Virus with six, seven, eight, nine, or 10 mutations had incrementally greater resistance of, respectively, 12-, 16-, 34-, 44-, and 53-fold. The Abbott team concluded that the number of defined mutations "constitutes a virtual phenotype for [lopinavir] and can be used to predict phenotypic susceptibility," and that "the accumulation of at least six to eight mutations might be required to achieve clinically relevant phenotypic resistance" to lopinavir.
 |
| Virus with as many as five protease mutations remained susceptible to lopinavir in a two-study analysis by Abbott [abstract 38]. Virus with six, seven, eight, nine, or 10 mutations had incrementally higher fold resistance. Other work by Abbott [abstract 89] showed that, among PI-experienced people with less than 10-fold resistance to lopinavir at baseline, 93 percent starting lopinavir/ritonavir (400/100 mg twice daily) had a 24-week viral load below 400 copies/mL, as did 78 percent with 10- to 20-fold baseline resistance, 67 percent with 20- to 40-fold baseline resistance, and 50 percent with more than 40-fold baseline resistance. But study participants were NNRTI naive and also took efavirenz when starting lopinavir. |
Together, this work by Sun and Kempf corroborates Kempf's earlier finding that lopinavir erects imposing barriers to resistance.6 For people without PI experience, that should mean a sustained response to lopinavir/ritonavir, if they take the drugs faithfully. People who have tried up to three PIs seem likely to enjoy a zesty virologic response even if they start with as many as five protease mutations--at least if they're naive to NNRTIs and add a potent nonnuke. Everyone looks forward to seeing how hardy lopinavir remains in PI-experienced people who have already played their NNRTI card.
Certainly no one could fault Kemp for lack of diligence. She began by constructing viruses that contained mutations linked to decreased susceptibility to tipranavir in Larder's study.7 This mutant backbone included the 10I, 20R, 36I, 71V, and 84V mutations. Although highly resistant to other PIs, this construct had less than 2-fold resistance to tipranavir. So Kemp added one or two heavyweight PI mutations, 82T and/or 90M. Resistance to this fearsome virus inched up to a 2.4-fold drop in susceptibility to tipranavir. Next she threw in changes at protease positions 13, 41, 53, and 73. Tipranavir shrugged.
So Kemp changed tacks. She began passage studies in which clinical isolates resistant to other PIs but sensitive to tipranavir get exposed to increasing levels of tipranavir. After 67 mind-numbing days, the passage studies selected the mutations 47V and 82L, but the drop in susceptibility was still only 13-fold. After 116 days, 85V and 91A mutations emerged. When Kemp added the 47V and 82L mutations to her backbone mutant construct, tipranavir resistance edged up to 4.5-fold. She proffered the preliminary conclusion that 82L or 82T seems to play a part in resistance to tipranavir.
At this point even the audience was exhausted at Kemp's unavailing odyssey. Thomas Merigan (Stanford University) proposed that a "blocking" mutation somewhere in Kemp's viral construct may be reversing resistance or that protease cleavage site mutations may figure in resistance to tipranavir. Others suggested that the site-directed mutagenesis approach Kemp used can amount to a fishing expedition. By this reasoning, starting instead with in vitro passage studies could let the virus--rather than the virologist--do the work.
Colonno sorted the viral isolates according to how many current PIs remained susceptible to them. Among 19 isolates with more than 5-fold resistance to a single other PI, one (5 percent) had more than 5-fold resistance to the BMS drug. Among nine isolates with more than 5-fold resistance to two other PIs, none proved resistant to 232632. Of 10 isolates resistant to three other PIs, three (30 percent) were also resistant to 232632. Eight of 12 isolates (67 percent) resistant to four other PIs had cross-resistance to 232632, and all 13 isolates resistant to the five marketed PIs had more than 5-fold resistance to BMS's PI.
These results led John Mellors to conclude, in a workshop wrap-up, that 232632 will usually keep its punch against viruses resistant to one or two other PIs. But a virus that can thwart three, four, or five PIs stands an escalating chance of cross-resistance to the BMS drug. If that trend holds true in practice, it could mean 232632 sidesteps cross-resistance less deftly than tipranavir or lopinavir. As already noted, nearly nine in 10 isolates resistant to four or more other PIs fell to tipranavir. And in the study Abbott's Eugene Sun presented, 42 of 56 people (75 percent) with a median exposure to three PIs got their viral load below 400 copies/mL after 24 weeks of lopinavir/ritonavir--albeit with a first-time NNRTI as well [abstract 89]. One advantage 232632 may enjoy over rival PIs is once-daily dosing, although the best dose for this drug remains undetermined.
New data suggest that several agents perking through the development pipeline--and one already on the market--could collar virus resistant to reverse transcriptase (RT) inhibitors. The investigational agents include a nucleoside (DAPD), a nucleotide (tenofovir), and an inhibitor of ribonuclease H (SP1093V) that frustrates NRTI- and NNRTI-resistant virus. The licensed agent that spoils the fun of multinucleoside-resistant virus turns out not to be an antiretroviral at all: hydroxyurea. Further down the pike--and not even in the pipeline--may lie novel agents that directly unhinge the mechanism of RT resistance.
How does this pattern pan out clinically? The compound remains to be tested in people infected with these mutants, but in treatment-naive people it knocks down viral loads swiftly. Steven Deeks and coworkers gave DAPD alone to 29 people with CD4+ counts between 307 and 551 cells/mm3 and viral loads ranging from 4 to almost 5 logs [abstract 9]. The highest doses studied, 300 and 500 mg daily, sliced viral loads by 1.5 and 1.46 logs. Fifteen days of DAPD monotherapy roused no detectable RT mutations.
Ten isolates tarnished by 184V alone succumbed to tenofovir a tad more readily than did unmutated virus. Another 10 viruses with a bunch of AZT mutations had only a 3.7-fold drop in susceptibility to tenofovir, compared with a 47-fold plummet to AZT. When Miller tested 10 more isolates with both the AZT mutations and 184V, he found that it too remained susceptible to tenofovir, with a mere 2.4-fold dimming of susceptibility.
Other isolates that failed to intimidate tenofovir were five with the 151M multi-NRTI-associated mutation (1.8-fold decrease in susceptibility), five with 151M plus 184V (1.6-fold), four 65R mutants, 2 of them with 151M (3.4-fold), and four 65R mutants with 184V (1.5-fold). In these studies only multi-NRTI-resistant virus conferred by codon 69 insertions proved resistant to tenofovir (23-fold). And when isolates with the 69 inserts also carried 184V, resistance retreated to the "intermediate" range (6-fold).
Early clinical study indicates that tenofovir tightens the rein on replication when added to an incompletely suppressive regimen, and it does so with more vigor than adefovir did in similar trials. But a few clouds recently darkened tenofovir's otherwise azure sky. Rats, dogs, and monkeys suffered bone thinning after receiving tenofovir doses 10 to 30 times those used in people. Although no hints of that side effect have emerged in human studies, the FDA now wants 48-week safety data on tenofovir, and that will push back the approval date--and the expanded access date--substantially. Gilead surmised that reduced clearance of tenofovir could explain the animal bone thinning, and that slower clearance may signal kidney toxicity. Nephrotoxicity played a big part in unraveling adefovir's approval hopes.
On top of that, serial passage studies using low concentrations of SP1093V have failed to spur the emergence of any resistance-conferring mutations of its own. Because Parniak suspected that mutations at the 501 position could generate resistant virus, he crafted an array of 501 mutants and subjected them to BBNH, the ribonuclease H inhibitor from which he derived SP1093V. Sure enough, all but one of the mutants destroyed BBNH's ability to shut down ribonuclease H activity. But those mutants gummed up reverse transcriptase. And the one mutant that didn't hobble reverse transcriptase remained sensitive to BBNH.
To be sure, SP1093V may trip over all sorts of toe-stubbers before it even makes it to big clinical trials. But the very discovery of such an agent reconfirms that drug-resistant virus can be tripped up too.
While testing hydroxyurea's effect on the 50 percent inhibitory concentration (IC50) of didanosine (ddI) against NRTI-resistant viruses, Race noticed that hydroxyurea seemed to flex some antiviral muscle of its own. To test that possibility, she doused a panel of viral variants with different concentrations of the drug. As expected, hydroxyurea did nothing to dampen replication of four NRTI-sensitive viruses. But against 13 viruses resistant to one to three NRTIs, the drug showed moderate activity, at a median IC50 of 156.5 µM, compared with IC50s above 500 µM against NRTI-sensitive viruses. Hydroxyurea proved even more active, at a median IC50 of 73.5 µM, against 31 viruses resistant to four or five NRTIs, most of them with multi-NRTI-associated codon 69 insertion mutations.
Race suggested her findings "could constitute the basis for a rational and targeted use of this compound in antiretroviral salvage therapy." Targeted is the operative word here. Clinicians who have given hydroxyurea to people with advanced disease know that it can contribute to viral load reductions, but often at the expense of limited CD4+-cell gains and potentiation of ddI-relate d side effects. Perhaps only people with multinucleoside resistance, presumably identified by genotyping, stand to benefit from this drug when HIV infection is advanced.
The not-so-good news on hydroxyurea emerged from a serial drug withdrawal study by Felipe Garcia (University of Barcelona). The title of Garcia's presentation said it all: "Hydroxyurea neither prevents viral load rebound nor facilitates further control of viral replication after three consecutive cyclic interruptions of therapy in chronic HIV infection" [abstract 189]. He randomized 18 individuals who reached viral loads below 20 copies/mL for at least 52 weeks while taking ddI, d4T, and indinavir to continue that regimen or to add hydroxyurea for another six months. Everyone then had three STIs. The two groups didn't differ in (1) viral doubling time during the STI viral rebounds, (2) development of HIV-specific cytotoxic T lymphocyte (CTL) responses, (3) emergence of resistance mutations, or (4) control of viral replication after the third STI.
This year Meyer returned with data indicating that foscarnet-induced mutations limit this process and so resensitize AZT-resistant virus to AZT [abstract 14]. That probably explains why foscarnet prodrugs being studied by Mellors and others reverse AZT resistance [abstract 3]. When Mellors added mutations evoked by the prodrugs to recombinant virus with an AZT-resistant backbone, the virus once more became susceptible to AZT. The advantage these produgs have over foscarnet is oral bioavailability. But they will also have to be less toxic than foscarnet to become viable antiretrovirals.
Work by Johan Lennerstrand (Virco) suggested that resistance to d4T, in some cases, relies on the same mechanism Meyer and Götte discovered for AZT [abstract 16]. When codon 69 insertions make virus resistant to d4T, Lennerstrand found, the AZT mechanism probably applies. But when a 75T mutation (rarely seen in humans) causes d4T resistance, the problem is probably decreased binding of d4T triphosphate to reverse transcriptase.
Bruno Canard (CNRS, Marseille) pursued this line of study by synthesizing novel versions of AZT and d4T that might counter this resistance mechanism [abstract 17]. Both agents interfered with pyrophosphorolysis, the process that repairs DNA chain disruption by NRTIs and so causes resistance. "Our results call for an increasing effort to design pyrophosphorolysis inhibitors," Canard concluded. "In that context, our novel analogues are excellent candidates to overcome [inadequate intracellular NRTI concentrations, incomplete inhibition of replication, and emergence of drug-resistant virus], and might have a generic value in fighting viral drug resistance."
