The orthographic plum was a sensible shift in the alphabetic ordering of ICAAC abstracts. For years HIV abstracts got grouped under I. This year the HIV crowd suffered a small jolt when flipping to the abstract book's familiar I section and finding -- nothing. What? No HIV abstracts this year? Hardly. HIV jumped one rung up the abecedarian ladder to H, replacing "Virology (other than HIV)," which tumbled to V.
But that wasn't the only HIV difference between ICAAC 42 and most every other ICAAC for the past decade. HIV attendees endured another little shock when they scanned the program for the opening night session, long devoted to the notorious retrovirus. Not this year. Instead the keynote session considered vaccines, with one HIV talk by Lawrence Corey, a CDC foray into "the impact of global immunization programs," and -- perhaps inevitably -- a talk on "vaccines for agents of bioterrorism."
Well, at least Tony Fauci is back with plenary lecture to update his take on HIV pathogenesis. What's this? Fauci is talking about what? Right. Bioterrorism. Perhaps the HIV set should be happy that the keynote session didn't focus entirely on treacherous plagues like West Nile virus or -- an epidemic still limited to the U.S. West, apparently -- mad deer disease.
But HIV hasn't gone away. Globally, no one needs convincing, though patterns may be mutating if we can believe the U.S. Central Intelligence Agency. The agency's National Intelligence Council predicts a watershed shift in the epidemic's course.1 The devastation of southern and eastern Africa will continue, according to CIA seers, but eight years from now most infected people will live in China, Ethiopia, India, Nigeria, and Russia. Things aren't so great in the United States either, where -- year after dreary year -- another 40,000 people open their lymph nodes to HIV, and reports of risky escapades by already at-risk people appear in the paper weekly. But teaching abstinence without promoting needle exchange will make things better, we hear.
It's easy to get cranky about this state of affairs. So far bioterrorism has killed fewer people in the U.S. -- five -- than Michael Gottlieb (four), Henry Masur (11), and Frederick Siegal (four) reported in their seminal AIDS case series of 1981.2-4 West Nile virus has sent 198 Americans to their graves since 1999,5 compared with 233 CDC-reported AIDS deaths in the epidemic's first three years here, 1979 through 1981.6 (And the counting wasn't so good back then, since no one knew the disease existed.)
Despite such gripes, ICAAC has served HIV well, and HIV researchers from near and far observe the yearly haj to showcase their studies there. ICAAC's 2002 edition featured 152 reports of new HIV research, not counting studies summarized in five symposia and three caffeine-fueled Meet-the-Experts exegeses. And there's plenty of room on ICAAC's sumptuous program for topics great and small. But any hints of drift from a stern focus on the retrovirus rouse concern.
The reasons for that are simple. One, HIV continues its largely unimpeded global stampede. And two, for all the superb science HIV has inspired, it still guards key secrets as closely as thick steel once secured lock boxes at the San Diego Trust and Savings Bank (at top). When newly wealthy San Diegans lined up to lock down necklaces and notes of credit in William Templeton Johnson's nouveau Renaissance cathedral of finance, they could not suspect those notes might be worthless a year later, when the 1929 crash ignited the Great Depression. The secrets HIV guards seem more likely to retain their value, and to resist turnkey attempts to retrieve them from their safe-deposit security.
Who would have thought that Jay Levy's CD8 antiviral factor (CAF), first described in Science7 only five years after Gottlieb et al. described AIDS, would elude discovery -- and even agreement about its existence8 -- for at least 16 years? Who would have imagined that in 2002 a good HIV vaccine might lie as far in the future as Gottlieb's report lies in the past?
Yet, as ICAAC showed, neither ugly possibility can be erased with confidence.
Who would have predicted, once puissant regimens rescued people from the grave's lip, that stopping successful therapy would be considered a reasonable -- even a wise -- option? Who could have foreseen, when a single protease inhibitor (PI) seemed destined to extract HIV from every T cell that mattered, that so much work would explore combining two of them, or three of them? Who would have guessed, once drug labs started spewing potent antiretrovirals, that bright clinicians who treat a deadly infectious disease would still have to debate when to start?
Yet in the Year of the Plague 2002, as ICAACers learned, those issues remain very much in play. Start with the starting debate, ably argued by Diane Havlir (University of California, San Francisco) at an ICAAC interactive session:
ICAAC has spotlighted some skilled educators at these sessions, none more skilled than Diane Havlir, who bravely dusted off that shopworn snarler -- when to start antiretroviral therapy -- and showed that it remains a vibrant and critical question. She zeroed in on the group's zeitgeist by posing a simple case and modifying it twice.
Vote, then see how your colleagues answered in Table 1.
If you're like seven of 10 people in the ICAAC audience, you don't treat a newly diagnosed 47-year-old woman whose CD4 count and viral load lie outside the GO box drawn by most current guidelines. Nearly the same proportion starts scribbling scrips -- usually for three drugs -- if that woman's viral load tops 100,000 copies/mL. And, even if her viral load is low (26,000 copies/mL), but her CD4 count precarious (205 cells/mm3), almost everyone starts treating.
Maybe the most surprising result of this survey is that nearly one third of respondents do start HAART when the T-cell tally still stands above 350 cells/mm3 and the RNA assay measures a mere 30,000 copies/mL. Maybe the woman's age, 47 years, has something to do with it. Or maybe this quick-trigger 31 percent doesn't buy the advice that discretion is the better part of antiviral prescribing.
Showing a cartoon depicting a woman pushing the when-to-start pendulum back toward earlier intervention, Havlir explains that the woman is she.
