22 Aug 2025
The current Alzheimer’s disease therapeutic antibodies carry a sizable risk of ARIA, especially for APOE4 carriers. Researchers believe the future of amyloid immunotherapies lies in targeted brain delivery that bypasses the larger blood vessels prone to this side effect, and several companies are working on this. At the Alzheimer’s Association International Conference, held July 27-31 in Toronto, scientists at Roche presented the latest data on trontinemab, the farthest along of these next-generation drugs.
- At 3.6 mg/kg, 91 percent of participants were amyloid-negative after six months.
- Trontinemab cleared plaque even in deep brain regions that other antibodies hardly reach.
- In blinded safety data, the ARIA-E rate was 1 percent.
- Roche will take this dose into Phase 3 trials to start this fall.
With Phase 2 data almost complete, the bottom line remains the same: Trontinemab clears nearly all plaque in six months, virtually without ARIA. Now it moves to Phase 3. Janice Smith at Roche described the design of two identical trials, Trontier 1 and 2, that will start this fall. She also announced plans for a study in preclinical AD, but no details yet.
The data generated excitement in Toronto. “I could barely contain myself from jumping up and whooping,” Reisa Sperling of Brigham and Women’s Hospital, Boston, said in a panel discussion. Stephen Salloway of Butler Hospital in Providence, Rhode Island, said he only regretted that it will take several years to get the Phase 3 results, noting, “If clinically effective, this will be the new paradigm. I’d like to roll this out for my patients now.”
Trontinemab marries the anti-amyloid antibody gantenerumab to a Fab fragment that recognizes the transferrin receptor. When endothelial cells internalize these receptors, they ferry bound trontinemab across the blood-brain barrier (Mar 2021 conference news). Luka Kulic at Roche previously showed that, at least in mice, this approach results in widespread distribution of antibody across the brain, even in deep brain regions, but triggers very little ARIA. Many scientists believe this is because the drug gets in through capillaries, avoiding the amyloid deposits coating the larger arteries penetrating the brain (Nov 2023 conference news).
A Phase 1a/2b trial tested trontinemab doses of 0.2, 0.6, 1.8, or 3.6 mg/kg monthly, and a part 2 extension added participants at the two highest doses (Mar 2024 conference news; Nov 2024 conference news). At the AD/PD meeting last April, Kulic presented completed data on all 76 participants in two 1.8 mg/kg trial arms, as well as from 38 people on 3.6 mg/kg. On this latter dose, 81 percent of people had become amyloid-negative within six months. At the time, there were no cases of ARIA-E, and one case of ARIA-H, in the blinded safety data for the 3.6 mg/kg dose, which includes placebo controls (Apr 2025 conference news).
In Toronto, Kulic showed nearly final data from the 3.6 mg/kg trial arms, with 73 of the 76 participants having completed six months of treatment. A fourth of them received placebo. Their average age was 72, two-thirds were female, 89 percent were non-Hispanic whites. Two-thirds carried an APOE4 allele, including 12 percent who were E4 homozygote. Half the participants had mild cognitive impairment at baseline, half mild dementia.
The numbers were much like those in the previous report, save for slightly greater plaque clearance and two additional cases of ARIA (see below for details on that). At six months, 91 percent of participants had become amyloid-negative, i.e., 24 centiloids or below. Strikingly, 72 percent were at 11 centiloids or below, suggesting more complete plaque removal. For comparison, at 1.8 mg/kg, 65 and 47 percent of participants ended up at or below 24 and 11 centiloids, respectively. Overall, the average amount of plaque cleared at 3.6 mg/kg was 99 centiloids, and the average final load was five centiloids (image above).
Gregory Klein at Roche noted that the five participants who remained amyloid-positive at 3.6 mg/kg had started with the heaviest loads, with some as high as 200 centiloids. They are now below 50 (image below). Importantly, all participants cleared amyloid at similar rates. “There were no nonresponders,” he said in Toronto. In previous immunotherapy trials, the number of amyloid-negative participants had tended to plateau around 80 percent.
