Answering the Big Questions in Progressive MS

“Once we put in place a system where academics, industry and patient representatives can collaborate and ask scientific questions, we’ll learn about the disease, disease progression and the patient journey,” explains Florian von Raison, MD, Clinical Development Head, Neuroscience at Novartis Development and Co-Chair of the Alliance Industry Forum. “This larger data set will provide valuable information that will help answer questions we need in order to stop the progression of MS.”

Read on to learn about the key players in this groundbreaking initiative, how artificial intelligence is changing the research game, and how the sharing of large volumes of data across academia and the pharmaceutical industry will help accelerate development of new treatments for progressive MS.

The key players

The Alliance has supported an unprecedented collaborative effort to collect data, according to lead investigator Douglas Arnold, MD, who serves as James McGill Professor of Neurology at McGill University in Montreal, in which each of the following players has been essential.

• The International Progressive MS Alliance is a nonprofit, global entity that funds and oversees the entire project. It provided a forum for pharmaceutical companies to come together and agree to contribute data from their individual clinical trials and now spearheads the coordination and framework for others in the MS research community to access this data. MS organizations that lead the Alliance create a collaborative, knowledge-focused space for industry leaders to have frank discussions about what is needed to develop new treatments. “The reason the Alliance is so valuable,” says von Raison, is that they’re “somebody neutral in the middle [of industry and academic research centers] to make resources available in a robust, compliant and transparent way.”
• Industry partners have donated their massive clinical trial data sets—which include MRIs, clinic visits and other data from nearly all recent MS trials from all the big pharmaceutical companies in the MS space—to be used by MS scientists. Not only is this huge volume of combined data an unprecedented resource for researchers in the field, but it’s also necessary in order for artificial intelligence to be able to draw valid conclusions. “The beauty is that we have a shared goal of finding solutions to progressive MS, a common passion that unites us,” von Raison says. The pharmaceutical companies in the Alliance’s Industry Forum will continue to contribute new data over time so the data resource can stay as up to date as possible.
• Academics, including the Imaging and Clinical Trials Collaborative Research Network, use the data with the aim of solving the various challenges of progressive MS. “Collaboration is becoming more important between clinicians and computer experts/AI experts,” Arnold says. “They each need to learn the other’s language and the other’s expertise so they can make progress in this field.”
• People with MS who have participated in clinical trials agreed to share their information with researchers (an essential resource), which researchers have made anonymous—using codes instead of names—to protect their privacy. “The project will amplify the contributions that people with MS made and have their data contribute to understanding MS better and inform clinical trials,” says Tim Coetzee, PhD, President and CEO of the National MS Society and Executive Committee Chair of the International Progressive MS Alliance.

The role of artificial intelligence

MS is a very complex disease that has no known cure. “We don’t have a good understanding of the mechanisms that drive progression yet,” says Tal Arbel, PhD, a Professor in the Department of Electrical and Computer Engineering at McGill University, where she is Director of the Probabilistic Vision Group and Medical Imaging Lab in the Centre for Intelligent Machines. Professor Arbel is also a CIFAR AI Chair at Mila (Quebec AI Institute). This knowledge gap has been the major roadblock to developing new therapies.

That’s where AI comes in. The nice thing about AI, or deep learning, according to Arbel, is that it can mine big data—including MRI sequences and clinical and demographic information, to build predictive factors. “AI learns what those predictive features are instead of using a pre-conceived set of known markers.”

“Our recent exciting developments in this space focus on models for image-based personalized medicine,” Arbel continues. How it will work: “You start with a baseline image when a patient first sees their doctor. The patient comes in; we have their clinical information and demographics. Our AI model uses an algorithm to predict outcomes or disease course of the patient in the short term based on different treatment options, including placebo, and uses that information to show what the patient’s response could be to certain drugs.”