Arguments over the clinical value of resistance testing are probably moot at this point, because most clinicians whose patients can pay for these tests seem to be using them. An electronic audience poll conducted at last fall's ICAAC by Charles Boucher (University of Utrecht) found that about three quarters of 600 clinicians who treat 50 or more people with HIV infection "occasionally," "frequently," or "always" use resistance tests when switching drug regimens.11 Those who don't use the tests cited "lack of reimbursement" as their reason more than twice as often as "not clinically proven" or "interpretation is difficult."
To be sure, these tests have not been clinically proven, though four (or maybe four-and-a-half) of six prospective randomized trials tie their use to better short-term control of viral load.12 Only a month before the Sitges meeting, the International AIDS Society (IAS)-USA Panel recommended testing after failure of a first regimen, after failure of multiple regimens, and for pregnant women.2 The panel suggested that clinicians should "consider testing" for people with primary HIV infection or established, untreated infection.
In an editorial riposte to these guidelines, Charles Flexner (Johns Hopkins) evinced high dubiety about the clinical value of genotyping and phenotyping.3 The tests are expensive, he wrote, and the turnaround is slow. Both "have fundamental problems with sensitivity and specificity." The IAS-USA panel, he noted, recommends or suggests considering resistance testing for every infected person except those who refuse treatment and those whose viremia remains fully suppressed. If a person needs these tests a few times a year, the thousands of dollars multiple tests would cost "might be more effectively spent, for example, on directly observed therapy managed by a visiting home nurse."
If Flexner just a party-pooper? He may have more misgivings than most, but he's hardly alone in harboring doubts. The IAS-USA panel itself spent plenty of time spelling out other limitations of resistance testing: the genetic diversity of HIV-1, interactions between mutations in the same virus, the fallibility of genotype-based algorithms, lack of "well-defined, clinically-validated cutoff points" for phenotyping, the impact of cellular mechanisms of resistance.2
More than a dozen presentations in Sitges grappled with the promise and problems of resistance testing, and skirmishes during attendant Q&A discussions grew more heated than at any other workshop session. Boiling these debates down to one or two questions would oversimplify these set-tos, but two questions did emerge as cardinal concerns: How closely do genotypic changes mirror phenotype (the actual waning of susceptibility to a drug)? And what, if anything, do phenotypic cutoffs--the 2.5-, or 4-, or 10-fold benchmarks--have to do with clinical response to the drug being assayed?
Contending with these conundrums wasn't made easier when Pierre-Marie Girard, a clinician at the Rothschild Hospital in Paris, overturned the blissful trend in resistance assay studies with results indicating, generally, that neither genotyping nor phenotyping does any better than standard-of-care history taking when planning a regimen for treatment-experienced people. Girard's unsettling analysis followed an update on VIRA3001, a study that showed an advantage for phenotyping. And preliminary results of another genotyping trial supported positive trends in earlier studies.
VIRA3001, a prospective phenotyping study, randomized 111 people to the phenotyping arm and 110 to the standard-of-care arm after failure of a first PI regimen [abstract 84]. No expert advice accompanied the phenotypic results. Everyone had already tried at least two NRTIs, but most had not taken an NNRTI. Calvin Cohen (Community Research Initiative, Brookline) reported equivalent baseline viral load and CD4+ numbers in the two groups.
People in the phenotyping arm switched more drugs, in every drug class, than those in the standard-of-care arm (P < 0.001), and that difference probably accounts for the better virologic response in the phenotyped group after 16 weeks. In an intent-to-treat analysis that included premature virologic failures, the median viral load dropped 1.23 logs in the phenotyped group and 0.87 log in the standard-of-care group after 16 weeks (P= 0.004). By the same analysis, 59 percent in the pheno group and 42 percent in the nonpheno group had a 16-week viral load below 400 copies/mL (P= 0.033). But in a more stringent missing-data-equal-failure analysis, those reaching sub-400 viral loads did not differ significantly--45 percent for phenotyping and 34 percent for standard of care (P= 0.099). The groups didn't differ in CD4+-cell response.
The number of active drugs in the new regimen proved the deciding factor in viral control. Among study participants with one active drug in the new regimen, none attained a 16-week viral load below 400 copies/mL. But for people taking two, three, or four new active drugs, 16-week sub-400 rates were 28 percent, 50 percent, and 55 percent, respectively. And people in the phenotyping group usually got more active drugs in their new regimen--that is, more drugs to which the virus remained susceptible (Table 1). It is interesting that the number of active nucleosides, rather than PIs, appeared to weigh more heavily in this difference between the groups. On the other hand, people in the standard-of-care arm got marginally more active NNRTIs.
Although this trial favors the phenotyping group by many measures, it is disappointing that the 16-week response came to only 45 percent in a strict intent-to-treat analysis in the phenotyping arm. That dim result may reflect the tendency not to use NNRTIs in either treatment arm (Table 1).
| Phenotyping arm (%) |
Standard-of- care arm (%) |
P | |
|---|---|---|---|
| >Two active NRTIs | 63 | 40 | <0.001 |
| >One active PI | 60 | 47 | 0.059 |
| >One active NNRTI | 38 | 49 | 0.108 |
| >Two active drugs from any class | 92 | 77 | 0.003 |
| NNRTIs = nonnucleoside reverse transcriptase inhibitors; NRTIs = nucleoside reverse transcriptase inhibitors; PIs = protease inhibitors. Source: Calvin Cohen, abstract 84. |
|||
Preliminary results from an ongoing Spanish study, the Havana trial, add to the string of successes with genotyping. This many-armed trial, presented by Lidia Ruiz (Germans Trias i Pujol University Hospital, Badalona), includes 315 people from 13 sites who have endured one, two, or three virologic failures (>1000 HIV RNA copies/mL) [abstract 59]. Their treatment experience ranges from NRTIs only to more than three PI-containing regimens. Ruiz and colleagues randomized everyone to a genotyping arm or a no-genotyping standard-of-care arm. The genotyping group was further randomized to get treatment advice from a software algorithm, or from the algorithm plus expert advice. And the standard-of-care arm was randomized to get expert advice or no advice.
So far, 274 people have been treated up to 12 weeks after randomization. Already the viral load decrease in the genotyping arm significantly exceeds that in the standard-of-care arm (-1.45 logs vs. -1.22 logs, P= 0.047). More people in the genotyping arm have reached a viral load below 400 copies/mL, though Ruiz did not report the numbers. She told the Journal that expert advice appears to be an advantage, but one that has not reached statistical significance.
If one considers only these two trials, and two reported earlier,12 all roads appear to lead to a role for resistance testing. But the 46-center NARVAL study (ANRS 088) could encourage renewed compass consulting. Pierre-Marie Girard and colleagues enrolled 525 people who had taken a PI regimen for more than three months and had a viral load above 1000 copies/mL [abstract 85]. They randomized them to have genotyping, phenotyping, or neither, and the three groups had equivalent baseline viral loads (medians of about 20,000 copies/mL) and CD4+ counts (medians in the high 200s). Clinicians treating people in the geno and pheno arms got expert advice on "first choice" drugs for the new regimen, "second choice" drugs, and poor choices.
People in this cohort had already taken a median of seven antiretrovirals, and most of them had used up more than two purportedly potent regimens. Although Girard did not break down his results according to how many active drugs people had available, the number must have been smaller, for example, than in Cohen's phenotyping study [abstract 84], whose enrollees had taken no more than one PI and most of whom had never tried an NNRTI. Most of the French patients had taken PIs for more than 30 months.
An intent-to-treat analysis showed no significant differences between the three groups in the primary endpoint--a 12-week viral load below 200 copies/mL: 41 percent in the genotyping arm, 34 percent in the phenotyping arm, and 34 percent in the standard-of-care arm hit that mark. At that point, study participants with viral loads still above 200 copies/mL could get phenotyping if they had been in the genotyping arm, and vice versa; people in the standard-of-care arm could get genotyping and phenotyping. After 24 weeks of follow-up, completed by 82 percent of enrollees, percentages with viral loads under 200 copies/mL at both weeks 12 and 24 were 29 percent for genotyping, 22 percent for phenotyping, and 17 percent for standard of care. The genotyping result was significantly better than the standard-of-care result (P= 0.008), but the phenotyping result was not (P= 0.255).
Deciding what these findings mean won't be easy. Despite criticism of the French trial design, including the short time to the primary endpoint, Girard proposed that "we should face the results and improve the tests." But Richard Harrigan (University of British Columbia) wondered if the fault lies with the tests or with the drugs. Although 80 percent of clinicians in the geno and pheno arms followed the test-based advice, John Mellors observed, many of them probably had no good "first choice" drug to turn to because of their extensive treatment experience.
But what struck François Clavel (Bichat-Claude Bernard Hospital, Paris), one of the investigators, was that phenotyping offered no help in this highly experienced cohort. A common assumption, he reminded colleagues, is that genotyping will yield a useless hash of data when people have worked their way through several regimens, and that phenotyping will prove a better option for them. NARVAL suggests otherwise.
And a phenotyping study presented earlier this year agreed with NARVAL on the value of these assays for people with moderate to heavy antiretroviral experience. In this 115-person US study, 75 percent had used two or more PIs.13 Even though phenotyping identified "active drugs" for 30 of 54 people (56 percent) who got phenotyping in that randomized trial, investigators rated only 15 percent as "virologic successes" after 16 weeks of a new regimen--worse than the standard-of-care group. About 75 percent in that study had taken two or more PIs before randomization. But treatment experience appears not to be a failsafe discriminator between good and poor results with genotyping. In the Havana study, for example, 179 of 315 people (57 percent) stratified by treatment experience had burned through at least two PI regimens, yet genotyping helped pick a better regimen regardless of previous PI use [abstract 59].
But NARVAL may not be all bad news, suggested Graeme Moyle (Chelsea and Westminster Hospital, London). Clinicians in the standard-of-care arm used three or more new antiretrovirals for 55 percent of patients, compared with 20 percent of patients in the genotyping or phenotyping arm who got that many new drugs (P < 0.001). And the standard-of-care participants certainly did not enjoy better viral control as a result. That means they squandered more antiretrovirals than people in the other groups, with nothing to show for it. In that sense, both genotyping and phenotyping may have been worth the cost.
More randomized resistance assay studies are in the works, but right now the bottom line is this: Despite the obvious difficulty in interpreting genotypic tests, and despite the caveats attending all genotyping studies, four of four such studies suggest at least one significantly better virologic result than standard of care.14 Among three studies of phenotypic testing, only one can make that claim, the one in which people had the least treatment experience.15
And Anne-Mieke Vandamme (Rega Institute, Leuven) noted that none of the resistance assay studies have figured how well participants adhered to prescribed regimens. In a study she conducted (one not reported at the workshop), people with baseline resistance to some drugs responded to the next regimen when they adhered completely. "Resistance testing," she insisted, "is not the holy grail."
One reason for differing virologic results with genotyping and phenotyping could be that the two tests don't correlate precisely. Although genotype certainly determines phenotype, there are enough gray pages in the annals of this young science to make a Manichaean weep. Beyond the sometimes looser than desirable fit between genotype and phenotype, studies have uncovered some variance between the output of one genotypic assay compared with another, and one phenotypic system compared with another.
Given the protean nature of the virus and the blurry continuum separating ice-clean poles of "wild-type" and "mutant" or "susceptible" and "resistant," it would be queer if results with different assays proved 100 percent concordant. So the question becomes, "How close do they have to be for comfort?"
