Read a PDF of the entire series.
This is Part 2 of a two-part series. See also Part 1.
5 April 2013. At the imaging symposium preceding the 11th AD/PD International Conference on Alzheimer’s and Parkinson’s Diseases, held last month in Florence, Italy, researchers debated how to put insights from longitudinal biomarker studies to use in the next wave of therapeutic trials, which increasingly include costly PET and MR scans.
When drug developers talk about looking for a “sweet spot” in testing their candidate drug, they mean an outcome that changes strongly and in linear fashion so that they can measure what the drug does to it. The natural history studies on AD have reinforced why trialists could sour on the late stages of this disease. As dementia worsens, the rate of amyloid accumulation declines to where it even dips naturally without a drug in the picture. In other words, it becomes biphasic. Biphasic outcomes are notoriously difficult to assess. “You would want to be doing your trials of an amyloid-reducing drug in a cohort where amyloid still accumulates, that is, from an SUVR of 1.5 to 2,” said Chris Rowe of the University of Melbourne, Australia.
Trialists are talking about trying to target drug trials to a "tipping point" in the progression, where people are about to become worse but can still be saved. They also talk about a "critical window," during which anti-amyloid drugs can still affect disease. Most researchers think the 17- to 20-year time span between amyloid positivity and AD represents this window (see Part 1 of this series).
What does this mean in practice? All three secondary prevention trials gearing up in 2013 will use brain imaging. The ADCS’ A4 trial expects to scan some 3,300 people 70 and older for β amyloid and then scan with structural MRI the expected amyloid-positive third among them. To investigate this critical window, the trial will allow—but not require—MRI evidence of neurodegeneration at enrollment and will study post-hoc if the extent of a person’s neurodegeneration affected the outcome, said Reisa Sperling at Brigham and Women’s Hospital, Boston. The A4 trial will stratify people by the extent of their amyloid load, as that appears to influence how quickly they progress (see Part 1) and scan everyone again at 18 and 36 months. The Alzheimer's Prevention Initiative and the DIAN trials of familial AD scan all participants at baseline and 24 months.
This is different from how the bapineuzumab and solanezumab Phase 3 programs used imaging, said Nick Fox of University College London, U.K. For inclusion, both required indirect imaging evidence in the form of an MRI or CT scan consistent with a diagnosis of AD. For outcome, the researchers measured volume changes in various brain areas as secondary endpoints. Both programs differed greatly from the upcoming trials, in that subgroups—not everyone—received outcome MRI scans, and only small subgroups had PET scans. The MRI data have been widely covered for their apparent paradox that the treated group lost slightly more, not less, volume than the placebo group. More analysis since last fall has confirmed that bapineuzumab generated a statistically significant MRI signal, and solanezumab, a trend. The signal consists of a small contraction in the hippocampus, coupled with an expansion of the temporal lobe and of the ventricular space, Fox said. Similar volume changes cropped up with the discontinued active vaccine AN1792 (see ARF related news story).
What does this signal mean? Fox suggested that it might reflect a combination of amyloid removal and fluid shifts that result from changes in associated inflammation. Astrocytes take up a large fraction of gray matter volume, so reduced astrocytosis alone could show up on MRI. This question should be studied, Fox said. Comparing the volumetric response of bapineuzumab with the response to approved antibody treatments in other neuroinflammatory conditions—for example, natalizumab in multiple sclerosis—might be a first step.
Who Is Right—The (Wo)Man or the Scan?
A second pressing research question to come out of the bapineuzumab and solanezumab programs is why a third of the enrolled ApoE4 non-carriers who had amyloid PET scans were negative on them. Call it the October Surprise of Alzheimer’s immunotherapy. This finding last fall startled the trial community. It not only influenced future trials, in which every participant now has to have evidence of brain amyloid by CSF or PET as a condition of entry, but also rippled through the research community with echoes of “we, too.” The same is true in ADNI, AIBL, and local clinics that have begun using florbetapir as a diagnostic aid in uncertain cases.
