The 16th iteration of the Clinical Trials in Alzheimer’s Disease conference, held October 24-27 in Boston, drew some 2,300 participants, of whom 1,500 crowded into a downtown hotel for the in-person experience while 800 followed proceedings remotely. The meeting started with a session on how efforts to diversify research participation in trials are beginning to yield data on biomarker positivity among different groups, and continued with a focus on which patients might benefit most from amyloid reduction.
To Recruit for Diverse Alzheimer Trials, Go to the People
Despite evidence that African American, Hispanic and Latino, and other minority groups have higher rates of Alzheimer’s and related dementias in the U.S., these populations are less likely to take part in clinical studies. This underrepresentation makes it difficult to understand differences in vulnerability to disease among populations, and whether treatments will work the same way in everyone. Stakeholders who grapple with these challenges reported making progress toward overcoming them at a conference called Enhancing Participation by Minoritized Groups in Alzheimer’s Disease and Dementia Research, which was held October 3 to 4 at Washington University School of Medicine in St. Louis. Similar issues resonated in Boston, at the Clinical Trials on Alzheimer's Disease Conference held October 24 to 27.
WashU’s Knight Alzheimer’s Disease Research Center hosted the former meeting. Director John Morris said that participants in AD and dementia research studies ought to reflect the populations most affected by these diseases. Not only is this the right thing to do, it’s also good science, he told Alzforum. The meeting drew more than 300 participants in person or virtually, including representatives from each of 37 National Institute on Aging-funded Alzheimer Disease Research Centers (ADRCs).
Consuelo Wilkins, a geriatrician and associate director of the Vanderbilt University’s ADRC in Nashville, Tennessee, argued that the onus of recruiting underrepresented groups falls on researchers. “Often when we think about barriers to participating in research and clinical trials, we put the blame on the people who are not participating,” Wilkins told the audience. “We cannot excuse ourselves from being the solution.”
Solutions, according to the speakers, start with community outreach and trust-building. Researchers noted that teams of outreach specialists familiar with the community at hand can lay the groundwork of connection and trust before potential participants are asked to join a study. Biomarker studies, including ADNI, have started to apply such community engagement methods to attract more Hispanic/Latino and black/African Americans, and reported that these approaches are paying off. (For a summary of biomarker data across ethnoracial groups, see Part 2 of this series.)
Funded by the National Institute on Aging, this year’s meeting was a follow-up to a prepandemic inaugural that had focused on African American participation (Oct 2018 conference news series). This time, the scope was expanded to consider Hispanic/Latino, Asian, Native Americans, as well as sexual and gender minorities. Inclusion of people with disabilities was also mentioned. Besides showcasing recruitment strategies, the meeting focused on social determinants of health that could contribute to dementia across underrepresented groups, as well as research participation.
Shifting Responsibility Johnathan Jackson is a cognitive neuroscientist who directs the Community Access, Recruitment, and Engagement (CARE) research center at Massachusetts General Hospital in Boston. He noted that 92 percent of participants in the Phase 3 pivotal TRAILBLAZER-ALZ-2 trial of donanemab, which recruited in 2020 and 2021, were non-Hispanic white. In the Phase 3 CLARITY-AD trial of lecanemab, that number stood at 76 percent, while 12 percent were Hispanic and 2.5 percent black. Owing to Eisai’s recruitment in Japan, 17 percent were Asian. Despite some remaining underrepresentation, especially for black and African American people, Jackson thinks recruitment of diverse populations in ADRD studies is outperforming that in other disease areas. Jackson heads a study called Foundations of Representative Engagement, Valid, and Effective Recruitment. Funded by the NIH, FOREVER is surveying attitudes about research participation in diverse communities at 21 ADRD research sites to learn about knowledge gaps among study staff, and recruitment challenges at each site. The idea is to help scientists engage with the communities they serve, Jackson said.
For her part, Wilkins leads the Recruitment Innovation Center at Vanderbilt, which develops tools to boost participant enrollment and retention in clinical trials. The center works with study leaders from across the country to develop community-based recruitment strategies tailored to each study’s target population. They specialize in equipping researchers with the knowledge and resources to connect with, recruit, and retain participants from underrepresented groups.
One such tool, called Faster Together, is a free, eight-module online course that teaches strategies for recruiting marginalized racial and ethnic groups. More than 4,000 people have enrolled so far, Wilkins said.
Building community partnerships that improve participation requires time and effort. Mark Gluck, Rutgers University, Newark, New Jersey, titled his talk, “There is no quick, easy, or cheap way to recruit older African Americans to aging and brain health research.” Gluck leads the Pathways to Healthy Aging in African Americans study of 500 cognitively healthy black/African American volunteers age 60 and older from the greater Newark area. Enrollees come for yearly, three-day visits of health assessments, cognitive tests, and brain scans.
The study has been ongoing for 17 years, but nine years of work in the community laid the groundwork before study recruitment began. A 16-person community engagement team connected with 40 local churches, 13 public/low-income housing areas, as well as local businesses, medical clinics, and senior centers. The team holds community events, such as dances, fitness classes, even a classic car show aimed at African American men, who are notoriously challenging to recruit, Gluck said. At these events, the scientists share information about the pathways study while also offering other resources of interest, for example information about diabetes or COVID vaccines.
Reinforcing Gluck's message that building trust and partnerships requires time and resources, Raina Croff, of Oregon Health and Sciences University in Portland, urged outreach efforts to address the broad health needs of the community. Croff noted that community outreach and study recruitment tend to get handled by the same staff, arguing that outreach should be distinct from recruitment. This would enable outreach specialists, who ideally have experience in the communities they serve, to interact with the community in a less transactional way. “We need funding to support such positions,” she said.
Morris agreed, adding that community outreach work is not specifically earmarked as part of the recruitment budget in most NIH-funded research studies. Funding for community-based recruitment strategies to spur more diverse ADRD studies is highlighted as a top priority in the NIA’s 2024 budget. Partly pursuant to a 2022 RfA about community-based networks to promote inclusion of underrepresented groups, such funding is increasing.
New IDEAS About Biomarkers
The advent of amyloid-targeted drugs has heightened the urgency behind calls to broaden access to, and ensure the accuracy of, Alzheimer's disease diagnostic biomarkers across populations. In St. Louis, Wilkins showed findings from a secondary analysis of the Imaging Dementia-Amyloid Scanning study. IDEAS asked if amyloid scans improve health outcomes of some 20,000 Medicare recipients with MCI or dementia (Wilkins et al., 2022; Rabinovici et al., 2019). About 90 percent of them were white.
The original study had not broken down the data by race or ethnicity, but when Wilkins and colleagues did so, they found that people who identified as Asian, Hispanic, or black were 53, 32, and 29 percent less likely to have a positive amyloid PET scan, respectively, than were white people. This finding raises questions about differences in the etiology or prevalence of AD across different racial and ethnic groups. It also implies that fewer members of these groups would be eligible for amyloid-targeted therapies, or be likely to benefit from them, Wilkins said.
To understand these differences, Wilkins co-leads “New IDEAS.” This sequel to the original aims to enroll 7,000 people, including 4,000 who are black/African American or Hispanic/Latino, plus 3,000 from any ethnoracial group, including non-Hispanic whites. As with the first iteration, New IDEAS includes Medicare recipients diagnosed with MCI or dementia, whose physician has referred them to a PET center. Wilkins said this poses a barrier, because fewer people from underrepresented groups are referred for PET, and some lack Medicare coverage or use Medicare Advantage, a form of Medicare many radiology centers do not accept. To remove this barrier, Wilkins and colleagues are securing funding to cover co-pays for PET scans and provide transportation to scanning centers.
To reach the desired groups in the first place, they are rolling out culturally adapted recruitment materials to targeted populations. They also have paused enrolling non-Hispanic white people to focus on minority groups. As a result, minority recruitment is on the rise, with 4,809 enrolled so far, including 1,123 African American and 1,018 Hispanic/Latino people, Wilkins told Alzforum.
Other biomarker studies are following suit. ADNI, the poster child of AD biomarker studies, between 2004 and 2020 screened 3,739 participants, of whom a combined 11 percent were black, Latino, or Asian even though these groups constituted a quarter of the population at the time (Ashford et al., 2022).
To help diversify ADNI, Godfrey Coker, of the Icahn School of Medicine at Mount Sinai in New York, joined the study. In St. Louis Coker described the ADNI Diversity Task Force formed in 2020. The group includes 15 engagement and recruitment specialists and 13 Diversity Recruitment Hub sites. Each site was awarded $150,000, plus the task force offered consultation, community engagement resources, and locally tailored social media campaigns. The task force dropped ADNI’s requirement for lumbar puncture, a procedure minority groups are particularly reticent to undergo. They also agreed that participants could learn their amyloid PET scan result, a common request among all research participants. In return, each site was expected to enroll five to 10 black or Latino participants. A San Francisco-based marketing firm was hired to create targeted Facebook ads and locally tailored ADNI landing pages (image below). The task force also involved local grass roots organizations, Coker said.
Custom Tailored. ADNI landing pages try to attract specific ethnocultural populations [Courtesy of Godfrey Coker, ADNI.]
Is it working? Coker reported that three-quarters of all new ADNI enrollees at participating sites came from underrepresented groups, bumping up the cohort’s overall proportion of non-whites by 20 percent.
The tools are also being used in ADNI4, which is just beginning, Coker said. This round of ADNI aims to screen 30,000 people, including 50 to 60 percent from underserved populations, which Coker said will include people of all ethnoracial groups with lower socioeconomic status and less education than the average ADNI participant.
Sid O’Bryant, of the University of North Texas in Fort Worth, considers less than 50 percent underserved groups inadequate for any ADRD study these days. Aiming higher will require changes to how clinical studies recruit, O’Bryant said. “We stay in our clinics, ignoring the research documenting that people from underserved communities do not present to specialty clinics,” O’Bryant said. To his mind, solving this problem boils down to this: “We need to meet the community where they’re at.”
O’Bryant works predominantly with Hispanic/Latino and black/African American people in the Dallas-Fort Worth area. He told Alzforum that going to the community works for any population not typically seen in a specialty clinic, including non-Hispanic whites who don’t fit the more highly educated, wealthier profile of the average memory clinic patient.
As the leader of the Institute for Translational Research at UNT, formed with an NIH grant in 2018, O’Bryant and his team frequently visit local churches, bingo games, community centers, and people’s homes, focusing on Hispanic/Latino and African Americans. To build the trust and partnerships that underpin study participation, O’Bryant said his team approaches a community not by asking them to join; instead, they first ask about needs and concerns, listen, and then try to serve each community’s needs. As does ADNI, O'Bryant lets participants see their data, including MRI and PET scans and cognitive test results.
O’Bryant’s team has used this approach to recruit participants in the Health & Aging Brain Study-Health Disparities (HABS-HD). Like ADNI, this multicenter, longitudinal study collects imaging, fluid biomarker, cognitive, and other health data on its participants. Unlike ADNI, HABS-HD was originally designed with community engagement recruiting to draw non-white participants. The study aims to recruit 1,500 Hispanic/Latino, 1,500 black/African American, and 1,500 non-Hispanic white participants. So far, 1,600 Hispanic, 1,000 African American, and 1,300 whites are on board. This success shows the power of community-based approaches, O’Bryant said.
In St. Louis, O’Bryant presented initial data on stark differences between amyloid PET positivity, and other AD biomarkers, across racial and ethnic groups. Researchers at CTAD did as well. For the highlights of these biomarker findings from both meetings, see Part 2 of this series.—Jessica Shugart
From St. Louis to Boston, Scientists Grapple with Ethnoracial Divide in AD Biomarkers
Evidence suggests that ethnoracial groups in the U.S. have different rates of age-related dementias, with Hispanic/Latino and black or African American people generally being more affected and Asian-Americans being less affected than white people. Socioeconomic status also plays a role. It now appears that the predictive value of imaging and fluid biomarkers for AD can also vary by race or ethnicity. The advent of AD-targeted therapies underscores the importance of understanding these differences, but doing so is difficult so long as most participants in AD research and drug studies are white. At Enhancing Participation by Minoritized Groups in Alzheimer’s Disease and Dementia Research, a conference held October 3 to 4 at Washington University School of Medicine in St. Louis, stakeholders debated community-based recruitment strategies (see Part 1 of this series).
Three weeks later, at CTAD in Boston October 24 to 27, scientists showcased findings on ethnoracial differences from biomarker studies. Counterintuitively, perhaps, they reported lower rates of amyloid PET positivity among black and Hispanic relative to white participants. Overall, there was a push for more research on the underlying health and sociodemographic factors that underlie these differences.
Amyloid PET
In St. Louis, Consuelo Wilkins of Vanderbilt University in Nashville, Tennessee, had reported ethnoracial differences in amyloid PET positivity in the Imaging Dementia-Amyloid Scanning, aka IDEAS, study of some 20,000 Medicare recipients with MCI or dementia, 90 percent of whom were white (Wilkins et al., 2022; Rabinovici et al., 2019). In a secondary analysis, Wilkins and colleagues found that Asian, Hispanic, or black people were 53, 32, and 29 percent less likely to have a positive amyloid PET scan, respectively, than were white people.
New IDEAS attempts to understand these differences. It aims to enroll 7,000 people, including 4,000 who are black/African American or Hispanic/Latino, plus 3,000 people who can come from any ethnoracial group, including non-Hispanic whites. Wilkins heads recruitment strategies for the new study (link to Part 1). At CTAD, Charles Windon of the University of California, San Francisco, reported demographic details of the current enrollees. Twenty-two percent are black or African American, 19 Hispanic or Latino; 57 are neither.
Relative to the original, New IDEAS more than quadrupled the percentage of participants with few years of education. This varied substantially by ethnoracial background, with 36, 53, and 21 percent of black, Hispanic, and white participants reporting less than a high school degree, respectively. Importantly, a third of New IDEAS enrollees reported a household income of less than $75,000 per year, the current median in the U.S. This baseline characteristic differed by only 5 percent across ethnoracial groups. Finally, Hispanic or black participants were more cognitively impaired than their white counterparts.
Windon shared a glimpse of amyloid-PET results from the first 3,707 scans of New IDEAS. At 60 and 61 percent amyloid positivity, black and Hispanic people were slightly less likely to test positive for amyloid than non-black or Hispanic participants, of whom 68 percent did. Men and women were equally likely to have amyloid across ethnoracial groups. This differed from the original IDEAS study, in which women were more likely to be amyloid-positive, Windon reported. He has not yet stratified the PET data by other sociodemographic factors, cognitive status, or comorbidities.
