The contract research organization (CRO) industry was born in the 1980s, and by the 1990s these organizations became indispensable to conducting clinical research. Sponsors began to see that outsourcing work was not only time-efficient but incredibly cost-effective, and the need for globalization of clinical trials was growing increasingly important. However, despite the decades that have passed and the advancements in healthcare and technology we’ve accomplished, the journey of bringing a product to market has become slower and more expensive. This phenomenon is described as Eroom’s Law. 

In efforts to combat this long-standing problem in drug development, modern CROs are leveraging more innovative technologies such as AI to assist in all facets of clinical trial execution. Between protocol writing and trial design all the way through data analysis and delivery, the use of AI in clinical trials can reduce trial length by up to 30%.

Faster Protocol Writing and Improved Trial Design

With all publically available information on historical and ongoing clinical trials, AI and machine learning can be leveraged to analyze trial outcomes and supporting literature to determine optimal biomarkers, optimal dosing, target patient populations, and endpoints for evaluation.

At Lindus Health, we combine deep therapeutic area expertise and KOL input with our home-grown protocol drafting AI platform, SPROUT, that crawls clinicaltrials.gov (containing over 53,000 study protocols) to effectively draft protocols for specific disease indications in ICH M11 format. This is then combined with another one of our in-house AI models that optimizes study design using interpretable machine learning-based to predict early trial termination.. Additionally, for adaptive trial designs that allow for ongoing modifications, AI can perform interim data analysis quicker than traditional methods. As a result, protocol adjustments can be implemented quickly, eliminating the laborious and costly process of protocol amendments.

Overcoming Recruitment Barriers

It’s estimated that 85% of clinical trials do not complete on-time due to patient enrollment, ultimately leading to increased costs and reduced statistical power. Through sophisticated algorithms and natural language processing (NLP), AI can swiftly sift through vast amounts of data to identify ideal candidates for clinical trials. Electronic health records (EHRs) and patient registries are valuable sources for information on demographics, diagnoses, and medication histories. These tools can also be used to monitor social media platforms and online discussion groups based on publicly available demographic information as well as expressed health concerns.

For studies running paid social media recruitment campaigns, AI can provide insights on online behaviors for tailored recruitment strategies, including the placement, timing and content of advertisements that should be implemented across various platforms. Once individuals see these ads or are made aware of a clinical trial another way, they can even be directed to an AI-powered chatbot that can quickly assess responses to pre-screening questions to further determine eligibility.

Enhanced Patient Retention & Engagement

These chatbots also have the capacity to promote patient retention and engagement. In one study, researchers developed ChatDoctor, a tool developed using large language models (LLMs) that provides more accurate medical information than tools like ChatGPT. This ultimately redefines the way patients can receive medical guidance, significantly reducing the time between asking questions and receiving responses from busy healthcare professionals.

Furthermore, the use of AI-powered predictive analysis tools can also combat patient dropout. By evaluating patient data against study protocols, individuals at risk of non-compliance can be identified early on. Retention efforts can then be tailored to better support participants, whether that be providing more frequent check-ins or additional guidance and educational resources. 

Smarter Remote Monitoring & Data Management

In efforts to garner more patient-centric trial experiences, Sponsors will often use hybrid or decentralized trial designs, allowing participants to perform more or all study activities remotely. Studies assessing the efficacy of or capturing data through wearable devices can highly benefit remote monitoring. By assessing vitals and other health data in real-time, machine learning techniques can identify patterns for Sponsors to better understand effects of the investigational product and even detect signs of adverse events to potentially prevent safety issues more quickly than traditional monitoring methods.

With the use of AI it is also easier for Sponsors and CROs to clean data while studies are ongoing, greatly reducing database lock timelines. Manual review of the EDC can be incredibly time-consuming. AI solutions can issue queries and detect potential safety issues much faster. An example of one of these technologies is Meteor, an application that analyzes eCRFs to identify possible documentation errors as well as possible protocol deviations and adverse events.

Conclusion

While it’s going to take more than just AI to break the spell of Eroom’s Law on a large scale, these technologies are becoming increasingly important to consider for overcoming barriers in drug development. Integrating AI solutions into clinical research is not only beneficial for running on-time, on-budget studies, but it is vital to enhancing patient outcomes.

‍