Symbol Swap matched or exceeded ADAS-Cog, MoCA, and MMSE across three independent studies
By Henry Baker
Published: Jul 15 2026, 3:39 PM EDT
On the final day of the world’s largest dementia research conference, a startup presented data Wednesday suggesting that a two-minute tablet task can do what a trained clinician typically takes 10 to 45 minutes to accomplish — and in some cases, do it better.
Cumulus Neuroscience unveiled two posters at the Alzheimer’s Association International Conference (AAIC) 2026 in London, showing that its Symbol Swap task, part of the company’s NeuLogiq® Platform, matched or outperformed three gold-standard clinical cognitive assessments for identifying participants most likely to carry the biological hallmarks of Alzheimer’s disease. The presentations landed at a conference drawing more than 12,000 researchers from 115 countries to ExCeL London — a record turnout — at a moment when Alzheimer’s trial infrastructure is straining under its own ambitions.
The most striking finding was not that the tool worked in people who already show cognitive symptoms. It was that Symbol Swap detected Alzheimer’s pathology in people who look completely normal on standard tests — the exact population that modern drug development needs to reach.
Alzheimer’s Trial Enrollment Is Broken — Here Is the Size of the Problem
Anyone who has watched an Alzheimer’s drug development program will recognize the bottleneck Symbol Swap is designed to pierce. Trials require participants who show early cognitive signs of decline and who also carry the biological hallmarks of Alzheimer’s — amyloid plaques and tau tangles — in their brains. Finding those people is expensive, slow, and riddled with failure.
Screen failure rates for prodromal Alzheimer’s trials run as high as 78%, and for preclinical trials they reach 88%. A single PET brain scan to confirm amyloid status in the United States can cost around $8,000. As biomarker-based eligibility criteria have grown more common — with 57% of Alzheimer’s trials in 2025 requiring biomarker confirmation — patients respond to a recruitment campaign, pass initial screening, and complete clinical assessments, only to be excluded after biomarker testing, extending timelines further.
The enrollment gap is equally stark. In 2025 alone, 182 Alzheimer’s treatment trials were actively recruiting. Of the roughly 50,000 participants needed to fully enroll them, only about 11,000 enrolled each year. That 39,000-person annual shortfall means promising drug candidates take longer to test, cost more to test, and sometimes fail to enroll at all.
A cheap, fast, automated digital prescreen that flags who is most likely to be biomarker-positive before any blood draw or brain scan is the field’s single most commercially attractive infrastructure problem. Wednesday’s data from Cumulus suggest the solution may be on a tablet.
What the Data Show
The centerpiece of the AAIC presentation was a poster titled “A Brief Digital Symbol-Coding Task Outperforms Clinical Benchmarks for Alzheimer’s Trial Pre-enrichment.” The primary tool, Symbol Swap, is a two-minute tablet-based implementation of the classic Symbol Coding task — formally known as the Digit Symbol Substitution Test — in which participants pair geometric symbols with numbers as quickly as possible.
Across three independent datasets, Symbol Swap matched or exceeded the pre-enrichment performance of the ADAS-Cog, the MoCA, and the MMSE in distinguishing control, MCI, and Alzheimer’s dementia groups. The ADAS-Cog is the primary cognitive scale used in most major Alzheimer’s trials and requires a trained clinician roughly 45 minutes to administer. Symbol Swap did comparably in two minutes with no clinician required — administered digitally and scored automatically.
Perhaps more consequential was what the task found in people who look cognitively normal. Symbol Swap showed robust associations with blood plasma pTau-217 — a biomarker that signals Alzheimer’s pathology in the brain — even in clinically unimpaired individuals who would typically score at ceiling on standard tests. This suggests the task could flag who is most likely to be amyloid- or tau-positive, enabling downstream plasma, PET, or CSF biomarker testing to focus on the highest-yield candidates rather than an unselected pool.
Why Processing Speed Detects What Memory Tests Miss
The technical reason Symbol Swap can identify preclinical Alzheimer’s pathology comes down to what cognitive domain it measures — and why that domain is sensitive earlier than the memory function tested by traditional scales.
Symbol Swap is grounded in processing speed: the efficiency with which the brain encodes, matches, and responds to new information under time pressure. Processing speed depends critically on the integrity of white-matter connectivity and the prefrontal-parietal networks that coordinate rapid information flow across cortical regions. Amyloid plaques and tau tangles — the defining pathologies of Alzheimer’s disease — disrupt these networks before they destroy neurons outright. A person with early amyloid accumulation may still pass an episodic memory test because their hippocampus has not yet been severely damaged, but their processing speed may already be measurably slower because their white-matter integrity has begun to erode.
Standard instruments like the MMSE, designed in the 1970s and 1980s, were built to detect overt dementia, not subtle processing-speed changes. Symbol Swap, by targeting processing speed via a gamified tablet interface, captures a signal that those instruments were never designed to see.
