FDA Antiviral Panel on January 11 Endorses 16-Week Viral Endpoints for HIV Salvage Studies and FDA Sends Additional Messages
     By Mark Mascolini 

After wading for seven and a half hours through the boggy arcana of factorial trial designs, modified factorial trial designs, and two-part hybrid trial designs, the FDAís Antiviral Drugs Advisory Committee pounced on the principal questions of the day:

In the January 11 meetingís last half hour, with hardly a hem or a haw, nearly all HIV physicians on the panel argued for 16-week viral load data. Thatís long enough, they maintained, to show whether a drug has any antiretroviral pop. And with time at a premium for people with lots of treatment experience, it doesnít make sense to wait longer.  

As for endpoints, the committee quickly swept aside a suggestion that RNA and CD4 numbers should be weighed together. The drugs in question are antivirals, most argued, and an antiviral that doesnít rein in virus needs another name. Whatís more, the percent of study participants with an undetectable viral load is probably too stringent a marker for a population with a hoary treatment history. Rather, the advisers urged, a 0.5-log drop from baseline viral load may do just fine.

Community members on the panel worried that setting a 16-week finish line for accelerated approval of rescue regimen drugs would let drug companies off the hook when it comes to coughing up long-term safety data. And the FDAís antiviral chief, Heidi Jolson, agreed that six- and 12-month side effect numbers should not be lost in a rush to quick approval. But she advised drug developers that 48-week safety findings neednít come from a 16-week viral endpoint study.

ìTensionî defines the day
The 24-member panel summoned by the FDA to brood aloud about trial designs for HIV-positive people ìwho have limited therapeutic optionsî agreed unanimously on at least one point: ìTensionî could be called the watchword of this day-long give-and-take. Not that the meeting was particularly acrimonious. Rather, ìtensionî defined the poles that pulled these FDA advisors between the smooth certitude of ice-rock data and the rough-and-ready urge to get new drugs to floundering patients fast. And which panel contingents gravitated toward what pole could not be divined with an antique 20th-century compass.

From the FDAís standpoint, said the agencyís Katherine Laessig, the clearest tension seemed the unlikely prospect that any single new drug ìwill suffice as salvage therapyî versus the FDAís mandate to discern, ìfor regulatory purposes,î the contribution of single drugs to an antiretroviral regimen. Carlton Hogan (University of Minnesota) saw similar forces at play when he pondered the tension between drug access and approval, and between public health and drug approval.  

Such diametrics informed  the dayís debate, but of course they remained unresolved. And it took only a tautology to explain why. ìThe issues [Hogan defined] arenít separable,î maintained Jonathan Schapiro (Tel Aviv and Stanford Universities), ìbecause theyíre not separable.î So the best approach, he advised, is to dispense with hand wringing and to tackle the questions posed by FDA taskmasters.

1. What type of information would you most like to see from studies conducted in treatment-experienced adults and children?
2. What study designs are best suited to trials of new antiretrovirals in people with heavy treatment experience?
3. What are the most appropriate study endpoints for this population?
4. Can studies in this population be shorter than the 24-week trials accepted for accelerated approval?

Where we are and how we got here
The FDAís Heidi Jolson began the day by posing an embarrassing question: How did so many antiretrovirals get approved with such scanty data from treatment-experienced people? Jolson gingerly shifted some blame to the drug industry, observing that the FDA ìhistorically . . . has recommended that [antiretrovirals] be studied in broad patient populations and has provided for flexibility in endpoint.î

But drug developers have planned most pivotal trials in treatment-naive or modestly experienced patients--in other words, people with the best chance of showing off the new drugís prowess. The evolving understanding of resistance and cross-resistance only made matters worse, as manufacturers typically avoided pitting their drug against a virus that had already strong-armed a drug in the same class. Even when Abbott investigators tested the merits of lopinavir (ABT-378) in people with protease inhibitor (PI) experience, they were sure to add a nonnucleoside and to exclude people with nonnuke experience. So now we have 15 antiretrovirals but skimpy data on how to sequence regimens.

Douglas Ward illustrated the impact of these trial-planning trends on his HIV practice in Washington, DC. Eight of the 12 trials (67%) offered to his group for 2000 and 2001 wanted only treatment-naive recruits. Another two trial designs imposed forbidding limits on the kind of antiretroviral experience people could have, accepting, for example, only a single PI failure, or insisting on non-nuke naivetÈ. The only two studies that invited out-and-out experienced people were the tenofovir expanded access program and a trial of pegylated interferon. ìThe trials we need,î Ward concluded, ìare not out there.î 

This state of affairs hasnít changed much since 1999, when Jolson stood up at the Second International Workshop on Salvage Therapy and affirmed that studies of new drugs in salvage populations are ìclearly important--I donít need to be convincedî[1]. She insisted that instructing new drug sponsors to come forward with data in experienced populations was ìthe direction youíre going to see in the future in terms of developing guidance for industry both in preclinical and clinical development.î 

Jolson soon started making good on her word. In October 1999 she penned a letter putting drug companies on notice that the FDA ìexpected that most new drug applications for antiretrovirals submitted under the accelerated approval mechanism will include some clinical data in patients with limited treatment optionsî[2]. In fact, she went on, ìcombining multiple investigational agents may be necessaryî for such populations. Although the agency still wanted trial results that could pinpoint ìthe contribution of the investigational agent(s),î factorial study designs could probably accomplish that.