What’s most important, Arbel adds, is breakthroughs in understanding the biological mechanisms underlying disease progression. “Different subgroups of people (e.g., age, gender, geographic location) might respond differently,” she says. “It’s not one size fits all.”

To that end, Arbel’s team is also addressing possible bias in AI, a concept called the “fairness gap,” in which a model works well for some subgroups of patients but not others. “In the context of MS, one bias is age. For example, brain atrophy is thought to be a marker of future progression. However, this marker is not predictive for everyone. As a result, the AI model will be biased towards older patients, for example, who typically have atrophy. We are developing models that identify these types of biases and account for them. Our group is working on explainability of our AI models, which will lead to the discovery of new markers for different patient subpopulations,” Arbel says.

How the Data Sharing initiative will speed clinical trials

Since there are currently no biomarkers to predict disease progression in progressive MS, phase II trials use relatively small numbers of patients to prove that a drug is working before going on to a big phase III trial. “This is a hurdle because large phase III trials cost hundreds of millions of dollars each, so if they don’t work, it’s a big hit for the company,” Arnold says. “It restricts innovation.”

AI tools, however, can help find clues that are predictive of success. “Then you could do a relatively small phase II trial enriched with patients who are likely to respond to a treatment,” Arnold explains. “For phase III, you’d expand enrollment to patients who are selected by the algorithm for their ability to respond, along with general patients, and see how well the treatment holds up.” Ideally, he adds, the new AI-based tools and other statistical methods should optimize trials, so they’ll be shorter, less expensive and hopefully successful.

AI models will also make future drug trials more effective. “As treatments become available, it becomes unethical to use a placebo (i.e., nobody should be denied a treatment for the sake of a trial), so drug companies need to design trials for new drugs against other already approved treatments, so they can gather sufficient data to draw conclusions,” Arnold says.

Lastly, a data sharing agreement will help smaller or new drug companies, and the MS research community at-large, conduct more informed research right out of the gate. “New start-up companies have no data with which to design their trials,” Arnold explains. “They can participate in this initiative and use other company’s data to inform their own new research and then share their findings with the community once they have them.”

How new tools will point out the best treatments

Ultimately, the initiative should help create tools that can be used in a clinical setting. “Because we have thousands of brain images, it means we can better understand how people with MS are going through different phases of the disease,” Coetzee explains. “Using imaging, we’re able to classify the damage in the brains of persons with MS into different profiles. As you follow people over time, does a particular profile predict that someone will have faster progression?”

Coetzee’s hope is that down the line, doctors will be able to say to people with MS, “‘Based on your latest MRI and what the software is telling me, your MS progression remains stable,’ or ‘Look, it suggests that the disease might be more active, so let’s switch to a different therapy.’”

Being able to compare a single patient to the experiences of thousands of patients should help point to treatments that will be most effective in managing the disease. To get there though, Coetzee adds, scientists still need to gather more images, test algorithms, and validate that AI is picking up signals that are accurate and specific. They’ll also need to consider more diverse samples—people of different races, ethnicities, and durations of disease, so the findings are more universally applicable to varying populations.
“These models will someday help doctors and patients have conversations about changes they might need to make to treatment,” Coetzee says. “AI is not going to replace the doctor-patient conversation, but it will be an important adjunct.”

Why data sharing is so critical

The Data Sharing Initiative promises to “unlock the contributions of clinical trial data,” Coetzee says. As he explains, the data comes from companies testing a particular treatment, which means there will be MRIs, careful assessments of how patients are doing and different blood tests—all captured over a three-year period, which is a lot of information on each patient, multiplied by 15 or 16K. “In the past, that data would sit on a proverbial electronic shelf,” he says. “This initiative brings all that data together and says, ‘what can this data tell us?’”

“We start to see patterns of how MS develops and how it progresses, including factors that contribute,” he continues. “You wouldn’t see that from one trial, but in aggregate, that’s when you start to see it.”

“It’s staggering how much data we have,” he adds, “and the incredible possibility it has for the future.”