At CTAD, Robin Wolz of Imperial College London presented amyloid-PET findings from Bio-HERMES. This biomarker study enrolled 956 participants with normal cognition, MCI, or mild dementia from 16 sites across the U.S. and Canada. White people comprised 76 percent of the cohort, while 10.7 percent were black or African American, and 11.4 percent Hispanic or Latino. Wolz reported that amyloid positivity on PET scans correlated with cognitive status across all ethnoracial groups. Among white people, 37 percent were amyloid-positive, compared to 26 percent of black/African Americans and 35 percent of Hispanic/Latino people. Interestingly, despite differences in amyloid positivity between blacks and whites as gauged by a central visual read, there were no differences in global cortical SUVRs across ethnoracial groups. To understand this discrepancy, Wolz suggested probing the relationship between regional SUVR and global amyloid positivity by race.
Amyloid-PET findings from the Health & Aging Brain Study-Health Disparities (HABS-HD) also revealed differences in amyloid positivity between groups. This biomarker study uses a community-based recruitment strategy to draw underrepresented groups, and has so far enrolled 1,600 Hispanic, 1,000 African American, and 1,300 white people (see Part 1). So far, 3,000 amyloid-PET and nearly 2,000 tau-PET scans are in the database, and in St. Louis, Sid O’Bryant of the University of North Texas in Fort Worth presented initial findings from the amyloid scans.
Jibing with other studies, O'Bryant saw that non-Hispanic whites were likelier to have positive amyloid scans than their African American or Hispanic counterparts. The difference was biggest among those with normal or mildly impaired cognition. For example, among those with MCI, 12 percent of African American and 10 percent of Hispanic people were amyloid-positive, compared to 33 percent of whites. Among those with dementia, only 43 percent of Hispanic or Latino people tested positive for amyloid, compared to 60 percent of African Americans and 63 percent of whites.
If It's Not Amyloid, Then What Is It?
O’Bryant also reported that MRI scans indicated a greater prevalence of neurodegeneration, starting at a younger age, in Hispanic/Latino than non-Hispanic white participants. He showed no tau-PET results but said that preliminary tau-PET findings point to Hispanic/Latino people being positive at younger ages than whites. O’Bryant suggested that the higher prevalence of neurodegeneration and tau pathology, despite lower prevalence of amyloid, among Hispanic or Latino people, may come down to co-morbidities. For example, preliminary findings from the HABS-HD study tied tau pathology to duration of diabetes. Hispanic/Latino people are likelier to develop diabetes than non-Hispanic whites, and do so about a decade earlier, O’Bryant said. He mentioned that this correlation between tau and diabetes duration was also seen in African Americans in this study.
At CTAD, Susan Landau of University of California, Berkeley, presented a variation on this theme. She compared the relationship between Aβ, tau, and cognition in two different cohort studies: ADNI, a clinic-based cohort that skews toward college-educated, wealthier, white participants, and U.S. POINTER, a community-based, more ethnoracially diverse cohort that preferentially recruits people with higher cardiovascular and other dementia risk factors.
Strikingly, while a similar proportion of age, sex, and CDR-matched participants from these two cohorts tested positive for amyloid, Landau found that they differed markedly on amyloid’s relationship with tau and cognition. In ADNI, amyloid positivity explained twice as much of the variability in tau accumulation, as measured by tau-PET, as it did in POINTER. What’s more, a combination of amyloid and tau accumulation correlated strongly with flagging performance on the preclinical Alzheimer’s cognitive composite (PACC) in ADNI, but had no relationship with these test scores in POINTER. Yet, POINTER participants scored lower on the PACC than did ADNI volunteers.
ADNI versus POINTER. Aβ and tau together explained some cognitive variability in the ADNI cohort (right), but not in POINTER (left). Vascular problems explained a small proportion of cognitive variance in POINTER. [Courtesy of Susan Landau, UC Berkeley.]
If neither Aβ nor tau were to blame for memory problems in POINTER, what was? Landau reported that cardiovascular risk scores explained a smidgen of cognitive impairment in POINTER, but not in ADNI, though this effect disappeared when she adjusted for sex, education, and ethnoracial status. High blood pressure and white-matter hyperintensities—both indicators of vascular problems—accounted for a small amount of cognitive variance in POINTER, but not in ADNI. Being part of an underrepresented ethnoracial group (URG) influenced cognitive performance in both studies, but had more sway in POINTER.
This still left a lot to be explained. ADNI director Michael Weiner, University of California, San Francisco, pointed to a big difference between the cohorts: 45 percent of POINTER participants reported smoking at some point in their lives, versus 9 percent of ADNI participants. Besides smoking history, both studies have collected information on myriad additional health variables, such as diabetes, BMI, antihypertensive drugs, as well as social variables. Landau told Alzforum that so far, her analyses have ruled out smoking and diabetes as major contributors, but there are many more to examine.
Plasma Aβ42/40: The People’s Marker?
At CTAD, WashU scientists Suzanne Schindler and Chengjie Xiong each reported findings on a related issue: If amyloid PET generates different results across ethnoracial groups, how about plasma biomarkers? Schindler had previously reported that plasma phospho-tau markers appeared less predictive of brain amyloid positivity among African Americans than whites, though the current, intense search for canonical p-tau blood tests is changing the state of the field rapidly (Schindler et al., 2022). At CTAD, Schindler reported that among volunteers at WashU’s Knight ADRC, plasma Aβ42/40 ratio, as measured by C2N Diagnostic’s Precivity-AD test, picked out amyloid PET positivity equally well in both groups. Its predictive value held up regardless of age, sex, hypertension, diabetes, and BMI, Schindler reported.
For his part, Xiong compared plasma Aβ42/40 ratios between white and black people enrolled in SORTOUT-AB, a longitudinal study to evaluate potential racial differences in AD biomarkers in participants from the WashU, University of Pennsylvania, and University of Alabama ADRCs. He found that, on average, black people had a slightly lower ratio at baseline, which was driven by a higher average concentration of Aβ40. However, longitudinal change in the ratio was similar across racial groups. Importantly, plasma Aβ42/40 tracked similarly with other AD measures, including amyloid PET and cognition. In all, Xiong said, while black people may have a lower rate of amyloid positivity than whites, the plasma ratio of Aβ42/40 is equally predictive of amyloid in both (Xiong et al., 2022).
To Schindler's mind, the findings support the idea AD biomarkers can be found that work equally across ethnoracial groups. Even for those that don’t, Schindler urged the field to resist the temptation of using race-based biomarker cutoffs, for example by setting a lower threshold for biomarker abnormality to enroll minority groups into trials. This could select people who are less likely to respond to AD-specific treatments, Schindler noted. Rather, she called for a deeper understanding of the sociodemographic and health factors that underlie ethnoracial differences.
Overall, biomarker data available thus far point to the idea that people from underrepresented groups are less likely to have brain amyloid but more likely to be impaired. For some, this could mean that they have a non-AD dementia. For some, it could mean that they have less cognitive resilience, hence become symptomatic at a lower amount of amyloid pathology. These types of differences could arise from comorbidities, and/or from social determinants of health (SDOH).
At both the WashU conference and CTAD, scientists called for a greater understanding of SDOH, defined as the conditions in which people are born, grow, learn, work, live, and age. Some think SDOH may underlie differences in disease, biomarkers, and also in research participation, among people from different backgrounds.
SDOH are manifold and difficult to capture. Researchers are trying anyway. Megan Zuelsdorff, of the University of Wisconsin, Madison, co-chairs an NIA panel tasked with developing a way to measure these social determinants. In St. Louis, she described work toward a uniform scale. In its current iteration, the questionnaire is a 10-minute, self-administered test that takes stock of socioeconomic status, access to transportation, financial stress, healthcare experiences, discrimination, social connections, activities, and environment.
Others at the meeting presented pilot studies of SDOH batteries, including WashU’s Marissa Streitz and UPenn’s Shana Stites. Both teams had created their own SDOH batteries, then harmonized them to become the Social and Structural Determinants Influencing Aging and Dementia (SS-DIAD). By incorporating SDOH scales into their ADRC studies, researchers plan to investigate how these social factors relate to AD risk and pathogenesis, biomarkers, and research participation, and to what extent SDOH explain differences between groups.—Jessica Shugart
Xiong C, Wolk DA, Lah JJ, Gleason CE, Roberson ED, Benzinger TLS, Schindler SE, Fagan AM, Hassenstab JJ, Moulder KL, Balls-Berry JE, Sperling RA, Johnson KA, Levey AI, Johnson SC, Luo J, Gremminger E, Agboola F, Grant EA, Ances BM, Gordon BA, Hornbeck RC, Massoumzadeh P, Keefe SJ, Dierker D, Gray JD, Andrews J, Henson RL, Streitz M, Manzanares C, Qiu D, Mechanic-Hamilton D, Stites SD, Shaw LM, Midgett S, Morris JC.
SORTOUT-AB: A Study of Race to Understand Alzheimer Biomarkers.
Alzheimer's & Dementia, 07 December 2022
Further Reading
No Available Further Reading
Treat Before ‘Aβ Bothers Tau,’ Scientists Say at CTAD
Some researchers have long argued for starting amyloid immunotherapy early, before tangles spread and neurons die all over the brain. At the 16th Clinical Trials on Alzheimer’s Disease conference, held October 24 to 27 in Boston and online, they added flesh to the bone of this idea. Despite coming from different antibodies—donanemab, lecanemab, gantenerumab—the findings paint a convergent picture. In short, participants at the earliest stages of the respective cohorts enrolled in each trial gained the most cognitively from treatment. The findings are preliminary, often involving post hoc analyses of small numbers of participants remaining from large trials. Even so, they stirred excitement in Boston.
In one striking tease, about two-thirds of participants with very early AD who took lecanemab actually improved on the CDR-SB over 18 months, compared with about one-third of a matched placebo group. Other findings offered the first concrete indication that amyloid immunotherapy may be able to prevent AD dementia. In the Dominantly Inherited Alzheimer Network secondary prevention trial, presymptomatic mutation carriers taking gantenerumab for eight years had half the odds of developing symptoms as did those on placebo.
“I think we will deliver on the promise of changing the course of the disease. That day is coming,” Randall Bateman of Washington University in St. Louis, who leads the DIAN Trials Unit, said after accepting this year’s CTAD Lifetime Achievement Award. As Reisa Sperling of Boston’s Brigham and Women’s Hospital quipped, “Who knew? Maybe earlier is better.”
Tangle Load Is Key. People with few tangles, below 1.10 tau PET SUVR (top row) fared better on amyloid immunotherapy than did those with intermediate (middle, 1.10 to 1.46 SUVR) and high (bottom) tangle loads. [Courtesy of Eisai.]
On Donanemab, Few Tangles, Young Age Independently Boost Efficacy
Pharma-led Phase 3 trials for all the large antibody programs have enrolled people with mild cognitive impairment or mild dementia due to AD, leaving open the question of whether the drugs might work better at even earlier stages. To guesstimate, researchers are parsing the data they have. In other words, they are comparing efficacy in people at the earliest-stage and the latest-stage ends of the cohorts they enrolled. Modeling is involved, too.
Mark Mintun of Eli Lilly presented one such post hoc analysis from Trailblazer-Alz2. This Phase 3 study of donanemab used tau PET screening to select people with intermediate tau levels, defined as 1.10 to 1.46 SUVR, for the primary analysis population. The trial also enrolled people with higher loads, to gather data on how that group responded. Because of this, all 1,736 participants, not just a substudy, had baseline tau PET scans. Lilly researchers reasoned that the intensity of the tau PET signal could serve as a proxy for disease stage, an assumption that has generated praise but also questions about why the company did not formally test donanemab against placebo in amyloid-positive people below 1.1 tau PET SUVR.
Fewer Tangles, Better Results. Modeling based on donanemab Phase 3 data predicts that people with the lowest tau PET SUVRs (x axis) will maintain their cognitive abilities the best (black bars). [Courtesy of Eli Lilly.]
When Lilly scientists correlated baseline tangle load with slowing of cognitive decline for each person, they found an inverse relationship. The fewer tangles a person had at the outset, the better that person maintained his or her cognitive abilities while on donanemab. Modeling based on these data predicted that for people with a tau SUVR of 1, cognitive decline would slow by 60 percent on the iADRS and 45 percent on the CDR-SB. At an SUVR of 2, by contrast, those numbers would be as low as 10 and 20 percent (see image above). This compares with measured values of 35 and 36 percent in the intermediate-tau trial population.
Results were similar when researchers measured disease stage in another way, by baseline plasma p-tau217. In the lowest tertile, donanemab worked better than in higher tertiles.
What about even earlier disease stages? In Boston, Mintun said that 1,053 people who did not meet trial screening criteria, due to having tau PET scans below 1.10, were enrolled into an open-label “addendum” study. There was no placebo control for these participants, hence the researchers could not calculate slowing of cognitive decline on donanemab. Even so, biomarker outcomes echoed those for other treated groups, with a rapid drop in amyloid plaques, plasma p-tau217, and the inflammatory marker GFAP. The data hint that this very-low-tau group would also benefit cognitively from treatment, Mintun said, though that remains to be shown.
Besides tangles, the researchers found that a participant’s age independently affected how well they fared on donanemab. Among the half of participants with lower p-tau217 at baseline, those younger than the trial’s median age of 75 slipped 43 percent less on the iADRS than did matched placebo controls. Those 75 or older benefitted less, slipping 30 percent less. For people in the upper half of baseline p-tau217, age made even more difference. Those younger than 75 had a 30 percent slowing of decline, those older, only 8 percent. Age and tangle load have additive effects on efficacy, Mintun concluded.
Because people joined this trial at different points in their disease, the whole cohort encompassed about 10 years of AD pathogenesis, Mintun noted. Lilly researchers used their data to model how a person’s disease stage at baseline might affect their drug response. The model predicted dramatic stage-based differences. For example: People who were two years less advanced than the cohort average could achieve as much as 88 percent slowing of decline on the iADRS, almost stalling their disease. On the other hand, those two years more advanced than the cohort average would gain but a negligible benefit of 6 percent slowing. One reason for this stark difference is that cognitive decline accelerates as disease advances, Mintun said.
