TAG at 20 – Part II: On a Darkling Plain—The Years of Despair

Before the Discovery of HAART

By Mark Harrington

And we are here as on a darkling plain
Swept with confused alarms of struggle and flight,
Where ignorant armies clash by night.
—Matthew Arnold, “Dover Beach”[i]

---

This is the second in a series looking back at the first two decades of TAG’s work to speed up AIDS research. In Part I: TAG’s early campaigns to reform the National Institutes of Health (NIH) AIDS research, boost the federal budget, and revitalize HIV basic science research. [ii] Here we look at the clinical science of AIDS before the discovery of highly active antiretroviral therapy (HAART) in 1995–96, and how TAG responded to the needs of people with AIDS.

---

For most of the early 1990s, it seemed that the science of HIV treatment was going backwards. In 1987, AZT became the first Food and Drug Administration– (FDA-) approved AIDS drug, giving hope to researchers and people with AIDS alike that this seemingly untreatable virus could be tamed. The NIH auctioned off two AZT-like drugs, ddI and ddC, to drug companies for development. In 1991, ddI (Bristol-Myers Squibb’s didanosine) was approved based on changes in a surrogate marker—a small rise in CD4 T cells in people receiving ddI, versus a continued decline among those on AZT—before the study’s final results demonstrated whether the CD4 cell increase predicted actual clinical benefit.

Early in 1992, the FDA formally issued regulations for accelerated approval to allow marketing of a drug for AIDS and other serious or life-threatening diseases based on such early changes in surrogate markers. It seemed like a great victory for the AIDS activists who had long struggled for such regulatory reform.

In April 1992, Hoffmann-La Roche submitted a new drug application (NDA) for accelerated approval of ddC (zalcitabine). I was an ad hoc community representative for the FDA Antiviral Drugs Advisory Committee hearing. It was not a pleasant task. The data on ddC were difficult to interpret. The drug had serious side effects including potentially crippling nerve pain. But a strong consensus for approval emerged among the AIDS activists at the hearing. I voted to approve ddC—not because it looked effective, but because of the desperate need for new treatments. A closely divided FDA advisory committee recommended a narrowapproval.

The benefits of these drugs were short-lived in the case of ddI, and nonexistent with ddC.

The difficulties of HIV treatment research in the early 1990s could be summarized as a vicious cycle caused by a combination of bad drugs, badly designed clinical trials, and inadequate measures or markers of anti-HIV drug activity.

Bad Drugs

First, the drugs. Virtually all the drugs under study in the early 1990s—ddI, ddC, as well as the subsequently approved d4T (stavudine, 1994) and 3TC (lamivudine, 1995)—were nucleoside reverse transcriptase inhibitors (RTIs). The nucleosides block the HIV protein reverse transcriptase from turning viral RNA into DNA by prematurely terminating the DNA chain, making it impossible for the virus to fully replicate. They differed, however, in their potency (anti-HIV activity), duration of effect, as well as in their toxicity.

AZT attacked the bone marrow, causing anemia. DdI, ddC, and d4T caused serious nerve damage, and sometimes pancreatitis. 3TC appeared the most benign of the first five nucleoside RTIs approved by FDA, but it was also the last to market.

By themselves, none of the nucleosides worked for very long against HIV. According to National Cancer Institute (NCI) virologist John Coffin, these drugs, used singly, caused “little” virus suppression. His measurements were based on viral RNA measurements—so-called viral load tests—which were not yet available in the early 1990s to most researchers, let alone to doctors and patients.

Badly Designed Clinical Trials

According to the well-known dogma of combination therapy for difficult-to-treat infections—when studies in the 1950s for tuberculosis made it possible for the first time to cure the disease—it is necessary to use two or more active agents against an organism that rapidly develops resistance to one drug used alone. As early as 1989, when ACT UP’s Peter Staley and I met with chain-smoking Burroughs-Wellcome virologist and AZT supremo David Barry at Research Triangle Park, North Carolina, he told us that the future of HIV treatment lay in combination therapy.

The federally funded researchers at the NIH’s AIDS Clinical Trials Group (ACTG) knew this. But most of them had worked only on herpes, a virus against which a single drug—acyclovir, discovered and brought to market in the early 1980s—worked for most people.

The ACTG’s early attempts at anti-HIV combination therapy would have been laughable had they not resulted in so many failed studies and wasted lives. Approaches such as one week of AZT alternating with one week of ddC were a recipe for the rapid emergence of resistance to both drugs. More often, they added a single new drug to an already-failing monotherapy. The most notorious of these studies, ACTG 155, compared AZT with ddC alone and with a combination of AZT + ddC in a group of people who had already been on—and most likely developed resistance to—AZT.

