Over the past decade, clinical trials have undergone significant changes, including new trial formats, endpoints and regulatory requirements. Coupled with external factors such as competitive pressures, it has led many to question whether clinical trials are becoming increasingly complex. A concern growing especially in the light of COVID-19.
In a scientific study titled Clinical trials are becoming more complex: a machine learning analysis of data from over 16,000 trials, Markey et al presented a prime insight into this complex dynamic. Their research, fuelled by machine learning algorithms, indicated that clinical trials are indeed growing in complexity, with oncology emerging as one of the most complex therapeutic areas. However, the average complexity of trials across all areas has seen a relaxation since the start of the COVID-19 pandemic.
Against this backdrop, we’ll discuss the role of machine learning in tracing the complexity of these trials. A tool developed by Fast Data Science is leading the charge by effectively predicting the complexity of a trial from the protocol text.
COVID-19 brought with it unprecedented challenges, ranging from faster turnaround times to managing clinical trials under social distancing norms. The intensity of the situation coupled with the urgency for solutions, the complexity of COVID trials naturally escalated.
Fast Data Science’s Clinical Trial Risk Tool uses machine learning as a critical solution to manage this increasing complexity. The tool uses a machine learning algorithm to automatically assess key features of these trials, similar to the Markey et al study. These include the number of endpoints, inclusion–exclusion criteria and more, combining these aspects into a ‘Trial Complexity Score’. This score correlates with the overall clinical trial duration, thereby indicating the level of complexity.
The power of Fast Data Science’s Clinical Trial Risk Tool lies in its ability to predict this Trial Complexity Score purely from the protocol text. In essence, it could be a powerful predictive tool for sponsors, CROs, and regulatory bodies to assess the expected complexity of a COVID-19 or other clinical trial.
This ability not only allows stakeholders to anticipate potential challenges but also refine the design of these trials to reduce complexity. By learning from the complexity scores of past studies, predictive models can be further refined to optimise future trials.
In conclusion, while the evolving necessities of medical research and the advent of COVID-19 have undeniably rendered clinical trials more complex, machine learning may hold promising solutions. By predicting trial complexity, we can better equip ourselves to navigate the trials of tomorrow.
