Today we heard the depressing news about cross-resistance developing between protease inhibitors. It made for a toxic cocktail.
As I left the auditorium I bumped into [Merck virologist] Emilio Emini.
Harrington: So what do you do if you fail Crixivan?
Emini: [sighs] We don't know what to do.
Harrington: Take two new nucleosides and nevirapine?
Emini: Yeah. And pray.
No one had yet assessed the healing effects of prayer on viral load. This was what we'd come to.
In the lobby of the Interior Department I ran into a colleague who was wild with fear and disappointment.
Colleague: I'm going to die.
Harrington: This is everything we were hoping wouldn't happen.
Colleague: I don't have anything to switch to. I'm going to stop everything.
Dawn Averitt tried to calm our colleague down.
Averitt: Why don't you wait for your next viral load?
Colleague: I'm already back at baseline!
Averitt: But your CD4 count is 250, and it was down to 60 in January.
Colleague: True, but my viral load is back to half a million.
Averitt: Why don't you wait until you get your latest numbers and see what your resistance profile is?
Colleague: I'm going to go outside and have a nervous breakdown.
We went outside. It was frigid, and fragile snowflakes swirled around in the wind. Sometimes the gap between how the researchers felt and how we felt becomes an abyss. They were excited about the endless possibilities opened up by the research advances of 1996; we were terrified about the limited treatment options facing people who had exhausted most of the current arsenal of antiretroviral therapy. What to do with those whose viral load refused to go undetectable? What to do with those who added a protease inhibitor to a failing two-drug regimen and now appeared doomed to develop resistance, or worse -- especially with ritonavir and indinavir -- cross-resistance to all other protease inhibitors? What to do with those who jumped aboard last year's [1995's] bandwagon, AZT+3TC, and now appeared likely to have developed 3TC resistance and, with it, cross-resistance to ddI, ddC, and possibly 1592 [abacavir]? The Chinese menu approach to antiretroviral treatment suddenly looked much less appetizing, and much less nourishing.
. . . Many of the researchers present did not share the activist sense that we were facing a crisis that, if handled improperly, might make things worse than before. This was not the prevailing view propounded by gun-happy virologists, drug-happy pharmaceutical companies, media captivated by a surprising good-news story and many people with HIV still struggling to absorb the complex developments of 1996. Just that week, back-to-back articles in The Wall Street Journal and The New York Times Magazine, both written by HIV-infected journalists, declared that the epidemic was virtually over. We were staring into the precipice while others were still climbing the hill.
Toxic Cocktail. TAGline 4:2, February 1997
Neither the chairs of the Health and Human Services guidelines panel nor the organizers of the Conference appeared especially eager to highlight the significant modifications that appeared in the February 2001 update of the Guidelines for the Use of Antiretroviral Agents in HIV-Infected Adults and Adolescents (hivatis.org/guidelines/AAFEB05S.PDF).
Prior to the Conference, I heard a rumor that the Conference organizers had turned down a request by guidelines panel co-chairs Anthony S. Fauci of NIAID and John Bartlett of Johns Hopkins University for a press conference to announce the new guidelines. So, at the opening press session I questioned Retrovirus chair Connie Benson of the University of Colorado whether they had indeed turned down a request by Drs. Fauci and Bartlett. "Not at all," said Dr. Benson, who is also a co-chair of the government's top HIV clinical research program. It turns out the only request she heard had come from a NIAID press person. "Do you think I would actually turn down Dr. Fauci?" she asked. And in fact, as the Conference opened, the only document NIAID had shown the Conference organizers was an anemic two-page press release; the Retrovirus press person had to download a copy of the new Guidelines from the Internet.
Why the low-key reception? Well, just as enthusiasm for the early use of AZT was deflated by the release of the Concorde study results in 1993, so do the 2001 treatment guidelines finally depose the ailing "hit early, hit hard" strategy. But for some, old strategies die hard.
Rather than trumpeting the changes by holding a press conference before the world's leading AIDS reporters in Chicago and admitting "Whoops! We were wrong! We screwed up! Maybe treating everyone with a viral load over 5,000 to 10,000 HIV RNA copies/mL wasn't such a great idea after all," the Guidelines panel co-chairs and the Retrovirus organizers all felt it was more opportune -- or more expedient -- to skate over the matter as casually as possible, as though the new Guidelines represented a simple and relatively uncontroversial recalibration of the standard of care.
When Mark Schoofs from The Wall Street Journal called me, he said, "This is a good thing. People don't have to go on therapy as early."
"Well, yes," I said, "but it would have been much better to have a clear answer from a randomized, controlled trial [RCT] by now."
"Well, Tony Fauci says that's unfeasible."
"Well, if he says so, it must be true," I replied, "but if they'd started one in 1996, we might know the answer by now."
"You sound angry."
"I am angry. Thousands of people were put on therapy too early, and some of them developed life-threatening side effects, or resistance, and they would have been fine if they had just waited."
Was failing to seek evidence for such a sweeping policy just an honest mistake? A gush of euphoria that the long, dark night of AIDS was over? Well, of course those promulgating the original standard of care were sincere -- but they defaulted from their obligation as scientists to prove that their expert opinions were actually correct -- and, considering the emergence of long-term side effects and widespread cross-resistance, it doesn't seem that they were.
