Alzheimer’s disease progresses at varying rates even among people who are at the same clinical stage. Because of this, trials need to enroll a large number of participants in order to see a statistically significant slowing of progression. Could enrolling people based on their degree of tau pathology make cohorts more homogenous, and trials more efficient? A new analysis in the March 13 JAMA Neurology supports this idea. Researchers led by Brad Hyman at Massachusetts General Hospital, Charlestown, and Rebecca Betensky at the Harvard T.H. Chan School of Public Health, Boston, used data compiled by the National Alzheimer Coordinating Center (NACC). It covers nearly 1,000 people with Alzheimer’s who were followed longitudinally during life and donated their brains after death. The authors calculated the actual rates of cognitive and clinical decline for individual people at different Braak stages, then plugged the data into hypothetical trial designs. In their models, trials that enrolled people all at the same Braak stage typically needed less than half as many participants to achieve the same power as a trial that did not take this staging into account.
“If you can make a cohort more homogeneous in terms of pathology, you reduce the noise. These data provide a rationale for using Braak stage as an entry point or a covariant for trials,” Hyman told Alzforum. For real-world trials, Braak stage could be determined by tau PET imaging, he added.
“I’m very enthusiastic about this paper,” Mark Mintun at Avid Radiopharmaceuticals, Philadelphia, told Alzforum. Avid developed the tau tracer AV-1451. Mintun noted that the data provide concrete support for something many researchers have assumed, namely that people with more tau pathology decline faster, and also estimate the magnitude of that effect. “That’s useful, because it allows you to make specific hypotheses,” Mintun said.
Some studies are already taking tangle pathology into account. “We will definitely look at responsiveness to treatment based on tau PET [status] in the A4 study,” Reisa Sperling at Brigham and Women’s Hospital, Boston, wrote to Alzforum (see full comment below). A4 enrolls people at the preclinical stage of the disease based on an amyloid PET scan, but a subset of the participants receive tau PET scans during the trial.
Tau PET patterns have been found to match Braak staging (see Mar 2016 news). Moreover, the tau PET signal correlates closely with cognitive decline and brain atrophy, unlike amyloid imaging (see May 2016 news; Aug 2016 conference news; Mar 2017 news).
To test the concept of selecting trial participants based on brain pathology, first author Jing Qian analyzed 984 people from the NACC dataset who had been seen at Alzheimer’s disease centers between 2005 and 2014. All were older than 50 at death, with the mean age being 82, and all had amyloid plaque accumulation, as would be the case in an Alzheimer’s clinical trial. The authors excluded people who had other neuropathological diagnoses, mixed pathology, or cognitive impairment due to other conditions, such as depression, alcohol use, or medications. To mimic the time frame of a clinical trial, the authors analyzed data from only the last two to three years of each participant’s life. In that time, participants completed an average of two clinic visits.
The authors stratified these participants according to whether their tangle pathology was low (Braak stage 1-2), moderate (3-4), or high (5-6). Based on cross-sectional data from other studies, the researchers assumed that the stage seen at autopsy roughly resembled that present for the two to three years prior to death. Even if there was some progression, the relative severity of disease among participants should remain unchanged, they reasoned.
As might be expected, the authors found that people at higher Braak stages had declined more rapidly in the last years of their life than did those at lower stages. This was evident on both the Clinical Dementia Rating-sum of boxes (CDR-SoB) and on a test of logical memory. However, the two measures revealed different patterns, with performance on the memory test bottoming out already at moderate Braak stages, while the CDR-SoB changed little between participants with low and moderate Braak scores, but worsened greatly in those with high pathology. “Different clinical endpoints are more sensitive at different points in disease,” Hyman noted.
In hypothetical clinical trials of 300 participants selected randomly from the NACC cohort, the CDR-SoB provided only 29 percent power to detect a 20 percent slowing of decline over placebo. If the trial selected 100 participants from each of the three Braak categories, however, the power rose to 69 percent, likely because the cohort had a better balance of stages than the randomly selected group. If all 300 participants had high pathology, the power topped out at 89 percent. By contrast, the logical memory test gave the greatest power, 68 percent, when all participants had moderate tangle pathology, compared to only 23 percent power when participants were not selected by Braak stage.
The authors performed separate analyses to determine how many participants would be needed to achieve 80 percent power to detect a 20 percent slowing of decline for each measure. Depending on the Braak stage and cognitive measure used, selecting participants by their extent of pathology lowered the total number needed by anywhere from 30 to 80 percent, they found. On the CDR-SoB, the number dropped from 1,176 to 230, 700, or 420 if all patients were at high, moderate, or low Braak stage, respectively. In general, even trials where all participants had low pathology, corresponding to preclinical disease stages, experienced this power boost. Many Alzheimer’s trials are now done in the preclinical population.
Lon Schneider at the University of Southern California, Los Angeles, called the analysis useful. “The take-home message is that tau pathology may help predict the expected placebo decline in any given clinical trial. That may affect the ability to detect a treatment signal,” he said.
Mintun noted that the next step will be to test the ideas presented in this paper by comparing baseline tau PET status with clinical decline in ongoing trials. “We now have to show it’s theoretically possible to improve power by selecting participants based on tau imaging,” he said. Designing a trial around this idea would require some thought, however, since selecting people based on their degree of pathology would be a very different approach from the clinical selection criteria used now. “We don’t necessarily know how to do those types of clinical trials yet,” Mintun pointed out.
While Braak stage provided a powerful way to reduce the variability in progression rates among participants, it did not account for all of the variability, Hyman noted. Even within Braak stages, participants fell into two distinct groups with slightly different patterns of decline. One of the groups worsened faster on the logical memory test, but slower on the CDR, compared to the other. It is unclear what biology drives this difference in progression, but the data demonstrate that there are still sources of heterogeneity among Alzheimer’s patients that remain to be discovered, Hyman noted.—Madolyn Bowman Rogers
- Qian J, Hyman BT, Betensky RA. Neurofibrillary Tangle Stage and the Rate of Progression of Alzheimer Symptoms: Modeling Using an Autopsy Cohort and Application to Clinical Trial Design. JAMA Neurol. 2017 May 1;74(5):540-548. PubMed.