The Death of Adefovir
"This is a very tough decision but I can't truly, in my heart, be convinced that there are enough data to support the safety or the efficacy of [adefovir dipivoxil] 60 mg," said Harlem Hospital's Dr. Wafaa El-Sadr at the FDA's Antiviral Advisory Committee's November 1st meeting reviewing Gilead's new drug application (NDA) for adefovir. The overwhelming majority of her colleagues agreed, voting 13 to one against approving the drug, the first time the advisory committee had ever said no to an antiretroviral. But their task had been daunting. The drug's NDA had been difficult to review because most of the data submitted pertained to its use at 120 mg per day, the dose that Gilead had initially planned to market. Tragically, however, toxicity to the kidneys had emerged in 30-40% of the patients treated for longer than six months at that dose. So in the eleventh hour of adefovir's clinical development, Gilead rewrote protocols to include a lower dose, 60 mg a day, in the hopes that it would be less nephrotoxic. Moreover, the company chose to seek approval for use only in people who had "failed prior therapy." But this only muddied the picture, as at least one of their supportive studies was not in that population. And the choice to switch to a lower dose was so late in the game that, when the NDA was under review, only 73 patients had been treated with the 60 mg dose for more than 48 weeks.
This led one activist, Michael Marco of the Treatment Action Group (TAG) to remark: "That is not enough. I cannot go back to my community and tell people that this is safe." Further complicating matters, many of the studies appeared to have conflicting or equivocal results, including the key study (Study 417) that sought to establish the equivalence of the 120 and 60 mg doses, at least in the FDA analysis. In the words of committee member Dr. John Hamilton of Duke University, "this study was confounded by a variety of circumstances, including [its] complexity, the possibility of drug interactions, the short term of the trial and missing data." "In the end," Dr. El-Sadr said, "I think this NDA was maybe prematurely submitted." Nevertheless, as committee chair Dr. Scott Hammer said before closing the meeting: "The sense of the committee is that this should not be a closed issue, that further studies and investigation should be performed . . . to establish the 60 mg dose as efficacious."
But little over one month later, Gilead announced that it was terminating development of adefovir for HIV in the US. "Although we respect the suggestions of the Advisory Committee and the FDA for further clinical research," said Gilead President and CEO Dr. John Martin in a company press release, "we believe it would be in the best interests of our shareholders and the HIV community if we allocate our resources to other development programs." The company halted new enrollment in all adefovir HIV studies (it will continue developing adefovir at a lower dose for the treatment of hepatitis B), and all patients currently in such studies were offered the option of receiving adefovir via the expanded access program. It will continue dispensing the drug to patients in that program indefinitely, but since most patients stop taking the drug within nine months (because of failure, toxicity or death), the program should effectively be over by the end of 2000.
Despite the fact that many activists and HIV clinicians had lobbied for approval, or at least further development of adefovir, the company announcement was met with barely a whimper from the HIV community. This was probably because few could question that Gilead's next candidate for approval, tenofovir, looks like a much better drug and that further development of adefovir for HIV could risk millions of dollars for what could turn out to be a very minor return. The money would be better spent getting tenofovir to market sooner. However, if this doesn't happen, or if the tenofovir early access program doesn't expand to fill the need no longer being met by adefovir, Gilead's relationship with the HIV community may grow more than a little rocky.
Since the initiation of clinical trials with adefovir, tens of millions have been essentially squandered on the drug's development. Thousands of people were exposed to a drug that, in the end, the FDA could not decide was more helpful than harmful. Gilead seemed to get so many things right -- studying its drug in salvage populations, setting up a large expanded access program and working closely with the community. There was also brilliant virological and resistance work; and when the kidney toxicity became apparent, the company rapidly disseminated information to health care providers about how to recognize the condition. How could it have gone so wrong? How could a company take a drug to the very brink of approval, and then simply write it off as a loss? How could the community, clinicians, federal government and advisory committees have spent so much time working with a company, developing studies and expanded access programs that enrolled thousands, only to have it produce an unconvincing NDA? And how can we all keep it from happening again?