Shoukat Qari (CDC, Atlanta) gave 529 pairs of plasma samples to Virco and ViroLogic to see how closely the two phenotypers' results would match [abstract 62]. The apparently reassuring result was that the two assays made the same call--either "sensitive" or "decreased susceptibility"--on 488 samples (92 percent). And among the 41 samples with discordant results, 26 (63 percent) had changes in susceptibility within 1 fold of either assay's cutoff.
But Qari's study was not a stern test of the two big phenotypers because only 32 percent of the plasma samples came from drug-experienced people, so only a small proportion had decreased susceptibility to antiretrovirals. Of the 488 concordant samples, 459 (94 percent) were rated sensitive and only 29 (6 percent) resistant. A study evaluating more plasma samples with a range of susceptibilities to different drugs would have been more compelling, and perhaps less reassuring.
How do genotyping tests compare with each other? Fairly well, but hardly perfectly, according to an analysis of Visible Genetics' TruGene and Perkin-Elmer's ViroSeq by Diane Descamps (Bichat-Claude Bernard Hospital, Paris) [abstract 68]. After testing plasma samples from 125 treatment-experienced individuals with both kits, she found 18 discordant calls on primary RT or protease mutations in 14 persons (11 percent). For example, ViroSeq spotted a 215 mutation in one individual that TruGene did not, while TruGene saw a 103 mutation in another person that ViroSeq missed. Looking at secondary RT and protease mutations, Descamps counted 24 discordant calls involving 18 persons (14 percent). Neither assay seemed more error-prone than the other.
Descamps figured that discordant calls would have led to incorrect genotypic interpretations in 11 of 125 people (9 percent). Five of these 11 discordant results persisted after a second genotyping run, but it's unlikely that a clinician would ask for a second genotype unless the first results smelled fishy. Although Descamps rated the concordance of the tests "good," misleading genotyping calls for nearly one in 10 people clearly leave much room for improvement.
Three groups compared genotypic and phenotypic results and again found less than a hand-in-glove fit. Philippe Clevenbergh (University Hospital, Nice) reported good overall correlation between genotypic calls and phenotypic findings (by Virco) among 90 participants in the VIRADAPT trial [abstract 90].12 But a drug-by-drug analysis showed several discordances, usually involving NRTIs. For AZT, the VIRADAPT genotyping algorithm predicted resistance about twice as often as it could be detected phenotypically. For ddI, zalcitabine (ddC), and d4T, the algorithm signaled resistance about 10 times more often than phenotyping. And for abacavir, genotyping suggested resistance nearly 30 times more often than phenotyping. On the other hand, the genotyping algorithm predicted that nearly 50 percent of isolates would be resistant to 3TC, but phenotyping called about 70 percent resistant. Genotyping also overestimated resistance to indinavir and amprenavir. Clevenbergh politely suggested that "the definition of phenotypic resistance should possibly be individualized for each drug."
Françoise Brun-Vézinet (Bichat-Claude Bernard Hospital, Paris) took a similar look at viral isolates from people in the NARVAL trial, and like Clevenbergh she found the greatest discordance between genotype and phenotype with NRTIs [abstract 100]. Genotyping (interpreted by an algorithm) sounded the resistance alarm for d4T and abacavir twice or more often than phenotyping (by recombinant virus assay) could verify decreased susceptibility. And again phenotyping showed resistance to 3TC when resistance- conferring mutations could not be detected. Amprenavir was the only PI on which the tests disagreed substantially: phenotyping indicated resistance in about 60 percent of isolates, whereas genotyping predicted resistance in fewer than 40 percent.
Brun-Vézinet noted that the French trial group's algorithm would be revised on the basis of these results. Clearly, any algorithm worth consulting would have to be updated regularly, because research routinely finds new mutations and mutation interactions.16 She also echoed another recurrent workshop theme: "Interpretation of phenotype will require the use of updated and clinically relevant cutoff values." But determining clinical relevance will take some work.
ViroLogic's Neil Parkin focused his pheno-geno comparison on clinical isolates of people who had taken protease inhibitors and once more found that algorithms can mislead [abstract 64]. He argued that primary protease mutations, like those at positions 82 and 90, don't always render a virus less susceptible to PIs because secondary mutations "may influence drug susceptibility in unpredictable ways."
Among isolates on ice at ViroLogic, Parkin found that 37 percent with a position 82 mutation remained sensitive to saquinavir and 8 percent were hypersensitive to that drug; 17 percent with an 82 mutation were sensitive to indinavir. Background mutations at positions 54, 71, or 77 correlated significantly with decreased susceptibility to saquinavir (P < 0.05), while a position 46 mutation correlated with increased susceptibility (P= 0.08).
This analysis reflects clinical reality, Parkin maintained, because he identified cases in which people with mutations at positions 82 or 90 had saquinavir-susceptible virus and responded virologically to treatment with saquinavir. Genotypic algorithms, he concluded, "may allow accurate predictions only in a subset of samples."
Some clinical investigators in the crowd remained unswayed by Parkin's data. William O'Brien (University of Texas, Galveston) noted that this kind of analysis takes a snapshot of the virus, but in infected people HIV keeps evolving cinematically. So a virus with a mutation at position 82 and a few secondary mutations will continue to mutate under drug pressure until it overwhelms the PI being taken. But Thomas Merigan saw value in Parkin's report because it points out that "if people look for just one amino acid change, it's not going to get them anywhere" in interpreting resistance.
An important obstacle to routine clinical use of phenotyping is the slow turnaround, often as long as three or four weeks, some clinicians say. Virco has exploited its vast storehouse of clinical isolates to determine a virus's "virtual" phenotype more quickly. And, in theory, the virtual phenotype should be easier to interpret than a list of mutations. The question is, "Does the virtual phenotype closely match the real phenotype?" Yes, argued Virco's Brendan Larder, on the basis of a 5000-sample comparison of virtual and actual phenotypes [abstract 63]. But first, what is a virtual phenotype?
Virco's viral stash now includes more than 13,000 isolates that have been both genotyped and phenotyped, Larder reported. That allows Virco to build a database of drug-susceptibility phenotypes for a wide array of genotypes. When a clinical sample comes in, technicians can determine the mutations then search the database to retrieve phenotypes of viruses with matching mutations. The computer calculates the average change in IC50 for those matched phenotypes. Five seconds later--voilà--the virtual phenotype of the submitted isolate.
Larder randomly picked 5000 isolates that had been phenotyped and genotyped since the beginning of 1999. Based on the genotype of each virus, he figured the virtual phenotype of that virus for all antiretrovirals. Then, by linear regression analysis, he compared the virtual phenotype of all isolates with their actual phenotypes, and the correlation was tight (r2 = 0.82). Virco's Werner Verbiest lent credence to that result by analyzing isolates from the VIRA3001 phenotyping trial and finding that the virtual phenotype correctly predicted the actual phenotype nine times out of 10 [abstract 81].
Larder also found that the genotype-based resistance algorithm used by the Resistance Collaborative Group17 called NRTI-sensitive virus resistant consistently more often than virtual phenotyping did, whereas that algorithm incorrectly called nelfinavir-resistant virus sensitive more often than virtual phenotyping. Virco's virtual phenotype outdoes genotype-based algorithms, Larder argued, because it is quantitative. In other words, it's based on a library of susceptibility scores rather than on consensus of an expert panel.
An obvious disadvantage of virtual phenotyping compared with real phenotyping is that it can't predict susceptibility to new drugs until Virco's database amasses enough clinical isolates from people with decreased susceptibility to that drug. Larder estimated a six- to 12-month lag between the time when isolates resistant to a new drug first enter the database and when enough data pile up to permit virtual phenotyping.
A more serious concern is one raised by Jonathan Schapiro (National Hemophilia Center, Kochav Yair, Israel) and other resistance-savvy clinicians: What do virtual phenotypes, or even real phenotypes--classified as "sensitive," "intermediate resistance," and "resistant"--have to do with clinical response to a drug? Schapiro noted one example from Larder's presentation in which an isolate with a 90M mutation was called sensitive to saquinavir. ViroLogic's Neil Parkin had reported the same thing [abstract 64, above]. But Schapiro's long experience with saquinavir--as an investigator and a clinician--convinces him that giving saquinavir to someone with 90M is a real bad idea. Phenotyping may say that a 90M mutant virus remains susceptible to saquinavir, but clinical experience says that susceptibility won't last for long--maybe not even until the clinician gets the phenotypic results back.
When Schapiro asked whether virtual phenotyping could replace the expert opinion urgently recommended in the IAS-USA resistance testing guidelines,2 Larder replied--not without reason-- "it depends who the experts are." The shortcoming of some expert-designed algorithms became apparent in the above-mentioned studies of Clevenbergh, Brun-Vézinet, and Larder. Jean-Claude Schmit (Centre Hospitalier de Luxembourg) compared interpretations of genotypic results by six clinicians with lots of HIV treatment experience, two with less experience, and three virologists [abstract 70]. Not only were their opinions often at odds (particularly for ddI, abacavir, efavirenz, and amprenavir), but the results didn't depend on the physicians' clinical experience.
Yet several resistance mavens worried aloud that the simple-seeming interpretation of phenotype--green for sensitive, yellow for intermediate, and red for resistant--is not simple but simplistic. "In the big picture," said Stanford's Mark Winters, "clinicians aren't going to have all the information at their fingertips" that they need to understand a phenotypic report. He suspected that more than a few clinicians "will react in a push-button manner" to the phenotypers' green, yellow, or red.
John Mellors suggested that one way to determine the clinical correlates of phenotypic readouts is to do linear regression analyses for each drug and for each assay. That job is complicated, he quickly added, by the differing activities of individual drugs in different combinations. But all these concerns seem unlikely to dissuade most clinicians from ordering up genotypes, phenotypes, or both if they think those tests will help hone their treatment decisions. Clinicians who do so thoughtfully may well teach themselves to use this technology to their patients' benefit, and they will surely also learn these assays' limitations. The wisest clinicians will exploit not only assay results and expert opinion, but also the growing number of interpretive tools at their disposal. And some of those tools are free! See "Irresistible resistance Web sites."
A glaring example of poor correlation between phenotypic "fold resistance" values and antiviral activity involves lopinavir (ABT-378). As Abbott's Dale Kempf demonstrated earlier this year, 48 weeks of lopinavir plus low-dose ritonavir produced the same virologic response rates in people with "reduced susceptibility" to lopinavir at baseline (>4-fold higher EC50) as in people with "sensitive" virus (<4-fold higher EC50).6 Glaxo's Randall Lanier showed that "decreased susceptibility" to abacavir on phenotypic assays didn't consistently predict a poor virologic response to that drug.18
At the Resistance Workshop two groups offered evidence suggesting that lowered susceptibility to NNRTIs doesn't always portend treatment failure in people starting their first regimen. But other findings on nonnukes stirred even more interest at the meeting. A few groups confirmed earlier findings of hypersensitivity to NNRTIs given to people with NRTI resistance. And ViroLogic scientists found that nonnukes can sometimes stimulate viral replication.
In a multivariate analysis, viral susceptibility to NNRTIs did not correlate with early virologic failure of nonnuke regimens. Two factors that did correlate with failure were female gender, which doubled the risk of failure (P <0.005), and a history of injecting drug use, which almost doubled the risk (P <0.05). Harrigan concluded that "the clinically significant level of resistance to NNRTIs may be greater than 10-fold," perhaps because people typically maintain high plasma levels of these drugs.