In spring 1993 came more bad news—the British-French Concorde study showed that early use of AZT (taken before the onset of symptoms and when CD4 T cells were still between 200 and 500 cells/mm³) didn’t provide any benefit over the longer study period.

In response, TAG intensified our work as a watchdog over the ACTG and industry studies of anti-HIV drugs, and deepened our expertise in clinical trial design, statistics, and HIV pathogenesis. We participated in and criticized the studies designed and carried out with public and private funds alike. We were members of key ACTG committees and met frequently with each company making a potential AIDS drug.

In June 1993, TAG and our colleagues witnessed the collapse of all the early hopes for combination therapy at the International AIDS Conference. As David Barr wrote for TAGline in 2003:

                  In Berlin, two central ideas at the heart of the treatment strategy were disproved. The first was that early use of AZT was beneficial. This was not a surprise, as the results from the Concorde study only proved what most people with AIDS on AZT found out the hard way: the drug stops working when used alone….The results of ACTG 155…showed that the two drugs did no better than one in helping people failing on AZT monotherapy. The bug was still mightier than the drug.

                  This was depressing enough. What made it even more depressing—and infuriating—was that NIAID [the National Institute of Allergy and Infectious Diseases] and the researchers involved in the study skewed their reports on the study results. Instead of presenting the results of the planned analysis, which showed the AZT+ddC combo as ineffective, an unplanned and statistically underpowered substudy analysis was performed which showed, in one of the arbitrary T-cell groupings, that patients on two drugs did “better” than the other patients.…Margaret Fischl….started presenting the post hoc substudy analysis, and we all lost it.

                  Mark Harrington, Gregg Gonsalves, Derek Link and I were all there, and we got up and just started to scream that this was a pack of lies. Dr. Fischl got all flustered. I got up from my seat and went to the microphone and started yelling that she was not telling the truth. She responded and had to admit that the planned results of the study were the exact opposite of what she was presenting. We continued to yell. The audience knew we were right and started applauding our comments. Finally, we all walked out of the auditorium.[iii]

As I wrote at the time:

                  After the Berlin conference,…[chief ACTG statistician] Steve Lagakos commented to one community member that “those activists wouldn’t be so mad [about the much-hyped “subset trend analysis” from ACTG 155] if the drugs were better.” Well that’s right Steve!—if the drugs were better, then the trials wouldn’t need to be bigger, or better designed, or analyzed more honestly—in fact, if the drugs were good enough, we might not need answers from randomized trials at all, as in the case of ganciclovir [DHPG]. But the drugs aren’t better—and that’s why we turn to statisticians in the hope that they will help design studies competently, and analyze them honestly, keeping in mind that the primary goal is the development of information useful to patients and their providers.[iv]

Indeed, in ACTG 155 it appeared that combination therapy was 50% more toxic than either monotherapy, but no more effective.

So by mid-1993 it was apparent that our hopes that early and limited CD4 cell increases would reliably predict clinical benefit were unfounded—since ddC had not shown the same benefits earlier seen with ddI—and that two drugs, at least in experienced patients, were no better, and possibly worse, than one drug.

No results were available yet on whether two drugs started at the same time—among people who had never taken anti-HIV drugs at all (combination theory postulated that two might delay resistance from emerging as quickly as it did with a only one drug)—were better than one. ACTG 175 was one such study, comparing AZT alone to ddI alone to AZT + ddI. Its results wouldn’t come in until late 1995.

In the meantime, d4T crawled forward as the next anti-HIV drug submitted for FDA approval, in mid-1994. The Alice-in-Wonderland quality of HIV drug development at that time was well captured in a New York Times story headlined, “F.D.A. Panel Recommends AIDS Drug Despite Incomplete Data”:

                  A panel of scientists recommended today that the Government approve a new drug to battle the virus that causes AIDS, but with strong warnings that doctors still do not know enough about the medicine to say who should take it.

                  An advisory committee to the Food and Drug Administration said the drug, d4T, or Stavudine, probably provided some benefit over the three existing AIDS drugs. But the panel could not say just who would benefit, just how safe it was or whether the manufacturer was on the right track to answer the many questions.[v]

TAG and its allies sent a delegation that was frankly split about the drug’s effectiveness, but in agreement that the data as submitted were impossible to interpret. TAG’s Gregg Gonsalves, who was the community representative on the FDA panel at that time, correctly predicted that the pivotal studies by d4T’s sponsor, Bristol-Myers Squibb, would be too small to prove that the drug worked. TAG’s Spencer Cox testified from the public, as did GMHC’s Derek Link. Neither could interpret the data clearly and both pointed out the flaws in the existing paradigm.