Bad Timing, Bad Luck, Bad Listening Skills
Adefovir, a little drug from a little pharmaceutical company, first arrived on the scene in 1994. At a time when the protease inhibitors and the concept of "HAART" were making front-page headlines, this drug with only modest antiretroviral activity elicited little interest. However, when heavily pretreated patients began failing their HAART regimens because they needed new background therapies, people began to look around for anything they could get their hands on. And there was adefovir, a drug that was, in vitro, active against virus resistant to previously used nucleoside analogs.
And indeed, early dose-ranging monotherapy studies in patients who were mostly treatment experienced, reported reductions in viral load ranging from 0.5 to 0.6 log from baseline. Though not dramatic, this effect was retained out to 48 weeks and appeared to be even greater against virus containing the M184V mutation -- common after 3TC failure. (The company would expend considerable energy to demonstrate that adefovir's antiviral activity was more impressive in patients with M184V-mutant virus.)
But all the rules for conducting clinical studies in HIV disease had changed (with the advent of HAART) and continued to be in a state of flux over the course of adefovir's development. The days of large monotherapy studies were over, and so was the time when a company could argue for dual versus triple therapy studies. Gilead had to find other ways to convincingly show that its drug worked in combination regimens.
Gilead planned two studies (Study 408 and CPCRA 039) to show that adding adefovir (the name practically spelled out the planned niche for the drug) to ongoing, but not maximally suppressive, HAART regimens would improve responses. At this time, the FDA still expected companies to have clinical endpoint studies that could demonstrate a drug's clinical effectiveness underway by the time of filing, so one of these studies (CPCRA 039) would have to be rather large (2,000 plus) and in the sickest patients most likely to suffer clinical endpoints.
It all seemed fairly simple. Surely no one could complain that Gilead, in contrast with other companies, was conducting the outmoded two versus three drug studies. But in meetings with activists, some of us did protest that patients in these studies could be receiving inadequate therapy. Indeed, if their viral loads were detectable on HAART, patients would need to do more than simply add adefovir to their regimen: they might have to change the regimen altogether. This still hadn't become standard practice though, and the studies were practically underway.
In fact, the first study (408) enrolled quickly and in the end appeared to demonstrate that the addition of adefovir had an antiviral effect. Nevertheless, very few patients' viral loads went undetectable, demonstrating the inadequacy of the strategy. This should have been expected given that the average baseline viral load in the study was 30,000 copies/ml. Treatment intensification with adefovir (which could perhaps reduce viral load by 70% or half a log) might make sense in a patient with a low but detectable viral load. But a patient with a viral load of 30,000 would need around a two log (100-fold) reduction in viral load to go undetectable.
The cards were stacked against the second study, CPCRA 039. First, it took a very long time to enroll, partly because of a protocol mandate to do more intensive monitoring and blood work on the first several hundred patients. In the time that it took to get this initial cohort enrolled, adefovir's kidney toxicity had been recognized, serving as a further disincentive to enroll. More crucial in terms of demonstrating a clinical effect for adefovir, healthcare providers were increasingly accepting that the goal of therapy should be achieving an undetectable viral load. This meant that treatments had to be changed long before clinical endpoints would be reached, which happened commonly in the more advanced and treatment-experienced patients that enrolled in this study.
Of course, Gilead did the right thing by allowing changes in background therapy when necessary. But frequent changes in therapy no doubt contributed to the low clinical event rate, which in turn led the trial's Data Safety & Monitoring Board to close the study early since it could not answer the clinical question. Allowing changes in therapy also created a chaotic environment in which to show modest antiviral activity. And there were significantly more changes in the placebo arm relatively early in the study. The result was a comparison between adefovir plus suboptimal therapy versus a complete change of regimens.