Diane Havlir's chief concern lies with the endpoints tallied in big cohort studies that find little advantage to starting treatment above 350 cells/mm3 versus starting between 200 and 350 cells/mm3. The much-sifted data from Robert Hogg,9 Timothy Sterling,10 and Matthias Egger11 all reckon descents to AIDS or death -- the endpoints -- then fix their compass on the same foreboding polestar -- 200 cells/mm3. On the one hand, using AIDS and death as endpoints makes sense because those are the two "outcomes" everyone hopes most to outlast. But Havlir argues that a blinkered focus on these two dire denouements overlooks other benefits earlier treatment may bring:
At the same time, as authors of cohort studies themselves observe, the typical two- or three-year follow-up will miss a longer-term advantage of earlier treatment, if such an advantage were to emerge. But perhaps history is the factor that most dilutes the reliability of the when-to-start cohort studies presented so far. They involve cohorts first treated from 1996 to 1999,9 1996 to 2001,10 and at a median date of December 1997 (interquartile range June 1997 to July 1998).11 Although Sterling10 tracked people into 2001, the median follow-up of 24 months means most of them started HAART in 1999 or earlier. So a large majority in all three studies began treatment with a musty 20th century regimen, not the more potent therapies prescribed up front today. As Havlir observed, 21st century first-line combos are also typically simpler and often more tolerable than 1998 standbys like stavudine (d4T), 3TC, and solo indinavir.
Havlir is not alone in making this point. In a long interview on antiretroviral strategies, veteran HIV clinician Howard Grossman observed that potent, simple, and safe regimens -- 3TC, abacavir, and tenofovir, for example -- should encourage a closer look at hit-hard-but-later fiat.12
At ICAAC Michael Saag (University of Alabama, Birmingham) accented the increasing potency of newer combinations in a slide comparing this year's standards of care with those of the late 1990s [abstract LB1]. The elder antiretroviral mixes came from a 2001 meta-analysis of 23 clinical trials completed in the second half of the last decade.13 This year's crop came from Saag's study of didanosine (ddI), emtricitabine (FTC), and efavirenz [abstract LB1] and a trial of d4T versus tenofovir plus 3TC and efavirenz [abstract LB2]. The simpler 2002 triple therapies won in a walk (Table 2).
Havlir's final pendulum push borrowed some muscle from two studies showing that HIV DNA -- a measure of virus at its ease in resting T cells -- offers another way to predict disease progression.14,15 In 130 people not treated with combination antiretrovirals, every 10-fold higher HIV DNA level in peripheral blood mononuclear cells raised the risk of AIDS 2.62 times and death 1.84 times, independently of age at seroconversion, baseline CD4 count, viral load, and T-cell-receptor excision circles.14 In 292 people evaluated before the HAART era, HIV DNA predicted disease progression and death independently of RNA load and CD4 count, and more robustly than either of those more familiar sibyls.15 More and perhaps better tools to predict progression, Havlir proposed, could justify earlier therapy in people who need it.
One oft-cited argument that Havlir mentioned only tangentially is the possibility that delayed therapy may allow ongoing immune wrack and ruin that treatment will not reverse. An enlightening study by Timothy Schacker and Ashley Haase, published after ICAAC, shows that one facet of immune decay can appear in the earliest days of infection.16 By this measure, treatment started at 355 instead of 205 cells/mm3 would make little difference. But the study bolsters Havlir's point that sturdy progression predictors remain to be discovered.
Scrutinizing lymph node biopsies from 11 people before they started potent antiretrovirals, Schacker found that baseline lymph tissue fibrosis predicted lymphoid CD4 levels and response to therapy. With collagen as a marker of fibrosis, he showed that the more collagen a person had before treatment, the fewer CD4 cells that person had in the paracortical T-cell zone, where 98 percent of T cells lie (r2 = 0.72, P < 0.0001). The amount of collagen did not correlate with pretreatment peripheral CD4 count, pretreatment viral load, or duration or stage of HIV infection. One person had acute HIV infection, three had been infected six months or fewer, six had CD4 counts over 200 cells/mm3 and no AIDS diagnoses, and one had AIDS. Lower pretreatment collagen also predicted bigger peripheral CD4 gains after six months of therapy in seven people who started antiretrovirals (r2 = 0.91, P < 0.0001).
Schacker and colleagues believe this lymph tissue fibrosis with HIV infection "is analogous to the pathogenesis of cirrhosis in chronic active hepatitis B and C infection, wherein ongoing viral replication in hepatocytes leads to a state of chronic inflammation and fibrosis." They suggest that a pretreatment lymph node biopsy could help clinicians stage HIV disease and "provide useful information on when to initiate therapy." The researchers also think their findings suggest the need to study adjunctive anti-inflammatory therapy as a way to limit, or even reverse, damage to the paracortical T-cell zone.
The study serves as a reminder that research can pry loose HIV's secrets, but that many others remain. It also hints that the standard of care in 2006 will differ more from today's standard than today's differs from 1998's.
The early group started with an average CD4 count of 506 cells/mm3, compared with 275 cells/mm3 in the later-starting group (P < 0.001). No one had an AIDS diagnosis before treatment. Although early and later starters did not differ in age (34 versus 35 years), viral load (4.8 versus 4.7 logs), or relative use of PIs and nonnucleosides (NNRTIs), the later group had three nearly significant and possibly related disadvantages:
After 12 weeks of antiretroviral therapy, 89 percent in both groups had at least a 1-log drop in circulating virus or fewer than 400 RNA copies/mL. After one year, 86 percent in the early group and 88 percent in the later group had a sub-400 viral load. The "clinical event" rate was higher in the delayed group, but not significantly so, and that group had a trend to more frequent immune reconstitution disease than the early group (50 versus 16.7 percent, P = 0.171).
The critical outcome difference proved to be the rate and reasons for regimen switching. In the early group, 57 percent switched their regimen over 18 months of follow-up, compared with 67 percent in the later group, a nonsignificant trend (P = 0.157). But nearly three times more later starters (36 percent) than early starters (13 percent) switched because of virologic failure (P = 0.005). In a multivariate analysis adjusted for the three variables bulleted above, a later start independently upped the odds of switching for failure by a factor of 5 (P = 0.003). But injection drug use, adherence below 90 percent, and HCV coinfection each raised that risk much more, at respective odds ratios of 50 (P = 0.016), 37 (P = 0.02), and 40 (P = 0.014).
Because of their lower pretreatment CD4 count, the later group also trailed the early starters in CD4 count three, six, nine, 12, 15, and 18 months after starting therapy (P < 0.0001 at every point). But at 18 months they had a good average count of 525 cells/mm3, compared with 786 cells/mm3 in the early group.