In Toronto, scientists speculated that the speedier plaque removal on this higher dose could make a difference clinically. Sperling noted that plaque acts as a reservoir for oligomers. “Getting rid of it quickly is a good thing,” she said. Gil Rabinovici at the University of California, San Francisco, was interested to learn if fast amyloid clearance would translate into a drop in tangles and a clinical benefit in future studies.
Roche did not collect tau PET data in this trial. Klein said they decided it was too burdensome for participants, who already undergo frequent amyloid PET, blood draws, and lumbar punctures. An audience member asked about clinical outcomes. Kulic said they are analyzing those data and will report the findings later. He minimized expectations, though, noting these cohorts were not powered to show a clinical effect.
Conventional Antibodies Can’t Go Deep
Importantly, at 3.6 mg/kg, plaque fell below the positivity threshold in all areas of the brain, even in deep regions such as the caudate and hippocampus, Klein said. This matters because those areas are crucial in Alzheimer’s disease, and other immunotherapies may not properly reach them.
In Toronto, Baayla Boon of the Mayo Clinic in Jacksonville, Florida, showed neuropathology data that supports this idea. Boon examined postmortem sections from the hippocampus, cerebellum, and frontal and occipital cortices of five people who had participated in the Engage Phase 3 trial of aducanumab. One, from the placebo arm of the trial’s blinded portion, had received almost no study drug before leaving the open-label extension due to ARIA. Two had received a cumulative total of 70 to 90 mg/kg aducanumab; they cleared between 10 and 30 centiloids, remaining amyloid-positive at 105 and 58 centiloids, respectively. The last two each had received a cumulative total of 200 mg/kg or more and became amyloid-negative by PET. All five had carried at least one APOE4 allele, two were APOE4 homozygotes, and two had carried a presenilin 1 mutation.
Despite their different treatment histories and genotypes, frontal cortex sections from all four of the participants who received aducanumab looked strikingly similar. By immunohistochemistry, all showed almost complete plaque clearance in layer 1. In lower layers, there was partial clearance, with plaques somewhat sparser than in untreated, matched controls. However, amyloid remained in deeper layers even in the brains that were considered amyloid-negative by the PET threshold, Boon said. She added that the other brain regions examined looked similar.
Do these findings mean that amyloid PET does not accurately reflect total plaque load? “We need to look into that,” Boon said, noting that higher-resolution PET technology is coming.
Overall, the data confirm that traditional antibodies predominantly reach superficial cortical layers, Boon said. Underscoring this, for two participants who had ARIA, inflammatory changes around vessels were confined to leptomeningeal and penetrating arterioles. This fits with the idea that antibodies primarily cross into the brain at the blood-cerebrospinal fluid barrier, travel over the brain’s surface in CSF, and enter along blood vessels. When those blood vessels have substantial amyloid deposits, the antibodies spark inflammation and ARIA there.
It is unclear if other traditional antibodies might reach deeper brain areas better. In the Engage trial, dosage was low and clearance overall was slow. Boon found no change in p-tau in these brain sections by AT8 staining, while in trials of lecanemab and donanemab, scientists did report a drop in CSF p-tau.
Consistent Picture From Fluid Biomarkers
Back to trontinemab, Klein also updated the Toronto audience on fluid biomarker results. Data for CSF A?42/40, p-tau181, total tau, and neurogranin, as well as plasma A?42/40, p-tau181, and p-tau217, remained nearly identical to the findings reported in March, with all markers normalizing.
Notably, the concentration in plasma of p-tau217, one of the most sensitive markers for plaques and p-tau in the brain, was halved. Henrik Zetterberg of the University of Gothenburg, Sweden, found this impressive. “That’s the biggest drop I’ve seen [in trials],” he said in Toronto. Klein noted that the trajectory of plasma p-tau181 and p-tau217 lowering mimicked amyloid reduction in this cohort.
New in Toronto, Klein added findings from synaptic marker CSF SNAP25, which edged down by 15 percent, and the astrocytosis marker plasma GFAP, which sank by 27 percent. Both represent an improvement. GFAP had an initial transient increase, which Klein speculated could indicate glial activation and phagocytosis. Plasma NfL stayed the same. Several immunotherapy studies have reported this marker tracks poorly with plaque removal (e.g., May 2025 conference news).