Most people in Harrigan's analysis took nevirapine, but a study of susceptibility and response among people starting efavirenz reached the same conclusion. DuPont's Lee Bacheler compared viral susceptibility to efavirenz and virologic response in 100 NNRTI- and PI-naive people beginning an efavirenz regimen [abstract 88]. She used baseline viral load and CD4+ count to match 50 people who endured virologic failures with 50 who maintained viral loads below 50 copies/mL.
Like Harrigan, Bacheler found that moderately decreased susceptibility to this NNRTI before treatment did not influence virologic response. She detected 4- to 10-fold pretreatment resistance to efavirenz among 26 percent in the virologic failure group and in 34 percent of virologic responders. And Bacheler cited the same reason that Harrigan did: "Efavirenz steady state Cmin plasma levels exceed the plasma protein binding adjusted IC50 values of baseline isolates with low-level (4-10 fold) resistance to NNRTIs."
But the DuPont team did find that more than 10-fold baseline resistance to efavirenz foreshadows virologic failure. Five of six people with pretreatment 10-fold resistance to efavirenz suffered a virologic rebound with that drug. Together, though, Harrigan's and Bacheler's studies show that 4- to 10-fold decreased susceptibility to NNRTIs--usually called "intermediate" resistance and coded yellow on phenotypic reports--does not rule out successful treatment with these drugs, at least in people taking their first NNRTI.
The NNRTI hypersusceptibility story began earlier this year when ViroLogic researchers reported hypersensitivity to NNRTIs among people with NRTI mutations and decreased susceptibility to NRTIs.19 They postulated that the phenomenon could explain the potency of NNRTIs in salvage regimens.
Shulman pursued these clues in her retrospective analysis of 30 people beginning efavirenz-containing salvage regimens. She counted 12 of 12 people with NNRTI-associated mutations at baseline as nonresponders (HIV RNA >500 copies/mL) at 12 weeks. Among 18 people without NNRTI mutations when beginning the salvage regimen, she found that eight had virus hypersusceptible to efavirenz (EC50 fold change <0.5 versus a wild-type control). Five of those eight (63 percent) were 12-week virologic responders. Among 10 people without NNRTI mutations and without hypersusceptible virus, only three were 12-week responders. Hypersusceptibility to efavirenz jacked up the chance of 12-week virologic success nearly four times, but that advantage was not statistically significant because of the small size of the study.
Richard Haubrich (University of California, San Diego) explored NNRTI hypersusceptibility in 164 people with NRTI and PI experience who were starting their first NNRTI [abstract 87]. Defining hypersusceptibility as an IC50 fold change less than or equal to 0.4, he found that 26 percent were hypersusceptible to efavirenz, 21 percent to delavirdine, and 18 percent to nevirapine. But hypersusceptibility to one NNRTI didn't ensure hypersusceptibility to another. For example, among 42 people with virus hypersusceptible to efavirenz, 60 percent were hypersusceptible to nevirapine and 69 percent to delavirdine.
Several factors correlated with hypersusceptibility to nonnucleosides, including more months of NRTI treatment before an NNRTI (P= 0.0001) and more individual NRTIs taken before an NNRTI (P <0.002). A history of treatment with AZT or ddC predicted hypersusceptibility to efavirenz (P <0.05), whereas treatment with ddI, d4T, or 3TC did not. However, decreased baseline susceptibility to d4T (IC50 fold change >2.5) correlated with hypersusceptibility to efavirenz (P <0.005), as did decreased susceptibility to AZT or abacavir (IC50 fold change >5 for both).
Predictive trends differed slightly in Shulman's analysis [abstract 99]. She found a Spearman rank correlation between number of NRTI mutations at baseline and greater susceptibility to efavirenz (r = -0.51, P= 0.01). Shulman also identified the AZT-associated mutations 41L and 215Y, and the 3TC- or abacavir-linked 184V mutation, in most people with efavirenz hypersusceptibility in her cohort.
Haubrich tracked viral load and CD4+ responses in 18 people with virus hypersusceptible to efavirenz and in 55 without efavirenz hypersusceptibility. Two months after starting efavirenz, the hypersusceptible group had a 1.5-log drop in viral load, compared with a 1-log dip in the other group (P= 0.03). At month four average viral load decreases had not changed too much in either group, but the difference between them lost statistical significance (P= 0.09).
People with efavirenz hypersusceptibility enjoyed higher and more sustained CD4+ cell gains than people in the nonhypersusceptibile group. At month four the hypersusceptible contingent had an average gain of 52 cells/mm3, compared with 5 cells/mm3 in the other group (P= 0.03). After 10 months of efavirenz therapy, people with hypersusceptible virus had an average gain of about 120 cells/mm3, compared with about 40 cells/mm3 in the control group (P= 0.008). The 10-month follow-up, though, included only seven people in the hypersusceptible camp and 16 in the control arm.
Haubrich's and Shulman's studies strongly suggest that hypersusceptibility to NNRTIs--or at least to efavirenz, the drug most studied in these analyses--can be traced to longer pretreatment with nucleoside analogs. Hypersusceptibility apparently translated into at least a short-term virologic advantage in these cohorts and, in Haubrich's hands, into a longer immunologic edge.
Haubrich contended that "these results should be considered when designing regimen sequencing strategies," apparently implying that a first-line NRTI-PI regimen could make more sense than a first-line NRTI-NNRTI combo. But the evidence is too skimpy so far--seven hypersusceptible people at 10 months in Haubrich's follow-up, and 12 weeks of follow-up in Shulman's study--to settle the argument over whether to start HAART with a PI or an NNRTI.
And hypersusceptibility to NNRTIs after NRTI treatment clearly is not a universal phenomenon. Remember that only eight of 18 people with NRTI experience and no NNRTI mutations (44 percent) were hypersusceptible to efavirenz in Shulman's study. Only 24 percent had efavirenz hypersusceptibility in Haubrich's larger cohort. And analysis of over 4000 isolates by Virco's Stuart Bloor found hypersusceptibility rates of only 2 percent for efavirenz, 4 percent for nevirapine, and 5 percent for delavirdine [abstract 169]. So anyone hoping to exploit NNRTI hypersusceptibility by holding nonnukes for a second or later regimen would have to phenotype a person's virus to verify hypersusceptibility.
To complicate things, sometimes more NRTI experience makes HIV less susceptible to NNRTIs. Among 64 NNRTI-naive but NRTI-experienced persons, Santiago Moreno (Ramón y Cajal Hospital, Madrid) correlated longer NRTI experience with 4- to 10-fold decreased susceptibility to nevirapine (51 vs. 37 months), although that difference only approached statistical significance (P= 0.09) [abstract 32]. And this decreased susceptibility resulted in only a marginally lower virologic response compared with people who had fully susceptible virus (-0.94 log vs. -1.15 logs, P= 0.62).
Esther Race reported that two NNRTI-naive individuals, treated only with ddI plus d4T, had virus that evolved to a nevirapine-resistant phenotype [abstract 33]. She identified two RT polymorphisms, 98S and 108I, that appeared to open the door to nevirapine resistance. In one person, 108I alone did not decrease susceptibility to nevirapine, but continued treatment with ddI/d4T coaxed 41L and 215Y mutations to emerge, yielding 31-fold resistance to nevirapine. In the other person, Race linked 98S alone to 6-fold resistance to nevirapine. When the evolving virus tacked on the multinuke mutation 151M, resistance to nevirapine shot up 120-fold.
"Our results warn that, in the context of particular RT genetic backgrounds and RT sequence polymorphisms," Race concluded, "the emergence of nucleoside analog resistance mutations can significantly decrease HIV susceptibility to NNRTIs in the absence of exposure to this class of antivirals."
Testing viral isolates from this person, ViroLogic confirmed that increasing concentrations of nevirapine boosted HIV p24 levels. In other words, Whitcomb explained, this virus "can replicate more efficiently in the presence of drug than in the absence of drug." The ViroLogic team went on to identify the same phenomenon in a few other individuals and confirmed that delavirdine and efavirenz can exhibit the same unsettling property. In a few instances this NNRTI-dependent stimulation revved viral replication as much as 400 percent.
Further work showed that a novel mutation, 230L, turned up regularly in virus that had this souped-up replicative capacity. ViroLogic's Wei Huang determined that the 230L mutation plus the standard NNRTI mutation 181C rendered virus more than 250-fold resistant to delavirdine and more than 780-fold resistant to nevirapine, far above the fold resistance values one might expect with 181C alone. And 230L plus 103N yielded virus with more than 250-fold, more than 270-fold, and more than 780-fold resistance to delavirdine, efavirenz, and nevirapine, respectively. Virus from three people with 230L plus 103N had 50 to 100 percent jumps in replication compared with the same virus tested in the absence of NNRTIs. But this NNRTI-dependent replication boost could not be documented in a fourth person with mutant 230L/103N virus.
What does all this mean clinically? For the first person in whom ViroLogic spotted this anomaly, it meant better virologic control without a nonnucleoside. But beyond that anecdote, the clinical relevance of NNRTI-dependent replication stimulation remains unclear. "I wouldn't want to see anyone stop using nonnucleosides because of this," Whitcomb stressed in an open discussion after her presentation. So far, she added, ViroLogic has discerned this peculiarity in only about 25 people, less than 1 percent of the ViroLogic database.
Still, Douglas Richman (University of California, San Diego) rated the finding "a very, very intriguing initial observation." Perhaps, he suggested, research should pay more attention to what happens when drugs get dropped from a regimen.
Transmission of drug-resistant virus, a hot topic at least year's Resistance Workshop, hasn't gone away. Four studies charted slow but steady upticks in resistance transmission across the US and the UK. But the sun is shining in Switzerland, where investigators traced a downtrend in new infections with resistant virus. This discrepancy, some workshop attendees noted, may reflect differing national treatment guidelines and, perhaps, better prevention and education in Switzerland.
Susan Little (University of California, San Diego) reported for a seven-city20 cadre of clinicians who compared resistance in people infected from June 1996 to April 1998 or from May 1998 to March 2000 [abstract 172]. Of 230 persons now tracked by Little and colleagues, 12 (5 percent) have more than 10-fold resistance to any antiretroviral. The rate with more than 10-fold resistance climbed from 2 percent in 1997 and 1998, to 11 percent in 1999, and to 22 percent in 2000. From the pre- to post-May 1998 periods, the 10+-fold resistance rate jumped from 1 to 5 percent for NRTIs and NNRTIs, and from 1 to 4 percent for PIs. The overall increase for more than 5-fold plus more than 10-fold resistance is statistically significant (P= 0.03).
Among 118 persons with primary infection in San Francisco, Robert Grant (Gladstone Institute) charted a waning of transmitted NRTI mutations from 15.4 percent in 1996 to 6.3 percent in 1999 (P= 0.08) [abstract 180]. But from 1997 to 1999, rates of mutations arousing resistance to NNRTIs or to PIs both rose from 3.6 to 6.3 percent (P= 0.36). Changes in viral susceptibility to the three drug classes generally reflected those trends.
