A consumer advocacy group has accused the authors of a study published in the Journal of the American Medical Association (JAMA) - including employees of Avid Radiopharmaceuticals in University City and the University of Pennsylvania - of withholding important negative information about an experimental dye meant to help diagnose Alzheimer's disease.

Public Citizen says the article in the Jan. 19 issue of JAMA left out information about variability in the way doctors read brain scans of patients who had been given the imaging agent Amyvid. Instead, it focused on the correlation between the doctors' median scores and brain autopsies.

The averaged scores were more likely to agree with the brain autopsies than were readings by individual doctors. In the real world, imaging tests likely would be read by individuals, the group argued, and variations among them could lead to patients' being given inaccurate information.

Michael Carome, deputy director of Public Citizen's Health Research Group, said an inaccurate test would obviously leave some people thinking they had a terrible disease when they didn't, and vice versa. In addition, for a society already struggling with its medical bills, he said, "you'd probably end up wasting a lot of money."

The U.S. Food and Drug Administration generated the individual data late last year, but it was not included in the JAMA article. Earlier this year, the FDA declined to approve the agent, citing problems with reading scans accurately.

Daniel Skovronsky, Avid's chief executive officer, said the study published in JAMA was never meant to look at the performance of individual doctors. It was designed to show whether scans of the brain using Amyvid showed the same thing as autopsies. Skovronsky, who is listed as one of the JAMA study authors and is an unpaid adjunct professor at Penn, likened it to using several drivers to test cars with a new kind of engine. This study was meant to analyze the performance of the engine itself, not how well the drivers were able to use it.

Whether different doctors can agree about what they see on the scans is a legitimate question for the FDA, he said, and it's one that the company is working with the agency to answer.

"Eventually, we'll write a paper about how doctors will use it," Skovronsky said, "but that's premature."

Last week, JAMA published a critical letter from Public Citizen and a response from Avid.

Avid, whose primary product under development is Amyvid, was purchased in November by Eli Lilly & Co. for $300 million. The deal called for up to $500 million extra if the company reached certain commercial and regulatory milestones, including FDA approval.

A Lilly spokeswoman said Avid did not receive the FDA's analysis of its 50,000 pages of study data until December, after the study results had been accepted for publication by JAMA. A JAMA spokeswoman said the Jan. 19 issue went to press Dec. 30.

Public Citizen, which has previously opposed FDA approval of Amyvid, pointed out that nine of the 20 JAMA study authors owned stock or stock options in Avid and were "highly motivated to portray these results in the best possible light."

That included lead author Christopher Clark, a memory-disorders expert who is a professor emeritus at Penn and medical director for Avid. Two other authors are professors at Penn. In a letter in JAMA responding to Public Citizen's criticisms, Clark defended the quality of the study and echoed many of Skovronsky's points.

Alzheimer's is a deadly, memory-robbing disease for which there is no dramatically effective treatment. Currently, brain autopsy is the only definitive way to diagnose it. Earlier diagnosis will be essential when treatments become available.

Amyvid, also known as florbetapir F 18 injection, is used to show amyloid deposits in the brain during positron-emission tomography (PET) scans. Amyloid is a marker of Alzheimer's disease, although it can also be present in people who do not have symptoms of dementia. One of the controversies among researchers is whether amyloid is a cause of Alzheimer's or a symptom.

Public Citizen's Carome thinks there's little that can be done to improve the accuracy of Amyvid. Patients outside of a clinical trial likely would have a wider range of disease than those in the study, and doctors probably would not be as well-trained.

"This study was done in a rigorous manner and probably under the best circumstances you're ever going to get," he said.

In the study, three trained physician PET-scan readers were asked to rate patients' amyloid deposits on a 1-5 scale. Carome said the FDA's analysis of study data found that, in 10 of 35 study subjects, at least one reader's global rating of the amount of amyloid differed from the other readers' by at least two points.

Carome said the results translated into false-negative rates of 10 percent to 45 percent for the three doctors and false-positive rates of zero percent for two of the doctors and 20 percent for the third.

According to the published study, the median scores for the PET images and results of autopsies agreed 96 percent of the time.

Skovronsky said the individual results were much better when nine new doctors were asked to look at the scans. They were trained to determine whether the study subjects had significant amyloid deposits or not. Eight of the readers had more than 90 percent accuracy, he said.

Skovronsky said that 15 percent to 20 percent of people now thought to have Alzheimer's disease actually don't have it. Amyvid could help doctors rule out Alzheimer's in some patients with memory problems.

It won't help doctors know how serious Alzheimer's is, though. Experts now think that amyloid begins building up in the brain long before symptoms are a problem. "By the time you get symptoms," Skovronsky said, "you've got a brain full of amyloid."

But some people also have a lot of amyloid and function just fine. "The amount of amyloid," he said, "doesn't correlate with symptoms at all."