The FDA’s Peripheral and Central Nervous System Drugs Advisory Committee today voted 13 to 3 to reject Avid Radiopharmaceutical’s New Drug Application for their amyloid imaging agent AV-45/florbetapir/Amyvid in its current form. They softened the blow minutes later with a second, unanimous vote that they would recommend approval provided the company strengthens its case that Amyvid scans are read consistently from one reader to another, in a previously acquired set of scans resembling the intended patient group, and using the actual reading method that is to be applied in community settings. The Food and Drug Agency usually follows the recommendation of its advisory committees.
The votes came at 4:20 p.m. after scientific presentations, searching deliberation, and appeals from audience members that began at 8:00 a.m. The amyloid imaging ligand’s sponsor, Avid, now owned by Eli Lilly and Company, had its day in court as FDA reviewers, Avid representatives, and a fleet of academic advisers with neurology, nuclear medicine, and statistics expertise sat in judgment of the data package Avid had submitted to the agency. There were occasional light moments, but like a jury, the advisers were admonished to refrain from discussing during breaks any of what was said in the proceedings.
Today's meeting followed a previous meeting by this group, on 23 October 2008, which had concluded with the agency requiring that Avid conduct a Phase 3 histopathological confirmation study as part of their new drug application (NDA). In this study, terminally ill people who had been given less than six months to live granted consent to undergo a florbetapir scan. Scientists then compared how well florbetapir measured within months before death matched up with the pathology quantified by two different established methods after the patient had passed away. Alzforum reported previously on this study, which appeared formally in print the day before the FDA meeting (see ARF related news story). Today, Avid garnered praise for the effort and speed with which it had pulled off this study. Alas, the reviewers found fault with some of the data Avid had submitted in hopes that this Phase 3 trial would be sufficient for a nod of approval today.
Avid’s NDA seeks an indication for Amyvid as a diagnostic agent for β-amyloid aggregates in the brain. It states that a negative scan would be clinically useful in ruling out the presence of pathologically significant levels of β-amyloid in the brain. “We see Amyvid as simply an added tool in the clinicians' armamentarium to help them find out what’s going on in a patient,” said Daniel Skovronsky of Avid. The NDA does not seek an indication for Amyvid as an Alzheimer’s disease diagnostic or as an AD screening tool in cognitively healthy or mildly impaired people. The question of what a positive scan means was not on the table at this meeting. That said, many practitioners are interested in amyloid imaging for these purposes, and worried patients are expected to demand it. Indeed, some advisory committee members asked pointed questions about what, in the real world, would stop Amyvid from being used broadly should it be approved.
Before the meeting, the FDA reviewers had already expressed some concerns in a briefing document. They agreed that the Phase 3 trial met its primary endpoints. Indeed Skovronsky went to great lengths in his presentations today to show that the company had played by the rules in providing the data required in the development plan the FDA had seen early on. The reviewers agreed that the data showed well enough that Amyvid sees what it purports to see, that is, fibrillar amyloid deposits in the brain (though they did quibble with the wide range of immunohistochemistry results for a given amyloid level). They agreed with neurologists who argued that ruling out amyloid was clinically useful in many cases. They saw no significant safety concerns. And yet, in digging deep into the raw data, the FDA reviewers also came up with what they considered significant limitations—and proceeded to move the goalposts a bit in light of that data.
Above all, FDAs statisticians criticized the way the Amyvid scans were read. The method used was semi-quantitative in that the raters looked at the scan and scored it on a scale from 0 (no amyloid) to 4 (a lot of amyloid). However, out in the real world of clinics across the country, raters would read the scans differently. Patients generally don’t want to know if they are a 2 or a 3 on a scale of 4; they want to know if they have problem or not. Hence, the rating method in the real world would be binary, yielding either an "amyloid-positive" or "amyloid-negative" verdict and requiring a cutoff point. Setting such thresholds is a hairy subject in itself, but the committee did not focus on that as much.
Rather, they criticized that the crucial dataset in Avid’s application—that of the Phase 3 autopsy study—did not include binary reads. The reviewers pointed out that in a previous Phase 2 study that had used a binary read, the agreement between the different blinded readers was poor. This was especially apparent to the reviewers when they subtracted those folks in whose scans getting it right is easy. Those are the young volunteers in their thirties and forties who are amyloid negative and whose brains are not atrophied. Those volunteers had been chosen as an agreed-upon “standard of truth,” and they do serve that purpose. Avid had, to their credit, admixed some 40 people from the autopsy study to make the task harder for the readers, and inter-reader reliability was okay. But those are still easy to spot because they are much older and atrophied. In short, the reviewers said, the sample disproportionately contained the two ends of the spectrum, and extremes are easier to call than intermediates.