What was surprising was that the earlier Study 408 had been able to demonstrate antiviral activity. In both studies, suboptimal therapy in some of the patients must have led to resistance to the more potent drugs they were on and subsequent jumps in viral load. This could have masked or obscured any effect adefovir had. For example, it would have been impossible to measure adefovir's activity in a patient experiencing a one to two log increase in viral load after becoming resistant to a protease inhibitor. Like trying to walk down an up-escalator -- you can take several steps down and still be at the top.
The failure of CPCRA 039 did not derail adefovir's development though, because in the meantime, the FDA had decided clinical endpoint data for antiretrovirals would no longer be necessary for full approval. Gilead increased the sample size of some of its other studies and moved on. Even so, the study went a long way towards discrediting the drug. Few stopped to think about its flawed methodology.
The kidney toxicity was unfortunate for everyone involved. Gilead's response to continue adefovir's development, albeit at a lower dose, deserves closer evaluation though. The case for dose reduction, which everyone seems to have accepted at face value, goes back to the very early dose-ranging studies in which no dose response curve was observed. The company first looked at doses up to 500 mg a day for two weeks and reported no dose response. The second compared 125 mg to 250 mg a day for 12 weeks and actually reported greater antiviral activity at the lower dose. This alone should have raised a red flag. The authors of the study (JID 1997;176:1517-23) suggested that perhaps there was less exposure to the drug in the 250 mg arm because there were more frequent dosing interruptions and increased time off the drug because of adverse events. That alone could have masked a dose effect, but another variable may well have confounded the ability of those studies to reach a reliable answer.
As noted earlier, many of the patients in the studies were nucleoside analog experienced. Most of that experience was on AZT. At the time, this didn't seem to be an issue, since adefovir was supposed to be active against AZT-resistant virus, but subsequently Gilead virologists discovered that patients who were highly resistant to AZT (conferred by the 215 codon mutation) did not respond well to adefovir treatment. Genotyping at baseline was not available for the patients in these studies, but the median duration of prior nucleoside therapy for those with treatment experience (over 70%) was over two years. If the patients who were AZT-resistant were excluded from the analysis, there may well have been a dose response. In fact, this was the conclusion communicated to this author by one Gilead scientist, who, before the kidney toxicity became apparent, felt that the company should re-evaluate higher doses.
It is interesting, then, that when Gilead performed the four-week placebo controlled study with adefovir 60 mg, it was in therapy naive patients. It is also telling that Gilead did not include a 120 mg arm as a comparitor in this study. The result? A 0.3 log reduction, which the company claimed was "similar" to the 0.5 to 0.6 log reduction seen with higher doses in the monotherapy studies in treatment experienced patients. Similar maybe, but it was also smaller.
This was probably the point at which Gilead should have re-evaluated its strategy. And one alternative to simply reducing the dose was known well enough. Their anti-CMV drug, cidofovir, is horribly and rapidly nephrotoxic, but this toxicity is minimized by increased hydration (IV saline) and the coadministration of probenecid. One poster at ICAAC (Abstract 1286) investigated the pathogenesis of the condition, technically proximal renal tubular damage (PRTD), and concluded that it was the same for adefovir and cidofovir. Both are substrates of (meaning they bind to) the human renal organic anion transporter 1 (hOAT1), which is expressed in high levels on the surface of renal tubular cells. High uptake of adefovir and cidofovir by these renal tubular cells leads to their death. But probenecid is an inhibitor of hOAT1 and protects these cells from cidovir and adefovir.
Gilead was understandably reluctant to study probenecid with adefovir because their experience using it with cidofovir had been miserable. It causes a litany of adverse reactions, including but by no means limited to headache, anemia, nausea, vomiting, rash and hypersensitivity reactions. But the dosage used with cidofovir was very high. Since adefovir is much less nephrotoxic than cidofovir, one would hope that the probenecid could be dosed much lower and perhaps intermittently, making it much more tolerable.
Treatment activists discussed this option with Gilead as early as two years ago, but resistance against probenecid was so strong that the company chose not to investigate it clinically. However, the company was preclinically investigating other potentially less toxic hOAT1 inhibitors. One was an over-the-counter nutritional product commonly used in Japan. There was some talk of perhaps using it with adefovir one day, but no real sense of urgency. Instead, Gilead bet that lowering the dose and only going for approval in the "desperate" treatment experienced population would be adequate.