In fact, at the imaging symposium in Florence, Murali Doraiswamy of Duke University in Durham, North Carolina, said, “We have scanned about 100 patients, and the first word that comes to mind is humbling.” He meant to say that even in cases where expert clinicians felt sure of their early AD diagnosis, the scan came back negative and forced the team to re-evaluate. Thomas Beach of Banner Sun Health Research Institute in Sun City, Arizona, recently published that the clinical diagnoses uploaded to the National Alzheimer’s Coordinating Center database upon autopsy proved to have had a specificity (i.e., had picked the correct disease) of only 70 to as low as 44 percent (Beach et al., 2012). “This means that even leading clinicians are often wrong,” said Doraiswamy.
Philip Scheltens of VU University Medical Center in Amsterdam, the Netherlands, noted that in an ongoing Dutch study of the value of amyloid scans for diagnosis, only 61 percent of the patients who had a clinical diagnosis of AD scanned positive, but so did 80 percent of those with a clinical diagnosis of dementia with Lewy bodies (DLB), and even 28 percent of those the clinician considered to have frontotemporal dementia (FTD). For a fourth of all patients, the diagnosis changed after the amyloid scan.
Scheltens is particularly interested in using amyloid imaging to try to make sense of what ails young patients who present with dementia without a certain cause, i.e., a known mutation. He is conducting the Dutch Flutemetamol in Young Onset Dementia Study. Flutemetamol is G.E.’s 18F amyloid PET tracer; its application for use in clinical practice is pending with the Food and Drug Administration and European Medicines Agency. The study will assess whether an amyloid scan can improve the diagnosis and healthcare management of such cases. The study currently scans four patients per week, and the first images have come back positive thus far, Scheltens said. In the case of a 61-year-old man who complained of progressive memory loss and had hippocampal atrophy, “I myself made an erroneous AD diagnosis,” Scheltens recalled. This man and his similarly affected sister were completely normal on their flutemetamol scan, and later proved to have a tau mutation.
In time, an amyloid scan may change clinical practice by cutting short the remaining assessments. In this study, a third of the clinicians said: “If I have seen the scan, I will not do a neuropsychological evaluation.”
Heading in the same direction, Stephen Salloway of Brown University, Rhode Island, presented data at AD/PD 2013 suggesting that flutemetamol PET scans were more accurate than the clinician had been in a 19-center autopsy study that formed part of G.E.’s marketing application. Among 68 cases of terminally ill older people who agreed to a flutemetamol scan and brain donation, visual up-or-down readings of the scans by five blinded readers matched the postmortem pathology used as the gold standard better than did the clinical diagnosis. In particular, the clinical diagnosis relative to NIA-Reagan Institute pathological criteria came in with a sensitivity of 45 percent and a specificity of 61 percent, whereas flutemetamol clocked a sensitivity of 84 percent and a specificity of 79 percent. “Flutemetamol performed better than the clinical diagnosis,” said Salloway.
Talks like Parkinson’s, walks like Parkinson’s, but isn’t. SWEDDs have something else. Image courtesy of John Seibyl
At least Alzheimer’s researchers are not alone in this problem, said John Seibyl. Their Parkinson’s colleagues, too, see patients in clinical practice and in trials who look like they have PD but scan negative for β-CIT/DAT. This SPECT tracer visualizes loss of dopamine transporters in the basal ganglia and is widely used to support a diagnosis of Parkinson’s. In PD research, these people have a name: "SWEDDs" have Scans Without Evidence of Dopaminergic Deficits (Hall et al., 2010). When researchers dug a little deeper, they found that the scans were right and the clinical diagnosis had been wrong. The SWEDDs turned out to display a "mimic" of parkinsonism that had nothing to do with progressive dopaminergic neurodegeneration. The underlying causes were vascular, a consequence of neuroleptic medication, or a form of tremor. The confusion was big enough for movement disorder specialists in the U.K. to hold a SWEDD meeting in 2010 (see SWEDD newsletter), and PPMI even added a special cohort to compare them side by side with true PD cases and controls (see ARF related news story).
For the completed bapineuzumab and solanezumab trials, the amyloid-negative AD cases remain a riddle. Their investigators are trying to dissect if amyloid-positive and -negative patients responded differently to the therapeutic antibody. Other scientists are wondering if the current PET tracers fail to bind other types of amyloid or if the amyloid-negative people truly did not have AD. For longitudinal studies, Mike Weiner of UCSF suggested that amyloid-negative participants could be offered a DAT scan to see if they have a dopaminergic neurodegeneration masquerading as AD. All clinical cases of DLB have abnormal DAT scans, said Ken Marek at the Institute for Neurodegenerative Disorders in New Haven, Connecticut.