But the NRTI dip in Grant's cohort was not reflected in a study of 41 newly infected persons referred to the Aaron Diamond Center in New York between April 1, 1999 and March 31, 2000 [abstract 167]. Viviana Simon found that mutation rates in this group were highest for primary mutations conferring resistance to NRTIs (15.3 percent), followed by primary mutations linked to NNRTIs (12.8 percent) and PIs (4.8 percent). The overall rate of resistance transmission in the study year, 24 percent, was substantially higher than the 16 percent rate logged in the preceding four years. Virus from 3 of 39 phenotyped isolates (7.6 percent) had more than 10-fold resistance to at least one drug. Two people in this cohort had multiclass-resistant virus.
A study of newly infected people from across the UK found only one isolate with primary NRTI mutations from among 40 individuals (2.5 percent) infected between 1995 and 1998 [abstract 127]. But among 16 people infected in 1999, three (19 percent) had primary mutations. Patricia Cane (PHLS Antiviral Susceptibility Reference Unit, Birmingham) reported that one of these people had mutations associated with resistance to all three drug classes. A second person had NRTI- and NNRTI-linked mutations. And the third had the 181C NNRTI mutation.
Things are different in Switzerland, where transmission of resistant virus apparently peaked in 1997 [abstract 185]. Sabine Yerly (University of Geneva) reported similar trends for transmission of primary NRTI and PI mutations in 198 persons with primary HIV infection. In 1996 a little over 5 percent of newly infected individuals had NRTI mutations. That rate climbed to 10 percent in 1997, then drifted down to 7 percent in 1998 and close to 3 percent in 1999. Primary protease mutations turned up in 3 percent of people infected in 1996, in more than 8 percent in 1997, then in only 2 percent in 1998 and 1999.
Yerly proposed that increasingly tighter control of HIV replication across the 4500-person Swiss HIV Cohort may contribute to this falloff in resistance transmission. In 1996 about 10 percent in the nationwide cohort had a viral load below 400 copies/mL. That rate rose to near 30 percent in 1997, exceeded 40 percent in 1998, and topped 50 percent in 1999. "Antiretroviral therapy, close monitoring of patients and education programs," Yerly concluded, "may reverse the trend to increased transmission of drug-resistant variants."
Infection with virus carrying primary resistance mutations correlated strongly with 24-week virologic failure in a study from several Italian sites [abstract 187]. Claudia Balotta (University of Milan) found that neither pretreatment viral load nor pretreatment CD4+ count predicted drug failure in a multivariate analysis. But primary reverse transcriptase mutations upped the odds of treatment failure more than 21 times (P= 0.03), and primary protease mutations did so more than 8 times (P= 0.06). The analysis involved 26 people who had completed 24 weeks of treatment.
Perhaps the workshop's most compelling presentation (if one is considering leading sacred cows to slaughter) involved analysis of participants in two indinavir trials who had brief bouts of plasma viremia--blips--while taking that PI. At last year's workshop, attendees learned that viremic spikes between 50 and 200 copies/mL led to a longer viral decay half-life--59 weeks in people with intermittent viremia versus 26 weeks in people whose viral loads stayed hunkered under 50 copies/mL (P= 0.007).21 The Aaron Diamond Center's Bharat Ramratnam proposed that every blip reflooded draining viral reservoirs and so prolonged the time it would take to eliminate the virus, if elimination ever proves possible. But do those blips prejudice a person's chance for continuing virologic control over the next few years? That's the question asked by Diane Havlir (University of California, San Diego) in her reanalysis of the Merck 035 study and ACTG 343 [abstract 112]. The answer, she found, was no. If that no holds true for diverse HAART-treated populations, it would undercut the tenet that durable viral suppression demands the tightest control of viremia (the holy bovine alluded to above).
The two studies whose data Havlir resifted both enrolled people with, at most, AZT experience. After people had a confirmed viral load below 50 copies/mL, a blip meant a 50-plus reading that then dove back into sub-50 territory. A virologic rebound meant two consecutive viral loads above 200 copies/mL.
Among Merck 035 participants followed for a median of 4.5 years, none of seven blipless people endured a viral rebound. But neither did six people who had blips during those 4.5 years. Blips became more and more infrequent as follow-up continued. In the ACTG 343 analysis, Havlir excluded people whose treatment failed when they scaled back their regimen to AZT/3TC or indinavir alone during the maintenance phase of this trial. After more than 60 weeks of follow-up, 96 of 241 people (40 percent) had at least one blip. Nine of these 96 blippers (9.4 percent) later had a confirmed rebound to more than 200 copies/mL, compared with 20 rebounders among 145 people (13.8 percent) who never blipped.
Viremic spikes did signal higher levels of viral replication, according to analysis of a subset of people whose viremia was measured by an assay that counts as few as 2.5 copies/mL. The median viral load was 23 copies/mL in spikers, compared with 2.5 copies/mL in those who had no spikes (P < 0.001). But Havlir charted at least one viral load above 2.5 copies/mL in everyone treated for as long as five years, whether they had blips above 50 copies or not.
What do these unexpected findings mean? Havlir stressed that they must be interpreted cautiously because she did not measure adherence or pharmacokinetics and because they may not apply to people taking other regimens or people with more treatment experience than those enrolled in Merck 035 or ACTG 343.
Another workshop study showed, though, that blips don't spell doom for people taking other regimens, a few of whom had slightly more treatment experience than those in Havlir's cohorts. Annemarie Wensing (University Medical Center, Utrecht) presented the cases of 11 people who reached viral loads under 50 copies/mL, then had transient spikes ranging from 50 to 462 copies/mL (median 72 copies/mL) [abstract 137]. Viral loads dropped back below 50 copies/mL in all but one individual (9 percent) while the same regimen continued. In most of these people resistance mutations emerged during the viremic spikes, but most of them had taken one or two NRTIs before HAART.
Havlir told the Journal it would be "cavalier" not to worry about a viral load that breached the 50-copy mark. As Thomas Merigan observed in a Q&A after Havlir's talk, no one can say whether a 70-copy viral load will fall back below 50 or build up enough steam to become a full-fledged rebound. But he added that "we have to have a more open mind about the switch point" at which one trades one regimen for another because "we may be using up drugs earlier than we need to."
In her analysis, Havlir said, viral loads that pushed above 200 copies/mL were "clearly on an ascending curve." And other workshop attendees suggested that--if the harbinger of a sustained rebound is not 60 or 70 copies/mL--it's not much higher than that. Right now, though, Havlir and Wensing are the ones with the data, and everything else is speculation. And how clinicians should navigate the waters between 50 and 200 copies is not immediately apparent. Havlir suggested that research should be able to pin down the clinical impact of blips in people starting their first regimen without much trouble. Deciphering blips will be dicier in people with more treatment experience.
Other workshop reports confirmed that HIV may go its merry way even when hounded by regimens more potent than the AZT/ 3TC/indinavir standby in Havlir's study. Hans-Jürgen Stellbrink (University Hospital Eppendorf, Hamburg) showed that HIV keeps multiplying in lymphoid tissue even while drugs keep viremia under 50 copies/mL [abstract 113]. His study involved 31 people taking d4T, 3TC, saquinavir, and nelfinavir, with or without interleukin 2 (IL-2), who had lymph node biopsies more than six months after viremia became undetectable.
In situ hybridization showed viral RNA in lymph nodes from 21 of these 31 individuals (68 percent). Whether they had taken IL-2 with their antiretrovirals appeared not to affect the chance of detecting RNA in lymph nodes. Ten of 16 (63 percent) taking IL-2 had lymph node RNA, compared with 11 of 15 (73 percent) not taking IL-2. People with detectable lymph node RNA generally had lower trough concentrations of saquinavir and nelfinavir, but this trend did not reach statistical significance. Histologic studies showed that both proliferating and resting lymph node cells kept spurting dribs and drabs of virus.
If three or four drugs can't clear virus after a few years of trying, how about five or six? Hanneke Schuitemaker (Academic Medical Center, Amsterdam) addressed that question in eight treatment-naive individuals and one treatment-experienced person who began regimens including up to three NRTIs, two PIs, and one NNRTI [abstract 115]. Unlike Stellbrink, her goal was not to see how often HIV hung on in lymph nodes, but how quickly industrial-strength antiretroviral combos scoured HIV from the resting T-cell reservoir.
Schuitemaker started plumbing that reservoir after plasma viremia sank below 5 copies/mL. She first attempted to isolate virus from resting cells after an average 66 weeks of treatment and continued to do so for an average additional 292 days. After two years of treatment, she counted as few as one productively infected cell among 10 million resting T cells. But Schuitemaker never failed to pluck virus from these patients' resting cells.
The mean half-life of the resting T-cell reservoir turned out to be 10.2 months in this study. That estimate comes close to the Aaron Diamond Center calculation of 6.2 months, figured in recently infected people taking three- or four-drug regimens.22 In contrast, Johns Hopkins researchers reckoned a much longer half-life of 43.9 months among chronically infected individuals.23 But because Schuitemaker's five or six drugs didn't shorten the average half-life compared with the Aaron Diamond's three or four drugs, she concluded that "intensification of therapy with currently available drugs may not necessarily result in an accelerated decay" of the latently infected T-cell reservoir.
Not all attendees seemed ready to swallow that conclusion. O'Brien pointed out, and Schuitemaker acknowledged, that nearly all study participants had to trim back their original six-drug regimens. If people had managed to stick with six drugs, he proposed, or if the future brings more potent smaller regimens, the resting reservoir may indeed evaporate more quickly.
Tight control of plasma viremia does clamp down on viral production by peripheral blood mononuclear cells (PBMCs), but virus persists inside those cells. Marek Fischer (University Hospital, Zürich) counted viral particles inside PBMCs and on the cells' surfaces after more than two years of potent antiretrovirals had lowered circulating virus to undetectable or vanishingly small levels in eight individuals [abstract 114]. He concluded that "HIV-infected cells prevailing during effective [therapy] may express low levels of HIV RNA, but they seem to scarcely produce viral particles." Notably, seven of the eight people in this analysis took only a protease inhibitor with 2 NRTIs. The eighth person took a PI, two NRTIs, and hydroxyurea.
If these three slide presentations are not enough to raise new questions about what constitutes "successful" therapy, two related posters should do the trick. A study of 35 people with well-suppressed viremia found that HIV continued to evolve in many of them. The evidence? Covert mutations that could confer resistance to drugs in their current regimens. Alejo Erice (University of Minnesota, Minneapolis) sequenced viral DNA from PBMCs of people who had maintained a viral load below 50 HIV RNA copies/mL for at least six months and whose median CD4+ count measured 524 cells/mm3 [abstract 146]. Twenty-nine people were taking a PI or an NNRTI with NRTIs, but a few were taking drugs from all three classes or only NRTIs.
Erice managed to sequence reverse transcriptase from 29 study participants and protease from 31. Of the 29 RT sequences, two harbored secondary RT mutations and 10 harbored primary and secondary mutations. Seven of the 10 primary mutations (at codons 41, 70, 75, 184, and 215) could have been elicited by components of the current regimens. Among the 31 protease sequences, 26 had secondary mutations and only one had secondary mutations plus a primary mutation, 30N, associated with resistance to nelfinavir.
Because Erice did not characterize the DNA species he analyzed, the mutants could represent archived or recently emerged virus. But the association of the RT mutations with NRTIs being taken at the time suggests ongoing viral replication. Erice said the clinical significance of his findings remains unclear. Just how unclear became obvious to anyone who strolled a few feet down the poster hall aisle to take in a report by Eoin Coakley (Beth Israel Deaconess, Boston) [abstract 144].
