The ability to visualize fibrillar Aβ deposits in the brains of living people has been hailed as a breakthrough for Alzheimer’s disease research and differential diagnosis. Now, as the first of a suite of positron emission tomography (PET) β amyloid imaging agents stands poised to receive a ruling from the Food and Drug Administration, seven scientists are questioning the validity of this technology. Led by Abass Alavi, a nuclear medicine professor at the University of Pennsylvania, Philadelphia, they penned an editorial that appeared online October 19, 2011, in the European Journal of Nuclear Medicine and Molecular Imaging. In it, they dispute the ability of any amyloid imaging agents, including the original Pittsburgh Compound B (PIB), to accurately label Aβ deposits. Joint first authors Mateen Moghbel and Babak Saboury, also at UPenn, and colleagues claim that none of the newer fluorine-18-labeled tracers deserve regulatory approval. This editorial has become the most downloaded article in the journal’s recent history, with around 1,200 downloads in the past 90 days. Based on it, the Washington-based consumer advocacy group Public Citizen has sent a letter to the FDA, urging the agency to reject Eli Lilly and Company’s pending application for florbetapir (see ARF related news story).
Scientists in the AD imaging field have fired back. In an 11-page rebuttal published in the same journal on January 5, 2012, they collectively challenged the claims made by Alavi and colleagues, detailing neuropathological evidence and basic PET physics that substantiate the reliability of amyloid imaging technology. Their editorial, signed by 24 leading AD and amyloid PET experts from around the world, is freely downloadable and will run in print in the February issue. It has racked up more than 1,000 downloads since it appeared. First author Victor Villemagne at Austin Health, Heidelberg, Australia, and colleagues in Japan, the U.S., and Europe argue that Moghbel et al. misunderstand the current state of knowledge in the AD field, and base their conclusions on misleading and inaccurate data. Villemagne et al. stress the importance of amyloid imaging for shedding light on AD processes, as it gives researchers a glimpse into the progression of the underlying pathology. The authors of the rebuttal editorial disclose that many of them consult for companies developing amyloid tracers, or have received research grant support from these companies. The authors of Moghbel et al. do not disclose conflicts.
Why the push for 18F-labeled tracers? Although carbon-11-labeled Pittsburgh Compound B (PIB) is the current gold standard for β amyloid imaging, its 20-minute half-life limits its use to state-of-the-art facilities that can synthesize the compound on site. As an alternative, several companies are developing 18F-labeled compounds, which have a 110-minute half-life and are expected to open up the use of the technology to many more centers. Florbetapir (trade name Amyvid®), first developed by Avid Radiopharmaceuticals, is the furthest along; others in the pipeline include florbetaben at Bayer Healthcare, flutemetamol at GE Healthcare (see ARF related news story and ARF update), and AZD4694 by AstraZeneca (see ARF related news story).
Lilly reported Phase 3 results for florbetapir (see ARF related news story), but the FDA initially denied approval, citing inter-reader variability in how scans are interpreted (see ARF related news story). Public Citizen had also picked up this issue in a previous letter to the FDA on February 21, 2011. Since then, Avid has developed and tested a reader training program as requested by the agency. It was presented at the Human Amyloid Imaging conference held 12-13 January 2012 in Miami, Florida. The agency’s ruling on the resubmitted application is expected this spring.
Alavi, who is a nuclear medicine physician, states technical and philosophical objections. In an interview with ARF, Alavi said that the AD community is going down the wrong path by developing β amyloid imaging agents. Alavi claimed that the physics of imaging precludes this technology from working, particularly in cases of mild cognitive impairment (MCI) and early AD. Alavi noted he is preparing another editorial in response to Villemagne et al., which he intends to write in collaboration with physicists. He questioned whether current agents bind to β amyloid in vivo, suggesting they are detecting non-specific proteins. Furthermore, Alavi told ARF he fears that the widespread use of β amyloid imaging would ultimately lead to clinical trials in presymptomatic people with amyloid deposits. He believes such trials would not be in patients’ best interests due to potential drug side effects.