“This gives us great urgency in thinking about how to diagnose and prepare patients for treatment in the future,” Mintun concluded.
Lecanemab: Did I Hear “Improvement”?
These donanemab analyses jibe with data from the Phase 3 Clarity trial of lecanemab. In Boston, Keith Johnson of Massachusetts General Hospital, Boston, presented an exploratory analysis relating the baseline tau PET signal in Clarity to its outcome. Johnson divided the trial’s tau PET substudy of 342 people into three groups: low tau, defined as below 1.06 SUVR; medium tau, 1.06 to 2.91; and high tau, above 2.91. These values were chosen based on the different patterns of tangle accumulation they represent: entorhinal cortex and hippocampus for the low-tangle group, those plus limbic regions for medium-tangle, those plus neocortex for high (see image at top of story). There were 141 people in low, 191 in intermediate, and 10 in the high groups, respectively. The latter two groups were combined for analysis. In each group, about half were on drug, half on placebo.
Because Clarity participants were not screened by tau PET, the cohort included a broader range of tangle accumulation than did Trailblazer-Alz2. The low-tau group in particular has not been analyzed in previous antibody trials, Johnson noted. Stephen Salloway of Butler Hospital, Providence, Rhode Island, asked whether this group had different clinical characteristics from higher tau cohorts at baseline. Johnson said, to his surprise, they did not.
How did their outcomes differ? On amyloid plaque clearance, the low-tangle group had much lower baseline amyloid than the intermediate/high group, 36 versus 94 centiloids. Because of this, they cleared plaque faster, falling below the threshold for amyloid positivity, here set at 30 centiloids, three months into the trial. This compared with 18 months for the intermediate/high group. At the individual level, 93 percent of low-tangle participants were amyloid-PET negative by the end of the study, compared with 57 percent of the higher-tangle group.
On tau biomarkers, however, treatment effects were bigger in the intermediate- than low-tangle participants. In the high-tangle group, plasma p-tau181 dropped by twice as much as in low-tau participants, and tangle progression slowed across every brain region examined. In low-tau participants, progression slowed only in medial temporal regions, likely because tangles had not reached the other regions yet.
What about cognition? On placebo, the low-tangle group lost less ground than higher-tangle groups, staying stable on the CDR-SB and slipping only slightly on the ADAS-Cog14 and ADCS MCI-ADLs over 18 months. In contrast, low-tangle participants on lecanemab stayed basically stable on the latter two measures, and bettered their baseline scores on the CDR-SB. Improvement was defined as scoring at least 0.5 points higher than before.
Overall, 76 percent of low-tau participants on lecanemab, and 55 percent on placebo, held their ground on the CDR-SB over 18 months. Remarkably, 60 percent of the cohort improved over baseline, compared with 28 percent of those on placebo. Percentages for the other two cognitive tests were similar. Johnson did not show cognitive data for the intermediate/high tau group alone.
The percentage who improved on drug grew over the course of the trial, agreeing with other data suggesting cognitive benefits take a few months to appear. The number of participants in these groups was small, with only about 50 people on lecanemab having 18-month data. Because of this, p values for many differences were nominal.
The findings caused a stir in the room and later in the hallways. One audience member asked what might account for improving cognitive scores. Michael Irizarry at Eisai speculated that it could be due to the mechanism of action, for example, lecanemab clearing protofibrils that were harming synaptic function. Sperling suggested that removing protofibrils before they have “bothered tau” in the early entorhinal/hippocampal areas might forestall tangle spread into further areas of the brain.
For her part, Sperling presented 24-month open-label extension data. In the OLE, low-tangle participants who switched from placebo to lecanemab appeared to stabilize on cognitive tests, losing no ground between 18 and 24 months. By contrast, the treatment group who had been on lecanemab since the beginning of the trial slipped on all three tests at 24 months relative to 18 months, nearly to the scores of the former placebo group. This type of convergence between groups would be expected for a symptomatic therapy, but not a disease-modifying one. Sperling noted that an early glimpse at 30-month data suggests the two groups are separating again; however, only about half the cohort has reached this time point.
Improving Cognition? On lecanemab (green), more people in the low-tau group stayed stable (left) or improved (right) on the CDR-SB over time than did those on placebo (black). Those who switched from lecanemab to placebo at 18 months also benefitted. [Courtesy of Eisai.]
By the numbers, 79 percent of the low-tau treatment group had not declined on the CDR-SB after 24 months; 50 percent had improved. For the group formerly on placebo, the percentage with no decline on the CDR-SB rose from 55 percent at 18 months to 62 percent at 24 once on lecanemab, and the percentage who improved rose from 28 to 40 percent (see image above).
Only a subset of the Clarity cohort had tau PET scans. To examine a larger group, Sperling used baseline amyloid of below 60 centiloids as a proxy for low tau. She chose this threshold because it encompassed 80 percent of the low-tangle participants. In this larger analysis of below-60 centiloid Clarity participants, the 190 who got lecanemab fared better on cognitive outcomes than did the 190 on placebo, declining by 51, 69, and 72 percent less on the CDR-SB, ADAS-Cog14, and ADCS MCI-ADL, respectively. For the whole Clarity cohort, i.e., low and medium/high tangles combined, these numbers were 27, 26, and 37 percent, again demonstrating more gain earlier in disease. Sperling did not show cognitive data for participants whose baseline amyloid was above 60 centiloids. Logic would demand that their cognitive benefit was below that of the combined group.
Altogether, Sperling said, the data support testing lecanemab in preclinical populations. This is being done in the AHEAD 3-45 and Trailblazer-Alz3 studies. Sperling noted that people with few tangles probably have less neuronal damage than those later in disease. “It’s encouraging that if you get there before neurons are dying, maybe you really can improve function,” Sperling said.
Tiny Shoots of Prevention Sprouting from the Data
Though these Phase 3 donanemab and lecanemab data show promise for earlier stages, all participants already had symptoms of Alzheimer’s dementia. What about catching people while they are still asymptomatic? Bateman offered a glimpse at such a future.
The DIAN Trials Unit secondary prevention study lasted 11 years, from 2012 to 2023. The Phase 2/3 study tested gantenerumab or solanezumab against a pooled placebo group. The double-blind period ended in 2019 with a negative result. Asymptomatic participants had not declined cognitively, thus no treatment effect could be measured. However, gantenerumab had strongly affected biomarkers including cerebrospinal fluid p-tau181 and total tau (Feb 2020 news; Apr 2020 news).
For this reason, the researchers offered gantenerumab to all participants in an open-label extension that lasted until August 2023 (see DIAN research update). For most participants, this meant a treatment gap of about two years was followed by two years of OLE. Of the 73 mutation carriers who completed the OLE, 22 were on gantenerumab for about eight years of exposure, while the remainder had switched from solanezumab or placebo. Bateman noted that the gantenerumab dose had been boosted fivefold during the trial, and was increased another threefold in the OLE. This is because initial trial results showed plaque is harder to budge in mutation carriers, who produce more Aβ, than in late-onset AD patients.
A pre-specified subgroup analysis of the people who were on gantenerumab throughout found they were half as likely to progress to symptoms as were those who switched from placebo or solanezumab. Those who did develop symptoms did so six years later than their expected age of onset (EYO). Curiously, people initially on placebo or solanezumab also had a delay in symptom onset, three years beyond their EYO. In the DIAN observational study, by contrast, symptoms started at the expected age. Because many participants remain asymptomatic, the number of people in this analysis was tiny: nine on gantenerumab, four on solanezumab, and seven on placebo. The findings are preliminary; the researchers are currently doing sensitivity analyses.
Pierre Tariot of the Banner Alzheimer Institute in Phoenix asked if these findings would generalize to late-onset AD. Bateman believes they will, because immunotherapy results for several antibodies have been similar in autosomal-dominant and late-onset populations. He said DIAN participants will continue to receive treatment with amyloid-lowering drugs, but did not say which drugs will be selected next.—Madolyn Bowman Rogers
Moving Forward: RNA-Targeted Attempts at Taking Down Tau, APP
As the proteopathic drivers of Alzheimer's disease, Aβ and tau have for decades kept scientists trying to understand which of their isoforms and fragments are most to blame for the memory-robbing pathogenic cascade—and which ones might be the best therapeutic targets. Two oligonucleotide therapies circumvent these thorny questions by taking aim at the transcripts encoding each troublemaking protein. At the Clinical Trials on Alzheimer’s Disease conference, held October 24-27 in Boston, exploratory findings from a trial testing an antisense oligonucleotide aimed at tau suggested that not only did the drug dock tangles, it may also have stemmed cognitive decline. A small interfering RNA that douses APP expression markedly reduced cerebrospinal fluid levels of APP cleavage products, including Aβ40 and Aβ42.
Both drugs were safe and well-tolerated. Both studies were far too small to draw conclusions, but they support the idea of starting larger ones.
Biogen’s latest glimmer of hope on its tau ASO, called BIIB080, follows a splashier reveal a few months ago, when the company reported that the drug reduced parenchymal tangle accumulation, as gauged by tau-PET (Apr 2023 conference news). That the drug reduced existing aggregates, as opposed to merely slowing their accumulation, was both exciting and unexpected to many in the field.
At CTAD, Biogen’s Melanie Shulman reported exploratory clinical outcomes in the same Phase 1b trial and its long-term extension (LTE). The trial's placebo-controlled portion enrolled 46 participants with mild AD. Four ascending doses of BIIB080 or placebo were administered in sequential cohorts via intrathecal injection over a three-month period. The two low-dose groups got 10 and 30 mg of BIIB080 every four weeks, respectively, while the two high-dose groups got 60 mg every four weeks or 115 mg every 12 weeks. Shulman noted an important protocol change: Midway through enrollment of one of the low-dose groups, Biogen changed the trial inclusion criteria from CDR 1 to now also enroll less-impaired people with a CDR of 0.5. Therefore, baseline CDR scores were lower for the two high-dose groups relative to the low-dose groups. Participants were tracked with three clinical tests—the MMSE, FAQ, and RBANS—for an additional six months.
For all three measures, Shulman reported a positive trend: At weeks 25 and 37, participants in the two high-dose groups had declined less than those who received one of the lower doses or placebo.
In the LTE, all participants received either 60 mg every four weeks or 115 mg every 12 weeks for a year, and were then monitored for an additional five months. Thirty-three participants enrolled in the LTE, but only 22 had available cognitive outcome data, including 16 who had transitioned from a high dose and six from one of the low-dose groups. While the high-dose participants transitioned seamlessly into the LTE, those who switched from the low-dose groups had a lag of five to 19 months. These low-dose switchers started the LTE with an average CDR of 4.9 sum of boxes, and their performance on cognitive tests was compared between the start and finish of the LTE. In contrast, cognitive changes in participants who had transitioned from one of the high-dose groups were compared across the entire two-phase study—from MAD baseline, when their CDR-SB scores averaged 3.1, to week 100.
To gauge potential drug effects in this open-label phase of the study, the scientists selected external placebo controls from TANGO, Biogen’s previous trial of gosuranemab (Jun 2021 news). Relative to matched TANGO participants, people who received a high dose of BIIB080 throughout both phases of the trial trended toward slower decline on the CDR-SB, MMSE, and FAQ between baseline and week 100 of the extension (image below). The researchers saw similar trends when they used matched ADNI volunteers as external controls, Shulman reported.
Tantalizing Trends. Volunteers in two high-dose BIIB080 groups (top and bottom) appear to have declined less on the CDR-SB (left), MMSE (middle), and FAQ (right) than external controls. [Courtesy of Biogen.]
Even among the six participants who switched from the low to the high dose in the LTE after a delay, Shulman still observed a whiff of benefit on the CDR-SB and MMSE, but not on the FAQ.
BIIB080 was safe and well-tolerated at the doses tested. Most adverse events, such as headache, back pain, and extremity pain, were deemed related to lumbar puncture, Shulman said.
“The results seem promising and are in the expected direction, but they are too preliminary to draw any firm conclusions,” commented Adam Boxer of the University of California, San Francisco. “The results of the ongoing Phase 2 BIIB080 trial will provide important insights into the mechanistic role of tau pathology in Alzheimer’s disease.” Shulman said the Phase 2 CELIA trial will compare the 60 mg dose of BIIB080 to placebo in 735 people with MCI or mild AD.
Einar Sigurdsson of New York University called the exploratory findings promising. “Collectively, this data, in light of some of the failed tau antibody trials, supports our view that targeting intracellular tau is likely to be more efficacious than focusing solely on extracellular tau,” he wrote to Alzforum (comment below).
Another program is taking aim at APP. At CTAD, Catherine Mummery of University College London presented interim safety and biomarker findings from an ongoing single-ascending-dose (SAD) study of ALN-APP. Developed by Alnylam Pharmaceuticals in Cambridge, Massachusetts, and Regeneron Pharmaceuticals in Tarrytown, New York, this small interfering RNA binds to APP transcripts, relegating them to destruction.
Mummery presented interim data from the first three cohorts, together comprising 20 people with early onset MCI or mild AD. Participants in each cohort were randomized 3:1 to receive placebo or a single dose of 25, 50, or 75 mg ALN-APP, also by intrathecal infusion. Depending on the cohort, participants had so far been followed for six to 14 months with safety assessments and lumbar punctures to measure APP fragments and Aβ peptides.
Previously, the researchers had reported that, within two months of treatment with the two highest doses of ALN-APP, CSF concentrations of both APPα and APPβ—cleavage products of α- and β-secretases, respectively—had plummeted. At CTAD, Mummery extended those findings out to six to 10 months, reporting that although both fragments had started to creep upwards after two months, they remained more than 30 percent reduced relative to placebo.
Mummery also showed Aβ peptide findings. Two months after treatment, the 50 and 75 mg doses reduced CSF Aβ42 and Aβ40 by 50 and 70 percent, respectively. Mummery said that phospho-tau is also being measured to look for amyloid-accumulation-dependent changes in tau processing, but she did not yet have that data.
The treatment appeared safe and well-tolerated at the doses tested, with most of the mild to moderate adverse events attributable to lumbar puncture. Mummery said that analysis of safety biomarkers, routine lab assessments, and preliminary measurements of CSF NfL revealed no concerning or significant responses to treatment. While this study continues to the higher-dose cohorts, a multiple ascending dose study is newly underway.
Michael Weiner, University of California, San Francisco, called the findings exciting. Recalling the cognitive worsening that sank BACE inhibitor programs, he asked Mummery if participants were being monitored for that. Mummery replied that cognition is being closely monitored for safety reasons, and no troubling trends have emerged thus far.