Coakley and collaborators have been tracking 17 people whose on-treatment viral loads have meandered below 1000 copies/mL for a median of 16 months (range 12 to 33 months). Seven are taking a PI regimen, five an NNRTI combo, and five only dual NRTIs. Maybe coincidentally, and maybe not, only four people are taking AZT. Most are taking 3TC (n = 14) and/or d4T (n = 12).
Now here's the really strange part. Viral sequencing in 16 of these 17 people turned up bountiful mutations apparently evoked by drugs in current or former regimens. These 16 folks had a median of two NRTI mutations (range 0 to 6) and 1 PI mutation (range 0 to 2). All five people taking an NNRTI had one major NNRTI-associated mutation. Four people had protease cleavage site mutations. Six-month follow-up of seven people showed new mutations in two of them, but viral loads still below 1000 copies/mL.
NNRTI mutations persisted for four and six months after two people stopped taking their NNRTI. Their viral loads changed modestly, and in opposite directions--from 247 to 61 copies/mL in one, and from 865 to 1005 copies/mL in the other. Peripheral neuropathy caused one person to stop taking ddI and 3TC, the only antiretrovirals he had ever tried. Two on-treatment NRTI mutations, 74V and 184V, persisted for 10 weeks into the withdrawal, and his viral load rose substantially from 174 to 7582 copies/mL.
What could all this mean? Go speculate. Coakley proffered no suggestions in his poster, but the diverse regimens and corresponding resistance patterns show that the phenomenon is not restricted to one or two drugs or one or two combinations. It could be that the resistant virus in these people is particularly unfit. Or perhaps some people recruited into this ongoing study can control replication without much help from drugs. The peak viral loads ever measured in 10 of the 17 never broke the 10,000-copy barrier, though five people had peaks above 30,000 copies/mL, and three of them had 100,000-plus readings. And the quick rebound in the man who stopped taking ddI and 3TC shows that those drugs were doing something.
Study of this Beth Israel Deaconess Ultrasensitive Sequencing Investigation (BIDUSI) cohort will continue. Coakley and coworkers have already started separate analyses of people who endured virologic failure after at least 12 months below 1000 copies/mL, and of those who keep their viral loads under 50 copies/mL for more than six months.
Two studies of different antiretrovirals turned up the same intriguing clue to when switching from a PI regimen to PI-sparing combos may work--and when it may not. Bernard Masquelier (CHU, Bordeaux) studied 38 people who had been taking a PI for more than six months and had a viral load below 500 copies/mL [abstract 119]. They all swapped the PI for nevirapine and kept the same NRTIs.
Twenty-two people had begun their PI as part of their first antiretroviral regimen. The other 16 had taken one or two NRTIs before starting a PI combination. After a median follow-up of 54 weeks, no one in the first group saw their viral load rebound above 500 copies/mL, whereas eight of the 16 who had taken one or two nukes before the PI regimen had a rebound after a median 20 weeks of follow-up. This significant difference in rebound rates (P <0.001) did not depend on how long the PI regimen had suppressed viremia before people switched to nevirapine.
Swiss and German clinicians collaborated to track down factors in virologic failure after participants in two trials traded in their PI for abacavir [abstract 120]. The simplified maintenance trial (SMT) randomized people to continue their PI regimen or switch to abacavir plus Combivir (AZT/3TC) after at least six months with a viral load below 400 copies/mL. Almost half of those enrolled in this study had begun antiretroviral therapy with a single NRTI or dual NRTIs. In contrast, nearly everyone in CNA30017 started treatment with two NRTIs plus a PI. After reaching a sub-50 viral load, they continued that PI treatment or kept the NRTIs and exchanged the PI for abacavir.
Once again early NRTI mono- or bitherapy proved crucial to the virologic response after switching from a PI. Milos Opravil (University Hospital, Zürich) reported that nine of 74 people in the SMT abacavir arm suffered a virologic rebound by 48 weeks, and six of those nine (67 percent) had taken AZT alone or with another NRTI before beginning a PI. It's harder to see much in the raw numbers of treatment failures in CNA30017, because so few people in that study had NRTI mono- or bitherapy and because none had endured a virologic failure before the switch. After 24 weeks of follow-up in this study, there were only five virologic failures in either arm, whereas most failures in the SMT study happened before 20 weeks, many before eight weeks.
One of three rebounds in CNA30017's abacavir simplification arm happened to someone who had AZT monotherapy before PI-based HAART. This one person with AZT experience had pre-switch AZT mutations in viral DNA of PBMCs, and those mutations popped up again in viral RNA after the rebound on abacavir. The only NRTI mutation that emerged during the other four rebounds in CNA30017 was 184V, the lightening-fast mutation linked to 3TC and abacavir.
Among rebounders in the SMT study, viral RNA analyzed after the rebound was loaded with AZT mutations in people who had taken AZT before PI-based therapy, regardless of whether they rebounded after switching to abacavir or while continuing their PI. This finding led Opravil to conclude that in the SMT trial "viral isolates with multiple NRTI-associated mutations at the time of virologic rebound in both [the PI] continuation arm and the simplified [abacavir] arm may be derived from archived isolates selected by prior mono and/or dual NRTI therapies." But screening PBMCs for mutants archived in viral DNA didn't always nose out those mutants, so Opravil couldn't recommend that strategy.
Masquelier and colleagues also detected abundant NRTI mutations among the eight rebounders who had taken one or two NRTIs before their PI combo. All rebounders also had one primary nevirapine-evoked mutation. So Masquelier suggested a conclusion similar to Opravil's: "Archived nucleoside resistance mutations selected before HAART are likely to facilitate the emergence of HIV-1" resistant to both NRTIs and NNRTIs when a "less potent" nevirapine regimen replaces a PI combination.
These two studies don't have the final word on why simplified PI-sparing regimens fail when they do. Lots of things can go wrong. But because both studies point to the same culprit--undercover NRTI mutations originally elicited by inadequate NRTI therapy--the studies suggest one factor to weigh when considering cashing in a protease drug for an RT inhibitor. And another study suggested that abandoning a PI may not always be necessary anyway, even in the face of detectable viremia.
This small study by Andrew Zolopa (Stanford University) and colleagues at three other centers added ritonavir to the regimens of 35 people taking standard-dose indinavir for more than three months [abstract 95]. The median viral load was about 1600 copies/mL, and the average CD4+ count 435 cells/mm3. Most people had been taking indinavir for a long time, for a median of 32.1 months.
Zolopa and colleagues cut the indinavir dose to 400 mg twice daily and introduced ritonavir at 200 mg twice daily for two days. They eased the ritonavir up to 300 mg twice daily on days three to five, and to the full 400 mg twice daily beginning on day six. No one had taken ritonavir before. All study participants kept the same NRTIs for at least three weeks after starting ritonavir.
Three weeks after beginning dual PI treatment, 21 of 35 people (60 percent) had reached a viral load below 50 copies/mL (n = 10) or had at least a half-log drop in viral load (n = 11). Two variables distinguished these responders from people whose viral loads didn't fall by at least a half-log: Three weeks of ritonavir boosted indinavir trough levels from 0.123 to 1.030 µg/mL in responders, compared with a smaller rise from 0.102 to 0.544 µg/mL in nonresponders (P= 0.04). And 15 of 18 responders (83 percent) had three or fewer protease mutations as baseline, compared with four of 12 nonresponders (33 percent, P= 0.009). The most common mutations in this cohort occurred at positions 10, 46, 54, 71, 82, and 90. Zolopa and colleagues concluded that people with low-level resistance to indinavir, characterized by three or fewer mutations, stand a good chance of responding to the familiar 400/400 mg twice-daily regimen of indinavir plus ritonavir.
Several workshop reports pointed to resistance as an implacable, though sometimes imperceptible, foe of structured treatment interruptions--STIs. Nonetheless, a single workshop study suggested that a carefully planned and executed series of STIs can prod the immune system toward spontaneous control of replication, at least for several months in a few people with already well-controlled chronic disease. Although this study didn't hit the STI jackpot--and although several attendees questioned its meaning--it did impress a few seasoned clinicians.
Felipe Garcia (University of Barcelona) gave d4T, 3TC, and indinavir or ritonavir to 10 treatment-naive people with CD4+ counts above 500 cells/mm3 and viral loads between 14,700 and 504,000 copies/mL [abstract 189]. After one year of treatment, all had viral loads below 20 copies/mL for at least eight months. Then Garcia began a series of three STIs. The first two lasted four weeks (weeks 0 to 4 and 28 to 32). The last STI, begun at week 52, could continue until the viral load climbed back over 10,000 copies/mL. One person stopped coming to appointments without explaining why, and another had to quit at the second STI when she became pregnant.
Viral load rebounded from under 20 copies/mL in ever case, at every STI. But the speed of the rebound--measured as viral doubling time--slowed from 2.23 days at the first STI to 3.38 days and 3.25 days at the second and third STIs (P= 0.05 for the difference between the first and third doubling times).
After viral load clambered nearly back to baseline during the second STI, it retreated spontaneously in four of nine persons to about 120 copies/mL in one person, to 20 copies/mL in another, and to about 6 copies/mL in the other two. Broad HIV-specific CTL responses could be detected in all four individuals, and all four had strong CD4+ cell responses against HIV antigens. Another person also had broad HIV-specific responses but failed to control replication without antiretrovirals. Nobody had measurable HIV-specific responses before the STIs. After the third treatment interruption, the four people who enjoyed spontaneous control of viremia during the second STI all maintained viral loads well below 10,000 copies/mL for 32 weeks without restarting their antiretrovirals. These new viral set points ranged from about 4 copies/mL to about 32 copies/mL.
The biggest question about this study--and one that can't be answered because it had no control arm--is whether these four responders had in fact responded to the STIs, or whether they would have controlled replication without HAART's help anyway. Garcia felt confident that everyone in the cohort truly had established chronic infection, because he had measured their viral loads on two occasions separated by at least three months. So it seemed unlikely to him that he had enrolled some people who had measurable viral loads that would have dropped without drugs.
Such unassisted slumps in viremia have been noted, most recently by Vancouver's Richard Harrigan, who reminded attendees of his findings after Garcia's talk. Harrigan tracked down records of 238 people who stopped their antiretrovirals, for any reason, for 90 days or more.24 Thirteen of them (5.5 percent) had viral loads below 500 copies/mL for at least 90 days without antiretrovirals. That 5.5 percent "success" rate doesn't come close to Garcia's 50 percent, but it shows that only a randomized trial will eliminate the possibility that such happenstance contributes to an apparently good result.
Hamburg's Stellbrink noted that Garcia's trial and other STI studies "heavily select" participants who stand a good chance of doing well. People in Garcia's group had never sampled an antiretroviral and had CD4+ counts around 600 to 700 cells/mm3--not far from what might be considered normal. And everyone had a first-rate and sustained response to unspectacular triple-drug therapy. Ignoring these silver-spoon baselines when considering the results, Stellbrink warned, could easily open such results to misinterpretation.
Richard Haubrich pressed that point in calling for an end to STI anecdotes in favor of well-planned randomized trials. Such trials are under way--including a big Swiss study and a quickly enrolling US study that will shoot for clinical endpoints. The study has no CD4+ entry requirements. But until results of such studies appear, Haubrich worried that HIV "boutique clinics" will offer STIs to all comers, sometimes with dire results.