Safer at Higher Doses
Safety remains a primary concern for amyloid immunotherapy. Here, too, trontinemab continues to look good, though safety data remain blinded, with placebo and treatment cohorts combined. Therefore, the incidence for trontinemab alone is not yet known.
In the 3.6 mg/kg arm, there has now been one case of mild ARIA-E, for an overall incidence of 1 percent. For comparison, the two approved immunotherapies produce 12 to 16 percent ARIA-E, while placebo controls in those trials had an ARIA rate of 1 to 2 percent. Regarding ARIA-H, there were two instances, both microhemorrhages. There was no concurrent ARIA-E and -H, no superficial siderosis, and no macrohemorrhages.
Oddly perhaps, the 1.8 mg/kg cohort had more adverse events. That arm has had three cases of mild ARIA-E and five instances of ARIA-H. The latter include two people with microhemorrhages and three with superficial siderosis. There was also a previously reported fatal macrohemorrhage in a woman with baseline superficial siderosis and CAA.
On trontinemab, infusion-related reactions were the most common adverse event. These caused symptoms such as chills, fever, headaches, and nausea, and occurred in half of people taking 1.8 mg/kg and 45 percent taking 3.6 mg/kg. They typically happened after the first dose, with the incidence below 10 percent on later doses.
In Toronto, Kulic reported that premedicating participants with steroids docked the incidence per infusion from 32 to 6 percent at 1.8 mg/kg, and from 18 to 7 percent at 3.6 mg/kg. By contrast, premedicating with NSAIDs did not help (image above). In immunotherapy treatment in general, doctors often prescribe steroids to treat ARIA, because these drugs tighten blood vessels and cool inflammation.
Anemia is also a concern with transferrin-based brain shuttles, because the receptor is abundant in red blood cells. Anemia rates were 18 percent at 1.8 mg/kg, and 10 percent at 3.6 mg/kg. Kulic said a 10 percent incidence would be expected in older adults due to frequent blood draws alone. He said Roche would continue to monitor anemia in Phase 3.
Trontier Trials: Expanding Trontinemab’s Frontier
With 3.6 mg/kg clearing the most plaque while producing the fewest side effects, it is no surprise that Roche chose that dose for Phase 3. Participants will receive 3.6 mg/kg monthly trontinemab or placebo infusions for six months, i.e., seven doses. Afterward, they will get 3.6 mg/kg trontinemab or placebo every three months as a maintenance dose. At the end of the 18-month trial, all participants will be invited into an open-label extension.
Each trial aims to enroll 800 people. Participants will have MCI or mild dementia, with an MMSE of 22 or higher. In accordance with current appropriate-use recommendations for amyloid immunotherapies, the trials will exclude anyone with more than four microhemorrhages at baseline, any siderosis, any history of macrohemorrhages, or with severe white-matter disease. However, there will be no restriction on APOE genotype, with APOE4 homozygotes allowed to enroll. Smith noted that homozygotes are excluded from clinical amyloid immunotherapy in Europe and the U.K. “We hope to see safety for them,” she said in Toronto.
CDR-SB will be the primary endpoint; secondaries include MMSE, ADAS-Cog13, ADCS-ADLs, and global CDR score. Participants will donate blood and CSF, and undergo MRIs and amyloid and tau PET.
Both trials are global. Altogether, they will take place in 18 countries on five continents: Argentina, Australia, Brazil, Canada, China, Denmark, France, Germany, Italy, Japan, Netherlands, Poland, South Korea, Spain, Switzerland, Taiwan, the U.K., and the U.S.
Roche will prescreen potential participants with blood-based biomarkers, and will use mobile research units to reach a broader community. The company will also collaborate with the Global Alzheimer’s Platform Foundation, which aids AD trial recruitment. Roche will assist participants with transportation and in other ways to allow a diversity of groups to participate, Smith said. The screening study, dubbed Traveller, has already begun in the U.S. and Canada. Interest is great. “We had huge numbers in the first week,” Smith noted.—Madolyn Bowman Rogers https://www.alzforum.org/news/conference-coverage/roche-spells-out-phase-three-plans-trontinemab