Reasonable people can disagree about what this study means. Perez-Elias and colleagues offered a binocular view. "It seems safe to delay initial [antiretroviral therapy] in non-AIDS patients until the CD4 count reaches a number between 350 and 200 cells/mm3, in terms of clinical outcomes," they concluded. "However, the higher risk of changes due to failure complicates further antiretroviral therapy management" in later starters. Indeed, the study slices some sinew from the stock argument that people who start treatment at high T-cell counts will expend their antiretroviral options faster than waiters. In this cohort, at least, the exact opposite could prove true.
Reasonable people can also turn the table and ask, if someone started treatment with a high CD4 count, as Perez-Elias's early group did, is it safe to stop if treatment keeps HIV under wraps? Before ICAAC two studies suggested that people can stop safely if they launched their antiretroviral argosy at CD4 and RNA sums no longer deemed grounds for intervention by the guideline givers -- generally, above 350 cells/mm3 or below 55,000 copies/mL.17,18 An ICAAC study reached the same conclusion.
Another Madrid group, represented by Manuel Fernández-Guerrero (Fundación Jiménez Díaz), tracked 49 people averaging 34 years of age who took their first antiretroviral with 300 to 500 CD4 cells/mm3 or 10,000 to 70,000 RNA copies/mL, and usually both [abstract H-1082]. The plan of this ongoing study calls for renewed treatment if CD4 cells tumble under the 300 line, if the viral load tops 70,000 copies/mL twice in a row, if an opportunistic disease develops, or if a person withdraws consent.
The group's average CD4 count when they stopped treatment stood at a robust 730 cells/mm3. That average dropped rapidly in the drug holiday's first four months, to 544 cells/mm3, then drifted to 525 cells/mm3 at month eight, 496 cells/mm3 at month 12, and 481 cells/mm3 at month 16. So far 12 people (24 percent) have resumed treatment, four in the first eight months, and eight during months nine through 20. But all restarters regained virologic control with the regimen they had stopped. Only one person endured a disease suggesting compromised immunity, herpes zoster.
Elena Seminari and colleagues at the Policlinico San Matteo in Pavia have nearly completed a one-year study that randomized people with viral loads below 50 copies/mL on treatment to continue therapy or to switch between treatment and no treatment every month. A subanalysis presented at ICAAC tried to figure why five people met the protocol criterion for leaving the study -- a CD4 count below 200 cells/mm3 after a full month on treatment following a break [abstract H-1745]. One person failed to climb back to 200 cells/mm3 at month two, two at month four, one at month six, and one at month eight. All five shared one trait, a T-cell nadir below 50 cells/mm3. But that deficit alone did not explain their poor CD4 recovery while on treatment, because 12 other study participants had nadirs under 50 and did not have to leave the study.
Comparing the five dropouts, the 12 others with sub-50 nadirs, and 45 study participants with nadirs above 50 cells/mm3, Seminari found no differences in age, gender, duration of treatment, duration of viral control below 50 copies/mL, or antiretroviral regimen. The CD4 nadirs in the 17 with sub-50 nadirs averaged 23 cells/mm3, compared with a nadir of 256 cells/mm3 in the other 45 (P < 0.001). Having a nadir under 50 cells/mm3 raised by 18 times the risk of failing to attain a count above 200 cells/mm3 after a full on-treatment month.
The first on-treatment month after the first holiday month held the clue to protocol-defined CD4 failure. Everyone in the 45-person comparison group regained all the CD4 cells lost during the first break during their first month back on treatment. The 12 people with nadirs under 50 cells/mm3 who did not have to leave the study regained 70 percent of their T cells in the first on-treatment month. And the five people who failed to climb back to 200 cells/mm3 at some point in the study regained only 37 percent of their T cells in the first on-treatment month. Seminari proposed that this "flattened regain curve" in the first treatment month after the first treatment break may predict later CD4 failure.
Not surprisingly, a CD4 count under 200 cells/mm3 emerged as one of eight predictors of death in a retrospective analysis of 67 people who died in a population of 639 with HIV infection [abstract H-1145]. But statistical analysis by Chiu-Bin Hsiao (State University of New York, Buffalo) rated a sub-200 CD4 sum as only a "minor" factor. The two major factors were a small surprise: thrombocytopenia and hyperglycemia (Table 3).
Hsiao's record review involved 417 males and 222 females seen between December 1998 and September 2001. The group included 292 African Americans, 229 Caucasians, and 90 Hispanics; 253 had HCV coinfection. After a median follow-up of two years, 47 males and 20 females had died. Forty-three of those who died (64 percent) had a CD4 count under 200 cells/mm3, and 42 (63 percent) had a viral load above 10,000 copies/mL.
A multivariate analysis determined that a platelet level below 130 x 103/µL inflated the risk of death 5.3 times, followed by hyperglycemia, a viral load above 10,000 copies/mL, anemia, high lactate dehydrogenase, HCV coinfection, a CD4 count under 200 cells/mm3, and age over 39 years (Table 3). Assigning a numeric value to each of the eight risk factors, Hsiao fashioned two schemes to reckon the risk of death.
A person with none of the risk factors has a 0.6 percent risk of death. Someone with all the risk factors has a 98 percent risk. Risk rates reflect the median follow-up of two years (range 12 to 1,245 days) for the cohort. In the 67 cohort members who died, a median of 360 days (range 12 to 904 days) elapsed between the date of data collection and death.
This kind of analysis comes with built-in limits. Although the cohort's make-up is diverse, it is relatively small and includes only 67 deaths. Substantial subsets were taking no antiretrovirals (151 or 24 percent) or only two NRTIs (66 or 10 percent). Hsiao did not report how many in those groups died, but he reported that different regimens did not influence mortality. Still, the surprise finding on thrombocytopenia and the high risk with hyperglycemia -- a PI side effect -- stress the value of tracking these lab values and should inspire further research.
19 Bristol-Myers Squibb hopes to change that by grooming atazanavir as an unboosted first-line PI with a unique resistance profile, a light touch on lipids, and a once-daily demand on memory. But an ICAAC report of a trial comparing the new PI with efavirenz will take some explaining before atazanavir catches on as prime-time primary therapy. And although triple nucleoside combos centered on abacavir rate first-line consideration among many clinicians, one abacavir medley rehearsed at ICAAC fell flat. Other studies pitted ritonavir-boosted PIs against each other and against efavirenz, and efavirenz didn't suffer in the comparison.