This is also why Symbol Swap’s correlation with pTau-217 in cognitively normal individuals is significant. The ALzPath Simoa assay used in the CNS-101 and Fastball studies measures phosphorylated tau at the threonine-217 position — a marker that reflects amyloid-driven tau pathology with accuracy exceeding 90% in some cohorts. If a two-minute processing-speed task produces a behavioral signal that tracks the same biology, it gives trial teams a scalable, low-cost first-pass filter that a blood draw can then confirm.
How the Studies Were Run
Symbol Swap was evaluated in three settings with different comparator assessments and participant populations.
CNS-101 was a year-long validation study in which people living with mild Alzheimer’s dementia and healthy controls used the NeuLogiq Platform at home, with Symbol Swap compared against the ADAS-Cog. Pathology status was defined by the AlzPath pTau-217 assay. The Fastball i4i study ran in-clinic with Symbol Swap compared against the MoCA, with the same pTau-217 pathology definition. Data presented from the Fastball i4i study are interim.
The Global Alzheimer’s Platform BioHermes-002 study, spanning 20 sites across the US, Canada, and Europe, compared Symbol Swap against the MMSE. Those data are also interim. The breadth and geographic range of BioHermes-002 — coordinated by GAP, a nonprofit trial infrastructure organization — gave the overall evidence package unusual real-world credibility.
Lammert Albers, GAP’s Chief Commercial Officer, said the data suggested that digital assessments including Symbol Swap may help identify individuals more likely to have underlying Alzheimer’s pathology, potentially supporting more efficient follow-up biomarker testing with the potential to shorten enrollment timelines and lower screening costs.
The Platform Behind the Task: EEG and AI in a Wireless Headset
Symbol Swap is one component of the broader NeuLogiq® Platform, which combines tablet-based cognitive tasks with a wireless dry-sensor EEG headset that captures electrophysiological signals during and between tasks. The EEG hardware is FDA 510(k)-cleared and UKCA-marked as a Class I medical device designed for self-setup at home, without the need for technicians or gel electrodes.
Traditional clinical EEG requires trained personnel to apply conductive gel and position wet electrodes precisely. Cumulus’s dry-sensor design uses a lightweight headset that participants set up themselves, enabling the kind of frequent, longitudinal brain function measurement that clinic-based snapshot assessments cannot achieve. The platform integrates AI-driven analytics applied to a real-world database of annotated, longitudinal patient data to generate multi-domain biomarker signals spanning cognition, electrophysiology, and sleep.
A second poster at AAIC, presented by Dr. James Rowe of the University of Cambridge (Principal Investigator of CNS-101), addressed the platform’s combined cognitive and EEG endpoints in the decentralized trial context. Results from CNS-101 confirmed that participants with mild Alzheimer’s dementia tolerated the multi-domain protocol well across a full year of at-home use.
Standard clinical endpoints like the ADAS-Cog and the Clinical Dementia Rating scale rely on infrequent, clinic-based snapshot assessments prone to rater error and with limited sensitivity to early change — a shortcoming that inflates the number of participants needed to detect a drug effect and drives up trial costs.
“In CNS-101 we found that multi-domain digital cognitive and EEG measures from the NeuLogiq Platform were well tolerated by people living with Alzheimer’s disease and sensitive enough to capture meaningful change,” Rowe said. “That sensitivity makes it possible to design smaller, more efficient trials — reducing the burden on patients while lowering the time and cost to test new therapies.”
The platform was designed with input from a pharmaceutical advisory group representing ten of the world’s leading pharma companies, and is already deployed in trials for Alzheimer’s disease, depression, and schizophrenia.
Read more: Oral Alzheimer’s Drug Enters Human Record With First Anti-Neuroinflammation Signal
What Digital Prescreening Could Do to Trial Sample Sizes
The implications of a reliable digital prescreening tool extend beyond reducing individual screening visits. Published research suggests the effect on overall trial scale could be substantial.
A February 2026 analysis of the Anti-Amyloid Treatment in Asymptomatic Alzheimer’s Disease (A4) Study — a major multinational secondary-prevention trial — found that combining a brief digital memory assessment with plasma pTau-217 reduced estimated sample-size requirements for a preclinical Alzheimer’s trial from 3,252 participants per arm down to 818, a 75% reduction. That reduction reflects the power of identifying, before a single intervention dose is given, the subset of cognitively unimpaired amyloid-positive people who will actually show measurable decline over the trial period.
Brian Murphy, Cumulus Co-Founder and Chief Scientific Officer, framed the AAIC findings in terms any biopharma executive who has watched a trial collapse under the weight of its enrollment funnel will recognize: “It is striking that a two-minute, patient-friendly task can match or beat assessments that take a trained clinician 10 to 45 minutes to administer — and that it picks up Alzheimer’s pathology even in study participants who look cognitively normal on standard tests.”