But industry has hardly battered down Jolsonís door with proposals for collaborative studies in people with few CD4 cells and fewer treatment options. In a review of trials for antiretroviral veterans, committee chair Roy Gulick (Cornell University, New York), could point to only one, Glaxoís study of their PI amprenavir and their nucleoside abacavir plus DuPontís non-nuke efavirenz--all of them unlicensed at the time.[3] And Glaxo planned that study long before Jolsonís letter.

Industryís appetite to study new drugs in heavily experienced people surely was not whetted by the FDA antiviral panelís 1999 dismissal of the nucleotide reverse transcriptase inhibitor adefovir, the first antiretroviral specifically slotted for a rescue regimen niche and studied intensely after other drugs had failed.

Meanwhile, the rolls of people seared by triple-class failures continue to expand. Douglas Ward figured that 12% of patients in his Washington HIV practice have worrisome viral loads and ìneed salvage.î But Ward canít piece together rescue regimens because these people have used up their treatment options. ìI donít have anything to offer them,î he lamented.

Wardís 12% looked mighty good to other clinicians sitting around the table. Michael Saag, who runs a 1000-patient clinic at the University of Alabama in Birmingham, counted 70 AIDS deaths in this cohort last year, an unhappy escalation in mortality from the clinicís best HAART years. Saag said he no longer worries about avoiding resistance or pushing viral loads under 400 copies/mL in lots of patients. ìThey already have resistance,î he explained. Now heís happy to get just a half-log viral load drop in the most experienced people, and to keep it.

Working at San Francisco General Hospital, Steven Deeks agreed that more and more people are sliding into the triple-class failure category. Some continue to do well clinically. But others ìare getting sick and dying,î and not from liver failure or lactic acidosis. Theyíre dying from AIDS.

To get a look at the big picture, Gulick reviewed recent cohort studies that tallied virologic successes and failures. In five cohorts including over 2700 people who began a potent regimen with little or no treatment experience,[4-8] proportions who remained above limits of detection ranged from 38% to 63% after 48 weeks to 2 years of follow-up. In three of the five cohorts, half or more had detectable viremia.

The news was grimmer in a study of 266 people with three-class experience in the EuroSIDA cohort.[9] All had taken two or more HAART regimens before starting a salvage combination. Only 40% pushed their viral loads below 1000 copies/mL, and only 30% kept them there for 6 months. After 1 year of follow-up, 55% had fewer CD4 cells than they started with, and 5% had a new AIDS diagnosis or died within a year.

One-drug add-ons and two-part hybrids
Against this cheerless backdrop, the FDAís Division of Antiviral Drug Product pressed its advisory committee--and sundry august guests--for advice on trials of investigational antiretrovirals in people with heavy treatment experience. Division chief Heidi Jolson made it plain that the agencyís main interest lay with ìregistrationalî trials rather than ìstrategyî trials. Registrational studies, she explained, are ìadequate and well-controlled pivotal trialsî that permit statistical analysis of an individual drugís antiviral action. But the FDA also sought advice on strategy trials--ìexploratory studies to assess efficacy of a regimenî--because such studies can contribute data to a drugís label.

Jolson also defined what she meant by the ìheavily treatment-experiencedî people these trials should enroll. Study participants must have sampled at least one agent in all three current antiretroviral classes, and they must have endured a ìloss or lack of virologic responseî to at least two regimens that could be called HAART.

Jolsonís colleague Katherine Laessig voiced the hope that study populations would include people with CD4 counts under 50 cells/µL and viral loads above 100,000 copies/mL. She added that definitions of treatment failure might also include problems with tolerability, adherence, and pharmacokinetics (drug absorption, distribution, metabolism, and excretion).

Laessig noted the general consensus that resistance testing should help pick ìoptimal background regimensî in these trials. Ideally, she urged, drug-drug interactions should be reckoned before new agents are combined. The University of North Carolinaís Joseph Eron evinced a knack for mnemonics in making this same point, saying trial planners should strive to eliminate ìwoops reactions.î

The advisory committee spent most of its time holding a hand lens up to trial designs, and here panelists tiptoed along the shore of a seemingly infinite sea. The species, subspecies, and quasispecies of clinical trial designs appeared limited only by the number of speakers panel chair Roy Gulick dared to recognize.