The study involved treatment-naive people in Asia, Africa, Europe, and North and Central America with a median baseline viral load around 4.9 logs (about 80,000 copies/mL). More than 40 percent had a viral load over 100,000 copies/mL. The median starting CD4 count measured 286 cells/mm3 in the atazanavir arm and 280 cells/mm3 in the efavirenz arm. The cohort was young, with a median age of 34 years. About one third of study participants were women.
The noncompleter-equals-failure analysis counted failure as stopping the assigned treatment for any reason, an AIDS diagnosis, or a confirmed rebound above 50 copies/mL after a response. More than 80 percent in both groups finished 48 weeks of treatment. But at that point only one third of study participants had a viral load under 50 copies/mL by the intent-to-treat analysis and only half had a sub-50 load in an on-treatment analysis (Table 4). The small differences between treatment arms did not reach statistical significance. Although the atazanavir group gained significantly more CD4 cells than the efavirenz group, Squires reasonably proposed that the modest difference probably has no clinical relevance.
One third in the atazanavir group had hyperbilirubinemia, but only 5 percent had to lower their atazanavir dose and fewer than 1 percent left the trial as a result. As in other atazanavir trials, fasting lipids changed little with the drug over 48 weeks. People taking efavirenz had bigger lipid jumps, but few people crossed National Cholesterol Education Program (NCEP) thresholds for antilipid therapy.
Some factor or factors clearly sabotaged the virologic response to both drugs in this study. Although research has yet to establish the potency of atazanavir, efavirenz has an unblemished record in head-to-head trials that enroll treatment-naive people. In a study of efavirenz plus 3TC with either tenofovir or d4T, for example, 48-week proportions with viral loads under 50 copies/mL measured over 80 percent in a missing-data-equal-failure analysis [see abstract LB2 below]. Reason dictates that efavirenz did not suddenly become a rotten drug.
A few factors could contribute to the dreadful virologic results. The protocol forbade dose reductions with AZT or 3TC or switches to different nucleosides. Those dictates may have spawned spotty adherence reflected in emergence of the 3TC-related M184V mutation in more than half of the study participants genotyped so far. Poor adherence seems even more likely when one considers that the nonnucleoside-linked K103N mutation cropped up in 65 percent of those genotyped, a rate that drew stunned murmurs from the crowd. Atazanavir's apparent signature mutation, I50L, appeared in only 16 percent so far. But, as much research shows, resistance will emerge first against 3TC and nonnucleosides when a regimen begins failing.
Poor adherence apparently followed a regional pattern. Squires told IAPAC Monthly that virologic responses in some regions considerably lagged those in others. European participants did worse than those in Asia and Africa, she noted, surmising that a higher proportion of injection drug users in the European population may partly explain those differences. At the same time, Squires added, the relative unavailability of antiretrovirals in Africa and Asia could inspire stricter adherence among people lucky enough to get into trials.
Finally, the trial used a tougher than usual definition of failure prescribed by the Food and Drug Administration. It counts any drug switch for any reason or two consecutive viral loads above 50 copies/mL as a failure. The AIDS Clinical Trials Group, for example, requires consecutive rebounds above 200 copies/mL to signal failure, figuring that touchy ultrasensitive assays could record some 50+ readings that do not signal abiding rebounds.
Clearer answers should emerge as the researchers analyze region-specific response and resistance patterns. (The trial did not include a formal adherence measure.) Even if poor adherence and a draconian failure definition explain this downfall of atazanavir and efavirenz, clinicians will likely want fresh proof of atazanavir's antiviral punch in the wake of this study. Few are likely to demand the same for efavirenz.
If atazanavir turns out to be a notch or two feebler than hoped, a ritonavir boost could be a fallback strategy. At this point atazanavir's maker still positions the drug as an unboosted first-line agent, but Bristol researchers did study the impact of a 100-mg ritonavir prod in 30 healthy volunteers [abstract H-1716]. S. Agarwala and colleagues gave them 300 mg of atazanavir once daily for 10 days, then added ritonavir on days 11 through 20. Atazanavir's geometric mean peak concentration jumped from 3,288 ng/mL on day 10 to 6,129 ng/mL on day 20, an 86 percent gain. The area under the concentration-time curve (AUC) for atazanavir vaulted from 16,875 ng · h/mL on day 10 to 57,039 ng · h/mL on day 20, more than a threefold surge. The trough concentration climbed about 10-fold. Ritonavir levels varied little over the course of the study.
As with any drug, the danger of increased exposure is increased toxicity. In earlier studies that used 400 or 600 mg of atazanavir daily, more people taking the higher dose had bilirubin elevations. And, as Bristol's Edward O'Mara reported in an analysis of 202 people enrolled in phase I studies of the PI, higher AUCs raised the risk of billowing bilirubin regardless of whether a person has a genotype favoring hyperbilirubinemia [abstract A-1253]. At the same time, a 7/7 genotype in the UGT 1A1 gene independently raised the risk that total bilirubin would exceed 2.5 mg/dL. (UGT 1A1 glucuronidates unconjugated bilirubin.) Eleven of 13 people (85 percent) with the 7/7 genotype had a total bilirubin topping 2.5 mg/dL, compared with 17 of 49 (35 percent) with a 6/7 genotype and eight of 62 (13 percent) with a 6/6 genotype.
Analyzing 950 isolates from people with PI experience but naive to atazanavir, Bristol's Richard Colonno found that most isolates resistant to one or two PIs remained susceptible to atazanavir when he used an arbitrary 3-fold drop in susceptibility to indicate resistance [abstract H-2049]. But most isolates resistant to three or more PIs also proved resistant to atazanavir. The analysis broke down like this:
Colonno also figured which mutations in these isolates conferred resistance to atazanavir. No single mutation did, but any five from the following familiar 14 did: (see Diagram 1)
An intent-to-treat analysis at week 48 considered all study participants who took their assigned drugs and had at least one on-drug RNA rating. By that measure, 51 percent taking saquinavir and 71 percent taking efavirenz notched a viral load below 50 copies/mL. The saquinavir group did better in an on-treatment analysis, but still lagged the efavirenz group, with 73 versus 93 percent under 50 copies/mL. The study tested the hypothesis that saquinavir/ritonavir is "noninferior" to efavirenz in pushing viral loads under 50 copies/mL. "Noninferiority criteria," Montaner reported, "were not met."