In response to a question from a reporter about what the field would do without Aβ imaging, Alavi said, “If I were the community, I would stick with fluorodeoxyglucose (FDG) PET.” FDG PET measures glucose metabolism in the brain; Alavi was among the pioneers of its use in AD patients in the 1970s. FDG PET is a tool widely used in AD research as an indirect measure of synaptic function. There is broad consensus among researchers in the AD field that it is not specific to the AD pathogenic process, whereas amyloid PET is being developed to visualize and quantify one of the two hallmark pathologies of Alzheimer’s disease. Alavi coauthored a Journal of Nuclear Medicine meeting abstract, which reported that florbetapir has superior sensitivity and specificity to FDG (around 95 percent for florbetapir, compared to high 80s for FDG) for distinguishing AD patients from controls.
Most of Alavi’s publications are in the areas of cancer research, infection, the effects of aging on peripheral systems, or studies of brain activity using FDG PET. Most of the coauthors are radiologists and faculty at UPenn. Curiously, coauthor Bengt Långström at Uppsala University, Sweden, is also a coauthor on some 20 papers on amyloid PET, beginning with the first reports on the technology in 2003 and 2004 (see also ARF related news story), and continuing with three papers to date in 2012. From 2002 up until three years ago, Långström was Chief Scientific Officer of GE Healthcare/Imanet, which develops flutemetamol. When contacted by ARF, Långström said he joined the editorial because he believes scientific issues regarding how β amyloid tracers work need to be resolved before taking the technology to the clinic. “I do hope that this discussion will bring the issue back to science, and the business [side] will have to wait,” he wrote to ARF (see full Q&A below).
Two of the 18F-labeled tracers, florbetapir and florbetaben, were originally developed by Hank Kung, a chemist in the Department of Radiology at UPenn. Kung chairs Avid’s scientific advisory board. In an interview with Alzforum, Kung said that Alavi is “motivated by professional jealousy. It’s unfortunate that we are being distracted by this.” Kung contends that Alavi’s objections to β amyloid imaging technology are not scientifically sound. He noted that the reasoning of Moghbel et al. would apply to any nuclear medicine procedure, many of which are used routinely in clinics around the world with great success. “What they are essentially saying is that all nuclear medicine procedures are not valid,” Kung said.
Alavi said he is concerned that β amyloid plaques cannot be imaged, based on physics principles. In their editorial, Moghbel et al. emphasize that PET gives poor spatial resolution, on the order of 5 mm, making it useless for visualizing amyloid plaques with an average size of 50 μM. Villemagne et al. counter that this claim is based on faulty logic. PET imaging of all sorts reflects the average signal from a brain region. It does not try to resolve small structures, they note. For example, FDG PET is based on cell uptake proportional to its glucose utilization, and the involved enzymes are smaller than plaques. Receptor imaging with PET is predicated on the tracer attaching to structures several orders of magnitude smaller than plaques. In general, the PET signal is derived from the tissue concentration of binding sites, not the size of the tracer’s target structure. In the case of amyloid PET, “Millions of fibrillar Aβ deposits produce a signal that is easily detectable in Aβ-laden parts of gray matter,” Villemagne et al. write.
Moghbel et al. further charge that the β amyloid burden in MCI would be too low to be detected. Their editorial notes that Aβ load in late-stage AD has been shown in histopathology slices to be around 6 percent of total brain area (see Clark et al., 2011; Bussière et al., 2002). The authors speculate that the load would be about 60 times less in people with MCI. No references are cited for this number; in answer to a question from Alzforum, Alavi said he heard the figure of 0.1 percent load in MCI in conversations with neuropathologists.