Attendees heard tell of the simple add-on study, the randomized add-on, the factorial and the modified factorial, the two-part hybrid, and the bridged phase I-III randomized controlled trial. A few undaunted clinicians proposed nameless variations on these heady constructs. When panelist Brian Wong (VA Connecticut Health Care System, West Haven) beseeched Martin Schechter (St. Paulís Hospital, Vancouver) to explain, once more, the Latin squares design, one sensed little sympathy for his curiosity.

But the committee shouldered this task bravely. And as soon as one got past the increasingly forbidding names, the trial concepts became clear to anyone who understands the ABCs--because thatís as many letters as it took to outline the toughest design.

The simple single-drug add-on design needs no letters at all. The study drug just gets tacked on to whatever a person is already taking. But this simplest of strategies is also the weakest, because the results can be compared only with what might have happened without adding the study drug. Clinical trial literature gives ìwhat might have happenedî a fancy name--historical controls. In other words, you compare study results with what happened in a similar group of people taking something similar to the baseline regimen without the add-on drug. Nobody likes this design, if it can even be called a design, because the comparison barely rises above guesswork.

The single-drug approach gains in sophistication when investigators randomize study participants to take the new agent or to take a look-alike placebo. Gilead used this strategy in testing both adefovir and tenofovir in antiretroviral-experienced individuals. The studies enrolled people taking a ìstableî regimen that wasnít choking off viral replication. Adefovir edged out placebo in lowering viral load,[10] but it stumbled on the way to approval because of side effects. Tenofovir opened up a wider virologic gap from placebo [11] and has not matched adefovirís uncomely side effects profile.

The most outspoken community representative at the meeting, the University of Minnesotaís Carlton Hogan, warned that most treatment-experienced trial candidates have no interest in more studies of this ilk. When someone is already taking a wobbly regimen, adding just one drug is ìinherently futile,î he argued.

  ìIt is our strong contention,î Hogan went on, ìthat any designs that offer only one new agent are unlikely to succeed and may compromise future use of that agent in a combination.î 

Committee clinicians proved less ready to toss add-on trials into the dust bin, at least not for the most desperate patients. After recounting the fast-advancing disease of some people he treats in San Francisco, Steven Deeks argued that these folks may benefit from fast access to even one new drug with a novel mechanism, such as the fusion inhibitor T-20. If thatís the best medical science can offer right now, he said, it should be offered. 

The FDAís antiviral chief, Heidi Jolson, cautioned that single-drug studies, if done at all, shouldnít be single-arm studies. Without a control arm, she warned, data from such studies are not what companies need to make ìlawful claimsî about their drug.

But Michael Saag noted that one-drug studies donít have to be one-arm studies, or even simple add-on studies. First the background regimen can be spruced up with the help of resistance testing. Then study participants can be randomized to take the study drug or placebo. The big caveat, Saag added, is that scrupulous follow-up must be possible so that placebo takers can be shunted to the study drug at the first hint of failure. 

Katherine Laessig of the FDA gave the agencyís take on just this type of trial. Randomization and blinding are ìpreferred,î she noted, and the risk to participants can be limited by an ìearly escape option,î just as Saag proposed. But Laessig left little doubt about how the FDA rates single-drug add-ons, calling such designs ìless desirable than modified factorialî plans. 

Ralph DeMasi, a statistician from T-20 developer Trimeris, ventured a variation on this theme, the so-called bridged phase I-III randomized control trial. That fearsome moniker means the study will incorporate a phase I dose-finding period, a phase II safety stint, and a phase III efficacy stage--all for a drug designed for rescue duty. The trial would randomize people to three arms, an optimized background regimen, dose A of the study drug, and dose B of the study drug. This phase I would last only 1 to 2 weeks, then everyone would add the best background regimen. At specified points, everyone taking the weaker dose would move to the stronger dose, and people taking just an optimized background could add the study drug. 

University of Minnesota pharmacologist Courtney Fletcher liked this idea. He rated himself a ìstrong proponentî of the design because it would let researchers do some crucial pharmacokinetic analyses early on--a big plus because such data are woefully lacking in people with heavy treatment experience. The drug concentration that translates into an effective dose in experienced patients, he explained, is not the concentration that works in the treatment naive. Without understanding that difference, Fletcher said, trials of drugs for experienced patients are ìdoomed to failure.î 

The University of Pittsburghís John Mellors had less enthusiasm for DeMasiís master plan. The 2-week two-dose lead-in may make pharmacokinetic sense, he allowed, but ìyou have to make absolutely sure you havenít done damageî to patients in that monotherapy phase. And the overall concept of a three-phase trial in one study struck Mellors as ìawfully ambitious.î 

A second add-on design broached by the FDAís Katherine Laessig was the two-part hybrid. The two parts that get hybridized in this plan, she explained, are an initial three-arm, 10-day randomization phase, followed by 22 weeks of prospective observation. As in DeMasiís bridged phase I-III concept, the two-part hybrid can directly assess antiviral activity during a brief lead-in. Hereís how it works: 

Participants get randomized to one of three arms: study drug plus their current regimen, just their current regimen, or a new optimized background regimen. After day 10, everyone gets the optimized regimen plus the study drug. The first 10 days allow investigators to gauge how much the study drug adds to current combinations, and follow-up through week 24 lets investigators measure durability.  