Gastrointestinal (GI) side effects proved the PIs' undoing in the intent-to-treat analysis. One third of the 81 people assigned to saquinavir/ritonavir had nausea, vomiting, or diarrhea, compared with 4 percent of the 80 people randomized to efavirenz. Twelve people (15 percent) stopped saquinavir/ritonavir because of side effects and 11 (14 percent) withdrew their consent to continue. Respective numbers in the efavirenz group were five (6 percent) and three (4 percent).
Total cholesterol rose from an average 171 mg/dL at baseline to 206 mg/dL in the PI group and from 167 to 204 mg/dL with efavirenz. Triglycerides climbed from 124 to 175 mg/dL with saquinavir/ritonavir and from 139 to 210 mg/dL with efavirenz. Because research shows that d4T boosts lipids [see abstract LB2 below], knowing which NRTIs people were taking may have enlightened this lipid analysis.
Montaner proposed that the saquinavir group's GI troubles may be traced to an ingredient in saquinavir's soft-gel capsule. He suggested that the venerable hard-gel formulation may be easier to stomach with a ritonavir boost.
Another study of saquinavir/ritonavir, this time at 1,000/100 mg twice daily in people with or without PI experience, pitted those PIs against indinavir/ritonavir at 800/100 mg twice daily [abstract H-172]. The regimens yielded similar virologic responses, but more people had to switch from the indinavir arm than from the saquinavir arm, and more taking indinavir suffered disease progression. Two features of the MaxCmin 1 trial design complicate interpretation of the results: First, as in the FOCUS study, individual clinicians picked the NRTIs to give with the PIs; they could also add an NNRTI if deemed necessary. But chief investigator Jan Gerstoft (University of Copenhagen) did not report the non-PI treatment assignments. Second, the study population can safely be called mixed. About 40 percent in each treatment arm had no PI experience. Among those who had tried a PI, about 60 percent had a viral load under 400 copies/mL when they entered the study.
After 48 weeks of follow-up, an intent-to-treat analysis involving study participants who took at least one assigned dose rated 20 percent taking indinavir and 18 percent taking saquinavir as virologic failures. But 41 percent randomized to indinavir/ritonavir had to stop taking those PIs compared with 27 percent randomized to saquinavir/ritonavir (P = 0.018). Gerstoft recorded clinical progression in 28 percent of the indinavir group and 15 percent of the saquinavir group (P = 0.006), with no deaths in either arm. Total cholesterol, low-density lipoprotein cholesterol, and triglycerides each rose about 20 percent with indinavir/ritonavir, but less than 10 percent with saquinavir/ritonavir (P < 0.05).
Another study by Jan Gerstoft appraised one familiar combination (saquinavir/ritonavir at 400/400 mg twice daily plus AZT/3TC) and two rarer hybrids (ddI/d4T/abacavir and AZT/3TC plus nelfinavir at 1,250 mg twice daily and nevirapine at 400 mg twice daily) in 178 treatment-naive people [abstract H-164]. The rationale offered for the triple-nuke regimen was that it may be less prone to cross-resistance than abacavir plus AZT and 3TC. The main outcome of the trial seems to be that half the people in each treatment arm bailed out of the study before week 48. With 60 people in each group, 33 stopped saquinavir/ritonavir, 24 nelfinavir/nevirapine, and 35 the three nucleosides.
Neurologic toxicity -- presumably neuropathy -- felled 16 people (27 percent) in the NRTI group compared with three taking saquinavir/ritonavir and two taking nelfinavir/nevirapine (P < 0.001). When compared by grade 4 toxicity rates, saquinavir/ritonavir proved the most tolerable regimen (four people), followed by nelfinavir/nevirapine (seven), and the triple nukes (eight). The NRTI regimen also lagged the others in a missing-data-equal-failure analysis of sub-20-copy viral loads at 48 weeks, with about 41 percent under that mark compared with 55 percent taking saquinavir/ritonavir and 62 percent taking nelfinavir/nevirapine. Few would contest Gerstoft's conclusion that "this triple NRTI regime cannot be recommended."
FTC trounced d4T by numerous measures of efficacy in a double-blind, placebo-controlled trial involving treatment-naive people also beginning once-daily ddI plus efavirenz [abstract LB1]. FTC -- a cytosine analog like 3TC -- needs only one 200-mg dose daily. Michael Saag (University of Alabama, Birmingham) reported that the 286 people randomized to FTC and the 285 randomized to d4T matched closely in pretreatment viral load (4.9 logs in both groups) and CD4 count (280 cells/mm3 in the FTC group and 300 cells/mm3 in the d4T group).
After 24 weeks of treatment, 81 percent taking FTC and 70 percent taking d4T had a viral load under 50 copies/mL (P = 0.002). FTC bettered d4T in 52-week Kaplan-Meier probabilities of efficacy, tolerability, and effectiveness, defined in Table 5. By week 52, people taking FTC had gained significantly more CD4 cells than those taking d4T.
If licensed, FTC would join ddI, 3TC, and tenofovir as nucleosides (or -tides) that can be taken once daily. Also like 3TC and tenofovir, it stifles hepatitis B virus as well as HIV. A 48-week interim analysis of another placebo-controlled trial showed similar potency with tenofovir or d4T in antiretroviral-naive people, but significantly fewer side effects with tenofovir [abstract LB2]. The 600 study participants had an average starting viral load of 4.9 logs and an average CD4 count of 279 cells/mm3. Everyone also took 3TC and efavirenz.