What’s a Trialist to Do?
For ongoing and future clinical trials of anti-amyloid drugs, large numbers of amyloid-negative participants are a huge problem, according to Samantha Budd of AstraZeneca in Cambridge, Massachusetts. “If you want to address amyloid therapeutically, then you want to ensure that amyloid is in the brain,” Budd said. She considers it “certain” that pathophysiology is ongoing years before the current diagnosis, and sees a growing need for imaging diagnostics, given the error rate of clinical diagnoses, especially at the early stages.
Budd spoke from the perspective of a developer of small-molecule drugs who is using imaging to give that drug the best shot in trials. Knowing that enough drug reaches its target in the brain to be able to be effective is critical, but this work sometimes gets short shrift in the race to the clinic. For example, in dissecting the results of the bapineuzumab trials, there is little discussion of the fact that the dose had to be dropped during the trial. “When you drop the dose, you walk away from something that, based on your prior data, you think was necessary,” Budd said, emphasizing how important it is to establish what is the dose and target engagement needed for efficacy.
For this goal, neuroimaging can be helpful. Microdosing—that is, labeling molecules with high-intensity radioactivity—enables scientists to tag a candidate drug, quantify its exposure in the brain, and decide whether to take it forward. If a therapeutic program is able to radiolabel the drug, then non-human primate PET can confirm that the drug gets into the brain sufficiently to support a dose range for human studies. If the program can further generate a PET ligand to the same target, this opens the possibility to look at target occupancy both in non-human primate and in human PET in parallel with the human Phase 1. “Target engagement is fundamental, especially in neuroscience,” Budd said.
In terms of letting only people with brain amyloid into anti-amyloid drug trials, CSF biochemistry and PET are somewhat equivalent, Budd said. Anne Fagan suggested as much in 2006, and subsequent research has confirmed this. As to using biomarkers as outcome measures, none has yet indicated a drug response that heralds a clinical effect; therefore, several candidates need to go into trials at this point. “We need to continue to explore the usefulness of multiple candidate outcome markers,” Budd said.
AstraZeneca is using an amyloid imaging agent it originally developed but in 2011 licensed to Navidea Biopharmaceuticals in Dublin, Ohio, and Andover, Massachusetts. Now called NAV4694, the low white matter binding of this 18F tracer enables it to detect small longitudinal changes and drug effects as sensitively as does PIB (see ARF related news story). Navidea started a multisite, three-year Phase 2b safety and efficacy study in people with MCI this March.
In Florence, Budd said that this tracer could be used to detect a reduction in amyloid as a result of treatment such as with BACE inhibitors (see ARF related news story). Since plaque is an "outcome" of amyloid pathogenesis, a BACE inhibitor should prevent further plaque deposition after some time. This was the case when six-month-old Tg2576 mice, which actively deposit amyloid, were treated with one of AstraZeneca’s BACE inhibitors for one month. “During a consistent reduction of Aβ monomer, we see a reduction in insoluble amyloid,” Budd said. In theory, this is true of solanezumab as well, though the smaller dynamic range and larger white matter binding of the current crop of 18F tracers may render them less able to quantify small changes in amyloid, Budd suggested.
The strategy of testing drugs only in people who have brain amyloid might have one unintended consequence. It may saddle drug developers with having to have a biomarker ready for clinical use as well. “What were those amyloid-negative patients doing in your trials? You need a tool to exclude them. But if you develop a drug that way, then the tool you use to exclude those patients from the trial must be available in clinical practice, too, to not expose those patients to that medicine,” said Cristina Sampaio, a neurologist and regulator who left the European Medicines Agency for CHDI in Princeton, New Jersey, but is still asked to share a regulator’s perspective to the AD/PD field. Some big pharma companies, including Merck, are developing such companion diagnostics alongside their investigational drugs.—Gabrielle Strobel.
This is Part 2 of a two-part series. See also Part 1.