The siren song of STIs also rings in the ears of clinicians trying to figure how to care for people with multiclass resistance. A little over a year ago, Veronica Miller (JW Goethe University, Frankfurt) sparked interest in this strategy when she found that resistance mutations faded from virus of some people who stopped all treatment before starting heavy-duty salvage.25 Those so-called wild-type reverters did better virologically once they began salvage than people whose mutations persisted during drug holidays.
But a few months before the Resistance Workshop, Miller reported that viral loads rebounded in 24 of 33 reverters (73 percent) within 78 days of restarting therapy. Apparently, as many had suspected, resistant virus lurked at undetectable levels in the reverters and promptly resurfaced when it sniffed out its favorite antiretrovirals in the salvage regimen.26 But because viral load rebounded in all the holiday takers whose virus didn't revert to wild-type, Miller argued that treatment interruptions may make sense for some heavily experienced people before a mega-HAART rescue regimen. Four studies at the Resistance Workshop addressed that question from different angles, and all turned up evidence against STIs for treatment-experienced people.
Vincent Calvez (Pitié-Salpêtrière Hospital, Paris) ran a small open-label study that tried to duplicate some of the conditions in Miller's original retrospective analysis [abstract 141]. He recruited 20 people with virus resistant to all three antiretroviral classes, a median viral load of 160,932 copies/mL, and a median CD4+ count of 77 cells/mm3. They had worked their way through a median nine antiretroviral regimens during 73 months of therapy.
Everyone stopped all their antiretrovirals for a median of eight weeks (range four to 24 weeks). Genotyping viral isolates before they started five- to nine-drug megaregimens, Calvez found that 11 of the 20 (55 percent) had virus that now had no mutations to one (n = 5), two (n = 2), or three (n = 4) drug classes.
Six months after mega-HAART began, however, these reverters enjoyed no virologic or immunologic edge over the nonreverters. Indeed, the nonreverters added an average 30 CD4+ cells/mm3 to their pre-STI baseline, while CD4+ counts of reverters languished at that baseline. But the CD4+ difference between the groups was not significant. Viral load fell an average of 2 logs in both groups and stayed there through six months of follow-up.
Calvez confirmed Miller's finding that STIs during advanced disease are not risk free. Miller had recorded stomach-churning CD4+ nose-dives averaging 120 cells/mm3 among wild-type reverters.25 And she charted 17 new AIDS diagnoses among 15 people during treatment breaks.26 CD4+ counts did not fall nearly as steeply in Calvez's cohort, though with a median baseline count of 77 cells/mm3, most of these people didn't have 120 cells to lose. T cells dropped no more among Calvez's reverters than among nonreverters, but five people did shed more than 50 CD4+ cells/mm3 when they stopped their drugs. Calvez counted four new AIDS events, including cytomegalovirus retinitis, and one aggravation of Kaposi's sarcoma during the STIs.
Although the Calvez study further erodes confidence in STIs before megadrug salvage, it did differ from Miller's original observation in one important way. The reverters in Miller's cohort lost mutations to all the antiretrovirals they were taking, whereas only three of 20 people in the Calvez study did so. Calvez reported that longer STIs led to more wild-type reversion, but STIs can last only so long in people with sub-100 T-cell counts who start getting new opportunistic infections.
What happens when people with some treatment experience--but not the heavy experience in the Calvez and Miller studies--undergo sequential STIs? Lidia Ruiz showed that mutations selected more than two years earlier can rapidly reappear during treatment interruptions [abstract 142]. This study focused on five people who maintained a viral load below 50 copies/mL for more than two years on a potent regimen. Viral DNA in their PBMCs--a suspected stronghold of drug-resistant mutants--showed no evidence of mutations before the STIs began. But NRTI-linked mutations popped up in virus of three of these five individuals during the STIs.
In one person who had taken AZT monotherapy 24 months before beginning HAART, AZT mutations at positions 67, 215, and 219 reemerged during viral rebounds that happened during all three STIs. The 184V mutation resurfaced in two other people who had taken 3TC with AZT before starting HAART. In one of these people the mutation popped up during the second and third STI rebounds, and in the other it reappeared during all three STI rebounds.
These three individuals always recontrolled viral replication when they restarted treatment after the STIs, even though their regimens included 3TC. And Ruiz noted that the two people with recrudescent 184V mutants had both taken IL-2 before their STIs. The IL-2, she suggested, may have awakened resting cells that sheltered mutant virus. Although these long-covert mutants did not hamper resuppression of viremia after the STIs, that might not be true for people with more treatment experience--and more mutations--who decide to try a few STIs.
Allan Hance (Xavier Bichat Hospital, Paris) used a supersensitive assay to confirm that apparently wild-type populations may include more than a few primary protease mutations after a three-month STI [abstract 117]. Hance tuned his real-time PCR-based assay to amplify and tote up two mutations shared by several PIs, 82A and 90M. Then he homed in on virus isolated from 13 people who had stopped treatment for two months or more. Garden variety genotyping said that virus was wild-type; Hance's assay found mutations in nine of 13, one of whom had not touched an antiretroviral for five months but still had the 82A mutation in 3 percent of the virus sampled.
Hance concluded that STIs will have to last longer than three months if one expects PI-resistant mutants to vanish.27 And in at least a few people, he showed, mutants will hang around even longer than that, despite withdrawal of all drug pressure. Analysis of secondary mutations that accompanied 82A or 90M in these viral isolates led Hance to propose that mutant virus spotted during STIs could have two origins. It may simply represent the emergence of genomes generated during earlier treatment. Or it may represent new viral species resulting from selective reversion and recombination of mutant populations.
Summing up these studies in a workshop overview, Mellors suggested "we were kidding ourselves" in thinking that STIs would lead to a meaningful reversion of drug-resistant virus. He labeled the opposite finding a key message of this meeting. No one disagreed.
Only eight years have slipped by since the first Resistance Workshop, held in The Netherlands the week before the Amsterdam international conference. Organized by Charles Boucher and Brendan Larder, the meeting drew several dozen experts to mull 40 of so presentations.
This year, workshop planners rejected more presentations--53--than attendees pondered at that inaugural meeting. This review touches on 59 slide or poster reports, only 30 percent of the meeting's total of 197. And readers can rest assured that those 138 reports not mentioned here were not fluff.
The study of resistance to antiretroviral drugs, always a daunting and humbling enterprise, is now more mazelike than ever. Sheer numbers suggest one needs more than a ball of string, or a genotype, to puzzle through this labyrinth. When Raymond Schinazi (Emory University, Atlanta) first compiled a comprehensive antiretroviral resistance table with Larder and Mellors in 1994, it listed 42 mutations.28 The 1997 update added 102 more. This year's edition counts 236 mutations involving 116 compounds. The number of multidrug resistance mutations soared from nine in 1997 to 50 this year.
But numbers tell only a tiny part of the story. Who would have suspected, when Larder and Richman described AZT resistance a decade ago, that the same mutations conferring resistance to that drug could crop up in AZT-naive people taking a later NRTI, d4T, and make that virus less susceptible to d4T? Organizers of this workshop devoted an entire poster subsection to that topic,29 which now has its own acronym, TAMs, for thymidine analog mutations.
Who would have guessed that some virus resistant to NRTIs would be hypersensitive to NNRTIs, or that increasing concentrations of NNRTIs could enhance the replication of certain viral strains? Who would have predicted that resistance mutations can emerge when drugs have a seeming stranglehold on the virus and keep viral loads under 50 copies/mL? Who would have anticipated the scramble to concoct drugs that throttle virus resistant to other drugs?
At least one resistance maven, Stanford's Thomas Merigan, did predict the clinical use of resistance testing and began talking and writing about it back when most clinicians accepted nucleoside monotherapy as the standard of care.30 Even he may have been surprised, though, by the speedy refinement of genotypic and phenotypic resistance assays, and their swift adoption by HIV clinicians.11 But any clinician who thinks using those tests will be a breeze must go to the blackboard immediately and write out, 100 times, "Clinically relevant phenotypic cutoff values for individual drugs remain unknown." And anyone who assumes that "expert opinion" will solve many an assay riddle may have thought twice if sitting at the Sitges workshop, where a roomful of experts routinely produced a roomful of opinion.
But no one should conclude that the resistance labyrinth will always be insoluble. And it's a safe bet that no clinician who's lasted this long in the epidemic is much fazed by complexity. HIV docs, indeed, the human species in general, appear to relish complexity more than revile it. What could have been more clear when two busloads of resistance researchers, given a half-day reprieve from the workshop, pulled up in front of Antonio Gaudí's still-unfinished masterwork, the Expiatory Temple of the Holy Family, or, Sagrada Familia?
Gaudí died in 1926, after 40 years of work on the Barcelona church had yielded a plan, a crypt, and one facade (Figure 2). In the intervening 74 years, other parts of this staggering monument have climbed skyward. But Gaudí's blueprints got burned during the Spanish Civil War, so succeeding architects have had a freer hand in imposing their own notions on Gaudí's framework.
 |
| On the workshop's second day, resistance experts tore themselves away from HIV's complications only to ponder the complexities of Antonio Gaudí's cathedral facade in nearby Barcelona. But careful study reveals the plot in Gaudí's cacophonous nativity facade. And a decade of resistance research has begun to trace the even more sinuous story of drug failure. |
The "nativity facade" Gaudí completed seems crafted in wax that has melted a tad under the brazen Catalan sun. The Holy Family huddles under molten pediments. Minor cast members crouch in stalactitic niches. Brash seraphim threaten to undo the drippy edifice with blared hosannas. Critics struggling to describe the effect rely on adjectives almost as vivid as the facade itself--sensuous, curving, furious, reptilian. Yet, somehow, this astonishing jumble coheres.
A few years back, contemporary architects and artisans completed another facade, depicting scenes from Christ's passion, in an utterly unsimilar style. It is fiercely cleancut, foursquare, noble, and more than a little boring. The verdict is already in. Beneath the "passion facade," a few spent and indifferent visitors lolled on the big granite blocks leading up to the pillar where Jesus is being flogged. Beneath Gaudí's bamboozling "nativity facade," virologists and other tourists clustered gape-mouthed to stare at the sensuous, curving, furious, reptilian, and nearly incomprehensible stone storybook.
Nearly incomprehensible. The nativity tale can be solved, and people clearly love sitting there solving it. Resistance, for all its gaudy complexity, can be solved too. It took nearly a decade, but hard-nosed research finally solved the mechanism of AZT resistance. Only a year later, the chase has turned to that even greater enigma, d4T resistance, and some headway was logged at this workshop. Such mechanistic solutions do not lead immediately to better drugs, though in time they could. And other work has yielded drugs that probably will better current agents--in tolerability, ease of administration, and ability to attain levels that swamp some resistant virus.
Gaudí started Sagrada Familia in 1883. The rosiest current projections see completion around 2050. The church consecrated 50 years from now will look only somewhat like Gaudí's dream. But tourists will still love the messy nativity facade most. And midcentury virologists, relieved somewhat by the development of effective HIV vaccines, will sweat over new puzzles, maybe applying resistance lessons learned long ago.
Resistance-specific sites on the Web grew in quantity and quality over the past year. Clinicians can now look up specific mutations, read timely reviews by leaders in the field, and even submit questions to top guns of the resistance world. Of course resistance assay makers also hawk their wares--and other useful information besides--on the Web. Here are some sites worth a browse to see if they suit your needs.