Joel Gallant (Johns Hopkins University) reported similar RNA and CD4 responses with tenofovir and d4T after 48 weeks. In a missing-data-equal-failure analysis, 82 percent taking tenofovir and 81 percent taking d4T had a viral load under 50 copies/mL. Responses did not differ in subgroups starting treatment with fewer or more than 100,000 RNA copies/mL. The tenofovir group gained an average 169 cells/mm3, compared with 167 cells/mm3 with d4T.
Rates of grade 3 or 4 side effects proved similar in the two groups -- bedeviling 19 percent taking tenofovir and 17 percent taking d4T. But the d4T group lagged tenofovir takers in three key lab measures and in peripheral neuropathy. Triglycerides rose an average 74 mg/dL in the d4T group but did not change in the tenofovir group (P < 0.001). Total cholesterol climbed by 53 mg/dL with d4T versus 25 mg/dL with tenofovir (P < 0.001). Whereas 7 percent taking tenofovir reached lactate levels above 2.1 mmol/L, 36 percent taking d4T did (P < 0.001). Peripheral neuropathy affected 2 percent taking tenofovir and 7 percent taking d4T (P < 0.001).
Among people starting tenofovir with a thymidine analog mutation (the TAMs are M41L, D67N, K70R, L210W, T215Y/F, and K219Q/E), only 11 percent attained a sub-50 viral load, whereas 49 percent without TAMs did so. Having the 3TC-induced M184V mutation also favored a good virologic response with add-on tenofovir; 49 percent with M184V at baseline got their viral load under 50 copies/mL.
In a multivariate analysis, Miller calculated the following odds ratios (OR) for a sub-50 copy reading according to baseline viral load or mutation array:
Intensification of a floundering regimen can be risky. If adding a new drug doesn't stop the slide, mutations can pile up, as Miller also demonstrated in this analysis. His findings give clinicians a few tools to help decide whether adding tenofovir is worth the risk. When tenofovir regimens did fail in this population, added TAMs proved the biggest culprit. The tenofovir-linked K65R mutation popped up in only 3 percent (eight people), and none suffered a virologic rebound.20,21 No one -- least of all people with time running out and drug options running down -- will look askance at the likely value of this novel antiretroviral. But one wonders how smoothly it will fit into real-world schemes.
Because T-20 is hard to make, it will cost a lot, almost certainly more than any current antiretroviral. It has to be mixed carefully a half-hour before each dose (see "T-20" in reference 12), and it must be injected under the skin twice daily. Almost everyone who uses T-20 gets injection site reactions, 98 percent in the two big randomized trials.20,21 Although only 3 percent in those studies quit as a result, more will almost certainly abandon the drug in practice. In the European-Australian trial,20 the overall 24-week dropout rate was 11 percent in both the T-20 group and the non-T-20 group. But in the North American-Brazilian trial,21 17 percent quit the T-20 arm compared with 5 percent in the control arm.
At ICAAC Julio Montaner reviewed results from the North American-Brazilian trial, looking at potential subgroup differences and at predictors of antiviral vim [abstract H-1074]. Compared with the optimized regimen alone, T-20 plus optimized background yielded a better virologic response in males and females, whites and nonwhites, people under and over 40 years old, people with baseline viral loads below or above 100,000 copies/mL, and people with starting CD4 counts below or above 100 cells/mm3. People with phenotypic or genotypic sensitivity to zero, one to two, or three to four antiretrovirals did significantly better with T-20 than with optimized background alone (P < 0.05). Those with phenotypic or genotypic sensitivity to five or more antiretrovirals got a better response with T-20, but not a significantly better response.
In a multiple regression analysis estimating the log change in viral load at 24 weeks, six variables mattered:
Replacing the phenotypic sensitivity score with a genotypic sensitivity score yielded similar results. Baseline viral load did not figure in the response to T-20. The interesting findings with lopinavir/ritonavir clearly indicate that T-20 does much better when it teams with a potent ally. Montaner noted, though, that the T-20 group did not differ from the control group in proportions pretreated with lopinavir or getting lopinavir in the salvage regimen.
The one randomized study of the three involved 37 people taking lopinavir and amprenavir with 100 or 200 mg of ritonavir twice daily [abstract H-1078]. Everyone in the study had used at least two PIs and one NNRTI. Median baseline numbers included seven protease mutations, 207 CD4 cells/mm3, and 4.7 log copies/mL (about 50,000 copies/mL). After 26 weeks, reported Gilles Raguin (Bichat Hospital, Paris), the median viral load had dropped 2.5 logs in the 200-mg ritonavir group and 1.4 logs in the 100-mg group (P = 0.02). Eleven people (61 percent) taking 200 mg twice daily and six (32 percent) taking 100 mg twice daily reached a viral load below 50 copies/mL (P = 0.07). CD4 counts rose about 120 cells/mm3 in both groups.
The extra ritonavir did not compromise tolerability in this trial. Four in each study arm had to stop treatment. Seven (39 percent) in the 200-mg group and 10 (53 percent) in the 100-mg group had a grade 4 side effect. In a multivariate analysis, the extra ritonavir (P = 0.007) and fewer baseline protease mutations (P = 0.02) predicted virologic success. Baseline resistance phenotype did not predict success.
A larger, noncomparative study in people with shorter antiretroviral résumés and less advanced disease documented a reasonable six-month response to indinavir/ritonavir at a dose of 800/200 mg twice daily [abstract H-173]. Pompeyo Viciana and colleagues in Madrid and Barcelona gave doubly boosted indinavir plus two NRTIs to 105 people who had already tried an average of 2.2 PIs. Their baseline viral load averaged 4.3 logs (about 20,000 copies/mL) and their starting CD4 count 284 cells/mm3.
After six months of indinavir/ritonavir, 41.9 percent had a viral load under 200 copies/mL in a noncompleter-equals-failure analysis. An on-treatment analysis showed that 54.4 percent had a sub-200 viral load. The average CD4 count climbed to 370 cells/mm3. Baseline detection of mutations conferring resistance to indinavir and ritonavir (V82A and L90M) did not rule out a good response to the salvage regimen. But fewer people with the nelfinavir-linked D30N mutation responded.
Ritonavir-induced gastrointestinal (GI) problems proved the most frequent grade 3 or 4 toxicity, vexing 16 people (15 percent). In all, clinicians reported a toxicity of any grade in 38 people (36 percent).