This designís feasibility rests on two assumptions, Laessig explained: that resistance to the new drug doesnít emerge during the initial 10-day combination with the current regimen, and that people continuing just their current regimen wonít suffer ìadverse consequencesî because they stick with a failing combo.  

Tel Avivís Jonathan Schapiro proposed that the two parts of the hybrid design should have different lengths for different drugs, depending on how quickly a drug elicits resistance-conferring mutations. The lead-in period might vary from 1 to 4 or 5 weeks, he suggested. But Schapiro doubted that non-nucleosides resembling current drugs in that class, with their half-inch-high resistance thresholds, could ever be studied safely in such a trial. If anyone mounts a hybrid trial, he added, it may pay to gauge the slope of viral decay during the lead-in phase to see if that yardstick can measure antiviral effect in heavily experienced individuals.

But the two-part hybrid drew raves from no one (and, yes, many found the name creepy). For the University of North Carolinaís Joseph Eron, the design smacked too much of a dissatisfying T-20 trial. That study tested T-20 alone in treatment-experienced people, then added an optimized background.[12] Clinicians who heard this trial presented remained unconvinced that it proved anything about T-20ís contribution to the added background therapy, Eron noted.

As talk of one-drug add-ons rolled along, community folks at the table--and then the  FDA--betrayed signs of impatience. Lynda Dee, from AIDS Action Baltimore, reminded co-panelists that add-on studies are ìthe most attractive to industry, but the least attractive to patients.î Except for people at the end of their CD4 rope, she said, no one will risk a study that may randomize them to sugar pills. 

Why even give industry the convenient escape hatch of a one-drug study, wondered the FDAís Jeffrey Murray, when you might press for a study of two or even three new drugs with a multifactorial design? Murray minced no words:

ìIíd hate this meeting to end with companies being off the hookî to put together two- or three-drug trials, he said. Companies with the will to try such trials could pick from a long list of collaborators with phase II antiretrovirals. ìIíd hate to go away from this table without companies hearing that,î Murray pleaded. 

How many factors can you fit in your factorial? 
The factorial designs discussed by the advisory committee came in two flavors, standard and modified. These trial plans can best be understood by plotting grids that spell out the separate factors being studied. For example, explained Vancouverís Martin Schechter, a 2 x 2 factorial design mixes four factors in hopes of divining how different mixes of factors--such as individual drugs, regimens, or strategies--will affect outcome. Schechter and his colleague Julio Montaner are planning a 2 x 2 factorial study called OPTIMA that will enroll people with multiple regimen failures. This grid lists the four factors to be studied:

OPTIMA: a 2 x 2 factorial design 


(4 or fewer drugs) 

(5 or more drugs) 

Treatment interruption

No treatment interruption


By randomizing a certain number of people to each cell (where the Xs are now), OPTIMA should tell Schechter and colleagues not only how mini-HAART salvage compares with mega-HAART salvage, but also how each salvage option works with or without a presalvage treatment break. As long as enough people get randomized to each arm, the advantage of this scheme is that it answers several questions in a single trial.

People familiar with the brief history of HIV salvage studies have already witnessed a 2 x 3 factorial design in ACTG 359.[13] In that trial Cornellís Roy Gulick and co-workers randomized indinavir-experienced, nonnuke-naive people to fill the following factorial cells:

Percent below 500 copies/mL after 16 weeks of therapy: ACTG 359ís 2 x 3 factorial design




Delavirdine + adefovir 

Saquinavir + ritonavir 




Saquinavir + nelfinavir 




The results told investigators that ritonavir and nelfinavir performed equally well as partners with saquinavir in this population. But delavirdine alone or delavirdine plus adefovir beat adefovir alone in lowering viral load. (comments from Jules Levin: findings from 359 were not very useful, but not necessarily due to use of factorial design but the drugs available at the time were limited. And more creative randomizations would be helpful).

Schechter maintained that factorial designs are ìideally suited when multiple therapies exist that can be given in different combinationsî and ìwhen different strategies can be combined.î Factors that could be thrown into a factorial mix, he suggested, include drug combinations, genotyping, treatment interruptions, adjunctive therapy, and immunomodulators. The symmetry and efficiency of factorial schemes, Schechter argued, give investigators ìmore bang for the buck.î (Comments: my concern is that patients don't get sub-optimal randomizations).