Young people, or old people near death, are not the patients Amyvid would be used for, the reviewers said. Amyvid would most likely be used on people along the spectrum of day-to-day clinical practice—subjective complainers, people with MCI, people with uncertain diagnoses, people with suspicion of mixed pathology. Many might have intermediate amyloid levels, which for some reason received few ratings in the autopsy study sample, and therefore might dance around the cutoff. Having a reliable assessment method for these people is important, said FDA reviewer Qi Feng. The agency’s Rusty Katz asked in whom the requested indication—a negative scan ruling out pathological amyloid—is useful. “In a normal 30-year-old, it is not. How robust is your data in early patients?”
The Advisory Committee asked Avid to provide data showing consistency between readers in such a group of people. They said that the Phase 2 study population of normal controls, MCI, and AD patients most closely reflected a clinically relevant sample among all Amyvid trials that are part of this NDA. “In approving imaging agents, we like to see that the agent is tested in the population and with the methods then used in the clinically applicable population,” said Captain Duane Reeves, who heads the medical imaging products division at the FDA. This echoed a goal Monique Breteler, formerly of Erasmus University in Rotterdam, The Netherlands, and now at the German Center for Neurodegenerative Disease (DZNE) in Bonn, had urged amyloid imagers to pursue during her keynote lecture at the 5th International Human Amyloid Imaging Meeting held in Miami, Florida, 14-15 January 2011. (An Alzforum report on that conference will be posted next week.)
What does this mean for amyloid imaging at large? In essence, it sounds as if sponsors of amyloid imaging PET agents are now asked to read positives and negatives in the same age bracket. Before, there was an age covariance: negative in young subjects and positive in old ones. “Removing that crutch sets the bar higher,” commented Bill Klunk of the University of Pittsburgh Medical School, a co-developer of amyloid imaging. “This was a surprise and a wakeup call for the whole field,” Klunk said. Showing reader consistency along the real-life spectrum of amyloid deposition is a challenge. Reanalyzing existing Phase 3 autopsy scans with binary reads is one thing, could be done in weeks, and the outcome is predictable. But doing so for the Phase 2 population—or for G.E. or Bayer or other companies to do in an equivalent cohort—is another. “That’s not a slam dunk; it’s more like a three-point shot. You put it up and hope for the best,” Klunk said.
Avid had anticipated concern about the reads, and begun to address them even before the meeting. As Avid’s Mark Mintun laid out, the company has since developed a binary read method that uses black-and-white versions of the scans together with a training protocol for routine clinical use. They tested this on nine additional readers from private radiology practice, who were inexperienced in neuro-PET or amyloid imaging. Eight of those nine did well with the training. The FDA encouraged Avid to expand this work and collaborate with professional organizations such as the Society of Nuclear Medicine or the American College of Radiology to develop professional education and training for amyloid imaging, perhaps within these organizations' CME programs. Indeed, representatives from these organizations were there and testified that they were ready and willing to develop this training.
But it was not enough to sway the committee. The Advisory Committee has not yet formally reviewed this new binary reading method, or the training, and based its decision today on the data in the NDA alone. They made their second "yes" vote contingent on such a training program that demonstrates reader consistency in a population resembling real-life use of the agent. They ignored pleas from audience speakers to recommend approval now. These included a personal one from Alzheimer's Disease Neuroimaging Initiative (ADNI) leader Mike Weiner of UCSF, who mentioned his own mother’s uncertain clinical diagnosis of dementia; also similar emphatic calls by clinicians Norman Foster of the University of Utah, Salt Lake City; David Brooks of Imperial College, London; and Reisa Sperling of Brigham and Women's Hospital, Boston, to hand them this tool to better serve their patients. Representatives of advocacy and professional organizations also spoke.
Importantly, the ruling does not affect the use of amyloid imaging agents—Amyvid, PIB, flutemetamol, and others—in research studies such as ADNI2, AIBL, and many others, or in therapeutic trials. The agents were available for this research before and will remain so. In fact, attendees highlighted the progress the field had made even since their last meeting in 2008. They noted studies in numerous labs around the world showing that mildly impaired, even cognitively normal, people who have brain amyloid are more likely to also have changes in other imaging or biomarker measures, are more likely to have subtle cognitive or functional problems, and are at higher risk of progressing toward AD. Scientists at the meeting called biomarker studies, in which amyloid imaging is a key piece, perhaps the greatest advance in the past few years in AD research as a whole. Today’s decision delayed Avid/Lilly's bid to make Amyvid available for clinical practice, though whether the delay is a question of a few months or longer is not yet clear.—Gabrielle Strobel.