The next drug to engage the attention of the community was the first in a new class, the protease inhibitors. This drug was saquinavir. Unfortunately, the sponsor was Roche. At that time, I was a member of the ACTG’s primary infection (HIV treatment) committee. The negotiations between Roche and the ACTG were nontransparent, to say the least.

Roche had failed to find a maximum-tolerated dose of saquinavir, so we did not know whether the dose they were moving into phase II was the best possible dose in humans. The company refused to provide the committee with its preclinical or phase I data. Instead, it took the two principal investigators (PIs) into the hallway, showed them the data, and the PIs came back and told the rest of the committee that the saquinavir dosing data looked fine. They proceeded to enroll a couple hundred patients into another badly designed study, this time comparing two- and three-drug combinations in people who were already failing AZT. ACTG 229 enrolled 302 people and randomly assigned them to AZT/ddC/saquinavir versus AZT/saquinavir versus AZT/ddC (the combination that bombed in Berlin).

Later it became clear that the dose of saquinavir was far too low. Ninety-six percent of the drug was excreted unchanged in the urine, meaning that just four percent of it got into the blood and the body’s cells to block HIV replication. Even at an insufficient dose, however, saquinavir could produce enough antiviral activity to lead HIV to become resistant, not only to saquinavir itself, but to all the first-generation protease inhibitors.

Despite the inadequate dose, small study, short duration, and shoddy data, Roche, undeterred, went to the FDA in midsummer 1994 to request that it consider saquinavir for accelerated approval.

When Bristol-Myers sought accelerated approval for ddI in 1991, there had already been over 35,000 people on the drug through expanded access since 1989, and thousands of people had been in ddI studies over a five-year period. By contrast, in mid-1994, saquinavir had been studied in fewer than a thousand people for less than a year. Neither the safety data nor the efficacy—such as they were—of saquinavir looked as good as those for ddI alone had three years earlier.

This time, TAG drew the line. We were determined that the failures to study and regulate the approval of the nucleoside analogues not be repeated with the protease inhibitors, which looked like a much more promising class of drugs, even if saquinavir itself appeared to be relatively weak.

We wrote a letter to and obtained a meeting with FDA Commissioner David Kessler—an impassioned reformer who was the intellectual father of accelerated approval—and the agency’s HIV drug-review staff.

It wasn’t a friendly meeting. TAG and our allies criticized the FDA for failing to hold Roche accountable for not fulfilling its promised postmarketing studies for ddC, and for allowing Roche and other sponsors to conduct studies that were both too small and too short to show whether the drugs worked in the majority of people with HIV.

TAG presented a plan to address these issues for the protease inhibitors as a class by combining the best of expanded access with a better clinical trial design that was both large and long enough to show whether adding a protease inhibitor—or starting with one—was better than adding or starting with a nucleoside. Modeled after a series of successful cancer and heart disease studies, we called this a large, simple trial (LST).

TAG had rushed into the FDA meeting and the LST proposal without sharing our ideas quickly or broadly enough with the wider community. The death rate from AIDS was still climbing and in our haste to change the direction of the ship we neglected to inform or collaborate with our fellow comrades in the crew.

This haste backfired on us later that summer when Barron’s business magazine published a provocative interview with TAG’s Spencer Cox criticizing the saquinavir trials with the even more provocative headline, “Do We Have Too Many Drugs for AIDS?”[vi]

Quickly, TAG became the most unpopular organization in the AIDS activist world. During the late summer and early fall, we retrenched, putting forward our position in a report, Rescuing Accelerated Approval: Moving Beyond the Status Quo,[vii] which we distributed at a heated FDA advisory committee hearing in September 1994, where the debate reached impassioned levels.

As HIV gastroenterologist Donald Kotler stated in a dramatic scene from that era, documented in David France’s How to Survive a Plague, ACT UP was talking about access. TAG was talking about answers. We were talking past each other—but all of us needed both.

Luckily we found allies as well as detractors. One of the most notable new allies was the brilliant HIV-positive activist and virtually self-taught statistician Carlton Hogan, a New Yorker in exile in Minneapolis, Minnesota, where he worked in the statistical center for the NIH-funded Community Programs for Clinical Research on AIDS (CPCRA), a community-inspired and grassroots rival to the more academically slick and sometimes self-satisfied ACTG.

