AI has been a transformative force in clinical research, and its impact continues to grow. By applying data-driven insights throughout the clinical trial lifecycle, from optimising study design to implementing advanced simulation and forecasting capabilities, AI enhances the planning and execution of clinical trials, saving time, reducing costs, and delivering significantly better outcomes for patients.
The growing need for AI
Clinical trials are growing in complexity, with increasing decentralisation, innovative designs coupled with a vastly expanding number of endpoints. An average of 3.6 million data points are now collected in Phase 3 clinical trials – a more than sevenfold increase over the past 20 years. Traditional methods are not enough to keep up with this pace of change. Instead, strategically implementing AI tools can help companies to navigate these complexities, increasing efficiencies, reducing costs and accelerating drug development timelines.
Practical uses of AI in clinical trials
Rather than opting to use a single AI tool to assist with a particular task, we are now at a stage where AI can be fully integrated across the trial lifecycle, ensuring that intelligent, data-driven decision-making is embedded throughout the process.
Optimising Trial Design
AI revolutionizes clinical trial design, drastically cutting timelines even before the first patient enrolls. AI models leverage historical data to rigorously test proposed inclusion/exclusion criteria, anticipating their impact on patient recruitment and retention. These models can also pinpoint patient subgroups at high risk of dropout or adverse events, proactively addressing challenges before they arise. This early integration of AI not only boosts the efficiency of entire clinical programs but also accelerates the advancement of the most promising molecules. In one instance, AI solutions helped to shorten an oncology trial timeline by at least one year, by using data to correlate early results with long-term survival. Additionally, many AI users I work with report gains in protocol design, site feasibility, and cohort identification.
Furthermore, AI ensures study designs align seamlessly with regulatory expectations. It identifies successful endpoints used by competitors and accepted by regulators, determines the most appropriate standard of care comparators, and suggests innovative surrogate endpoints. This significantly increases the likelihood of drug approval and reimbursement.
Mitigating Cost Issues
AI has dramatically driven down the marginal cost of generating critical insights. A few years ago it might have taken a couple of human experts months to extract even a single meaningful insight from a large dataset. Today, AI copilots, endowed with a deep understanding of both data and historical context, can empower the same team to perform multiple analyses in the same timeframe with minimal additional cost. While human expertise remains paramount, AI supercharges its productivity value.
Forecasting Potential Issues
Mid-study changes, if mishandled, can unleash significant time and cost penalties on a clinical trial. Trial delays caused by these issues cost companies around $40,000 per day, and contribute to approximately $500,000 in lost sales per study. However, sophisticated “what-if” simulations, powered by AI, can precisely evaluate the potential impact of these changes on factors like study enrollment and timelines, ensuring their implementation is strategic and maximally efficient.
Moreover, AI delivers real-time performance oversight, benchmarked against forecasts or similar past trials, and predicts potential safety events. This proactive vigilance allows for early identification and mitigation of issues, minimizing their downstream impact, and ensures that companies are making the right decisions today that will prevent issues tomorrow.
Human oversight of these tools remains imperative, as the intention is for AI to complement human judgement, rather than replacing it entirely. By identifying potential issues at an earlier stage, study teams can devote more of their time to critical thinking and decision making, rather than dealing with issues after they have arisen.
The importance of good data
Data quality remains a formidable hurdle in applying AI models to clinical trials. A model can only be as good as the data it is trained with, and therefore it is absolutely imperative that any AI model is trustworthy, accurate, and truly representative of the task at hand to be effective and to circumvent bias—a paramount concern for major regulators. For instance, bias in training datasets (i.e., lack of demographic representation for certain subgroups) may lead to an AI model that makes recommendations that do not apply in certain populations or groups.
Having a large repository of historical trial data provides a comprehensive body of information demonstrating successes and failures across a variety of therapeutic areas, geographies and study size, as well as patient-level information.
Through relentless and responsible innovation, AI holds the transformative power to mitigate future risks in clinical trials and across the healthcare ecosystem, perpetually enhancing their efficiency while delivering a safer and superior patient experience.