Villemagne et al. write that this 0.1 percent figure is a “surprising and unsupported assumption.” They note that a milestone study of 79 postmortem brains, using quantitative enzyme-linked immunosorbent assays (ELISA) to measure Aβ, found that people who die at the MCI stage have about half the typical amyloid burden of people with AD (see ARF related news story on Näslund et al., 2000). Villemagne et al. point to a recent paper coauthored by Bengt Långström that demonstrated ample binding sites available for PIB in postmortem brain tissue. The paper concludes, “This radiotracer is, therefore, very suitable in the early diagnosis of AD” (see Svedberg et al., 2009).
Moghbel et al. draw attention to the finding that the brain’s white matter retains β amyloid tracers even though it contains no Aβ plaques. This nonspecific signal may spill over into nearby gray matter and overwhelm the Aβ signal, the authors speculate. In answer, Villemagne et al. acknowledge that the nonspecific signal from white matter is a limitation of all β amyloid radiotracers, probably due to a slower rate of clearance from white matter. However, they note, this nonspecific binding does not differ between AD patients and controls. Villemagne et al. charge that Figure 2 of Moghbel et al. is misleading. It uses data from FDG PET images of lung tissue (see Hickeson et al., 2002), where differences in uptake by various tissue types are huge, to represent the differential tracer uptake by white and gray matter of the brain. In a typical brain containing amyloid, the signal intensity from gray matter is threefold that from white matter, Villemagne et al. write, and white matter spillover does not interfere with the ability to read the gray matter signal.
Alavi and colleagues cast doubt on the specificity of radiotracer binding. As evidence, they point out that all existing β amyloid tracers produce a high signal from the frontal lobes, claiming that this contradicts neuropathology studies showing that Aβ deposits are highest in the occipital and temporal lobes, and lowest in the frontal lobes (see Arnold et al., 1991; Braak and Braak, 1997). “Clearly, there is something else that this agent is binding to that is heavily concentrated in the frontal lobes,” Alavi told ARF.
Villemagne et al. respond that the statement about frontal lobes containing little Aβ “is inconsistent with the current state of knowledge regarding the neuropathology of AD.” They write that it is contradicted by numerous more current studies by many of the same authors, which show early, heavy AD deposition in the frontal lobes (e.g., Thal et al., 2002; Arnold et al., 2000, and even the paper by Braak and Braak, 1997 cited by Moghbel et al.). A recent paper describes good correlation between amyloid tracer uptake and Aβ load in the frontal cortex, as determined by histopathology (see ARF related news story on Wolk et al., 2011). In addition, the quantitative study by Näslund et al. reported that frontal lobes, in fact, contain two- to fourfold higher concentrations of Aβ than do other brain regions.
Villemagne et al. point out that different forms of AD show different patterns of tracer retention (see, e.g., ARF related news story on Klunk et al., 2007), which would argue against nonspecific binding. They cite nine neuropathology studies that show strong correlations between quantitative measurements of Aβ and Aβ radiotracer retention. At the January 2012 HAI meeting, four additional studies were presented, all showing the same strong neuropathology correlation. Villemagne et al. conclude, “The functionality, sensitivity, and specificity of Aβ plaque imaging agents has by now been demonstrated in a level of detail and reliability that has not been required or provided for most other imaging tracers clinically used today.”
More globally, Moghbel et al. raise concerns about how β amyloid imaging might be used. They refer to a positive scan in a cognitively healthy person as a “false positive,” and suggest that a high rate of this finding would diminish the value of the technology for diagnosis and clinical practice. Villemagne et al. disagree, noting that this idea reflects a “conceptual misunderstanding.” Mismatches between neuropathology and symptomatic expression of AD do not reflect a problem with the imaging technology, which visualizes pathology, but instead highlight the slow progression of AD, where amyloid deposition is widely thought to precede dementia by about a decade, they write. In support of this, several studies have shown that MCI patients with Aβ deposits are more likely to continue to decline cognitively than are their peers with amyloid-negative scans (see, e.g., Forsberg et al., 2008; Okello et al., 2009). On this issue, too, ongoing longitudinal cohorts in the U.S. and Australia continue to add evidence that MCI patients with brain amyloid have a higher risk of progressing to AD dementia within the next few years.