A second noncomparative study in people with even better baseline barometers did not record a better response to the same dose of indinavir/ritonavir than did Viciana [abstract H-175]. David Parr from the University of Texas Medical Branch at Galveston, and coworkers in Georgia and Florida, treated 63 people with a mean CD4 count of 360 cells/mm3 and a mean viral load of 3.85 logs (about 7,000 copies/mL). Forty-eight of them (76 percent) were taking their first failing single-PI regimen, while 15 (24 percent) were on their second PI or their first dual PI combo.
The initial study lasted 24 weeks, during which 16 people dropped out (seven because of toxicity). Of the 47 people who finished 24 weeks, 20 did not enroll in a 24-week extension because of the timing of the extension approval. Of the 27 who did continue, eight dropped out before week 48 (five because of toxicity).
In a 24-week noncompleter-equals-failure analysis, 35 people (56.5 percent) had a viral load under 400 copies/mL and 23 (37.1 percent) were under 50 copies/mL. In the 48-week analysis excluding the 20 people who could not re-enroll, 10 (23.3 percent) were under 400 copies/mL and nine (20.9 percent) under 50 copies/mL. A last-observation-carried-forward analysis charted mean CD4 gains of 96 cells/mm3 at week 24 and 131 cells/mm3 at week 48.
Not counting those unable to sign up for the 24-week extension, 24 people dropped out of the study at some point. Nine of the 24 (37.5 percent) quit because of side effects. But some of the other reasons for bowing out could reflect poor tolerance, such as withdrawn consent (five people) and loss to follow-up (three people). Eight of the toxicity-related dropouts (representing 13 percent of the original study group) involved GI intolerance, a possible consequence of the doubled ritonavir dose.
Clinicians considering the extra ritonavir boost have to weigh the virologic benefits recorded in studies like these against the risk of increased toxicity. And toxicity rates will surely vary depending on disease stage and other variables. In a case series at the Lipodystrophy Workshop a week before ICAAC, Düsseldorf clinician Stefan Mauss reported that eight of 11 people could not tolerate 200 mg of ritonavir twice daily with lopinavir (400 mg twice daily) and amprenavir (600 mg twice daily).22 In an earlier study of lopinavir plus amprenavir with 100 mg of ritonavir twice daily, Mauss reported only one discontinuation among nine people because of GI toxicity.23
Alice Tseng (Toronto General Hospital) measured fasting lipids in 21 people starting a lopinavir salvage combination [abstract H-1916]. After a median nine months of follow-up, the median total cholesterol rose from 4.46 to 5.60 mmol/L (172 to 217 mg/dL, P < 0.001) and median triglycerides from 2.22 to 4.71 mmol/L (197 to 417 mg/dL, P < 0.09). The median lopinavir trough concentration measured 4.13 µg/mL in people with high triglycerides and 2.64 µg/mL in people with low triglycerides (P = 0.05). Among people with high total cholesterol, the median lopinavir trough stood at 4.08 µg/mL versus 2.87 µg/mL in people with low total cholesterol, but that difference was not significant. Lopinavir peak concentrations did not correlate with hyperlipidemia in Tseng's analysis.
Studying 22 people taking lopinavir in salvage, Sergio Padilla (University Hospital, Elche, Spain) figured a mean lopinavir plasma level for each person based on three to nine troughs [abstract H-1915]. Percent increases in cholesterol and triglycerides from baseline to weeks 12 and 24 correlated positively with mean lopinavir levels (r = 0.55, P = 0.008, for triglycerides at week 12; r = 0.55, P = 0.007, for cholesterol at week 24). At week 48 cholesterol was significantly higher in people with higher lopinavir troughs (8.88 versus 5.67 mmol/L [343 versus 219 mg/dL], P = 0.017). Subcutaneous and adipose abdominal tissue measured by CT increased over 48 weeks. But fat changes did not correlate with lopinavir concentration.
How much PI experience a person has when starting lopinavir could make a difference. Among 70 of 100 people who took lopinavir/ritonavir for 204 weeks as part of their first antiretroviral regimen, 12 (17 percent) had cholesterol levels between 240 and 300 mg/dL and one (1 percent) had between 300 and 400 mg/dL [abstract H-165]. Robert Murphy (Northwestern University, Chicago) reported that nonfasting triglycerides stood between 400 and 750 mg/dL in 18 (26 percent) and above 750 mg/dL in four (6 percent). One of 100 study participants dropped out because of high lipids.
A group of Spanish clinicians not affiliated with Virco found that Virco's VirtualPhenotype did better than standard phenotyping in picking more potent rescue regimens [abstract H-1079]. (The Virtual-Phenotype matches a submitted isolate's genotype with phenotypes of database viruses that have the same genotype.) This double-blind trial randomized 276 people from five outpatient clinics to have VirtualPhenotyping or standard phenotyping as a guide for their clinicians to pick new regimens. Most people (60 percent) had tried drugs from all three classes, and the average viral load hovered around 10,000 copies/mL.
Maria Jesus Perez-Elias reported that people in the Virtual arm got an average of 2.9 drugs rated "active," while those in the phenotyping arm got an average of 2.8. The 24-week missing-data-equal-failure analysis defined failure as a regimen switch because of virologic failure -- but not because of intolerance. By that standard, the median viral load dropped 1.3 logs in the Virtual arm and 0.92 log in the phenotyping arm (P = 0.008). Respective proportions with a viral load below 400 copies/mL were 56.5 percent and 46.5 percent (P = 0.01).
At the 2002 Resistance Workshop, Neil Parkin plumbed ViroLogic's vast viral database to propose novel genotypic pathways promoting resistance to lopinavir.24 At ICAAC he did the same favor for amprenavir. Analyzing more than 4,000 isolates, he found frequent discordance between genotypic resistance or sensitivity determined by four rule sets and phenotypic sensitivity or resistance determined with a conservative fold-change susceptibility cutoff of 2.5. For example, according to the ANRS algorithm, 15 percent of isolates called genotypically sensitive proved phenotypically resistant, and 7 percent called genotypically resistant proved phenotypically sensitive, for a total discordance score of 22 percent. All the rule sets yielded total discordance rates between 22 and 25 percent.