Lon Schneider of the University of Southern California in Los Angeles noted that some scientists believe monomeric forms of Aβ could be neuroprotective. Might ALN-APP counteract those potential benefits? Mummery acknowledged that ALN-APP does reduce all products of APP processing, including Aβ monomers, and that the effects of such a broad reduction are not yet known in humans. APP and its metabolites do have physiological functions, she said, and the drug partially lowers the protein. Careful dosing studies will be essential to address this issue, she agreed.—Jessica Shugart
After Long Wait, Aβ Oligomer Detangler Poised for Phase 2
While the trials, tribulations, and successes of Aβ immunotherapy were commanding Alzheimerologists' rapt attention, a different anti-amyloid approach has been quietly moving forward on the sidelines. PRI-002—a small molecule that disassembles Aβ oligomers—is about to start a sizeable Phase 2 trial. The announcement, along with a hint of a cognitive signal in a small Phase 1b trial, were presented at the Clinical Trials on Alzheimer’s Disease conference, held October 24 to 27 in Boston.
It is too early to draw conclusions about whether PRI-002, or the strategy of dismantling Aβ oligomers, will derail the early phases of AD pathogenesis. But the time is ripe to find out, according to Dieter Willbold of Heinrich-Heine-Universität Düsseldorf, Germany, who has ushered development of this drug for two decades. Unlike amyloid-targeted antibodies, which are costly, burdensome to administer, and come with risks such as ARIA, PRI-002 is a small molecule that can be taken as a capsule. Plus, it does not depend on the double-edged sword of the immune response to do its job, Willbold told Alzforum.
Identified in a screen more than 20 years ago, PRI-002, also known as Contraloid or RD2, is a 12-residue D-enantiomer peptide that interferes with assembly of Aβ42 oligomers (Wiesehan et al., 2003).
In 2017, Willbold and others co-founded Priavoid GmbH to develop the drug. Preclinical studies have shown that the compound readily crosses the blood-brain barrier and disassembles Aβ oligomers on the verge of prion-like mischief (e.g., Schemmert et al., 2018; May 2022 news; Kutzsche et al., 2023).
At CTAD, Oliver Peters of the Charité in Berlin presented an exploratory clinical outcome from a Phase 1b study. Top-line results had been reported at AAIC back in 2022. The trial was small, with a simple design: Seventeen people with MCI and two with mild AD were randomized to take daily 300 mg doses of PRI-002 or placebo for four weeks, then followed for another month. The drug appeared safe, at least over such a short period. Not only were there no serious adverse events, but more people in the placebo group had a mild adverse event than those on drug.
Relative to participants on placebo, those who took PRI-002 performed better on the CERAD word list, a measure of memory function. The effect reached significance one month after treatment stopped, Peters claimed. Memory scores improved in all nine participants on treatment; they improved in four of the 10 who received placebo.
More Words on Drug? CERAD wordlist score improved for all participants who received PRI-002, compared to half who got placebo. [Courtesy of Oliver Peters, Charité Berlin.]
The trial also measured change on the CDR-SB. There were no differences between groups on this measure. Willbold said that the CDR-SB would not be expected to change after such a short treatment period.
A Phase 2 trial is slated to begin in early 2024. It will enroll at least 270 participants with MCI or mild AD at 40 trial sites in seven European countries. The trial will randomize volunteers to take daily doses of 300 mg or 600 mg PRI-002, or placebo, for one to two years. To generate information about treatment duration effects while also shortening the trial, Willbold said that earlier enrollees will be treated for the full two years, while later recruits will only go for a year. Change in CDR-SB will serve as the primary efficacy endpoint. Though this measure did not budge in Phase 1, it would be expected to do so in a longer trial, and is a standard outcome in future, pivotal studies, Willbold said. The trial is funded by SPRIN-D, the German Federal Agency of Disruptive Innovation.
After Peters' talk, Benjamin Wolozin of Boston University asked if similar cognitive benefits had been observed among healthy volunteers in earlier dosing studies, as this could address the question of whether the effects of the compound are due to its interaction with Aβ oligomers, or perhaps some other target. Priavoid conducted a single-ascending- and a multiple-ascending-dose Phase 1 trial prior to this one, but Peters said cognition was not tracked in healthy participants. Peters noted that the drug did lower CSF Aβ oligomers in the Phase 1b trial, suggesting that it hit its target. Echoing Wolozin, Robert Rissman of the University of California, San Diego, noted that PRI-002’s apparent CERAD benefit emerging after a short treatment period could imply that it exerts a more general rather than a disease-specific effect. This key question will be addressed in the longer, larger Phase 2 trial, Peters said.
Willbold told Alzforum that he would not expect to see a benefit in cognitively healthy people, because no relevant off-targets of PRI-002 had been identified in receptor screens.
Despite years of drug development effort, there are few FDA-approved D-peptide drugs. One is Octreotide, a synthetic form of the hormone somatostatin that has 2 D-amino acids in its structure. Additional experimental ones are being explored in indications including Covid, cancer, and Alzheimer's, for example, one against tau (Aillaud et al., 2022).—Jessica Shugart
Current amyloid immunotherapies poorly enter the brain, with only one in a thousand antibodies getting through. For several years, researchers have been exploring alternate delivery methods that could allow for lower dosing and higher potency. Some fruits of that labor were on display at last month’s Clinical Trials on Alzheimer’s Disease conference, held October 24 to 27 in Boston and online. Roche scientists debuted the first efficacy and safety data for their new antibody trontinemab, which links gantenerumab to a “brain shuttle” that ferries it across the blood-brain barrier. In a small study of 44 people, trontinemab exceeded Roche’s expectations, completely clearing plaque from three-quarters of participants within six months.
Another talk presented a different way to pierce the barrier—using ultrasound to stretch vessel walls in local areas, making them leaky enough for aducanumab to pass. This strategy cleared half of deposits in those regions within six months.
Importantly, preliminary results suggest these methods may cause less ARIA than conventional antibody delivery. Why? Scientists showed that circulating antibodies end up in cerebrospinal fluid long before they reach the brain’s parenchyma. CSF communicates with perivascular spaces around large arteries, where cerebral amyloid angiopathy abounds. Thus, vascular amyloid may be the first form that conventional antibodies encounter. Delivering antibodies directly through the blood-brain barrier avoids this danger, because smaller vessels have less CAA. Other talks in Boston hammered home that ARIA represents an inflammatory reaction to CAA, an idea catching fire in the field (Aug 2023 conference news).
Per-Ola Freskgård of BioArctic called the brain shuttle data fantastic. He previously worked on the program at Roche, and is now developing a similar shuttle at BioArctic. “It’s a game-changer, not only in Alzheimer’s disease but for the use of biologics for brain disorders in general,” he told Alzforum. “I think the whole field will move into this space of active transport of antibodies.”
Conventional versus Shuttle. In mouse brain, conventional antibodies (green, left) linger in the choroid plexus and ventricles five days after infusion. A brain shuttle (right) delivers them evenly throughout brain parenchyma. [Courtesy of Roche.]
Meet the New Kid—Trontinemab Struts Its Stuff
Roche’s brain shuttle is a Fab antibody fragment that binds the transferrin receptor. After binding this receptor on vascular endothelial cells, the shuttle and its cargo are internalized, passing through the blood-brain barrier via active transport. Previously, Roche had reported data from a Phase 1 single-ascending-dose study, which found eightfold higher brain uptake of trontinemab than gantenerumab (Mar 2021 conference news). Analyzing pharmacokinetic data, the scientists predicted that 3.6 mg/kg monthly intravenous trontinemab might lower plaque load by a quarter after six months (Dec 2021 conference news).
In Boston, Roche’s Luka Kulic reported interim data from a multiple-ascending-dose study. It turned out trontinemab handily beat that estimate. The Phase 1b/2a trial has so far enrolled 44 participants who have mild cognitive impairment or mild dementia due to AD. Participants were 70 years old on average, with an average MMSE of 20 and between 90 and 100 centiloids of plaque at baseline. To date, the trial has tested doses of 0.2, 0.6, and 1.8 mg/kg; each dose cohort consists of about 15 people, of whom three receive placebo. For comparison, the gantenerumab Graduate trials administered roughly 15 mg/kg monthly, eight times more than the highest dose here.
Going, Going, Gone. Six months of trontinemab lowered amyloid dose-dependently, with the highest dose of 1.8 mg/kg (dark blue) completely clearing plaque. [Courtesy of Roche.]
The results? After six months, people receiving 0.2 mg/kg had lost about 20 centiloids, those on 0.6, 31. At the 1.8 mg/kg dose, plaque clearance accelerated dramatically, with 62 centiloids having vanished by three months and 84 centiloids by six. At this dosage, three of four participants dropped below the amyloid positivity threshold of 24.1 centiloids (see image above). A fourth and final dose cohort getting 3.6 mg/kg is now fully enrolled; those data will be presented next year, Kulic added.
Based on these data, Freskgård estimated trontinemab’s efficacy might be 40- or 50-fold that of gantenerumab.
The treatment produced side effects, but none so far that are showstoppers. Most common were infusion-related reactions; they occurred in three of four participants on the 1.8 mg/kg dose. This typically happened at the first infusion, and symptoms were usually mild, consisting of fever and chills. The researchers now give acetaminophen to participants beforehand to ameliorate this reaction.
In addition, 19 percent of people on 1.8 mg/kg developed mild anemia. This could be a result of the brain shuttle construct interfering with iron uptake in red blood cells. It could also be a result of the frequent blood draws in this study, because the placebo group had mild anemia as well, Kulic noted. Because anemia was most common in people whose iron levels were low at baseline, these participants are now being pretreated with iron supplements to bring them up to normal before starting trontinemab. Anemia went away after trontinemab treatment stopped.
Trontinemab also induced anti-drug antibodies. Curiously, ADAs were most pronounced at lower doses, curtailing trontinemab exposure by 70 percent at 0.2 mg/kg, 60 percent at 0.6, and 25 percent at 1.8. Kulic told Alzforum that this phenomenon of bigger ADA effects at lower doses has been seen with other drugs, for example in oncology, but is not fully understood. ADAs were not associated with ill effects in this study, he added.
Now for the big question—what about ARIA? Here the news was good. At the lower two doses, there were no incidents. At 1.8 mg/kg, one person developed ARIA-E, and one ARIA-H, for an incident rate of 7 percent each. The ARIA-E case was mild on MRI, affected the person’s ability to pay attention, and went away after a month. The ARIA-H was asymptomatic and consisted of two occurrences of the leakage called meningeal siderosis.
If the findings hold up in larger cohorts, this would be the lowest rate of ARIA-E yet seen with amyloid immunotherapy. Other antibody programs have reported rates ranging from 12 to 33 percent. Gantenerumab, which forms the backbone of trontinemab, triggered ARIA-E in 22 percent of people in its Phase 3 trial.
Pinning the Blame on Vascular Amyloid
Why does trontinemab cause less ARIA? Possibly, because it enters the brain differently than do gantenerumab and other standard anti-Aβ antibodies. At a task force meeting before CTAD, Kulic presented mouse data pinpointing where in the brain fluorescently labeled antibodies accumulated five days after dosing. A conventional IgG antibody directed against a neuronal target was found mostly in the choroid plexus, ventricles, and subarachnoid space—in other words, in CSF. That same antibody conjugated to the brain shuttle, by contrast, had broadly diffused throughout the brain, including deep brain structures (see image at top of story).
Freskgård believes that conventional antibodies enter the brain through the choroid plexus, because this tissue’s job is to filter blood. Though a blood-CSF barrier exists, consisting of tight junctions between the ependymal cells that line the choroid plexus, he thinks this hurdle is easier for large molecules to cross than is the blood-brain barrier. Once in CSF, antibodies presumably circulate through ventricles and the subarachnoid space, eventually entering the brain through the perivascular spaces that surround large penetrating arteries. These arteries tend to be laden with the most CAA. Because of this, conventional antibodies likely encounter vascular amyloid long before they see any parenchymal plaques, Freskgård speculated.
At CTAD, Cynthia Lemere of Brigham and Women’s Hospital, Boston, showed unpublished data that supports this idea. When her team injected amyloidosis mice with a standard anti-amyloid antibody once weekly for three weeks, then immunostained for it, they found nearly all the signal around blood vessels in the brain. When they treated for 13 weeks, they found immunoreactivity in plaques as well as blood vessels. The data suggest that antibodies associate with vascular amyloid first, and do so for quite some time, before they reach the parenchyma. That may explain why ARIA develops early in treatment, and why it associates with CAA, Lemere suggested. “Binding of antibodies to CAA is critical for ARIA,” she said in Boston.
Trontinemab, by contrast, is taken up by transferrin receptors in capillaries that pervade the brain everywhere, giving it broad, rapid access to parenchyma. Capillaries carry much less vascular amyloid than do leptomeningeal arteries, potentially limiting the antibody’s exposure to such deposits. This may be why trontinemab causes less ARIA than does gantenerumab, Kulic noted. He added that Roche is exploring other possibilities as well, such as a lower systemic dose or a different pharmacokinetic profile playing a role.
Other talks at CTAD reinforced these ideas. Lars Lannfelt of Uppsala University, Sweden, compared the binding characteristics of lecanemab and donanemab. Both antibodies bound synthetic fibrillar amyloid spiked with pyroglutamated Aβ with similar affinity and, as expected, donanemab did not bind when there was no pyroGluAβ. When it came to CAA deposits, however, donanemab bound with about twice the affinity of lecanemab, as seen by three different assays: immunoprecipitation, surface plasmon resonance, and a type of electrochemiluminescent ELISA. The relative CAA binding strength of each antibody correlates with the amount of ARIA-E it triggers—12 percent for lecanemab, 24 for donanemab—again suggesting a link, Lannfelt said.
Meanwhile, Gioacchino Curiale of Biogen reinforced the evidence that ARIA is an early phenomenon. He presented data from Embark, aducanumab’s ongoing OLE study. Among 1,578 people with MRI safety data, 586 had previously received at least one dose of 10 mg/kg aducanumab in a trial, 784 had never reached the maximum dose, and 208 had previously been on placebo. Curiale found dramatic differences in ARIA rates between these groups, especially comparing participants with full dosing and those with none. ARIA-E rates were four times higher in the treatment-naïve group, 40 percent versus 11, and ARIA-H rates were twice as high, 38 percent versus 18.
