The cost to create an innovative drug within the U.S. ranges between $1-2 billion, and pharmaceutical companies are now spending more than 10 times what was spent on drug R&D in the 1980s according to the Congressional Budget Office. That’s because 90% of medicines fail at some point during the clinical evaluation pathway, spanning preclinical research through Phase 3 trials to determine if a medicine is safe and effective. To reduce development costs, associated risks and streamline the process, there’s a growing trend to use AI-generated digital twins in drug research.
Before you roll your eyes in digital disbelief, ClinicalLeader.com estimates that AI will be used in 60-70% of clinical trials by 2030, saving the biopharma industry $20-30 billion annually thanks to speedier trial timelines and cost savings. While AI digital twins are likely to be only part of that growth, the ability to virtually test the safety and efficacy of new medicines would be a major leap forward for drug makers, physicians and patients.
AI Digital Patient Twins Are Here – But What Are They?
It’s important to note that AI generated, patient-specific digital twins are not science fiction – they’re available now. Digital patient twins are being evaluated by regulators and quietly integrated into trial designs across the industry.
Put simply, a digital twin within healthcare is a computer-generated version of a real patient that predicts how that person’s health would change over time without receiving the new treatment. It’s built using past medical data and helps researchers compare what actually happens to a patient with what likely would’ve happened if they got the standard treatment or a placebo. And that’s already changing the game.
“Our AI-generated digital twins are comprehensive, individualized predictions of a patient’s future health outcomes based on their baseline characteristics. These predictions span the multitude of assessments, laboratory tests and clinical events that define a patient’s health trajectory,” said Aaron Smith, founder and machine learning scientist at Unlearn.ai in an email response.
“Digital twins are powerful tools to understand disease progression and answer key clinical research questions – one of the most important being how to make clinical trials more efficient.”
How AI Digital Twins Cut Clinical Costs And Time
Unlearn is one of the leading companies using digital twins to improve trial efficiency. Their proprietary “Digital Twin Generators” use disease-specific neural networks to predict how a patient’s condition would evolve over time without the new treatment.
That opens up two major use cases, first as simulated controls in single-arm clinical trials and also as prognostic covariates – basically a set of facts about a patient to predict health outcomes – in randomized controlled trials.
Both approaches are gaining acceptance from regulators, including the FDA and European Medicines Agency. The EMA has already issued a formal qualification opinion for Unlearn’s PROCOVA method – an advanced form of covariate adjustment that relies on digital twin data to improve trial power without increasing sample size.
“In randomized controlled trials, digital twins enhance the data measured in both the treatment and control arm, increasing statistical power without sacrificing statistical rigor,” Smith explained. “Alternatively, digital twins can reduce the number of control participants required while maintaining the original power, making studies smaller, faster and more efficient.”
The implications are significant. According to Smith, even a 10% reduction in sample size for a Phase 3 trial could cut enrollment time by four months and save tens of millions of dollars. For patients with rare diseases or aggressive cancers, those months matter.
AI Digital Twins As A Compassionate Alternative
The digital twin technology is especially compelling in situations where traditional control groups are difficult or ethically questionable – think pediatric trials, late-stage oncology or ultra-rare conditions. However, within each of those instances it would not be right to withhold potentially lifesaving medicines to patients in the control arm of the study.
Most control group participants are dosed with placebos or sugar pills – without their knowledge or the clinical investigators’ – to create a reliable baseline for the real medicine to perform against. For such dicey ethical dilemmas, digital twins provide a realistic and regulator-accepted alternative to real-world control arms.
Smith noted that a common misconception with the AI tech is the belief that digital twins only work as synthetic controls. “While that is one valid application, a completely separate and arguably more scalable application is the use of digital twins in randomized controlled trials,” he said.
“In that setting, digital twins don’t act as replacements for patients. Instead, they provide additional prognostic information about each enrolled participant.”
That extra layer of data sharpens the findings and makes trial results more reliable. But it also raises questions around transparency and explainability – concerns that regulators haven’t overlooked.
“We take transparency seriously, documenting every aspect of our models – from how they’re trained to how they’re validated in their specific context of use,” Smith said. “That trust is key to adoption.”
The Value Proposition Of AI Digital Twins Is Clear
Given increasing costs, delays and patient recruitment challenges, digital twins could ease pressure across the clinical research sector. A peer-reviewed position paper published by Unlearn, stated that the use of digital twins could reduce the burden on real trial participants by reducing placebo assignments and shortening overall trial duration. That’s not just a technical improvement – it’s a human one.
However, digital twins are only as good as the data that train them, and healthcare datasets remain fragmented and noisy. Bias, inaccuracies and missing values can all weaken model reliability. That’s why Unlearn emphasizes rigorous validation within each trial’s specific context. “Validation in the context of use is an essential step in applying these methods,” Smith said.
Looking ahead, scaling the use of digital twins will require more than AI innovation – it will demand a cultural seachange. Trial sponsors, clinical research organizations and regulators must get comfortable with machine-learning outputs shaping critical clinical decisions. But if recent history is any guide, that transition is already underway.
“We’re at an inflection point,” Smith said. “Digital twins are no longer theoretical. They’re working, they’re trusted and they’re delivering value today.”
And for an industry in constant search of smarter, faster and safer ways to deliver treatments, AI’s evolution toward digital twins could be just what the doctor ordered.