But factorial studies of HIV therapy have been almost as rare as henís teeth. Schecter turned up only three in a search of the literature. Why? Industry hasnít bought the idea, for several reasons summarized by the FDAís Katherine Laessig:

-- Unanticipated drug-drug interactions can upset the results.
-- Overlapping toxicities can complicate planning.
-- The right types of drugs have to be available at the right time.
-- If the trial is planned to license one or more drugs, when itís all over, who owns the data?

Laessig countered that ìindustry concerns are valid but not insurmountable.î Factorials would allow researchers to fashion randomized trials ìthat participants are more likely to complete.î And, if two or three companies agree to cooperate, costs can be shared and no one has to fret over getting placebos from a competitor.  

But the panel struggled with several dangers they saw hiding behind those tidy factorial grids. Colleen Cunningham (State University of New York, Syracuse) suggested that treatment-experienced people with some ìcruise timeî might be willing to enroll in a 2 x 3 trial like ACTG 359. But folks with more advanced disease, figuring they need as many new drugs as they can find, might not be keen about a trial in which they could get two new PIs and only delavirdine or two new PIs and only adefovir, for example, when what they really want is two new PIs and two other new drugs.  

Pittsburghís John Mellors seconded this opinion, driving home his point with an overhead chart. In a 2 x 2 placebo-controlled factorial trial sizing up two new drugs, he figured, study participants have only a 25% chance of getting randomized to the two-drug arm: 


New drug A 

New drug A placebo 

New drug B 

Drug A +
drug B 

Drug A placebo +
drug B 

New drug B placebo 

Drug A +
drug B placebo 

Drug A placebo +
 drug B placebo 

Those odds are too steep for people with dwindling options, he argued, and they get steeper in 2 x 3 designs and 2 x 4 designs. 

But the University of Minnesotaís Carlton Hogan, who has seen his CD4 count bottom out at 1 cell/µL, argued that salvage studies shouldnít necessarily aim to ìpile on as many new drugs as possible.î As someone who has witnessed the darkest depths of immunosuppression, he said, his goal is to take the minimum number of drugs that will work (and lower the risk of multidrug toxicity at the same time). Factorial designs, Hogan insisted, offer the best hope of finding the minimum number of effective drugs. 

With ACTG 359, for example, that proved true. As a group, people who got the two PIs plus delavirdine and adefovir did no better virologically than those who got the PIs and just delavirdine. But Mellors remained unswayed. If you get too fine with your trial design and end up undershooting the number of new drugs a study participant needs, he said, ìthatís it.î (Comments: yes, the percent undetectable in all the arms of 359 was very disappointing and again likely due to the lack of adequate available new effective drugs). 

Speaking for the Intercompany Collaborative (ICC) for AIDS Drug Development, a consortium that includes nearly all the key antiretroviral makers, Gileadís Jim Rooney also backed away from factorial designs in registrational trials. He reiterated several of the concerns Laessig summarized (above), and added that factorial studies make it difficult to attribute side effects to individual drugs. Rooney said, though, that the ICC endorses combined studies of two unlicensed agents developed by different companies. 

If factorial designs lack industry and academic proponents, will they ever get done? Modified factorial designs may be a better bet. The panel focused on a modified factorial design spelled out by the FDAís Laessig, an ambitious scheme that includes three investigational agents (A, B, and C below) and no placebo controls. Treatment-experienced people would be randomized to four arms:

1. Optimized background regimen (OBR) + A + B
2. OBR + A + C
3. OBR + B + C
4. OBR + A + B + C 

Michael Marco, representing New Yorkís Treatment Action Group (TAG), advanced an even more ambitious modified factorial plan that would also weigh the merits of three unlicensed agents, but with placebo controls. Marco dispensed with the ABCs and suggested a few real antiretrovirals nearing the end of the development pipeline:

1. OBR + T-20 + tenofovir + tipranavir
2. OBR + T-20 + tenofovir + tipranavir placebo
3. OBR + T-20 + tenofovir placebo + tipranavir
4. OBR + T-20 placebo + tenofovir + tipranavir

Right away you can see that investigators lacking ambition need not apply for the principal investigator post of such trials. The advantages, on the other hand, are obvious. Laessig figured that the modified factorial she proposed would take one third the number of participants needed if the makers of drugs A, B, and C conducted separate studies. Another big plus is that everyone would get at least two new drugs, and 25% would get three. 

