Understanding and predicting COVID-19 clinical trial completion vs. cessation

Area of Science:

• Clinical trial analysis
• Machine learning in healthcare
• Pharmacovigilance

Background:

• Over 5,193 COVID-19 clinical trials were registered by March 2021.
• A significant number of these trials (191) were terminated, suspended, or withdrawn, while 909 were completed.

Purpose of the Study:

• To investigate underlying factors differentiating completed from ceased COVID-19 clinical trials.
• To develop predictive models for accurately forecasting trial completion or cessation.

Main Methods:

• Utilized a dataset of 4,441 COVID-19 trials from ClinicalTrials.gov.
• Engineered four feature types: administration, eligibility, study information, criteria, drug types, study keywords, and machine learning embeddings.
• Employed predictive modeling to analyze trial success factors.

Main Results:

• Drug features and study keywords were identified as highly informative.
• All four feature types collectively proved essential for accurate prediction.
• Achieved an Area Under the Curve (AUC) score exceeding 0.87 and a balanced accuracy of 0.81 in predicting trial outcomes.

Conclusions:

• Computational methods effectively identify differences between completed and ceased COVID-19 trials.
• Predictive models demonstrate satisfactory accuracy in forecasting trial status.
• Findings can aid stakeholders in optimizing trial planning and reducing costs.