ARIA was more severe in those new to treatment, with 22 versus 12 percent of ARIA-E cases symptomatic, and 7 versus 0 percent severe on MRI scans. Those new to treatment were also more likely to have ARIA recur during the three-year study, at 17 versus 3 percent. Participants who had previous low aducanumab exposure fell in between these groups on all measures.
Within the subgroup of those previously treated with aducanumab, however, people who had ARIA-E before were three times as likely to develop it in the OLE as were those who’d never had it, suggesting some underlying vulnerability.
Dennis Selkoe of BWH called these data important, and wondered if the lower ARIA rates in those with previous exposure were due to their starting the OLE with lower baseline amyloid. Curiale said this was likely. This would fit with a model in which removal of vascular amyloid during initial treatment made these participants less susceptible to ARIA.
ARIA Risk Factors. Machine learning identified six variables that affected a person’s odds of ARIA with immunotherapy, five risk and one protective. [Courtesy of Steven Greenberg.]
Further tying ARIA to CAA, Steven Greenberg of Massachusetts General Hospital, Boston, reported that factors that contribute to ARIA risk are all associated with vascular amyloid deposits. Using data from the donanemab Phase 2 and 3 studies, he applied machine learning to parse the contribution of 42 different variables, finding six that correlated with ARIA-E. No surprise, APOE genotype was the strongest risk factor, with two copies of the E4 allele tripling the odds of ARIA-E. Microhemorrhages were the next strongest; having even one at baseline boosted the odds by 40 percent, while two to four microhemorrhages more than doubled risk. Most trials allowed people with up to four baseline microhemorrhages to enroll. Likewise, the presence of one instance of superficial siderosis at baseline doubled the risk of ARIA-E. CAA is more common in E4 carriers, and frequently causes small brain bleeds.
The fourth risk factor was a surprise, however. Having a mean arterial pressure greater than the population norm of 93 raised the risk of ARIA-E by 40 percent, Greenberg reported, noting this is a new finding. Conversely, the use of antihypertensive medication dropped the risk by 40 percent, suggesting hypertension is a modifiable risk factor. Finally, high baseline amyloid of more than 108 centiloids also nudged up risk by 30 percent.
Microbubbles Make Plaque Go ‘Poof’?
Another talk at CTAD proposed a different way to foil the blood-brain barrier. Ali Rezai of West Virginia University, Morgantown, presented a pilot study of whether ultrasound could get more antibody in. Previously, low-intensity ultrasound directed to spots in the brain had temporarily opened the barrier in AD patients without obvious ill effects. The method uses microbubbles injected into the bloodstream; when an ultrasound pulse hits a vessel, the bubbles inside expand, pushing endothelial cells apart. These openings reseal within 24 to 48 hours (Aug 2018 news).
Rezai tested whether this method would boost uptake of aducanumab, the only approved amyloid immunotherapy at the time. For six months, three AD patients whose plaque load reached 200 centiloids received a monthly aducanumab infusion and lay for two hours of ultrasound in an MRI scanner. Ultrasound beams were targeted to brain regions of 40 mL in size; the same region of the opposing hemisphere served as control. Aducanumab was titrated up from 1 to 6 mg/kg during the study, below the effective dose of 10 mg/kg established in Phase 3 trials.
The procedure mopped up plaque quite well. Each participant cleared about half their baseline amyloid in the targeted regions within six months. No ARIA was seen, but APOE4 carriers were excluded. The three people will be followed for five years for long-term safety and efficacy.
Rezai is now enrolling a larger study that will infuse lecanemab and direct ultrasound to larger brain regions. One way or another, scientists have their eyes on conquering the blood-brain barrier.—Madolyn Bowman Rogers
Second-Generation γ-Secretase Modulator Heads to Phase 2
Can γ-secretase modulators stage a comeback? Scientists led by Irene Gerlach at F. Hoffmann-La Roche, Basel, Switzerland, are banking on it. At this year’s CTAD meeting, held October 24-27 in Boston, they reported that RG6289, a second-generation GSM, appeared safe in a Phase 1 trial and shifted Aβ production toward smaller, less-sticky peptides.
Dennis Selkoe, Brigham and Women’s Hospital, Boston, called the data compelling. “As the AD field desperately needs orally available, small-molecule therapeutics to lower amyloid, and potentially prevent AD progression if used early enough, I look forward to the continued development of Roche’s molecule and other well-validated GSM candidates,” he wrote (comment below). Roche plans to start Phase 2 next year.
The trial could re-energize a field that has suffered stinging setbacks. First there were γ-secretase inhibitors, such as semagacestat and avagacestat, which sank once researchers realized that scuttling a protease that processes more than 100 transmembrane proteins was bound to interfere with cognition (Dec 2012 news; Aug 2010 news).
The field pivoted to modulating γ-secretase on the idea of tweaking the enzyme's active site such that it still runs and, in the case of Aβ, runs more efficiently, processing amyloidogenic Aβ42 and Aβ43 into smaller, innocuous peptides such as Aβ37 and Aβ38. In 2001, some non-steroidal anti-inflammatories were reported to act as such, if at low potency (Weggen et al., 2001). Derivatives followed in the 2010s, including those by Chiesi, Forum Therapeutics, and Neurogenetics. None made it to Phase 2.
Scientists then discovered second-generation GSMs, mostly compounds containing an aryl-imidazole group, including Eisai’s E2212, Bristol Myers Squibb’s BMS-932481, and Pfizer’s PF-06648671, which were tested in Phase 1 (Yu et al., 2013; Soares et al., 2016; Ahn et al., 2020). BMS's compound was toxic to the liver. None made it to Phase 2 (for a review see Nordvall et al., 2023).
What, after all this, rekindled interest? The realization that shorter Aβ peptides are not only less toxic than Aβ42 but protective, paired with clinical benefit seen with amyloid-reducing immunotherapy (Cullen et al., 2022; Jan 2022 news; Jul 2023 news). “Since we now know that Aβ-immunotherapies work, I think the next step is to also have small molecules against Aβ generation, to eventually be used in maintenance therapies after immunotherapy,” wrote Harald Steiner, University of Munich. “One could also think of prevention therapies with GSMs,” he added.
To hear Gerlach tell it, Roche never gave up on GSMs. “There was a time when the field believed there was no need for them because BACE inhibitors had arrived, but we never took that view,” she told Alzforum.
In 2020 Gerlach and colleagues reported that RO7185876, a novel class of GSM with a triazolo-azepine structure, shifted proteolysis of Aβ toward Aβ37 and Aβ38, at the expense of Aβ42 and Aβ40, both in vitro and in mice (Ratni et al., 2020). This structure avoids the imidazole group that might have caused toxicity of some earlier GSMs, though Gerlach told Alzforum that RG6289 is yet a different molecule. She revealed no further structural details.
Indeed, Roche has divulged little about RG6289 but, at a talk at CTAD, Agnes Portron said that it stabilizes the substrate-secretase interaction and prolongs cleavage, perhaps by preventing the substrate from slipping out of the membrane, as recently reported (Nov 2023 news). In vitro, RG6289 boosted Aβ37 and Aβ38 about sixfold at sub-micromolar concentration. It all but eliminated Aβ40 and Aβ42, without affecting total Aβ production. The compound did not interfere with processing of Notch, a γ-secretase substrate that regulates neurogenesis and other cell proliferation pathways.
Roche began a Phase 1 trial in 2021, and Portron presented some of the results. This study enrolled 127 healthy volunteers. It is using an adaptive design that increased doses contingent on pharmacokinetic and pharmacodynamic data as it emerged over the course of the trial. Roche split the trial into six parts. Portron showed data from four: single-ascending doses in 18- to 45-year-olds; multiple-ascending doses over two weeks in 18 to 64-year-olds; multiple doses over two weeks in people 64 to 85; and a study on the effect of food intake on pharmacokinetics.
Overall, the drug came with mild adverse events, such as headache, nausea, and diarrhea, many of which were more common in the placebo group. The incidence of adverse events did not go up with dose.
Given by mouth, RG6289 rapidly entered people's plasma, where the drug's concentration fell steadily over the next five days or so, whether it was taken with or without food. Pharmacokinetics suggested a half-life of 13 to 21 hours, which favors dosing once per day. With this daily regimen, RG6289 reached a steady-state concentration in the blood after about seven days. About 3 percent of it reached the CSF, which Portron said was equivalent to the amount of free compound coursing through the blood. CSF samples were collected once from people in the younger multiple-ascending-dose group, once the concentration of the compound in the blood had reached a steady state.
Analysis of plasma and CSF Aβ42 suggested that the compound had reached its target. Plasma Aβ42 fell in a dose-dependent manner, though Porton did not say what those doses were. Secretase activity outside the brain also releases Aβ from APP, so the measured drop in plasma could be in keeping with modulation of peripheral γ-secretase.
In the multiple-ascending-dose regimen, Roche recorded a dose-dependent tapering of Aβ42 and Aβ40 in the CSF, dropping by 80 percent at the apparent highest dose. Again, Porton disclosed no dose information, nor did Gerlach. Along with these Aβ42 and Aβ40 reductions, CSF Aβ37 and Aβ38 rose in parallel, suggesting modulation of γ-secretase processivity.
Roche plans to start a Phase 2 trial next year, selecting doses based on the pharmacokinetics of the Phase 1 trial. Gerlach said this would be in people who test positive for amyloid, whether unimpaired or with mild cognitive impairment due to AD. She said details of study design will be presented at a conference early next year. In a poster at CTAD, the scientists hint that the Phase 2 doses will aim to lower plasma Aβ42 by 27 to 66 percent. The authors write that they believe this range defines a positive benefit-to-risk ratio.
One risk people in the field worry about is altered processing of Notch and other substrates. Recently, scientists in Bart de Strooper’s lab at KU Leuven, Belgium, found 85 of these substrates in microglia alone. Indeed, γ-secretase inhibitors (GSIs) profoundly limited microglial responses associated with amyloidosis. Many of these transmembrane proteins signal through their intracellular domains (ICDs), which γ-secretase clips off in an endoproteolytic step that is blocked by GSIs. Modulators, on the other hand, are designed to accelerate carboxypeptidase processing once ICDs have been released.
“There is no concern about negatively affecting signaling functions of γ-secretase as GSIs do, since GSMs do not affect cleavages that generated ICDs,” Steiner explained.
Still, Gerlach and colleagues are testing for effects on the wider γ-secretase proteome. “In a proteomic study using a cell line that expresses substrates endogenously, we have confirmed that the GSI blocked processing of serval substrates, but saw no such effect on any of the investigated substates using our series of GSMs,” Gerlach told Alzforum. Of course, we haven’t investigated all the substrates, but our data show very consistently no effect on other γ-secretase substrates (non-APP), as expected for a modulator.”—Tom Fagan
Gotta Get Rid of It All: Total Plaque Clearance Key for Clinical Benefit
As researchers parse what makes amyloid immunotherapy work, one answer keeps bubbling up: Take out all plaque quickly, so that clinical benefits have time to show up. At last month’s Clinical Trials on Alzheimer’s Disease conference, held October 24 to 27 in Boston and online, several talks added heft to this emerging view. Scientists at Eli Lilly correlated amyloid negativity early on in their Phase 3 donanemab trial with slowed tangle buildup by its end. This partially answered the nagging question of how the study could have achieved a therapeutic effect without budging tangles overall. Data from other programs reinforced that what matters is not just how much plaque is removed, but how low it gets, and how soon.
“Amyloid removal to the normal range is the best predictor of efficacy, and is directly proportional to the magnitude of the effect,” Randall Bateman of Washington University in St. Louis said in Boston. Noting that tau PET findings have been less consistent so far, he added, “We still have an enormous amount to learn.”
Meanwhile, a meta-analysis found that amyloid immunotherapies work equally well in APOE4 carriers and noncarriers, while Eisai scientists said a subcutaneous formulation gets more lecanemab into the blood than does the intravenous version, though at the price of correspondingly higher ARIA. These data reflect the field's current effort to hone anti-amyloid therapy toward use in clinics around the country.
Tangles Lag Amyloid Removal. In the donanemab Phase 3 trial, people free of plaque within one year (green) had fewer tangles throughout the brain six months later than matched placebo controls (gray). [Courtesy of Eli Lilly.]
All Signs Point to Plaque Clearance
Tangle accumulation has slowed in all successful anti-amyloid antibody trials so far—except in the Trailblazer-Alz2 study of donanemab. That trial posted the highest efficacy reported to date for amyloid immunotherapy, holding back decline on the CDR-SB by 36 percent, yet without an apparent effect on tangles (Jul 2023 conference news). Many scientists were nonplussed by this (Aug 2023 conference news).
In Boston, Lilly’s Sergey Shcherbinin offered post hoc evidence that donanemab does curb tangles if you wait long enough after plaques are gone. Shcherbinin analyzed data from Trailblazer-Alz2’s “combined tau” cohort, i.e., all participants in the trial’s placebo-controlled portion. Initially comprising 1,736 people, this group had baseline tau PET scans above 1.10 SUVR and, after one year on donanemab, two-thirds were below the amyloid positivity threshold of 24.1 centiloids. Shcherbinin wanted to compare the magnitude of their treatment effect at 18 months with that of the whole cohort.
The researchers had a problem, however. They could not simply compare the “rapid clearance” group to the placebo group, because people whose plaque vanishes quickly may have different baseline characteristics than those whose plaque is harder to budge. Indeed, compared with the trial’s overall cohort, “rapid clearers” were a year older, slightly less likely to carry APOE4, and started with less plaque, 95 centiloids rather than 103. Their baseline tau PET averaged 1.31 SUVR, versus 1.34 in the cohort overall. To account for these differences, researchers matched up the 390 rapid clearers with 390 placebo controls of similar age, APOE genotype, and amyloid and tau load at baseline.
How did rapid clearers measure up? On cognitive tests, they eked out about the same small benefit as did the Trailblazer-Alz2 cohort as a whole, declining 34 percent less on the iADRS and CDR-SB than did their matched controls. On tau PET, by contrast, a difference emerged. In the whole trial cohort, the treatment group had only 9 percent less accumulation than placebo on a composite tau PET measure, ditto in frontal cortex, with neither change statistically significant. Rapid clearers, however, had 33 percent less tangle growth on the composite measure, 44 percent less in lateral temporal regions, and 41 percent less in parietal cortex. There was also a trend toward slowing in the frontal cortex, with 29 percent less growth (see image above).