CEO Tina Sampath described the two AAIC posters as telling a connected story: Symbol Swap as a front-end enrichment filter, and the CNS-101 multi-domain endpoints as enabling smaller trials once the right participants are enrolled. “Together, that’s a path to substantially reducing the cost and timeline of every study where they’re deployed,” she said.
What Still Needs to Happen Before This Changes Standard Practice
The data presented Wednesday remain largely interim for the BioHermes-002 and Fastball i4i studies, and independent peer-reviewed publication of the three-dataset findings will be necessary before Symbol Swap is likely to be adopted as a standard front-line filter in pivotal trials.
There is also a regulatory qualification gap. The NeuLogiq EEG headset has FDA 510(k) clearance as a medical device — a finding of substantial equivalence to an already-marketed device. But Symbol Swap itself has not been qualified as an FDA Drug Development Tool for cognitive enrichment, which would be required before sponsors could use it as a primary cognitive endpoint in an FDA submission. Without that qualification, Symbol Swap can function as an enrichment tool — a way to identify likely biomarker-positive candidates before a more expensive confirmatory test — but cannot independently carry a trial’s primary efficacy claim. Cumulus has not announced any ongoing FDA qualification process for Symbol Swap.
The competitive landscape matters too. Linus Health, Cogstate, and Neotiv are all active in the digital cognitive assessment space for Alzheimer’s trials, with Cogstate’s Brief Battery already widely deployed in pivotal studies and Linus Health presenting data at CTAD 2025 on its own brief digital prescreening approach. Symbol Swap’s three-dataset evidence package and its specific pTau-217 correlation data give Cumulus a well-documented validation profile, but independent head-to-head comparisons have not been published.
Still, the trajectory of the evidence is notable. Symbol Swap has now been evaluated in three independent study populations with consistent results across different comparators and settings. The company describes the platform as scalable across in-clinic and at-home workflows — an important consideration as the field moves toward decentralized trial designs that reduce the geographic and physical burden on participants. If the final results of BioHermes-002 and the Fastball i4i study confirm what Wednesday’s data suggest, Symbol Swap may represent a meaningful shift in how Alzheimer’s trials handle their most expensive and failure-prone phase — before a single PET scanner is switched on.
Frequently Asked Questions
What is the Symbol Swap task and how does it measure brain function?
Symbol Swap is a two-minute tablet-based test derived from the Symbol Coding paradigm — a classic neuropsychological task in which participants match geometric symbols to numbers as quickly as possible. It measures processing speed, which reflects how efficiently the brain’s networks communicate under time pressure. Processing speed depends on white-matter integrity and prefrontal-parietal connectivity — circuits that Alzheimer’s pathology disrupts before it causes the overt memory loss detectable by traditional clinical tests. By targeting this earlier signal, Symbol Swap can identify cognitive changes in people who still score normally on tests like the MMSE.
Why does Alzheimer’s trial enrollment fail so often, and what would digital prescreening change?
Alzheimer’s clinical trials screen patients through a cascading series of tests — clinical assessment, cognitive screening, and finally biomarker confirmation through blood tests, spinal fluid, or PET brain scans — before anyone is enrolled. Screen failure rates run as high as 88% for preclinical trials, meaning most screened patients are excluded after significant time and cost. A reliable digital prescreen administered at home or in a primary care office before a blood draw or $8,000 PET scan could redirect biomarker testing toward only those most likely to qualify, reducing the number of failed screening visits and the costs they generate.
Could digital cognitive tests allow trials to enroll dramatically fewer participants?
Published evidence suggests yes, at least in combination with blood-based biomarkers. A 2026 analysis of the A4 Study found that combining a brief digital memory assessment with plasma pTau-217 testing reduced sample-size estimates for a preclinical Alzheimer’s trial by 75% — from more than 3,000 participants per arm to fewer than 900. The mechanism is straightforward: enriching a trial with people who will actually decline during the trial period — rather than cognitively stable amyloid-positive people who will remain flat — dramatically increases statistical power without adding participants. Symbol Swap has not yet been evaluated specifically in this dual-enrichment framework with pTau-217, but its correlation with pTau-217 in cognitively normal individuals positions it as a candidate for exactly this kind of combined approach.
Has Symbol Swap been approved or qualified by the FDA for use in Alzheimer’s drug trials?
Not as a primary clinical endpoint. The NeuLogiq EEG headset has received FDA 510(k) clearance as a Class I medical device, meaning the FDA determined it is substantially equivalent to an already-marketed device for measuring electrophysiological brain signals. However, Symbol Swap itself has not yet been qualified through the FDA’s Drug Development Tool program, which would be required before it could serve as a primary cognitive endpoint in a pivotal trial supporting a drug approval. Sponsors can currently use Symbol Swap as an enrichment and exploratory tool. FDA DDT qualification would require additional validation work and formal submission to the agency.