But will such a complex design allow investigators to pinpoint the antiviral contribution of individual drugs? Yes, argued the FDAís Tom Hammerstrom, even in Laessigís no-placebo plan. Each of the first three arms, he explained, lacks one of the three study drugs. If enough people enroll in each arm, group-to-group comparisons should tease out the antiviral impact of A, B, and C. But that assumes there are no big drug-drug interactions between any of the matched agents, noted the University of Alabamaís Michael Saag.

John Mellors reiterated Rooneyís concern about toxicity: Will a design like Laessigís, with no placebo controls, let investigators assign toxicities to specific drugs? Yes, suggested statistician Victor DeGruttola (Harvard School of Public Health, Boston), again because each of the first three arms lacks one drug. So if a certain toxicity turns up in the A + B arm and the A + C arm, but not in B + C, drug A will look like the culprit. But he agreed that ìyou clearly lose something [in the ability to assign toxicities] by lopping off the placebo.î And throwing out placebos, DeGruttola added, chisels away some statistical power to compare the antiviral punch of each drug.

Although the panel didnít specifically address TAGís placebo-controlled factorial plan, adding placebos apparently would make the trial statisticianís life easier. But placebos sure wouldnít make the study participantís life easier, because it would inflate the pill count, or, in the case of twice-daily subcutaneous T-20, compel some people to inject placebo.

Something else bothered clinical investigators sitting around the table more than pill burden. Some of them simply couldnít conjure the image of three pharmaceutical companies signing up for a complicated study with two rivals. Even if one company has two new drugs ready to go, as Glaxo did with amprenavir and abacavir, youíd still have to get a second company to play ball.

ìMy problem with [three-drug] modified factorial designs,î said San Franciscoís Steven Deeks, ìis that I just donít see it happening.î As for the standard factorial, he added, ìIíd prefer to let patients cruiseî if theyíre not desperate for an antiviral lift. And for those with fewer than 50 CD4 cells, Deeks still favored a trial comparing optimized background therapy (with a quick escape option) versus optimized background and a new drug.

But a modified factorial design ìdoesnít preclude a quick escape,î said the FDAís Hammerstrom.

ìThat presupposes that you can switch quickly enough,î rejoined John Mellors.

ìThatís not a problem with the design,î countered Hammerstrom.

ìBut thatís a problem with the disease,î said Mellors.

Michael Saag attempted to sort through some of these conundrums in what may have been the most complex plan of the day, yet one that might appeal to people with different levels of treatment experience, and to their clinicians. Saag started with the entry criteria spelled out by the FDAís Heidi Jolson: virologic failure of two HAART regimens and experience with at least one drug in each antiretroviral class. Then everyone gets a resistance test and the cohort gets split in two:

People with virus susceptible to at least three current antiretrovirals get randomized to four arms:
1. Optimized background regimen (OBR) + A
2. OBR + B
3. OBR + C
4. OBR + some placebo

People with virus susceptible to fewer than three current antiretrovirals get randomized to four arms:
1. OBR + A + B
2. OBR + B + C
3. OBR + A + C
4. OBR + A + B + C

This strategy allows people who still have reasonable options among licensed drugs to take quick advantage of one new antiretroviral. People with fewer options among approved agents would get at least two new drugs. But again the design rests on the sizable assumption that three drug makers will step forward with three investigational agents.

Panel endorses 16-week RNA endpoint
What endpoints should investigators aim for in trials of new agents that enroll experienced people? W. Christopher Mathews (University of California, San Diego) called for both viral (RNA) and immunologic (CD4) markers, because accumulating data indicate that people often maintain CD4 gains well after their viral load starts rebounding. But how to mesh those markers in studies of rescue regimens was a tough question. And the other clinicians who addressed this issue werenít sure that question even needed an answer, because they believe a viral load marker alone will suffice as a primary endpoint in such studies. 

The University of North Carolinaís Joseph Eron argued that requiring both an early RNA and CD4 response isnít necessary in this population. When peopleís CD4 counts go up after they start a rescue combo, thereís always some RNA drop, he said. So the studies being discussed could keep it simple and focus mainly on RNA, perhaps after 8 or 16 weeks of the new regimen. 

In a hint of FDA sentiment, the agencyís Jeffrey Murray agreed. ìIn my mind I would really be looking for a drug with antiviral effect,î he said, if a company planned to position that drug for rescue regimens. Even if an antiviral did boost CD4s in some people without punching down the viral load, such a result would be ìpretty murky.î 

But what viral marker makes sense? Should a new regimen in experienced people flatten viral loads below the point of detection? Everyone, beginning with the FDA, agreed that would be too stern a test. Summarizing pre-meeting input from industry and the community, the agencyís Katherine Laessig suggested that measuring the ìproportion undetectable may not be feasible,î except perhaps ìfor multiple investigational or highly potent agents.î Instead, many who responded to the FDAís pre-meeting request for comments proposed that rescue trials might focus on the mean change from baseline RNA, or the proportion of participants with a certain log drop in viral load. 