Shcherbinin called this “conceptually analogous” to the tau PET effect seen in Phase 2, where accumulation slowed between 32 to 59 percent in lateral temporal, parietal, and frontal lobes. In Phase 2 as well, people who became amyloid-negative benefitted the most (Mar 2021 conference news).
In Phase 3, it appeared that the lower a person’s amyloid load was by one year, the more that person’s tangle growth slowed, with a weak correlation of r=0.2 for the cohort. Lower plaque load at one year also associated with slower cognitive decline, though that relationship was weaker still, at r=0.1. To Shcherbinin's mind, the findings imply that plaque needs to be absent for a few months before tangle growth starts to slow.
Others at CTAD joined the chorus of the benefits of complete plaque removal. John O’Gorman of Biogen showed a post hoc analysis of open-label extension data from the Phase 3 Emerge and Engage trials of aducanumab. In their amyloid PET substudy, the researchers separated those who had cleared the amyloid positivity threshold of 20.2 centiloids by the end of the 18-month study from those who had not. After a year of OLE aducanumab, the 41 people in the amyloid-negative group had slid about a point less on the CDR-SB than had the 99 who were amyloid-positive at the trial’s end. Effects were similar on other outcomes, with about 1.5 points less decline on the MMSE, three on the ADAS-Cog13, and two on the ADCS-ADL-MCI. The findings again point to cognition improving months after plaque removal. O’Gorman cautioned that sample sizes were small, hence subject to bias.
Delayed Cognitive Benefit? In toto, learnings from immunotherapy trials thus far imply that a person needs to be amyloid-negative for months before they get better. At CTAD, Paul Delmar of Roche presented a meta-analysis of eight immunotherapy trials that supports this. The study comprised Phase 3 aducanumab and lecanemab, and Phase 2 and 3 donanemab and gantenerumab studies. Delmar wanted to know what factors correlated best with efficacy, and when. Prime suspects were amount of amyloid removed, final load, and percentage of participants who became amyloid-negative. For each, Delmar considered two time points: the end of the trial, and a year before the end.
The amount of plaque removed turned out to be a poor measure of efficacy. For example, both the Clarity trial of lecanemab and the Graduate trials of gantenerumab cleared around 60 centiloids, yet the former slowed decline on the CDR-SB by 27 percent, the latter by 7.
The best measure? Plaque load a year before the trial end. This measure correlated most strongly with the final CDR-SB score, with a Spearman coefficient of r=0.85. Similarly, the percentage of amyloid-negative participants a year before the trial end performed well, with a correlation of 0.70. Both were more predictive than amyloid load or negativity at the end of a trial, again highlighting that cognitive benefits lag behind amyloid removal.
Meanwhile, Guoqiao Wang of Washington University, St. Louis, discussed data from the Dominantly Inherited Alzheimer Network’s secondary prevention trial of gantenerumab, which began in 2014. Although the trial was negative, gantenerumab normalized biomarkers in this population (Feb 2020 news; Apr 2020 conference news). The study enrolled people at various stages of autosomal-dominant AD, where some were already symptomatic but most were not, hence treatment effects varied from person to person. In particular, because gantenerumab dosing started low, and mutation carriers deposit plaque quickly, some people continued to accumulate amyloid on drug.
To find out if any group had a cognitive benefit, Wang divided the 52 people taking gantenerumab into the 20 who responded with lower plaque, the 18 with no change, and the 14 with continuing amyloid deposition. These three groups did not differ from each other at baseline in disease stage, amyloid load, or cognitive scores, he noted.
After six years of treatment, the responder group had declined 47 percent less on the CDR-SB than had the continuing deposition group. Similarly, other cognitive tests showed between 32 and 57 percent less slippage. Curiously, the no-growth amyloid group had the least cognitive decline of all, as well as the least change on MRI biomarkers, behaving much like non-progressors seen in other Alzheimer’s trials. Cerebrospinal fluid p-tau181 and total tau stayed relatively stable in the no-growth and amyloid-growth groups, and dropped in the amyloid-lowering group.
The size of therapeutic benefit in the amyloid-lowering group was similar to that seen in the positive trials of lecanemab and donanemab, Wang noted. He believes the findings represent a step toward establishing plaque removal as a surrogate endpoint for trials.
Never Mind APOE? A meta-analysis of amyloid immunotherapy trials found no consistent difference in the cognitive benefit between APOE4 carriers and noncarriers. [Courtesy of Eli Lilly.]
APOE4 Carriers: Same Small Benefit, Might Need Earlier Treatment
Does a person's APOE genotype affect how much immunotherapy helps them? Intuitively, the assumption might be “yes,” but previous studies generated mixed results, with APOE4 carriers sometimes doing better and sometimes worse than noncarriers. Indeed, in Boston, Lilly’s Cynthia Evans made a case for genotype having no overall effect. First, she analyzed Trailblazer-Alz2 data, where E4 carriers made up 72 percent of the cohort. Carriers and noncarriers benefitted nearly identically on all clinical tests, both in the primary analysis population of people with low-to-intermediate tangle loads, and in the full population that included those with more baseline tangles. The APOE genotypes also cleared plaque at the same rate—after taking into account more treatment pauses due to ARIA in E4 carriers, that is.
Next, Evans presented a meta-analysis of Phase 2 and 3 donanemab trials along with Phase 3 aducanumab and lecanemab trials. Despite variation in individual trials—for example, noncarriers fared worse on drug than placebo in certain trials—there was no consistent difference overall between carriers and noncarriers on the CDR-SB or ADAS-Cog, Evans reported (see image above).
What is different? In Trailblazer-Alz2, APOE4 carriers were two years younger than noncarriers, but still had the same plaque load at baseline. This agrees with previous work showing that plaque accumulates at younger ages in carriers (Reiman et al., 2020). In addition, APOE4 carriers in Trailblazer-Alz2 trended toward more tangles, with a baseline tau PET SUVR of 1.22, versus 1.20 in noncarriers.
A recent paper sheds light on this relationship between APOE4 and tangles. Researchers led by Nicolai Franzmeier of Ludwig Maximilian University in Munich analyzed amyloid and tau PET scans from two observational studies, ADNI and Avid-A05, Lilly’s tau imaging study (Pontecorvo et al., 2017). The former comprised 237 participants, the latter 130. In ADNI, 45 percent were E4 carriers, in Avid-A05, 35 percent. Participants ranged from cognitively healthy to mild cognitive impairment.
First author Anna Steward compared longitudinal tau PET scans done 1.5 to two years apart. In both studies, amyloid-positive E4 carriers had faster cortical tangle spreading, at lower plaque loads, than did noncarriers. The speed at which tangles accumulated started diverging between the groups at plaque loads as low as 12 to 15 centiloids, when amyloid PET scans are still considered negative.
This means APOE4 might lower the threshold at which “Aβ starts bothering tau” (Nov 2023 conference news). If so, this would also mean that immunotherapy should start earlier in E4 carriers than noncarriers, the authors concluded. APOE4 was previously found to promote tau pathology in animal models (Sep 2017 news).
Getting Under Your Skin
Most efficacy data so far comes from trials of antibodies infused into people's veins. However, companies are investigating injection under the skin as an easier, more scalable delivery method. How do the two compare? Quite well, according to Eisai’s Michael Irizarry. In Boston, he showed interim data from the company’s subcutaneous lecanemab formulation, which was added to the open-label extension of the Phase 3 Clarity trial.
Out of 394 OLE participants, 322 had previously gotten IV lecanemab. The other 72 were new to the drug, either because they had been on placebo, or because they enrolled directly into the OLE. Both groups started treatment at about 77 centiloids. Participants received 720 mg of lecanemab weekly, administered as two doses. The trial used two methods of getting the antibody in, auto-injector pens and syringes filled from a vial. They performed about the same, Irizarry said.
At an interim, six-month time point, subcutaneous dosing resulted in 11 percent higher exposure in blood than did IV, Irizarry reported. Comparing people new to lecanemab with the results from Clarity, it also removed 14 percent more amyloid, in keeping with the higher exposure. The trial will continue to 12 months.
Subcutaneous safety was similar to that of IV lecanemab, with local injection reactions such as redness and swelling in 8 to 15 percent of people, few systemic reactions, and no skin rashes. Alas, subcutaneous lecanemab did appear to produce more ARIA than the IV version, 17 versus 13 percent ARIA-E, and 22 versus 17 percent ARIA-H, also in keeping with the higher exposure.
Notably, dosing made lecanemab blood levels more stable, without the high peak or drop seen with IV administration. For IV dosing, peak exposure had correlated with ARIA-E, leading researchers to hope that lowering this would lessen ARIA. However, for SC lecanemab thus far, total exposure correlated with ARIA-E, and overall rates were comparable to IV.
Eisai has said it will submit a Biologics License Application to the U.S. Food and Drug Administration for the subcutaneous formulation by March of 2024. It is likely other drug makers will follow suit with their own SC formulations, hence fewer people may need to travel to infusion centers for amyloid immunotherapy in the future.—Madolyn Bowman Rogers
Gauging Cognition From Your Couch? The Digital Age Has Arrived.
When COVID forced much of the world into virtual work, dementia scientists sped up their ongoing adaptation of cognitive assessments to digital versions (Dec 2021 conference news). How are they performing now? At the Clinical Trials in Alzheimer’s Disease conference, held October 24-27 in Boston, scientists described smartphone apps that flagged amyloid positivity after just a few days, discerned early AD as well as in-clinic exams, and pinpointed areas of brain atrophy from narrated stories. Others optimized automated scoring algorithms to the point where they beat humans at rating cognitive tests.
“Many digital tools are not ready for prime time, but they are getting there,” Dorene Rentz, Massachusetts General Hospital, Boston, told Alzforum. “Within a year or two, we will see an explosion in this area.”
Why digitize testing? The energy for the push across the field toward doing so comes from indications that asking a person to complete tasks on their personal computing device in their home, and doing so regularly, might better reflect their state of mind and memory than what paper and pencil can reveal during rare, stressful clinic visits. Rentz envisions using digital tests in clinical trials to probe cognition more frequently than was typical with in-person tests, so researchers can more accurately detect change over time.
Sussing Out Preclinical Alzheimer's in Days
How well an older person learns seems to signal the presence of AD pathology and subtle cognitive decline. That’s what Kate Papp at Brigham and Women’s Hospital, Boston, had figured out when she asked cognitively healthy people to take the tablet-based Computerized Cognitive Composite (C3), which includes the Cogstate brief battery, face-name associations, and an object-recall task. Participants took the C3 each month for a year, and at CTAD 2021, Papp reported that all got better at it, showing practice effects. However, after three months, people who were amyloid- or tau-PET-positive, or who slipped fastest on the in-clinic Preclinical Alzheimer's Cognitive Composite (PACC5), improved less on the C3 than did amyloid- or tau-negative participants or people who declined slowly on the PACC5. To Papp, such diminished practice effects signal a failure to consolidate memories.
Might daily testing spot this more quickly? To find out, Papp and colleagues developed a smartphone app called the Boston Remote Assessment for Neurocognitive Health. BRANCH includes face-name association, digit-symbol pairing, and grocery item and price recall tasks (image below; Papp et al., 2021).
At this year’s CTAD, Papp reported results from 164 cognitively normal adults enrolled in the Harvard Aging Brain Study (HABS), of whom 36 had positive amyloid scans. For one week, they spent an average of 12 minutes each day completing the same BRANCH tasks. “We often try to limit practice effects by using alternate forms or increasing the time between assessments,” Papp said at CTAD. “With BRANCH, we want to embrace the practice effect and use it over a short interval to detect individuals with subtle memory consolidation difficulties.”
Practice Makes Perfect? People using the BRANCH app repeat the same tasks each day for a week to measure how well they learn. [Courtesy of Roos Jutten, Massachusetts General Hospital.]
Participants found the app easy to use, with 97 percent completing all seven days of testing (Weizenbaum et al., 2023). This means older people can fit these assessments into their daily lives.
Lo and behold, after only four days, a subtle difference between amyloid-positive and -negative participants emerged, when the former completed tasks about 8 percent less accurately than the latter.
Repeated testing on the BRANCH app accomplished what a single in-person test could not, as baseline PACC5 scores did not discern amyloid status. This suggests that the app might become more useful for clinical trial recruitment than a single in-clinic assessment.
Poor learning on BRANCH also predicted future cognitive slippage and worsened learning problems, albeit with a relatively weak R-value of 0.54. People who had learned the least after a week of using the app declined the most on the PACC5 one year later. “Not only can we see differences in seven days of performance, but learning curves may also change from baseline among those with elevated AD biomarkers,” noted Rebecca Amariglio of MGH, who co-developed BRANCH.
Roos Jutten, a postdoc in Papp’s lab, found that when amyloid-positive people repeated new versions of the BRANCH tasks six months and a year after their original sessions, they were less accurate each time. “Ongoing research into how these learning curves change over time will determine whether they can be used to track short-term cognitive changes and be useful as an outcome measure in AD secondary prevention trials,” Jutten told Alzforum.
To validate BRANCH in larger, more diverse cohorts, Papps' team added the app to the Alzheimer Prevention Trials webstudy. They plan to give BRANCH to people who finished the U.S. POINTER study and are being followed in an extension, and they are testing a Spanish version of BRANCH.
In-Clinic Tests—No Longer Gold Standard?
In another effort to track cognitive change remotely, Jutten and Rentz are testing how the NIH’s Mobile Toolbox app stacks up against the pencil-and-paper PACC5. Through eight tasks, the MTB tests episodic memory, processing speed, executive function, spelling, and vocabulary. The PACC5 uses five tasks to gauge episodic memory, processing speed, semantic memory, and global cognition. The scientists thought four MTB tasks—face-name associations, picture sequence memory, number-symbol match, and a vocabulary test—might make for a fair PACC5-MTB comparison.
At CTAD, Jessa Burling of MGH presented preliminary results from 80 cognitively healthy HABS participants who spent 30 minutes taking the entire MTB on their smartphones. She compared their results to their PACC5 scores and amyloid and tau PET scans. PACC5 scores modestly correlated with composite scores of the four “PACC-like” MTB tasks and those plus combinations of the other tasks, clocking coefficients of 0.29 to 0.40. The PACC5 and all MTB composites correlated weakly with tangles, but not with plaques. This aligns with Papp’s prior finding that the PACC5 does not discern a person's amyloid status.