But the 24 million dollar question was whether reducing the adefovir dose would reduce the side effects. In Study 417, the preliminary answer appeared to be yes, but for how long? At six months, four percent of patients in the 60 mg arm and 14% in the 120 mg arm discontinued therapy due to adverse events. Serum creatinine elevations, prognostic for PRTD, were also lower in the 60 mg arm: 29% had elevations over 0.5 mg/dl in the 60 mg arm, compared to 42% who were taking 120 mg. However, most PRTD cases appear after six months on treatment. It is not clear whether this difference should be reassuring considering the short follow-up in this study and the very high drop-out rate. Too short a term of follow-up was also the problem in an analysis of the safety of the lower dose in the expanded access program. The cohort enrolled on adefovir 60 mg was sizable (over 3,000), but there were data past six months for only 604 patients, and data past 48 weeks for a mere 43.
"Regarding safety," said FDA reviewer Dr. Kimberly Struble, "there is insufficient information on long-term administration of adefovir 60 mg. From the results in Study 417, it appears that the time to onset for creatinine and phosphate abnormalities is delayed for the 60 mg compared to the 120 mg group." However, there were no statistically significant differences for the frequency or resolution of nephrotoxicity between these doses. "In the expanded access program . . ." she continued, "approximately 40 percent of patients on the 60 mg arm compared to 50 percent of patients on the 120 mg arm will develop creatinine and phosphate abnormalities." But she questioned whether those differences were significant and noted that again, there were no data on the reversibility of the toxicity.
Gilead, meanwhile, thought it had characterized the syndrome well enough so that, with monthly monitoring, clinicians could identify the predictors of more serious liver toxicity. A number of HIV specialists, several of whom spoke at the hearing, concurred. Dr. Charles Farthing of the AIDS Healthcare Project noted, "We had no serious nephrotoxicity leading to dialysis, and I found it reasonably easy to instruct our 16 providers on how to monitor their patients, supplement their patients and discontinue when necessary." New York's Dr. Howard Grossman added, "There have been a number of people who have voiced concern about the ability of doctors in the community to administer this drug properly. I think it is time we all realized that all the drugs we are giving are as toxic and as difficult as cancer chemotherapy. We have a lot of responsibility when we are giving these drugs to know what we are doing."
Still her colleagues' experiences did not convince Advisory Committee member Dr. Judith Feinberg that adefovir was worth approving. "A lot of people spoke from personal experience, both the physicians and patients, but I too have given this drug to a lot of people, both in controlled clinical trials as well as in expanded access, and my experience is not quite as cheerful. I was almost starting to believe that the six patients with clinically significant problems -- that maybe half of them were mine. It is not an inconsiderable kind of toxicity . . . and I have not seen ready reversibility, so I am certainly concerned that we know that these drugs work if we are going to be offering them to patients."
But Dr. Feinberg seemed implicitly to agree with Dr. Grossman about one thing: doctors prescribe far more toxic drugs. The critical question was whether adefovir at 60 mg was active enough to justify its potential danger. Such a small virologic effect as a 0.3 log drop falls within the standard deviation of a viral load assay, and it would be difficult to know whether the drug was working, or had stopped, within an individual patient, not to mention a clinical trial. Said Dr. Hammer, "there probably is activity to the 60 mg dose [which] probably just crosses the threshold of biologic and assay variability . . . but can I say it has been conclusively demonstrated? The answer is no."
At 60 mg, adefovir's activity was little, virtually undetectable. Its toxicity, however, was not.
What Was Learned?