If a study uses mean change from baseline RNA, Eron wondered, is the traditionally cited 0.5-log drop enough to suggest some benefit? Harvard number-cruncher Victor DeGruttola cautioned that the 0.5-log benchmark derives from early antiretroviral studies and could bear verification with more recently approved antiretrovirals. But Birminghamís Michael Saag doubted that a 0.5-log drop with AZT, for example, means something different from a 0.5-log drop with some newer antiretroviral. The drugs may have changed, he said, but ìthe biology is the same.î Saag considered a 0.5-log decline sufficient for the studies being discussed. 

Steven Deeks reinforced this argument with a sensible observation. Most rescue combos canít make viral loads vanish in people with the heavy experience defined by the FDA. HIV docs have understood this for some time. So the goal with these people is just to get their viral load down. If a viral load drop is a reasonable goal in the clinic, he observed, itís a reasonable study endpoint too. Deeks added, though, that drug trials in experienced patients should also try to pinpoint predictors of virologic failure, changes in viral fitness, and correlations between RNA and CD4 changes. 

No one challenged these arguments for falling viral load as a primary study endpoint. But several panelists suggested secondary endpoints. Clinical endpoints--new or recurrent AIDS conditions and death--are reasonable in any population with advanced HIV disease. Deeks himself made this point in an editorial on Julio Montanerís mega-HAART study.[14] Writing with his colleague Jeffrey Martin, Deeks noted that the ìhigh clinical event rateî in this study (two AIDS diagnoses and seven deaths among 106 people about a year) ìsuggests that controlled trials of multiple drug rescue therapy with clinical outcomes may not require unduly large sample sizes.î[15] 

The FDAís Laessig observed that the agency hopes for ìbetter collection and adjudication of clinical endpoints, regardless of the primary endpoint or patient population.î Some who responded to the FDAís pre-meeting request for comments suggested that new CDC class C events, such as MAC or CMV infection, could be counted in such trials. 

Three panelists, NIAIDís Carla Pettinelli, the University of Minnesotaís Carlton Hogan, and NATAPís Jules Levin called for close scrutiny of drug toxicity, if not as a precise endpoint, then at least as a way to define risk-benefit ratios. TAG representative Yvette Delph urged that adherence with therapy be measured, and that adherence interventions be planned before and during trials of third- and fourth-line therapies. And committee chair Roy Gulick suggested weighing some quality-of-life measure.

With general consensus on at least a 0.5-log viral load drop as a primary study endpoint, discussion turned to the FDAís final question: Do trials of rescue therapies have to last 24 weeks--the standard for accelerated approval--before viral endpoints can be confidently evaluated? The FDAís Katherine Laessig asked the advisory committee to ìconsider earlier assessment of antiviral effect for this population with longer term safetyî data. And thatís exactly what clinical investigators on the panel recommended. 

University of Minnesota pharmacologist Courtney Fletcher advised that 8 to 16 weeks is long enough to tell if a drug ìbehaves as an antiretroviral.î But a reasonable handle on safety, he said, would come only with longer follow-up. And how long is a hard question because people with heavy treatment experience generally need higher antiretroviral concentrations to challenge often-resistant virus. So benchmarks set in trials of treatment-naive people donít apply. 

Pittsburghís John Mellors cautioned that getting good safety data from a salvage study would be even tougher than that. In fact he called it ìincredibly difficult.î The problem is that valid safety results can be collected only when enough people stay in each treatment arm. But in people with advanced disease, an ethical trial has to promise fast switches to other drugs if the assigned treatment doesnít work. So investigators may well find themselves settling for shorter-term safety data.  

This dose of candor didnít go down well with community representatives on the panel. Baltimoreís Lynda Dee voiced some anxiety about a 16-week viral load endpoint, because if an antiretroviral gets licensed as a ìrescue drugî with 16-week data, nothing can prevent clinicians from using it in first-line combos if they want to. But Dee said she might be able to stomach 16-week endpoints if drug makers could be compelled to track safety data for a good while longer.  

TAGís Yvette Delph was less accommodating. She opposed 16-week endpoints, saying it would be better for treatment-experienced people to get new drugs through expanded access programs. Faster access may be a worthy goal, Fletcher countered, but it doesnít yield much useful data. And the University of Alabamaís Michael Saag added that even 48-week safety data wonít reliably turn up some of the most troubling side effects.

FDA antiviral chief Heidi Jolson sided with the community. ìIt would be enormously difficult to conceive of a situation where we donít have some [safety] data past 16 weeks,î she warned. But she had a card up her sleeve that drug makers should like. Yes, ìsome percentageî of study participants has to be followed much longer than 16 weeks to track side effects. But that ìcertain percentageî doesnít have to come from the same study that provides the 16-week RNA endpoints. 

