Neurodegenerative diseases like Alzheimer’s, Parkinson’s, Huntington’s, and ALS are among the most complex and devastating conditions to study in clinical trials. These diseases progress unpredictably, affecting multiple aspects of a patient’s abilities over time.

To measure whether a treatment is working, researchers rely on composite endpoints—scoring systems that combine multiple symptoms into a single measure. But these endpoints don’t always reflect the true impact of a disease or a treatment, making it harder to detect meaningful effects.

At Unlearn, we’re advancing AI-powered solutions that help pharmaceutical companies make smarter, faster, and more confident decisions in neurodegenerative disease trials. In our latest research, which will be presented at ISCTM 2025, we show how our Digital Twin Generators (DTGs) can optimize composite endpoints to deliver clearer, more actionable evidence—helping to accelerate drug development and improve patient outcomes.

‍‍

The Problem: Noisy Endpoints in Neurodegenerative Trials

Neurodegenerative diseases affect multiple biological systems, leading to complex symptoms that evolve over time. To capture disease progression, researchers use well-established composite endpoints such as:

• Alzheimer’s Disease: Clinical Dementia Rating Scale Sum of Boxes (CDR-SB)
• Parkinson’s Disease: Movement Disorder Society - Unified Parkinson’s Disease Rating Scale (MDS-UPDRS)
• Amyotrophic Lateral Sclerosis (ALS): Revised ALS Functional Rating Scale (ALSFRS-R)
• Huntington’s Disease: Composite Unified Huntington’s Disease Rating Scale (cUHDRS)

These endpoints play a critical role in regulatory approval, but they also come with inherent challenges. They typically assign equal weight to all symptoms, even though some may have a greater impact on disease progression and patient quality of life. This can introduce unnecessary variability, making it harder to detect meaningful treatment effects.

For example, a recent global initiative in ALS research has highlighted the need to improve the ALS Functional Rating Scale (ALSFRS-R), as it treats all 12 symptom categories equally, even though some decline faster and have a greater impact on patient quality of life. The same challenge exists across neurodegenerative trials—introducing noise that can obscure treatment effects and make it harder to determine if a drug is working.

‍‍

The Solution: Optimizing Endpoints with AI

To address this challenge, we’ve developed a machine learning-based method that reweights composite endpoints by assigning data-driven weights to their individual components. Instead of treating all symptoms equally, this approach ensures that the most clinically meaningful and sensitive measures carry more weight—helping trials detect treatment effects more reliably.

‍

How DTGs Make This Possible

A Digital Twin Generator (DTG) is an advanced, disease-specific machine learning model that generates individual forecasts of disease progression under control conditions for each participant in a clinical trial.

‍‍

Here’s how it works:

1. DTGs are trained on thousands of patient-level data points from past clinical trials and observational studies.
2. After testing and validation, the DTG uses baseline data from a trial participant to generate their unique digital twin—a model of how their disease is expected to progress without treatment.
3. These AI-driven forecasts help refine composite endpoints, ensuring that each component is optimally weighted to maximize trial power without increasing patient enrollment.

The Impact: Helping Pharma Make Smarter Decisions in Neurodegenerative Trials

For pharmaceutical companies developing disease-modifying therapies for neurodegenerative diseases, clear decision-making is everything. This method helps answer some of the most important questions in clinical trials:

• During trial design: Should optimized endpoints be prespecified in our SAP as sensitivity analyses or exploratory endpoints to better understand treatment effects? 
• At the end of a trial: Would optimizing endpoint weighting have led to detecting a clearer treatment effect? Does this optimized endpoint give us clearer interpretation of treatment benefit to patients?

What We Found: AI-Optimized Endpoints Improve Trial Power

In our study, we applied this method to key neurodegenerative disease endpoints. The results showed that AI-optimized composite endpoints consistently improved statistical power, making it easier to detect real treatment effects.

• Optimizing endpoints in ALS and Huntington’s disease trials led to substantial power gains, meaning a higher likelihood of detecting a true drug effect.
• In Alzheimer’s trials, the optimization confirmed that some commonly used endpoints are already well-weighted while others could be improved.
• In Parkinson’s trials, we saw variability in power gains, suggesting the need for further research into endpoint optimization.

The Bigger Picture: More Efficient, More Reliable Trials

Neurodegenerative diseases are among the most challenging conditions to study in clinical trials, and inefficient endpoints only make it harder to bring new treatments to patients. AI-driven trial optimization changes that by helping pharmaceutical companies:

• Reduce uncertainty in decision-making
• Provide impactful evidence to key stakeholders
• Increase the chances of detecting a meaningful treatment effect

Better endpoints mean better trials, better decisions, and, ultimately, better treatments for patients.

Want to learn more? 

Get in touch to see how Unlearn can help optimize your neurodegenerative disease trials.