Antiviral Drug Resistance Online
(http://www.viral-resistance.com/). This is the Mother of All Resistance Tables, researched and updated by Raymond Schinazi, John Mellors, Brendan Larder, Jennifer Hammond, Amy Juodawlkis, and Tamara Barnett. Organized by drug name, the tables list the RT or protease sites involved, amino acid changes, whether resistance has been reported in vitro and in vivo, fold-resistance, and cross-resistance profiles. But you'll have to remember that 141W94 became amprenavir, and the last time we checked the online table didn't include lopinavir (ABT-378), for example, or BMS-232632. It's only a matter of time, though, because resistance findings on those drugs were published in the May 2000 International Antiviral News, in a table by these same authors.
HIVresistance.com
(http://www.hivresistance.com/). This new site features a star-studded editorial board, and they're not just window dressing. Resistance luminaries like Mark Wainberg, Charles Boucher, and Jonathan Schapiro contribute regular articles and field thoughtful questions from users. The site includes mutation charts, a glossary, and "interactive, at-a-glance compendiums of drug- and mutation-specific knowledge."
Antiretroviral Resistance Mutations
(http://hivinsite.ucsf.edu/medical/resistance/2098.474c.html). Multiview lists of primary and secondary resistance mutations easily let you compare mutations from one drug in a class to another. And you can download snappy, printable pdf files of these tables, which are prepared by the International AIDS Society-USA Panel, promulgator of the recent clinical resistance testing guidelines.2
Los Alamos HIV Drug Resistance Database
(http://204.121.6.64:581/resist_db/default.htm).This searchable mutation database lets users choose between "simple" and "advanced" search tools and shows the 3-D structures of HIV reverse transcriptase and protease. You can track down all protease inhibitors, for example, that call forth mutations at codon 84. Then you can link to the MEDLINE records of articles reporting that mutation for different drugs.
Stanford HIV RT and Protease Sequence Database
(http://hivdb.stanford.edu/). Created and maintained by Stanford's Robert Shafer, this site is more than a sophisticated tool for researchers (described at the Resistance Workshop in abstract 82). Clinicians, and anyone interested in resistance to antiretrovirals, will find the "Notes" sections on NRTIs, NNRTIs, and PIs an easy-to-understand review of key concepts in this field, such as AZT and d4T cross-resistance, NNRTI mutation interactions, and differences between "primary," "secondary," and "accessory" PI mutations. See http://hivdb.stanford.edu/hiv/notes.pl.
1. Eliot G. Middlemarch. New York: Bantam Books. 1992:12.
2. Hirsch MS, Brun-Vézinet F, D'Aquila RT, et al. Antiretroviral drug resistance testing in adult HIV-1 infection: Recommendations of an International AIDS Society-USA Panel. JAMA 2000 May 10;283(18):2417-26.
3. Flexner C. HIV genotype and phenotype--arresting resistance? JAMA 2000 May 10;283(18):2442-4. Comment on: JAMA 2000 May 10;283(18):2417-26.
4. We're speaking metaphorically here, of course. A study of lopinavir's effects on triglycerides and cholesterol will be presented at the International Conference in Durban.
5. Benson C, Brun S, Xu U, et al. ABT-378/ritonavir (ABT-378/r) in protease inhibitor-experienced HIV-infected patients: preliminary 24 week results. Antiviral Ther 1999;4(suppl 1):6.
6. Kempf D, Xu Y, Brun S, et al. Baseline genotype and phenotype do not predict response to ABT-378/ritonavir in PI-experienced patients at 24 and 48 weeks. Presented at: 7th Conference on Retroviruses and Opportunistic Infections; January 30-February 2, 2000; San Francisco. Abstract 731.
7. Larder B, Bloor S, Hertogs K, et al. Tipranavir is active against a large selection of highly protease inhibitor-resistant HIV-1 clinical samples. Antiviral Ther 1999;4(suppl 1):5.
8. Bazmi HZ, Hammond JL, Cavalcanti SC, et al. In vitro selection of mutations in the human immunodeficiency virus type 1 reverse transcriptase that decrease susceptibility to (-)-ß-D-dioxolane-guanosine and suppress resistance to 3'-azido-3'-deoxythymidine. Antimicrob Agents Chemother 2000 Jul;44(7):1783-8. The 65R and 74V mutations are spotted, but only occasionally, in people taking ddC or ddI.
9. Meyer PR, Matsuura S, Mian M, et al. Elevated nucleotide-dependent primer unblocking by zidovudine-resistant HIV-1 reverse transcriptase leads to increased removal of zidovudine-MP from blocked primers. Antiviral Ther 1999;4(suppl 1):17-18.
10. Götte M, Arion D, Parniak MA, Wainberg MA. Mechanisms of HIV-1 resistance to zidovudine and lamivudine. Antiviral Ther 1999;4(suppl 1):18.
11. Boucher CAB. The value of resistance assays when changing therapy. Presented at: 39th Interscience Conference on Antimicrobial Agents and Chemotherapy; September 26-29, 1999; San Francisco. Abstract 1368.
12. Two genotyping studies showed better virologic results with genotyping than with standard of care drug switching: Durant J, Clevenbergh P, Halfon S, et al. Drug-resistance genotyping in HIV-1 therapy: the VIRADAPT randomised controlled trial. Lancet 1999 Jun 26;353(9171):2195-9; Baxter JL, Mayers DL, Wentworth DN, et al. A pilot study of the short-term effects of antiretroviral management based on plasma genotypic antiretroviral resistance testing (GART) in patients failing antiretroviral therapy. Presented at: 6th Conference on Retroviruses and Opportunistic Infections; January 30-February 2, 1999. Abstract LB8. Very early results of a third genotyping study, Havana, confirm the VIRADAPT and GART findings [abstract 59]. One phenotyping study presented at the Resistance Workshop showed a virologic benefit for this test [abstract 84], whereas another study of phenotyping and genotyping had mixed results [abstract 85]. Another phenotyping study presented earlier this year13 found no value in that assay for heavily experienced individuals. See "Two yeses and one no" in the text for discussion of these latter four studies.
13. Melnick D, Rosenthal J, Cameron M, et al. Impact of phenotypic antiretroviral drug resistance testing on the response to salvage antiretroviral therapy (ART) in heavily experienced patients. Presented at: 7th Conference on Retroviruses and Opportunistic Infections; January 30-February 2, 2000; San Francisco. Abstract 786.
14. VIRADAPT,12 GART,12 HAVANA [abstract 59], and NARVAL [abstract 85].
15. VIRA3001 [abstract 84].
16. At this meeting alone, four studies suggested five new mutations or polymorphisms involving NNRTIs [abstracts 30b and 34], PIs [abstract 60], or multiple antiretrovirals [abstract 51].
17. DeGruttola V, Dix L, D'Aquilla R, et al. The relation between baseline HIV drug resistance and response to antiretroviral therapy: re-analysis of retrospective and prospective studies using a standardized data analysis plan. Antiviral Ther 2000;5:41-48.
18. Lanier ER, Melby T, St. Clair MH, et al. Potential clinical impact of small differences between Virco Antivirogram and ViroLogic PhenoSense assays for abacavir in 3TC-experienced patients. Presented at: 7th Conference on Retroviruses and Opportunistic Infections; January 30-February 2, 2000; San Francisco. Abstract 788.
19. Whitcomb J, Deeks S, Huang W, et al. Reduced susceptibility to NRTI is associated with NNRTI hypersensitivity in virus from HIV-1-infected patients. Presented at: 7th Conference on Retroviruses and Opportunistic Infections; January 30-February 2, 2000; San Francisco. Abstract 234.
20. The seven cities are Birmingham, Dallas, Los Angeles, Montreal, San Diego, Seattle, and Vancouver.
21. Ramratnam B, Mittler JE, Zhang L, et al. The decay of the latent reservoir of replication-competent HIV-1 is inversely correlated with the extent of residual viral replication during prolonged anti-retroviral therapy. Nat Med 2000 Jan;6(1):82-5.
22. Zhang L, Ramratnam B, Tenner-Racz K, et al. Quantifying residual HIV-1 replication in patients receiving combination antiretroviral therapy. N Engl J Med 1999 May 27;340(21):1605-13.
23. Finzi D, Blankson J, Siliciano JD, et al. Latent infection of CD4+ T cells provides a mechanism for life-long persistence of HIV-1, even in patients on effective combination therapy. Nat Med 1999 May;5(5):512-7.
24. Harrigan PR, Whaley W, Dong W, et al. No detectable HIV RNA in thirteen individuals months after stopping antiretroviral therapy. Presented at: 7th Conference on Retroviruses and Opportunistic Infections; January 30-February 2, 2000; San Francisco. Abstract 351.
25. Miller V, Rottmann C, Hertogs K, et al. Mega-HAART, resistance and drug holidays. Antiviral Ther 1999;4(suppl 1):27-28.
26. Miller V, Sabin C, Hertogs K, et al. Antiretroviral treatment interruptions in patients with treatment failure: analyses from the Frankfurt HIV cohort. Antiviral Ther 2000;5(suppl 2):22.
27. A retrospective analysis of 24 heavily experienced patients who stopped treatment because of multidrug resistance confirmed Allan Hance's finding that a two- to five-month STI can be too brief to allow reversion to a wild-type population [abstract 143]. Corinne Amiel, MD (Rothschild Hospital, Paris) counted only six reverters in this group, and two of those six still had protease mutations. Two people with apparently complete reversion to wild-type restarted treatment, and their preholiday resistance pattern emerged in one to two months.
28. Schinazi RF, Larder BA, Mellors JW. Mutations in retroviral genes associated with drug resistance: 2000-2001 update. Int Antiviral News 2000;8:65-91. Available at: http://www.viral-resistance.com/.
29. Abstracts 38 through 45.
30. Merigan TC Jr. Refining markers of disease progression to improve therapy. HIV Adv Res Ther 1993;3(2):11-14. "Decisions on when to change a therapeutic plan must ultimately be based on in vitro resistance patterns of individual patients' viruses," Merigan wrote, "just as such decisions involving antibiotic strategies are now modified for patients with bacterial infections."
*Mark Mascolini writes about HIV infection (mailmark@ptd.net).
000810
IA000801
Copyright © 2000 - International Association of Physicians in AIDS Care. All rights reserved. http://www.iapac.org
ÆGiS is made possible through unrestricted grants from Roxane Laboratories, Inc., iMetrikus, Inc., the National Library of Medicine, and donations from users like you. Always watch for outdated information. This article first appeared in 2000. This material is designed to support, not replace, the relationship that exists between you and your doctor.
ÆGiS presents published material, reprinted with permission and neither endorses nor opposes any material. All information contained on this website, including information relating to health conditions, products, and treatments, is for informational purposes only. It is often presented in summary or aggregate form. It is not meant to be a substitute for the advice provided by your own physician or other medical professionals. Always discuss treatment options with a doctor who specializes in treating HIV.
Copyright ©1990, 2000. ÆGiS & the Sisters of Saint Elizabeth of Hungary. All materials appearing on ÆGiS are protected by copyright as a collective work or compilation under U.S. copyright and other laws and are the property of ÆGIS and the Sisters of Saint. Elizabeth of Hungary, or the party credited as the provider of the content.