Gilead's problems with adefovir highlight some of the hurdles that drug companies with potentially useful but sometimes problematic drugs must now face to get their drugs to the HIV market. In order to qualify for accelerated approval, companies need to show that they offer meaningful benefit over existing treatments (and today, existing treatments for first line therapy work fairly well, at least over the first 24 weeks). On one hand, one would think that it would be easiest to demonstrate benefit in patients who are unresponsive or intolerant to HAART. But it can be very difficult to convincingly show that a drug with modest activity works in the jumbled context that is salvage therapy. "It is not so obvious to me how, in treatment advanced patients, you can cleanly do a placebo-controlled study that isn't going to have the same problems that the 417 study did," said Dr. Chris Mathews of UCSD, the sole advisory committee member to vote yes for adefovir's approval. "You are going to be comparing it to other drugs that are also active, [with] confounding drug interactions. I am just at a loss to see how that would work."
One rule of thumb is to know the potential drug interactions before launching major studies (see Gilead vs. FDA below). Granted, in Gilead's case, there was good reason for these to be unexpected since adefovir is not metabolized by CYP3A4, but it would have been better for Gilead to have been able to convincingly prove its case. Another word of advice is to try to keep the large pivotal studies as simple as possible in order to limit the number of possible confounding variables. A final thing to remember is that the standard of care in HIV disease is rapidly evolving. If it looks like it will be changing in the near future, it probably already has. It makes sense to use tomorrow's standard of care when planning clinical studies to ensure a relevant result when the trial is complete.
Finally, working on the forefront of technology can be a heady experience. Biotechs and other young companies spend half of their time selling themselves to investors -- everyone in the company has stock options and a vested interest in seeing that the company succeeds. After a while, they begin to believe in their own marketing hype.
But these companies should be more cautious. Things will go wrong. Despite brilliant bench science and better chemistry, any new drug will have side effects. Often, this comes as a complete surprise to people working at the company -- they don't fully believe it and they try to downplay it. But side effects aren't statistics to manage, they are people suffering.
Gilead had warnings that it was too soon to file and that its strategy would not work, but the company seemed to be blinded by its own self-confidence. Gilead got lost in the details of adefovir's development when it should have listened to the skeptics who saw the big picture and who told them it looked bleak. At least, the company might have been better prepared for the worst -- the FDA statisticians.
A Hypersensitive Virus? Or Was It Just Having a Bad Day?
Much of the case for using adefovir in experienced patients was based on research suggesting that the 3TC-associated mutation M184V made the virus more sensitive to adefovir.
A presentation at last fall's ICAAC by German researcher Veronica Miller confirmed and expanded upon these results (Abstract 433). Her team investigated adefovir's activity against a wider range of viral genotypes: 1) wild-type; 2) AZT highly resistant; 3) M184V only; 4) AZT highly resistant + M184V; 5) K65R (which confers resistance to adefovir) + M184V; 6) Q151M complex (which is a super-NRTI resistant strain) + M184V; and 7) T69S/insertion (associated with d4T) + M184V. They obtained 10-20 clinical samples of each strain and assayed them through Virco's phenotypic assay. They found that only the AZT-resistant and the K65R mutant viruses were resistant to adefovir and that the presence of the M184V mutation negated this resistance. All other strains were susceptible or hypersensitive to adefovir if that M184V mutation was on board.
A following paper struggled to explain this resensitization (Abstract 434). In vitro tests could detect no increased binding of adefovir diphosphate (the active intracellular adefovir metabolite) to M184V mutant reverse transcriptase. Instead the authors noted that the mutant virus simply replicated less well and concluded that that probably explained adefovir's enhanced effect. Which begs the question: is this really an adefovir effect, or a 3TC/184 mutation effect?
The FDA's analysis of the 408 virologic substudy raised a similar issue. In contrast to Gilead's analysis, the FDA found that those with the 184 mutation did better whether they were on adefovir or placebo. Furthermore, "the presence or absence of the 184 mutation did not affect the treatment difference for adefovir compared to placebo in patients with high level AZT resistance," said the FDA's Dr. Struble. "Perhaps the clearest result from the substudy is that patients with high-level AZT resistance alone will demonstrate cross-resistance to adefovir."
Back to the GMHC Treatment Issues Winter 1999-2000 contents page.