A ìclear statementî from the FDA?
FDA operatives picked a prime site for this meeting on salvage trials, the luxurious Versailles Ballroom of the Holiday Inn, Bethesda. If the rapt industry-heavy audience managed to overcome its awe at the glitzy chandeliers, plush rugs, and Louis Quatorze accouterments, it could not have missed several crystal clues to the FDAís leanings.

Jeffrey Murrayís efforts to focus discussion on cooperative multi-agent trials, and Heidi Jolsonís roadmap to gathering safety data were only the most obvious examples. There was also this nugget in Katherine Laessigís ìRegulatory Conclusionsî slides:
Important to consider strength of NDA package: One controlled, traditional study plus other well-designed studies in H.T.E. [heavily treatment-experienced patients] may be preferred over 2 identical studies in naive and less rx experienced patients
Whether that ìmay be preferredî becomes ìwill be preferredî is a worthy topic for discussion in the research units of all companies developing antivirals. Those researchers will also want to consider whether TAGís Michael Marco got it right when he said, ìToday is the day when industry can stop saying that the FDA doesnít give us a clear statementî on salvage studies.

Mark Mascolini writes about HIV infection (mailmark@ptd.net).

1. Comments by Heidi Jolson transcribed by this reporter at the Second International Workshop on Salvage Therapy for HIV Infection, May 19-21, 1999, Toronto.
2. Jolson H. Untitled letter to antiretroviral developers. Available at: http://www.fda.gov/oashi/aids/ind.html. Accessed January 13, 2001.
3. Falloon J, Piscitelli S, Vogel S, et al. Combination therapy with amprenavir, abacavir, and efavirenz in human immunodeficiency virus (HIV)-infected patients failing a protease-inhibitor regimen: pharmacokinetic drug interactions and antiviral activity. Clin Infect Dis 2000;30:313-318. (2007 is I think the study number) 

4. Wit FW, van Leeuwen R, Weverling GJ, et al. Outcome and predictors of failure of highly active antiretroviral therapy: one-year follow-up of a cohort of human immunodeficiency virus type 1-infected persons. J Infect Dis 1999;179:790-798.

5. Valdez H, Lederman MM, Woolley I, et al. Human immunodeficiency virus 1 protease inhibitors in clinical practice: predictors of virological outcome. Arch Intern Med 1999;159:1771-1776.

6. Lucas GM, Chaisson RE, Moore RD. Highly active antiretroviral therapy in a large urban clinic: risk factors for virologic failure and adverse drug reactions. Ann Intern Med 1999;131:81-87.

7. Ledergerber B, Egger M, Opravil M, et al. Clinical progression and virological failure on highly active antiretroviral therapy in HIV-1 patients: a prospective cohort study. Swiss HIV Cohort Study. Lancet 1999;353:863-868.

8. Deeks SG, Hecht FM, Swanson M, et al. HIV RNA and CD4 cell count response to protease inhibitor therapy in an urban AIDS clinic: response to both initial and salvage therapy. AIDS 1999;13:F35-43.
9. Mocroft A, Lundgren JD, Phillips AN. Response to salvage therapy in patients exposed to all three classes of antiretrovirals: the EuroSIDA study. Antiviral Ther 2000;5(suppl 2):4. Abstract 4.
10. Kahn J, Lagakos S, Wulfsohn M, et al. Efficacy and safety of adefovir dipivoxil with antiretroviral therapy: a randomized controlled trial. JAMA 1999;282:2305-2312. 
11. Schooley R, Myers R, Ruane P, et al. Tenofovir disoproxil fumarate (TDF) for the treatment of antiretroviral experienced patients: a double blind, plaebo-controlled trial. Presented at: 40th Interscience Conference on Antimicrobial Agents and Chemotherapy. September 17-20, 2000. Toronto. Abstract 692.
12. Bolognesi D, Matthews T, Kang MC, Hopkins S. Development and clinical evaluation of T-20: the first member of a novel class of anti-retroviral agents that inhibit membrane fusion. Presented at: 7th Conference on Retroviruses and Opportunistic Infections. January 30-February 2, 2000. San Francisco. Abstract S16.
13. Gulick RM, Hu XJ, Fiscus SA, et al. Randomized study of saquinavir with ritonavir or nelfinavir together with delavirdine, adefovir, or both in human immunodeficiency virus-infected adults with virologic failure on indinavir: AIDS clinical trials group study 359. J Infect Dis  2000;182:1375-1384. 
14. Montaner JSG, Harrington PR, Jahnke N, et al. Multiple drug rescue therapy for HIV-infected individuals with prior virologic failure to multiple regimens. AIDS 2001;15:61-70.
15. Deeks SG, Martin JN. Reassessing the goal of antiretroviral therapy in the heavily pre-treated HIV-infected patient. AIDS 2001;15:117-119.