Not all efforts to digitize in-person assessments have worked out. Rachel Nosheny of the University of California, San Francisco, reported that the electronic Clinical Dementia Rating (eCDR) underperformed relative to its in-clinic counterpart (Nosheny et al., 2023). She compared scores from 163 cognitively healthy older adults and 43 with MCI from UCSF's Brain Health Registry and three clinical sites across the U.S., who had taken both versions within two weeks.
Of 52 items shared between the in-person and eCDR, 46 correlated with greater than 70 percent concordance. Almost all discordant items fell into the memory domain, a key component of assessing incipient AD. Yan Li at Washington University had previously reported similar results (Dec 2021 conference news). This discordance translated to high rates of false positives and negatives, 8 and 55 percent, respectively. The CDR's detailed interviews with the participant and his or her study partner may be challenging to render online, Nosheny concluded.
The eCDR's global score modestly correlated with its in-person counterpart, with an area under the curve of 0.79 out of 1. Nosheny cautioned that an optimized version would be needed before it is ready for clinical use.
Scary Story Time
The way you tell a tale can reveal to an astute assessor whether your verbal cognition is starting to slip. Researchers at London-based Novoic have developed a web app that aims to capture this. Called Storyteller, it uses an automated recall task, where a proband listens to a 120-word story and immediately retells it, does the same with a different story, and then recalls the first story. Storyteller records and transcribes the narrations, scores them based on how closely they match the original version, and sums the recalls into a final score.
In an early study called Amyloid Prediction in Early Stage Alzheimer's Disease From Acoustic and Linguistic Patterns of Speech (AMYPRED-UK), Storyteller pegged people diagnosed with mild cognitive impairment as well as did the in-clinic PACC5 (Dec 2021 conference news).
Since then, Novoic has deployed new AI methods to improve the app. At CTAD, Caroline Skirrow of Novoic pitted man versus machine at scoring an in-person verbal story recall task. Participants included 84 cognitively normal adults and 89 people with MCI from AMYPRED-UK and AMYPRED-US, a sister study based in California. For each, she compared a consensus score set by a group of raters—considered the gold standard for this study—to the score given by one of 15 highly trained raters and an automated score generated by the new Novoic software. Called AccuRater, it rates the participant’s recorded, uploaded, and auto-transcribed response based on standardized scoring instructions.
Compared to the individual rater scores, AccuRater scores correlated a bit more tightly with the consensus score, bumping the correlation coefficient from 0.94 to 0.98 (image below). This was because raters tended to make more mistakes and give slightly lower scores than the software.
AI Beats Human. For a story recall task, automated scores calculated by Novoic’s AccuRater software (right) correlated more tightly with scores assigned by a committee of raters than did scores by individual raters (left). [Courtesy of Caroline Skirrow, Novoic.]
Skirrow thinks AccuRater will enable consistent detection of cognitive impairment for screening clinical trial participants or tracking changes over time. She said Novoic is incorporating the AccuRater algorithm into Storyteller to improve its scoring. The app is already being used to screen 20,000 participants into the latest iteration of the Alzheimer’s Disease Neuroimaging Initiative (ADNI; Oct 2022 news).
Rentz asked if AccuRater works on audio from people speaking with accents or in languages other than English. Skirrow replied that Novoic's transcription software is multilingual and performs better than Google Speech-to-Text. As to applying the scoring system to other languages, Skirrow said this is possible but hasn't been done yet. “We lack validation data,” she said.
A different speech analysis storytelling app by Roche is starting to fill this gap. Story Time is a smartphone app where people narrate a story depicted in cartoon strips, then retell it immediately, and again a day later without seeing the pictures. The app measures word count, noun rate, and speech duration. In a study shown at CTAD, 32 cognitively normal adults, 31 with subjective cognitive decline but no amyloid plaques on PET, 30 with amyloid-PET-positive SCD, and 30 with early AD completed nine Story Time sessions over a month. Half of the participants spoke Spanish.
Irma Kurniawan and colleagues at Roche assessed how well the app detected early AD and how storytelling performance mapped to brain atrophy on MRI. Compared to controls, people with AD used more words, and paused less, while telling a story the first time, yet spoke fewer words when recalling it. Participants who used the most words during the first narration had the most atrophy in their left angular gyrus and middle temporal lobe. People who recalled little of the story the next day had atrophy in different areas, such as their left perirhinal cortex and hippocampi (image below). Notably, the left hemisphere regions are involved in language processing, while the hippocampus is crucial for memory, possibly giving a functional explanation for the storytelling deficits.
Anatomy of Narration. Talking more and pausing less while telling a story correlated with atrophy in the left angular gyrus and middle temporal lobe (left) while poor recall associated with atrophy in other areas (right). [Courtesy of Thanneer Perumal, Roche.]
Thanneer Perumal of Roche told Alzforum that the company plans to use the Story Time app in upcoming clinical trials, but would not disclose which ones.
Cognition isn’t the only component of dementia that can be better measured by technology, Rentz noted. She also wants to see digital measures of passive activities, such as sleep, gait, and of physical activity, with the ultimate goal of using them as outcome measures in clinical trials. Cognition drives function but the reverse can also be true. For example, restored physical energy, perhaps thanks to better sleep, in turn can benefit cognition. “Cognition is important, but function is the be-all and end-all,” she told Alzforum.—Chelsea Weidman Burke
Papp KV, Samaroo A, Chou HC, Buckley R, Schneider OR, Hsieh S, Soberanes D, Quiroz Y, Properzi M, Schultz A, García-Magariño I, Marshall GA, Burke JG, Kumar R, Snyder N, Johnson K, Rentz DM, Sperling RA, Amariglio RE.
Unsupervised mobile cognitive testing for use in preclinical Alzheimer's disease.
Alzheimers Dement (Amst). 2021;13(1):e12243. Epub 2021 Sep 30
PubMed.
New Alzheimer’s Diagnostic Criteria Remain ‘Research Only’
The latest effort to rewrite the diagnostic criteria for Alzheimer’s disease has stirred more than a little controversy. While the proposed revisions have garnered praise from researchers for more accurately reflecting the underlying biology of the disease, they also have sparked criticism for advancing toward clinical use what were previously research criteria. In particular, the American Geriatrics Society called the move to the clinic premature, noting that many uncertainties remain about how well biomarkers predict symptom onset. The AGS also questioned the extent of the National Institute on Aging’s involvement in the working group. The NIA has now removed its name from the criteria, though its representatives continue to advise the group.
At last month’s Clinical Trials on Alzheimer’s Disease conference in Boston, representatives of the Alzheimer’s Association Workgroup discussed how they have revised the draft criteria to incorporate community input. The group’s leader, Clifford Jack of the Mayo Clinic in Rochester, Minnesota, emphasized that the new criteria should not be considered guidelines for clinical practice at this time; rather, they represent a bridge between research and practice. Jack also noted that tau biomarkers have been split into two categories, to better distinguish between soluble and fibrillar forms, and to distinguish the different roles of each of these forms in diagnosis. This change was suggested by biomarker scientists.
The Concept of Preclinical AD Draws Flak
The working group first presented the draft criteria at last summer’s AAIC in Amsterdam, and solicited feedback from the field (Aug 2023 conference news). This generated 64 public comments from practicing neurologists, pathologists, and organizations such as the Gerontological Society of America. There were also comments from advocacy groups, including the National Down Syndrome Society and the National Task Group on Intellectual Disabilities and Dementia Practices, and from industry, such as Biogen and C2N Diagnostics.
Among the commentators was the AGS, a nationwide organization that represents more than 6,000 healthcare professionals who care for the elderly. Their comment was influential. The geriatricians noted that the NIA had been a co-equal convener in the 2018 NIA-AA criteria but was taking merely an advisory role for the 2023 update. The 2018 criteria were articulated as a research framework, in keeping with NIA’s mission as a research agency. It would be unusual for the NIA to take part in developing clinical criteria, the AGS argued.
The National Institutes of Health has now directed NIA to remove its name (see Forbes article). Jack told Alzforum that this was done to conform to NIH guidelines regarding the use of institute names. He noted that NIA involvement in the project has not changed, including neuroscience division director Eliezer Masliah holding a position on the four-member steering committee.
The AGS also criticized the extent of industry ties of the working group, in which 16 of the 22 members either work for companies or have received consulting fees from pharma or the Alzheimer’s Association. The panel includes highly regarded biomarker scientists, but few clinicians who regularly see patients.
Some Alzheimer’s researchers remain troubled by the idea of defining people with plaque buildup in their brain, but no symptoms, as having AD. Many others have been moving in this direction for years. Writing for the group Methods for Longitudinal Studies in Dementia (Melodem), Deborah Blacker of Massachusetts General Hospital, Boston, noted that a diagnosis of AD can still induce dread, and might confuse patients and family members. “Many of us would prefer a term like brain amyloidosis or the neuropathologists’ ‘Alzheimer disease neuropathologic changes’ to stress the difference,” she wrote (see comment under Working Group Meetings on Melodem page).
Two Types. Updated biomarker scheme distinguishes between markers of soluble and fibrillar tau. The former are more closely tied to amyloidosis than tangles, and are diagnostic of the disease, whereas the latter are more useful in staging.
In Biomarkers, Tau Doesn’t Equal Tau
Many updates to the draft criteria were technical, part of a continuing refinement of diagnostic and staging biomarkers as new pathophysiology data roll in. The biggest change is that tau biomarkers have now been split into two categories, T1 and T2 (see table above). The former comprise markers of soluble phosphorylated tau, such as p-tau217, p-tau181, and p-tau231; the latter mark fibrillar tau, such as cerebrospinal fluid MTBR-tau243 and tau PET.
T1 biomarkers are considered “core 1” markers, capable of diagnosing disease on their own. The updated criteria specify that such biomarkers should have a demonstrated accuracy of 90 percent or better; at the moment, this would include only amyloid PET, CSF markers, and some plasma p-tau217 assays. Jack emphasized the crucial role of clinicians’ judgment in interpreting biomarker results for the patient before them, particularly when the person has other medical conditions or co-pathologies that may affect the assays.
T2 biomarkers are “core 2,” not intended to be used as stand-alone tests. Instead, they should be combined with core 1 markers to stage disease, Jack said. The updated criteria now suggest staging be done only with amyloid and tau PET, as fluid biomarkers are not yet well enough established for this. This change is also a response to comments from the field.—Madolyn Bowman Rogers
Plasma p-Tau-217 Assays Work Well, But No Home Run for Diagnosis
Feel like your head is spinning? Like a little blood sample in a centrifuge, perhaps? No need to panic. It just means you are straining, like the rest of us, to keep up with the Alzheimer's disease plasma biomarker development field. New markers, new assays, new protocols seem to arrive at a dizzying pace. A flurry of those was on display at this year’s CTAD meeting, held last month in Boston. “The field is moving very rapidly,” noted Andreas Jeromin, ALZpath Inc.
Scientists reported results of a round-robin comparison of more than two dozen plasma phospho-tau assays. (Hint: p-tau 217 tests outdid the rest of the flock.) A clearer picture emerged of how plasma markers might be used for diagnosis. (Another hint: Two cut points work better than one). Plus attendees saw early data from two new commercial tests, ALZpath’s p-tau 217 assay and PrecivityAD2 (see Part 12 of this series)..
Straight Answers from a Round-Robin? Nicholas Ashton, University of Gothenburg, Sweden, presented results of the round-robin of the Global Biomarker Standardization Consortium, an initiative originally started by the Alzheimer’s Association to standardize and validate biomarker tests. The GBSC had previously reported poor correlation among 11 different plasma Aβ42/40 assays conducted at 11 different sites (Aug 2019 conference news; Pannee et al., 2021). At least for p-tau217, the picture now looks a little rosier.
Ashton and colleagues compared 26 different p-tau tests by 12 different vendors, 11 commercial ones plus UGot. Participation was free of charge, but also blinded, meaning no handler knew who the samples came from, and no vendor knew how any other vendors’ assays performed. They joined anyway. “The response was remarkable,” Ashton told Alzforum. Vendors included AbbVie, ADx Neurosciences, Alamar, ALZpath, Fujirebio, Janssen, Lilly, MagQ, Meso Scale Diagnostics, Roche, and Quanterix (image below). An exception was C2N Diagnostics, St. Louis, which was invited to participate but did not.
For the test samples, Jonathan Schott, University College London, provided Ashton with 1 mL of plasma from each of 40 individuals with suspected AD. It turned out that 25 of these had AD, based on cerebrospinal fluid Aβ42/40 and p-tau181. Ashton aliquoted their plasma into smaller samples he shared with each participating lab. Each then determined who among the 40 donors likely had AD, and who didn’t, based on plasma p-tau181, p-tau212, p-tau217, or p-tau231. Accuracies were determined by the area under the curve, or AUC, of sensitivity versus specificity plots—an AUC of 1.0 being perfection and 0.95 or better typically deemed highly desirable for routine clinical work.
Most of the 26 assays identified AD donors from controls with high accuracy (image below). Among them P-tau217 stood out for consistently high AUCs, and a greater fold increase between AD and controls. Ashton acknowledged that the sample size of 40 was small. That said, he was encouraged by the tight agreement among the 11 p-tau217 tests—between-assay correlations ran from 0.85 to 0.97.
“This gives us hope that the field does not have to decide on one [p-tau217] assay, but can convert between them,” said Ashton. “This suggests that it is possible to combine and interpret data from large research cohorts, or from different clinical trials, that are based on different assays,” he added. The Alzheimer's fluid biomarker field has long struggled with such comparisons because of the number of different tests in use.
Ashton also noted that immunoassays running on Fujirebio's Lumipulse readers emerged on top, performing on par with a mass spectrometry test run by U Gothenburg. This is good news for the field because Lumipulse machines are already being widely used in clinical practice, often being the diagnostic workhorse in hospitals. They are fully automated, meaning the sample never gets processed by human hand. They are also fast, taking about 20 minutes to run a test. “This is what the future of AD clinical testing will look like,” predicted Ashton.
P-tau217 Wins Round 1. In a round-robin analysis of 40 plasma samples, tests for p-tau217 best distinguished people with AD from healthy controls (left). They also record the biggest fold differences (right). [Courtesy Nick Ashton, University of Gothenburg.]