Because the genotypic rules relied on defined amprenavir mutations -- V32I, I50V, I54L/M, and I84V -- Parkin figured that genetic pathways not including amprenavir-selected mutations must contribute to slumping susceptibility. To scout out the mutational milestones on these shadowy pathways, he analyzed mutations at all protease positions that turned up in more than 1 percent of his samples. Then he used a univariate analysis to pick mutations that may contribute to reduced susceptibility to amprenavir, and he tested those mutations for resistance when combined with each other and with a second list of canonical protease mutations. This exercise yielded a suite of mutations and mutation sets that shave susceptibility to amprenavir:
If Parkin added the first set listed above to the mutations in current algorithms, total phenotypic-genotypic discordance dropped from 24.5 percent to 20.6 percent. If he then tossed in 33F + 82A, total discordance dwindled to 17.9 percent. Piling on the third set narrowed total discordance to 15.1 percent. Finally, Parkin verified the refurbished algorithm against a validation set of 1,634 resistant isolates not included in the set used to track down the newly implicated mutations. Along the way to these findings, he also showed that resistance conferred by amprenavir's signature mutation, I50V, can be reversed by the N88S substitution.
Retroviral researchers of every stripe have now stalked HIV through dusky forests for 20 years. And some did frame fearful symmetries -- X4 and R5 viruses, unraveling error-prone RNA strands, and life-saving two-target therapies come to mind. But the talk of HIV research as this year's ICAAC unfolded was whether Linqi Zhang and David Ho (Aaron Diamond AIDS Research Center, New York) had managed to frame another symmetry by completing a coup with CAF -- Jay Levy's long-since discovered, but never defined, CD8 antiviral factor. As Fate's immortal hand would have it, ICAAC organizers had already slated Ho to talk during a pathogenesis plenary. And by the time he took the stage, everyone knew the "Host factors that suppress HIV replication" in his talk's title were alpha-defensins [presentation 1192].
Never mind that few -- if any -- attendees had ever heard of these pint-sized proteins before Zhang and Ho's paper went online.26 Many could appreciate the neat symmetries here. Levy published his CAF discovery in Science in 1986,7 but diligent work in his lab never managed to nab the culprit molecule. Over the years Levy emerged as an outspoken critic of Ho's eradication essays,27 not to mention his hit-hard-hit-early tenet.28 Now, 16 years after CAF's first sighting, Zhang and Ho's paper, again in Science, claims their findings "show that alpha-defensins-1, -2, and -3 collectively account for the anti-HIV-1 activity of CAF that is not attributable to beta-chemokines."26
Alpha-defensins, which Zhang identified with a novel gene chip, turned up in CD8 cells from three of three nonprogressors, none of four rapid progressors, and 11 of 15 volunteers without HIV infection. Zhang showed that antibodies against alpha-defensins eliminate CAF activity, and that adding synthetic alpha-defensins to HIV-infected cells inhibits HIV replication.
Although the discoverer of alpha-defensins called this work "very convincing,"8 not everyone buys the conclusion that they account for the tiger's share of CAF's anti-HIV action. The study population was small; the antiviral activity weak; and no one could explain how alpha-defensins, which breach the walls of noxious bacteria, also bring down HIV. (For more detailed reviews of Zhang's study, see references 8 and 29.)
But there's a bigger question about this research: What does it mean for the millions of people with HIV who are not long-term nonprogressors? The most optimistic answer is that -- if Zhang and Ho are right about alpha-defensins' antiviral role -- this work could lead to novel antiviral strategies. But, appropriately, Ho himself downplayed that possibility in press interviews. Whether synthetic alpha-defensins could be turned into a tolerable drug, and -- if they can -- whether they would stop HIV in humans without unhinging healthful doings, remain weighty riddles.
No. The talk that preceded Ho's framed a symmetry that means more to people destined to become long-, short-, or mid-term progressors. Back when CAF came to light in Levy's lab, virologists thought about two types of HIV-1 -- the type found in North America, and the type found in Africa, explained Thomas Quinn of Johns Hopkins University [presentation 1191]. A scant six years later, viral orthographers had named five subtypes -- or clades -- of the retrovirus, A through E.30 That was only the beginning.
Now we are up to subtype K, and there is talk of L. But wait. It gets more complicated. Some of the original subtype designations turned out to be wrong. Today's subtype E, for example, really represents a circulating recombinant form (CRF) of A and some older E. Subtype I actually denotes a CRF that recombines four parental strains. So far researchers have tallied 14 CRFs, and they haven't stopped to catch their breath.31 what most troubled experts was that the superinfected person's immune system recognized the new virus and mounted a response. But this man's strong and broad T-cell responses did not protect him from a second virus of the same subtype (B) that differed from his first virus by only 12 percent at the protein sequence level. As Duke University's Kent Weinhold told IAVI Report, "These data challenge our notions that magnitude and breadth translate into protection."32
IAVI Report goes on to explain why this case of superinfection may not spell doom for vaccine research. The index case had HIV to begin with, and even minimally compromised immunity could make him more susceptible to reinfection than a freshly vaccinated, uninfected person. His immune response to the new virus was cell based, and no one expects a CTL-based vaccine to block infection. But, to tell the truth, everyone would be happier if this man's immune system had fought off the new virus.
As Quinn noted, leaders in this field disagree when estimating the impact of viral diversity on vaccine development. Andrew McMichael (Oxford) thinks vaccine developers have to trek from region to region, sample the locally prevalent virus, then use it to brew a local vaccine. Bette Korber (Los Alamos National Laboratory) believes they must time-travel all the way back to the parental HIV-1 -- via computer calculations -- if they want to make a vaccine that works. Quinn himself proposed that a vaccine derived from subtypes A, B, C, and D, plus CRF01 and CRF02, which account for 85 percent of worldwide infections, holds the most promise.
So far, of course, no one seems close to a single-subtype vaccine.
Those fearfully symmetric correlates of protection, which a potent vaccine would mimic, remain unframed.
Mark Mascolini writes about HIV infection (firstname.lastname@example.org).
Back to the December 2002 issue of IAPAC Monthly.