Inadequate Surrogate Markers

There was no immediate resolution to the crisis, because there still weren’t the scientific tools to better measure the anti-HIV activity of different drugs, combinations, and strategies. CD4 T-cell levels, while showing some correlation with antiviral activity, were an indirect marker, and different drugs in the same class appeared to have consistent effects on CD4 counts while clinical results diverged.

Early methods to measure HIV activity directly yielded conflicting results. One of the easier methods was to measure blood levels of the p24 protein, a component of HIV that was sometimes measurable at high levels. In many people, however, p24 was hard to measure, probably because it was bound to p24 antibodies, gumming up the test.

An even more demanding technique was quantitative co-culture—taking blood from the cells of a person in a study and measuring how fast that blood could infect cells in culture. This was difficult, varied from lab to lab, and could not be commercially standardized.

But in the earlier phases of research there was, suddenly, promise.

The polymerase chain reaction (PCR) method of measuring DNA or RNA sequences by multiplying their binding to a target genetic sequence was discovered by Kary Mullis and colleagues in 1983. By 1989, early forms of PCR were being applied to HIV research. In January 1995, David Ho and colleagues from New York’s Aaron Diamond AIDS Research Center (ADARC) showed in a study of quantitative PCR that unlike the nucleosides, more potent protease inhibitors such as Abbott’s ritonavir could reduce HIV levels by 99% (two logs of 10) within two weeks.[viii] (The weakest nucleoside, ddC, reduced HIV by only one-half log.) However, when protease inhibitors were used alone, drug resistance soon emerged as well.

Now researchers had a tool to measure—and to try to prevent—the rebound in viral replication that occurred so quickly after an initial decline on any single drug.

In February 1995, at a hearing of the National Task Force on AIDS Drug Development, TAG and GMHC delivered a series of recommendations to the manufacturers for each protease inhibitor. We recommended that Roche double the size of its pivotal saquinavir efficacy trials. We were scathing about Merck’s studies of L-735-524 (MK639, later indinavir), calling them “poorly controlled, badly designed, inadequately powered, and unlikely to provide useful information on the drug’s clinical utility.”[ix]

However, the activists praised Abbott for adopting a “novel ‘standard-of-care’ control arm [in its studies of ABT-538 (ritonavir)]…..[allowing p]atients to take any nucleoside analogue they wish, with the possible exception of 3TC…[and] then be randomized to receive either ABT-538 or a matching placebo.”[x] This control arm had originally been proposed by Spencer Cox of TAG.

Later in 1995, the first study demonstrating clear clinical benefit to combination therapy was announced. It was a by-now-retrograde comparison of AZT alone, ddI alone, AZT + ddI and AZT + ddC. For once, the researchers studied the regimens in people who were not yet already AZT-resistant. The results of the study, ACTG 175, showed that the combination of AZT and ddI was clearly better than AZT alone. However, ddI alone also appeared to be better than AZT alone, and the study wasn’t big enough to conclusively show that two drugs were better than one.

Another study with early results showed something even more promising. Abbott presented some tantalizing early results of ritonavir as monotherapy and in combination. As TAG reported at the time:

D. Norbeck.…described a French triple-combination study…Participants were given AZT/ddC/Ritonavir. Their CD4s went up by 110 and their plasma RNA went down by 2.5 logs at 20 weeks. Over the subsequent weeks, he claimed, an increasing proportion of participants became viral culture negative—which is to say, they could not culture infected cells from the blood. “Some became PCR and culture negative, which suggests that the viral reservoir was empty.”[xi]

We were so used to unsubstantiated or later-to-be-discredited industry—and, for that matter, academic and NIH—hype that we were instinctively incredulous at Abbott’s claim for the unprecedented ability of a triple combination including two mediocre drugs—AZT and ddC—plus the superpotent but also new and untried ritonavir, to render viral cultures negative or viral load in the blood undetectable.

The year ended with a whimper, not a bang. The FDA gave stavudine (d4T) full approval, despite the inadequately powered studies. It approved Glaxo Wellcome’s new me-too drug 3TC. And it gave accelerated approval to Roche’s saquinavir, the first protease inhibitor approved, and also the weakest.

In the United States, the wave of AIDS-related illness and death crested in 1995.

By 1995, just fourteen years after the disease was identified, the New York Times reported that AIDS had become the leading cause of death among Americans ages 25–44.[xii] In that year, cumulative U.S. AIDS deaths (311,381)[xiii] as reported by the CDC surpassed the total U.S. battlefield deaths in World War II (291,557).[xiv]

More Americans died of AIDS in 1995 alone (50,798)[xv] than on the battlefields of Vietnam during the entire course of that war (47,434).[xvi]

It didn’t seem like drug development was making a dent in the relentless piling up of bodies.