Unlearn in Alzheimer's disease
The scientific intelligence layer for clinical development
In Alzheimer’s disease, where trial failure rates are high and every patient matters, Unlearn turns data into a compounding advantage — so each trial informs the next, decisions get sharper earlier, and new therapies reach patients faster.
.svg)
The Unlearn Platform
Plan, monitor, analyze — one connected platform organized around the key decisions that shape every clinical trial.
Plan
Trial Planning and Simulations
One workspace for upstream trial design
Unlearn brings together the critical components required for trial design — where decisions are iterative, assumptions evolve, and rationale must remain clear and defensible across review cycles.
Continuously search, structure, and summarize relevant literature and regulatory precedent from sources like PubMed, ClinicalTrials.gov, and drugs@FDA — all in one place. Eliminate scattered searches and align on precedent in days, not weeks.
Explore harmonized clinical trial and real-world datasets to validate clinical and statistical assumptions. Assess population characteristics, endpoint behavior, and benchmarks.
Build and compare trial-design scenarios to evaluate endpoints, inclusion/exclusion criteria, sample size, and constraints. Every scenario is reproducible, and linked to underlying evidence — supporting informed design trade-offs before protocol finalization.
.avif)
.avif)
.avif)
Monitor
Advanced Trial Monitoring
Unlearn’s advanced trial monitoring solution turns blinded signals into curated insights that support trial decision-making and keep execution on track.
Detect anomalies as they emerge: unexpected values, off-trajectory responders, and multivariate signals benchmarked against historical patient trajectories rather than generic, population-wide cutoffs.
Surface patient- and site-level issues ahead of the standard data-review cycle, informing decisions such as:
- which patient observations warrant medical review and attention
- which multivariate clinical data outliers require investigation
- which sites to escalate for in-depth monitoring visits
In a reanalysis of the publicly available ADCS DHA Phase 3 Alzheimer’s trial (402 patients, 51 sites), Advanced Trial Monitoring flagged anomalies on the ADAS-Cog Word Recognition subtest roughly two months into enrollment, surfacing coding errors about seven months before the formal Data Safety Monitoring Board review.
Analyze
Trial Analyses with Digital Twins
AI-generated digital twins of trial participants are forecasts of an individual’s control outcomes at every future time point. By predicting how each participant would progress under control or standard of care, they serve as the powering technology for more rigorous clinical analysis.
Design and run smaller RCTs that maintain power or boost it without additional study participants. Unlearn’s digital twin methodology has been validated across multiple Phase 2 and Phase 3 AD trials, demonstrating consistent sample size reductions, cost savings, and accelerated timelines. This approach is qualified by the EMA and aligns with current FDA guidance.
When a concurrent placebo arm isn’t possible, digital twins serve as an AI-generated externally controlled arm (ECA), enabling credible treatment comparisons where randomization is infeasible. Estimate treatment effects with greater rigor than standard external control approaches, and strengthen high-stakes go/no-go decisions through improved statistical sensitivity.
Improve sensitivity in interim looks and subgroup analyses to catch signals traditional methods may miss. Re-evaluate historical trial data using regulatory-aligned methods to support learning across programs.
Technology
Digital Twin Generators
Digital Twin Generators (DTGs) are machine learning models that produce individualized forecasts, called digital twins, of each trial participant’s expected clinical outcomes under placebo. These forecasts are generated using baseline data and include predicted values for item-level assessments, labs, vitals, and other clinical measures. DTGs are powered by Neural Boltzmann Machines, a proprietary machine learning architecture optimized for probabilistically modeling complex, multivariate, clinical time-series data.
.webp)
Regulatory Acceptance
Paving the regulatory path for AI in clinical trials
Unlearn’s methods have been recognized and supported by both U.S. and European regulators.
Our digital twins-based method was officially qualified by the European Medicines Agency for use in Phase 2 and 3 trials with continuous outcomes.
U.S. FDA provided positive feedback on PROCOVA, supporting its use in covariate-adjusted analyses across clinical development.
FDA recommends that sponsors adjust for covariates that are anticipated to be most strongly associated with the outcome of interest…it may be useful to use previous studies to select prognostic covariates or form prognostic indices.
In a trial that uses covariate adjustment, the sample size and power calculations can be based on adjusted or unadjusted methods.
Case Studies from retrospective analyses
Phase 2 — Tilavonemab trials
Reduced overall sample sizes by up to 15% and control arm sizes by up to 25%
Participants' digital twins demonstrated a 10–15% reduction in overall sample sizes and an 18–25% reduction in control arm sizes when estimating the treatment effect on CDR-SB at Week 96.
Boost your chance of observing a treatment effect by up to 35%
Digital twins demonstrated increased power by up to 7% (from 80% to 87%) on ADAS-Cog11 at Week 96 using PROCOVA™. A 7% increase in power translates to a 35% reduction in type 2 error.
Phase 3 — Bapineuzumab trials
Reduced control arm size by up to 33%
Participants’ digital twins, combined with observed outcomes, demonstrated a 30-33% reduction in the sample size in the control arm, depending on the treatment arm and APOE e4 allele status, when estimating the treatment effect on ADAS-Cog11 at 18 months.
Boost your chance of observing a treatment effect by up to 50%
Participants’ digital twins, combined with observed outcomes, demonstrated increased power by up to 10% (from 80% to 90%) when estimating the treatment effect on ADAS-Cog11 at 18 months, using *PROCOVA. A 10% increase in power translates to a 50% reduction in type 2 error.
AbbVie utilized Unlearn's digital twins in their completed Phase 2 AWARE study (NCT02880956), a randomized, double-blind, placebo-controlled trial in early AD. The trained AD-DTG was pre-specified before conducting retrospective analyses. Prognostic scores derived from participants' digital twins were used to estimate treatment variance for the clinical outcomes.
A leading global pharmaceutical company used digital twins in three randomized, double-blind, placebo-controlled completed Phase 3 trials in mild-to-moderate AD and mild cognitive impairment to demonstrate the potential of participants’ digital twins in reducing sample sizes or increasing study power using ADAS-Cog11 and CDR-SB clinical outcomes. Efficacy was assessed using ADAS-Cog11 score at 18 months.
- APOE e4 allele carriers (NCT00575055)
- APOE e4 allele non-carriers (NCT00574132)
Unlearn x AbbVie
Conference
Vienna, Austria
Evidence
Driving Impact Across Clinical Development
Explore how our partners are accelerating their clinical development programs with us.
Reduce sample sizes while maintaining power or boost power without adding participants
Strengthen evidence and confidence in early-stage studies
.avif)
