Overview

With the continuous advancements of AI/ML technologies, the number of ways that AI/ML can be applied keeps growing. Many industries are already adopting AI/ML tools in their key processes and the pharma industry is no exception.

AI/ML helps clinical execution, from study design’s start to submission’s end, spanning the entire value chain for efficiency. This article will highlight some of the AI/ML Use cases in clinical trials with the largest potential.

Site Selection

Many companies often choose study sites based on personal connections and quick approvals. However, that might not always lead to the best location for the study. Manually searching through platforms like ClinicalTrials.gov or EudraCT can be slow and inefficient, yielding limited results.

Relying on outdated historical data for site selection may not align with the specific needs of the current study. Opting for unsuitable sites can adversely impact patient enrollment and cohort diversity, influencing patient outcomes and study endpoints.

Machine learning models assess vast datasets, including site history, performance, patient demographics, infrastructure, and competing trials, enhancing the evaluation process. The model then generates a ranked list of candidates that the study can select from, providing a more informed approach to selecting sites. This not only increases the number of site matches but also ensures quality matches that live up to the performance criteria. Data-driven site selection minimizes bias, promoting objective decisions and diverse patient enrollment, fostering more effective and impartial selection processes.

Transformation & Mapping to CDISC Standards

Clinical trials generate a huge amount of data that needs to be organized into standard formats. Data transformation and mapping are vital for integrating and analyzing the data while meeting regulatory standards. However, doing this manually takes a lot of time and skill, and there is a risk of human errors.

AI/ML models provide an automated solution, mapping raw data to standard formats like SDTM, CDISC ODM, or FHIR. These models identify and categorize data points, ensuring they fit appropriately in the defined SDTM domains. By constantly learning from new data, AI can adapt to different trial designs and data structures. As a result, AI keeps the transformation process accurate as trials evolve.

Automating the data transformation and mapping ensures faster access to data for programmers and other stakeholders. While some human intervention will still be necessary, AI tools can significantly minimize the labor required in data transformation and mapping. It can minimize human errors and data inconsistencies, thereby improving the data quality and reliability.

Patient Retention and Medication Adherence

Consistent patient participation and retention is important for the validity and reliability of clinical trial results. Patient drop-out can lead to increased trial duration and costs, as researchers might need to recruit additional participants to maintain the study’s power. Identifying patients at risk of dropping out is crucial for trial efficiency, safety, and continuity, ensuring participant well-being.

One of the reasons for patients dropping out can be non-adherence to the treatment regimen. Patients who don’t take medication as instructed may be at increased risk of experiencing adverse events and therefore dropping out. Non-adherence can introduce variability into trial outcomes, making it difficult to determine the actual effects of an intervention.

Utilizing historical data from past clinical trials and current participant data, AI/ML models can forecast which patients have a higher likelihood of drop-out and medication non-adherence. Such predictions can be based on various factors, from the complexity of the regimen, patient demographics, and disease severity, to logistical challenges like distance from the trial site.

Based on data collected from wearable devices, the researchers can also be notified in case the model identifies any early signs of disengagement or potential drop-out indicators. Some apps can even use facial recognition to monitor the patient’s facial expressions during medication ingestion to detect any potential non-adherence or discomfort.

With better predictive insights, researchers can pay special attention to patients who may need additional support or interventions to keep them engaged and compliant in the trial. Ultimately, this can help improve patient adherence and patient retention in clinical trials.

Conclusion

In addition to the three use cases mentioned, AI/ML tools can help optimize many other processes in clinical execution. For pharma organizations at the beginning of their AI/ML journey, it is recommended to select a few high-impact and low-risk pilot use cases to test the potential and build a technical foundation.

To ensure a successful AI/ML implementation, take important preparatory steps such as assessing resource needs, checking data availability, and understanding regulations.