Amyloid deposition does not equate to a clinical diagnosis, Villemagne et al. stress, nor is it intended to. Instead, it is a tool for increasing the understanding of AD pathology and progression, and could be combined with clinical and fluid biomarkers to improve the diagnosis. In fact, amyloid imaging may clarify some cases of clinically diagnosed Alzheimer’s as not being due to AD but to something else when the patient turns out to have a negative amyloid scan.
“Aβ imaging has been repeatedly held up as one of the major successes of the past decade in the fight against AD,” the rebuttal notes. “The inability to obtain the information provided by Aβ imaging would most certainly slow down the urgently needed progress in understanding the basics of neurodegeneration and in the development of new approaches aiming to treat these devastating disorders.”—Madolyn Bowman Rogers
Q: What motivated you to coauthor the first editorial?
A: The reason for me to join the editorial in the first place was based on the following points:
- Development and validation of a PET tracer molecule is a long and tedious journey. In this case, I believe that there is value in having an amyloid biomarker available in such a way that we could use a tracer molecule which could perform an in-vivo visualization of the brain pathology, giving similar information as that obtained by a pathologist who is performing various staining techniques at autopsy for the same subject.
- The first human PIB study—in which I was involved—started actually with another compound, PIA, but was based on our previous experience of how good in-vivo PET tracers behaved. PIB was selected for the human study because this change in the molecular structure resulted in similar characteristics as several of our best PET tracers had. So we used a similar paradigm for the selection of the tracer as we had been using when developing tools for receptor expression or enzyme function.
- The PIB story also contained another paradigm: the microdose concept which allowed human applications at microdose even with limited toxicology information. That was the main reason why we were not able to carry out a PIB study using a dose escalation, or as an alternative, a competition displacement study, in order to determine if the binding was changed due to mass dose effects, or whether binding was similar in various brain structures, or changed in various patients and healthy aging controls.
- This is a fundamental problem with the existing "amyloid tracers": We don't know if the binding in various patients and aging healthy controls really is of the same type, that is, whether binding is the result of brain amyloid loads. That information might have been sorted out if we had carried out the dose/binding investigation. This is something which definitely should be carried out before moving this into clinical trials. The various amyloid tracers give somewhat different results, which also could be an indication that they are binding to various tissue targets.
In my estimation, this is a missing factor. The paper by Sabbagh et al., 2011 emphasizes that pathology staining might not always be in correlation with the in-vivo method which we now are discussing.
There are also other parameters which we should explore in more detail before we perform clinical trials with the intention to commercialize these tracers. It would be very important to explore in detail the impact of blood flow on the in-vivo amyloid imaging. That may increase the potential to perform quantification. Furthermore it would be valuable to understand what type of amyloid aggregates, or whether other tissue targets (e.g., enzymes) are targeted in vivo by these PET tracers. Such work is in progress.
Q: Do you believe that β amyloid imaging agents such as PIB and the new 18F-labeled tracers are specific and reliable?
A: There is literature evidence that some of them are not specific for ”amyloid aggregates.” I don’t know if this is entirely true in humans, but I certainly would like to see that we—as a scientific community—focus on performing rigorous scientific investigations to answer these questions before this is taken over by business.
In studies with a large number of patients, we see patterns. However, there are fundamental differences between the amyloid tracers compared with other structural tracers targeting functional tissue proteins, because amyloid aggregates are what I would claim is a residual (garbage) protein.
Q: Do they accurately reflect Aβ deposits in the brain?
A: Again, I don't know, but we certainly need to perform more scientific, unbiased, and professional studies to validate what the imaging observations may indicate. There is a huge need for these kind of studies, and I do hope that these will be made by the whole scientific community.
Q: What do you think of the rebuttal editorial by Villemagne and colleagues?
A: To join the first editorial was, for me, a must, since I have at many occasions been making the point that there is a need for more fundamental studies which we should perform before making precipitous conclusions about the value of these "amyloid tracers." I do hope that this discussion will bring the issue back to science and the business will have to wait.