SILVER SPRING, Md. – A Food and Drug Administration advisory committee voted Jan. 20 against immediate approval of the beta-amyloid imaging agent florbetapir, saying that inter-reader variability and lack of a single reading technique in the existing data must still be addressed.
But after voting 13 to 3 to deny approval, the Peripheral and Central Nervous System Drugs Advisory Committee – in an unusual motion – took an additional stance on the compound. The committee voted unanimously to approve florbetapir if Avid Radiopharmaceuticals Inc., the company developing the compound, would train readers in a consistent technique, and then re-evaluate scans from both a recent phase III and a previous phase II trial. Eli Lilly & Co. acquired Avid in November 2010.
"What we need as the underpinning for approval is a re-read and re-study of these scans based on a training program," said committee member Dr. Peter Herscovitch. "[Without that] we cannot say there is consistency in the reading" of patients with mild cognitive impairment – the population seen as the most likely to utilize florbetapir-PET imaging.
The action of a second vote surprised Capt. Rafel Rieves, director of FDA’s Division of Medical Imaging Products. But he said there was no prohibition against the committee voting on a motion of its own making, as long as it answered FDA’s primary question – whether or not to approve florbetapir based on the existing data.
The committee based its decision partially on Avid’s latest trial, published Jan. 19 (JAMA 2011;305:275-83). The study comprised 35 end-of-life patients, whose brains were also examined at autopsy, and 74 young, healthy controls. It found that florbetapir-PET imaging was 93% sensitive and 100% specific for identifying beta-amyloid plaques in the brain – a finding necessary for a diagnosis of Alzheimer’s disease.
But committee members pointed out that different readers examined the cohorts and used different techniques for each one. The autopsy cohort, read by three radiologists, rated amyloid burden on a scale of 1 to 5. The control cohort, read by three other radiologists, used a binary rating of positive or negative.
Reproducibility seemed good in these patients, who appeared to represent two ends of the amyloid spectrum – young adults almost universally expected to have negative results, and elderly patients close to death, who were much more likely to have positive results. None of the patients in the trial had a "moderate’ amyloid load, although the committee said that these patients would represent the bulk of those evaluated in clinical practice.
Avid’s phase II trial comprised 79 healthy controls, 45 patients with Alzheimer’s disease, and 60 with mild cognitive impairment (MCI). Among this MCI group, reader variability was high, indicating readers’ difficulty in judging the visual cut point of a clinically significant amyloid burden, the committee members said.
Re-evaluating the scans in both groups, with more readers specifically trained in florbetapir imaging techniques, would provide the definitive answers needed for approval, the committee agreed.
Panel members have been cleared of potential conflicts of interest by the FDA prior to the meeting, but occasionally the FDA grants a waiver to a panelist with a conflict of interest.