[Other significant changes in the new Guidelines include the addition of indinavir/ritonavir and lopinavir/ritonavir (Kaletra) to the "preferred" initial regimens, extensive new sections on adherence and drug-related adverse effects, and a section on women and viral load.]
The "evidence" to which Dr. Fauci referred is a random grab bag of observational studies which are contradictory and hardly definitive. One can conclude whatever one wants to from these observational studies -- if you favor early treatment, you can find studies that do too, and if you favor later therapy, other studies will back you up.
In British Columbia, AIDS care is free and antiretrovirals are centrally distributed to all eligible HIV-positive individuals. Over 5,400 participants have enrolled; currently over 2,600 are on antiretrovirals. Robert Hogg and colleagues looked at all HIV-positive men and women over 18 years old and totally antiretroviral naive who were first prescribed triple therapy (containing either a PI or an NNRTI plus two NRTIs) between August 1996 and September 1999. Analysis was by (slightly censored) intent to treat. They assessed rates of mortality by different CD4 and HIV RNA thresholds.
One thousand, two hundred and nineteen HIV-positive adults fit the study criteria. Nine hundred and nine of these (74.6%) received a PI-based regimen, while 310 (25.4%) received an NNRTI. Median follow-up was 20 months. 1,034 participants (85%) were men, and 185 (15%) were women. At baseline the median age was 37, the median CD4 count 280 cells/mm3, and the median plasma HIV RNA was 120,000 copies/mL.
By the end of January 2000, 86 deaths had occurred (7%). "Fourteen were not attributed to AIDS and were censored as non-events at the time of death; the remaining 72 deaths gave a crude mortality rate of 5.9%." Cumulative
12-month mortality was 3.2%.
The investigators stratified the results by various CD4 and RNA break points. For example, in early 2000, they used the "new" threshold of 350 cells and a viral load over 30,000. No difference was found in mortality between those who started with a CD4 count over 350 and those with CD4 between 200 to 350. By contrast, mortality was higher in the group that started with a CD4 below 200. Baseline viral load appeared to make little difference when cutoffs of <5,000, 5,000 to 30,000, and >30,000 were used.
Therefore, they used their own data to identify "natural" break points (see chart, below).
After adjustment by multivariate analysis, no baseline viral load level was associated with greater or lesser survival. A baseline CD4 count below 200 predicted worse survival. There was little difference between those starting with
CD4 counts between 200 to 350 and those starting with >350.
Having a baseline CD4 count below 50 cells/mm3 at baseline increased the risk of mortality by seven-fold compared with those starting with over 200 (p<0.001), while starting with a CD4 count between 50 to 199 had a three-fold greater risk (p<0.001). Having a baseline viral load over 200,000 copies/mL appeared to increase the risk by about 1.75-fold, but the result was not statistically significant.
The BC investigators concluded that "the effectiveness of antiretroviral therapy on survival is independent of plasma HIV-1 RNA levels, AIDS, and protease inhibitor use at baseline, but dependent on CD4 levels. . . . The effectiveness of therapy on survival becomes compromised in patients [who] start with CD4 counts below 200mm3" (Hogg 2001).
Three caveats should be recognized about extrapolating from the BC cohort study to, for example, the USA:
Although not presented at Retrovirus, a reassuringly complementary study -- this time assessing virologic rather than clinical outcomes -- was presented at the 5th International Congress on Drug Therapy in HIV Infection at Glasgow in October 2000. This was a meta-analysis by Andrew Phillips and others of virologic outcomes from three European HIV cohorts -- the Swiss HIV Cohort Study, the Frankfurt HIV Cohort, and the EuroSIDA study. They looked at virologic outcomes by baseline CD4 count and HIV RNA in 2,742 individuals who were HAART-naive.
In summary, neither baseline CD4 nor RNA appeared to significantly affect risk of rebound in these three observational cohorts.
Phillips and colleagues concluded that, "So long as the CD4 count remains above 200mm3 and the viral load remains below 100,000 copies/mL, there is no evidence that lower CD4 counts and higher viral loads are associated with poorer responses to antiretroviral therapy" (Phillips 2000).
Median follow-up times were lopsided, with 2.1 years among cases and only 1.3 years among controls. The event rate was low. Four cases (1.1%) and 16 controls (4.4%) developed an AIDS-defining condition, while four cases (1.1%) and 12 controls (3.3%) died.
Among the bizarre features of this study was the fact that the investigators didn't even appear to look at lower CD4 thresholds, such as 200. Of course, starting above 350 is likely to be better than not starting at all. Moreover, the overall event rate (20 cases of AIDS, 16 deaths) was quite low -- 2.75% for AIDS and 2.2% for death. Finally, the case-control methodology is riddled with potential for bias. These deficiencies failed to deter the Retrovirus conference organizers from highlighting this paper as a late-breaker, and hailing it in their helpful notes to the press as a study that "supports [sic] the increasing [sic] body of evidence that earlier treatment is immunologically and clinically beneficial . . ." [Highlighted Abstracts from the 8th Annual Retrovirus Conference, p. 20].