P-tau181 and p-tau231 also performed well. Correlations among the 10 p-tau181 assays ranged from a low of 0.04 to 0.9. Ashton puts this down partly to the assays and partly to plasma. Unlike p-tau217, which can only be detected in plasma after amyloid pathology has begun, p-tau181 and p-tau231 are found in plasma of healthy controls. “There seems to be a basal level of noise that accounts for some of the variability,” said Ashton. AUCs for p-tau231 assays ranged from about 0.8 to about 0.95, but correlations between assays were not shown, perhaps because there were only four of them.
Finally, one assay from UGot of p-tau212 predicted AD with an AUC of about 0.9. It is too soon to pit the accuracy of this relative newcomer against that for p-tau217, but p-tau212 might have an edge because it may be one of the earliest markers of amyloid pathology.
To Ashton’s mind, a caveat emerged from the round-robin. That is, correlations between plasma and CSF were not terribly strong for any given marker. For Fujirebio’s p-tau217 test, the correlation coefficient was 0.67; for U Gothenburg’s test, only 0.48. “This is a really important take-home message,” said Ashton. “Once you get into the positive range, the correlation is almost nonexistent.”
He said this should be a consideration for the proposed new NIA-AA diagnostic criteria, which lump plasma and CSF together to predict the initial stage of AD (Aug 2023 news). “I don’t think this is the way to go,” said Ashton. “You cannot directly stage people when mixing plasma and CSF,” he said, though he said other assays might have tighter correlations.
Suzanne Schindler, Washington University, St. Louis, noted as much for assays that use the p-tau217/tau217 ratio instead of p-tau217 absolute values. She told Alzforum that among 529 people studied at the Knight Alzheimer’s Disease Research Center, the correlation between the percentage of p-tau217/tau217 in plasma and CSF, as measured by mass spectrometry, was about 0.85. Scientists hope that measuring the percentage phosphorylation normalizes for disconnects in the plasma/CSF equilibrium caused by worsening pathology or by co-morbidities. In effect, the non-phosphorylated tau would act as an internal standard.
Research Versus Implementation
In her CTAD talk, Schindler emphasized that some of the blood-based biomarker tests, particularly those for p-tau217, are now as accurate as CSF tests. “We are at the point where we need to think about whether we even need to follow up blood tests with CSF or PET,” she said. Still, she thinks that even with AUCs of 0.95, the tests are not good enough. In a clinical population where 61 percent are amyloid-positive, a typical scenario at memory clinics, such tests would still misclassify one in 10 people, she calculated.
From a treatment perspective, this would not work, agreed Oskar Hansson, Lund University, Sweden. The problem, Hansson told Alzforum, is that shifting from treating markers as continuous variables, as scientists do in research settings, to a yes/no diagnostic requires too much simplification of the data. “The values that are close to the yes/no cutoff are always uncertain,” he said.
He sees a solution in setting two cutoffs. Values higher than the top cutoff would identify people who are sure to be positive, and values lower than the bottom cutoff would distinguish those sure to be negative. Anyone who fell in between could be sent for further testing. “Clinics are doing that for prostate specific antigen tests right now,” Hansson said. Men whose tests are uncertain go for follow-up MRI scans.
At CTAD, Hansson showed how this approach could work for AD. Collaborating with Nicolas Barthelemy and colleagues at WashU, who have perfected an MS-based assay to measure %p-tau217 in plasma, the scientist tested samples from the Swedish BioFinder2 study and the Knight ADRC at Washington University, scientists in his lab were able to bump up the test's positive and negative predictive values by using two cutoffs. About 20 percent of the cohort fell into the uncertain zone. Indeed, these people had less amyloid plaque as seen on PET scans than the clear p-tau217 positives.
The strategy also worked for identifying people with neurofibrillary tangles. Basically, the two-cutoff approach upped both the positive and the negative predictive values from 90 to 95 percent, the magic threshold clinicians want for a practical diagnostic test. Hansson and colleagues saw similar improvements in diagnostic performance of Lilly and Janssen's p-tau217 plasma assays when they used two cutoffs (Brum et al., 2023).
Schindler said she likes this approach. In clinical practice, however, she envisions two tests being used: a triage test and a confirmatory test. The former would identify people who likely have AD, but would need to be followed up with amyloid PET, especially if the person were going to receive immunotherapy based on the outcome. The confirmatory test would have to be good enough to stand alone.
A blood-based biomarker working group convened by George Vradenburg at UsAgainstAlzheimer’s recommended the triage and confirmatory tests. The group’s goal is to speed up and facilitate treatment of AD, said Schindler, who is a member.
How is a triage test useful if it requires a follow-up test? For one, it could help clinicians figure out who goes to a specialist. “Dementia clinics are inundated, and it can often take a year or more to get an appointment,” said Schindler. “A blood test could help us identify who needs to be seen soonest.” For another, a triage test could rule out people who do not have AD. It would need to be highly sensitive. “In populations where the prevalence of AD is low, a negative result is almost certainly negative if the test has high negative predictive value,” said Schindler. “We think a triage test will be helpful for people who are ‘worried well.' Then we could look for other explanations for their memory concerns," she told Alzforum. For more on the newest p-tau217 tests, see the next story in this series.—Tom Fagan
Pannee J, Shaw LM, Korecka M, Waligorska T, Teunissen CE, Stoops E, Vanderstichele HM, Mauroo K, Verberk IM, Keshavan A, Pesini P, Sarasa L, Pascual-Lucas M, Fandos N, Allué JA, Portelius E, Andreasson U, Yoda R, Nakamura A, Kaneko N, Yang SY, Liu HC, Palme S, Bittner T, Mawuenyega KG, Ovod V, Bollinger J, Bateman RJ, Li Y, Dage JL, Stomrud E, Hansson O, Schott JM, Blennow K, Zetterberg H.
The global Alzheimer's Association round robin study on plasma amyloid β methods.
Alzheimers Dement (Amst). 2021;13(1):e12242. Epub 2021 Oct 14
PubMed.
The blood-based p-tau marker field is getting pretty busy. In addition to existing tests by Fujirebio, Lilly, Janssen, and Meso Scale Discovery, C2N debuted a new CLIA-approved mass spectrometry-based test last August, while ALZPath Inc. still plans to offer its immunoassay for clinical use in 2023. Data presented at CTAD in Boston indicate that both new tests identify people with amyloid pathology with accuracies that rival the current crop of assays. A round-robin study, also shown at CTAD, suggests that plasma p-tau217 assays have the best chance of becoming stand-alone diagnostic tests for AD (see Part 11 of this series).
PrecivityAD2, C2N’s new test, combined plasma Aβ42/40 and percentage of tau phosphorylated at position 217 to identify people who have 25 C2N’s centiloids or more on amyloid PET scans, which would include people with relatively low levels of amyloid pathology. At CTAD, C2N’s Tim West described how the test performed in two cohorts: 224 people in the Plasma Test for Amyloidosis Risk Screening study, aka PARIS, and 359 people in MissionAD, Eisai’s Phase 3 program for its discontinued BACE inhibitor elenbecestat.
The latter enrolled people with early AD. PARIS, on the other hand, is a sub-study of the Imaging Dementia–Evidence for Amyloid Scanning study. IDEAS recruited Medicare beneficiaries to test if amyloid imaging adds value in primary care settings (April 2015 news). “In terms of cohorts we can obtain samples from, this is as close as you can get to the real world,” said West. Since PARIS collected no blood samples, C2N sent phlebotomists to volunteers’ homes. “This worked out great,” West said. “I don’t think we could have recruited that many patients if they were required to go to a blood collection center.”
West had previously reported that the percentage of tau phosphorylated at 217 identified amyloid-positive people in PARIS with an AUC of 0.95 (Aug 2022 conference news). In Boston, West showed that for the 583 people in PARIS and MissionAD, the AUC for %p-tau217 was almost the same, 0.94.
C2N combined this tau marker with the Aβ42/40 ratio to predict the likelihood that a person's amyloid PET score surpasses 25 centiloids. West and colleagues determined that this so-called Amyloid Probability Score 2 had a sensitivity and specificity of 88 and 89 percent, respectively, in the PARIS/MissionAD cohorts.
This was better than the original APS score based on Aβ42/40, ApoE proteotype, and age. That earlier test had delivered a specificity of 86 percent, but only by excluding about 14 percent of people who fell into a no-man’s land, i.e., the intermediate zone where the APS score could not determine if they were indeed positive or negative for plaques. This was a problem with the test (image below). At CTAD, West showed that, in contrast, APS2 scores fell at the extreme ends of the range, i.e., well away from no man’s land, with an optimal cut point at 47.5 on a scale of 0 to 100. “With the new test, the prediction score is binary,” he told Alzforum. “People are either positive or negative.”
Better Precivity? In PrecivityAD, which was based on Aβ42/40, ApoE proteotype, and age, about 14 percent of people fell into an intermediate zone, meaning their Amyloid Probability Score (APS), failed to predict whether they were amyloid positive or negative. In PrecivityAD2 (not shown), APS2 scores concentrate at the extremes of the scale, allowing for a single cut point. [Courtesy of Hu et al., 2022].
How might the test perform in the real world? As with all tests, this depends on the prevalence of the disease, noted West. In a population where the dementia prevalence was 65 percent, PrecivityAD2 had positive and negative predictive values of 94 and 80 percent, respectively. In a population with mild cognitive impairment, where prevalence clocks in at around 50 percent, those numbers fell to 90 and 87 percent. These numbers seemed stable, unaffected by demographic variables, including race, ethnicity, sex, age, and ApoE4 status.
ApoE isoform, a test variable in the original PrecivityAD, is absent from PrecivityAD2. West told Alzforum that it didn’t budge the AUC enough to warrant including it. “And this way, we avoid having to disclose genetic information to people who take the test,” he added. Such revelations can worry some volunteers. C2N will continue to offer a separate ApoE test, West said. In 2019 it received Breakthrough Device Designation from the FDA for its amyloid blood tests, which cover PrecivityAD2, West told Alzforum, but the company has not filed for FDA approval yet. He said they have a set of prospectively collected PARIS samples on hold for this purpose.
A Different Path
For its part, Carlsbad, California-based ALZPath Inc. hopes to roll out its p-tau217 test for routine clinical use later this year. An immunoassay that uses proprietary antibodies and the highly sensitive single molecular array platform, aka SIMOA, ALZpath pTau217 is commercially available through Quanterix, a diagnostic company based in Billerica, Massachusetts.
Scientists have tested the ALZPath assay in more than 40 clinical cohorts worldwide. At CTAD, Hongming Zhang and colleagues from Frontage Laboratories, Exton, Pennsylvania, a contract research organization (CRO), reported on a validation study in which the assay had a lower limit of quantification of 9.77 fg/mL. Yes, that’s femtograms, and makes this test perhaps the most sensitive commercial p-tau217 assay. In a recent head-to-head comparison, which did not include the ALZpath test, the Janssen assay had a lower limit of detection of 13 fg/mL: Limits of detection are less stringent than limits of quantification (Janelidze et al., 2022). In Zhang’s hands, the minimal volume required—which can be in the mL range for some tests—was 33μL for plasma and 5μL for CSF. The variation between different runs of the same sample were low, indicating good precision.
Recently, scientists led by Kaj Blennow at University of Gothenburg, Sweden, reported that within-subject variation in the ALZpath assay was slightly worse than for Aβ42/40 assays, though not as bad as for p-tau181, but that the large fold change in p-tau217 between AD and control samples would balance out this variability (Brum et al., 2023).
At CTAD, Ahmed Chenna and colleagues at Labcorp-Monogram Biosciences, another CRO based in South San Francisco, reported that among 200 volunteers in the Australian Imaging, Biomarkers, and Lifestyle cohort, the ALZpath pTau217 plasma test best identified people who had amyloid in their brains per PET scan. It outperformed plasma p-tau181, Nf-L, GFAP, and Aβ42/40, returning an AUC of 0.95.
The next best was a Lumipulse assay of Aβ42/40 which rang in at 0.88. In a paper recently uploaded to medRxiv, scientists led by Nicholas Ashton at UGothenburg reported testing the ALZpath assay on samples from 786 participants in three cohorts, Translational Biomarkers in Aging and Dementia (TRIAD), Wisconsin Registry for Alzheimer’s Prevention, and the Sant Pau Initiative on Neurodegeneration. The assay identified amyloid pathology with AUCs of 0.92 to 0.96, and tangle pathology with slightly better accuracy, AUCs of 0.93 to 0.97 (Ashton et al., 2023).
Similarly, scientists at UGothenburg led by Andrea Benedet compared how the ALZpath and Janssen assays performed in the TRIAD cohort. The assays fared equally well, distinguishing people with AD from those with other neurodegenerative diseases with AUCs of O.95 and 0.96, respectively. The p-tau217 levels determined by both also correlated with amyloid and tangle PET SUVRs, though the Janssen assay tracked slightly better with tau PET than did the ALZpath one. Their preprint was uploaded to The Lancet on September 19 (Therriault et al., 2023).
In a separate study, scientists, including Jeromin, led by Ashley Price, Butler Hospital, Providence, Rhode Island, tested if ALZpath pTau217 and three other markers were able to identify people in preclinical stages of AD. They tested plasma from 143 cognitively healthy people, between the ages of 55 and 80, of whom 62 and 81 were deemed to have low and high risk, respectively, of getting AD. Eleven of the high-risk group carried one copy of APOE4, the other 70 had two copies.
First author Peter Snyder reported that only ALZpath pTau217 differentiated the low- and high-risk groups. Plasma Aβ40/42, NfL, and p-tau181 showed no difference. Among 25 people in the cohort who had amyloid scans, both Aβ40/42 and ALZpath pTau217 distinguished amyloid-positive from -negative, but the latter did so more robustly. The AUC in this population was 0.85. NfL, and p-tau181 correlated with age, but neither p-tau217 nor Aβ40/42 did, indicating these are less susceptible to “noise.” In fact, this might be why p-tau217 outperformed all other markers in the round-robin presented at CTAD. Other tau markers turn up in the plasma even in healthy controls.
As CTAD began, Quanterix announced it would partner with Janssen to commercialize Janssen’s p-tau217 assay. “This confused some of our customers,” ALZPath Inc’s Andreas Jeromin told Alzforum. “Quanterix will continue to support the ALZPath assay in parallel,” he assured Alzforum.—Tom Fagan
Comments
No Available Comments
Make a Comment
To make a comment you must login or register.