FDA Advisory Panel Conditionally Endorses Clinically Useless Amyloid Tracer
I’m scratching my head over an FDA advisory panel’s unanimous recommendation to endorse Lilly’s amyloid tracer, florbetapir. While the panel did not recommend approval of the tracer at this time, it did unanimously (16-0) endorse conditional approval on the basis of a big stipulation: Namely that Lilly must run a trial confirming the reliability of scan interpretations. But this stipulation entirely misses the point of having a reliable (meaning, useful and not confusing) test for Alzheimer disease.
Regardless of the accuracy of PET interpretation, an amyloid scan for AD is worthless without clinical context. And in the context of normal cognition in the elderly, scan results are arguably irrelevant—because such a large majority of cognitively normal, older individuals will demonstrate brain amyloid. In the context of clinical dementia (which is the context in which a scan would be ordered, I would think), a positive scan may be supportive of the diagnosis of AD. But again, if amyloid is found in such a large majority of elderly brains anyway, what’s the point of the scan—which costs thousands of dollars—when AD is already suspected clinically?
Perhaps a negative scan in the context of clinical dementia might steer a clinician away from the diagnosis of AD and onto another (and potentially reversible) cause of dementia. But again, what’s the point of such an expensive imaging test, if the clinician is already looking for reversible causes of dementia (as she will), and the scan results are unlikely to substantively change recommendations for care—like if the patient has another irreversible dementing illness, say Pick’s disease?
In short, I don’t know what diagnostic niche Lilly (or the FDA advisory panel) is hoping for here: The proposed indication for a florbetapir-enhanced PET scan is to rule out “the presence of pathologically significant levels” of amyloid in the brain. The operative words here seem to be “pathologically significant,” and that’s where variabilities in scan interpretation become important (and worrisome). What we don’t really know from the submitted trial results (which were published this week in JAMA) is the frequency of clinical dementia in the 29 end-of-life patients (or at least these data aren’t provided in the abstract). About half of these patients’ brains showed pathologic evidence of AD at autopsy. The presence of scan-detected amyloid correlated well with autopsy findings. But that’s it. There’s no clinical (ie, cognitive) correlate as far as I can tell. [01/24/11: See an important update below.]
ABC News and MedPage Today interviewed at least 2 AD experts—neurologist Richard Caselli and scientist William Thies—who are equally at sea about what to do with such an amyloid scan in practice. In my previous discussions on the subject, here, here, and here, I viewed florbetapir only as an adjunct to the clinical investigation of anti-amyloid treatments for AD,* which is how Lilly’s been using the agent. But the company’s evidently pushing for primetime use and a chance to recoup something on its, in my opinion, misguided investment.
* Which, on a related note, have not panned out terribly well to date.
01/24/11 update: A review of the published article indicates that what the authors are really arguing is that B correlates with C, in which B is AD findings at autopsy, and C is the detection of amyloid by florbetpair-enhanced PET scans. The correlations are respectably robust—0.71 and 0.78 (depending on the autopsy staining techniques).
What’s importantly missing is the A-to-B-to-C correlation (or A-to-C correlation), in which A is dementia during late life. While Table 1 of the published article, which list the features of the patient enrollees, gives the cascading impression that normal cognition (or dementia) in late life directly relates to amyloid-related autopsy findings (and therefore PET amyloid findings), any calculations of such an A-to-C correlation are notably absent.
There is also precious little acknowledgement of the high prevalence of brain amyloid in the general population of cognitively normal elderly. Importantly the authors do recognize the lack of an absolute relationship between clinical cognition in late life and brain findings at autopsy, citing that 10%-20% of patients clinically diagnosed with AD don’t show AD findings at autopsy. But, in head-spinning fashion, they talk out of the other side of their collective mouth, emphasizing the importance of amyloid detection during life,
There is a growing body of evidence that the presence of [beta]-amyloid aggregates in individuals prior to developing AD is a significant and independent risk factor for cognitive impairment and eventual development of AD.
The authors’ hedging follow-up conclusion, however, is one that I agree with—namely that florbetapir-enhanced PET imaging is merely, at present, an investigative adjunct to clinical trials of anti-amyloid drugs (which don’t seem to be panning out particularly well).
[B]rain florbetapir-PET imaging of [beta]-amyloid aggregates has the potential to improve selection and monitoring of patients considered candidates for studies of disease-modifying AD treatments.