In early 2000, after considerable activist pressure, the NIAID Division of AIDS appeared to recommend $42 million in funding for a when-to-start trial, and held several workshops to discuss methodology and feasibility issues. These initiatives were smothered in the cradle by AACTG leadership and community representatives in their thrall who dismissed the feasibility of long-term research. They were finally buried at the NIAID Council meeting in January 2001.
As the press started to get hold of the new treatment guidelines, it was natural to ask Dr. Fauci why there were no clinical trials to answer what appeared to be such an important question. Even though the NIAID-sponsored feasibility studies had yet to be completed, Fauci told ABC News, in a story aired on January 31, 2001, that such a study would be "logistically impossible to do. . . . No one has yet been able to come up with a protocol" (Eisner 2001).
Some have suggested doing a WTS study in a developing country. But in places that can barely afford HAART or the necessary infrastructure, treating people with CD4 counts over 350, or even over 200 cells/mm3, may be a luxury they can't afford, even if some so-far undetected benefit actually accrues to such a strategy.
At the Retrovirus conference, one of the leading figures in the Adult ACTG told me that, having dismissed the idea of a randomized WTS study with clinical endpoints, the AACTG is now exploring the feasibility of (1) smaller randomized WTS studies looking at viral load, CD4 counts, and other laboratory parameters, and (2) establishing a larger observational cohort which, supposedly, could shed light on the question.
There are several problems with this approach. The smaller randomized study with surrogate markers simply wouldn't answer the question of whether people who start treatment earlier live longer or not. Obviously CD4 counts would be higher, and RNA levels lower, in the group that was treated earlier. But, this might not affect longer-term outcomes. The prospective observational study would be as expensive as a "real" trial, wouldn't answer the question any sooner than a randomized controlled trial, and would suffer from all the limitations of the observational studies described above.
It appears unlikely that any of the NIH-funded trials networks will do a controlled clinical endpoint study looking at WTS. It appears even less likely that such a study could be carried out in resource-poor developing country settings. Perhaps the Europeans, in conjunction with other developed countries such as Canada or Australia, may do such a study but clinicians in those countries already tend to start treatment later.
More likely, we'll have to continue to rely on observational studies with accumulating inferences, hints, and clues from smaller randomized studies, as well as new and emerging insights about the predictors and correlates of various drug-related adverse events, to guide the standard of care for the next few years.
As someone who has spent the last 12 years of my life trying to encourage more and better AIDS research, and worked with community groups to help push Congress and three presidents to provide more resources for that research, it's profoundly disappointing that the leaders of the research effort -- both at NIH and in the lavishly funded AACTG and the smaller, but still substantial CPCRA -- have signally failed to do anything more to address the question of when to start than to reluctantly replace one set of guidelines based on expert opinion with another.
|B.C. Cohort Mortality Data -- Effort to Find "Natural" When-to-Start Break Points|
|N (%)||24 months|
|Intermediate||50 to 199||301 (24.6%)||91%|
|Intermediate||50,000 to 199,999||443 (36.3%)||97%|
|European Cohorts -- Baseline Characteristics|
|Baseline CD4 Level|
|N||<200||200 to 349||<350|
|Swiss HIV Cohort Study||1,492||50%||25%||25%|
|Frankfurt HIV Cohort||701||45%||29%||26%|
|Gender (% female)||25%||26%||25%|
|Age at starting||37||35%||34%|
|Viral load||52 log 10||4.8 log||4.6 log|
|Phillips and colleagues assessed virologic outcomes by two measures -- the relative hazard of achieving an "undetectable" viral load (RNA <500 copies/mL) by 32 weeks, and, among those who achieved an undetectable viral load, those whose viral load went back above 500. (They did not appear to distinguish between cases where the RNA rise was a blip or when it was a breakthrough.)|
|European Cohorts -- Virologic Response by Baseline CD4 and RNA|
|Relative Hazard (RH) of RNA <500 copies/mL by 32 weeks|
|By baseline CD4|
|CD4 200 - 349||1.07 (0.96 - 1.20, p=0.23)|
|CD4 <200||0.88 (0.79 - 0.99, p=0.03)|
|By baseline RNA|
|RNA 10,000 - 99,999||1.03 (0.90 - 1.18, p=0.82)|
|RNA >100,000||0.70 (0.61 - 0.80, p<0.0001)|
|Only a baseline CD4 below 200 or an RNA above 100,000 was significantly predictive of a lower likelihood of becoming undetectable at 32 weeks.|
|European Cohorts -- Virologic Rebound (among Responders) by Baseline CD4 and RNA|
|Relative Hazard of VL >500 after suppression|
|By baseline CD4|
|CD4 200 - 349||1.12 (0.87 - 1.15, p=0.37)|
|CD4 <200||0.96 (0.74 - 1.24, p=0.74)|
|By baseline RNA|
|RNA 10,000 - 99,999||0.99 (0.72 - 1.36, p=0.94)|
|RNA >100,000||1.04 (0.76 - 1.43, p=0.81)|