Using an Artificial Intelligence Platform to Accurately Predict Pharmacokinetic Sampling Outcomes in Clinical Trials
In this poster you will learn:
- Remote patient monitoring is critical in ensuring optimal drug exposure and accurately measuring drug intake
- It is estimated that 30-50% of CNS studies fail because subjects are not taking the study drug
- High rates of non-adherence in clinical research and the lack of a continuous and accurate measure of adherence interfere with hypothesis testing and the interpretation of trial results
- Current measures of adherence such as pill counts are inaccurate and underestimate non-adherence
- PK (pharmacokinetic) sampling, while accurate, is used sparsely in ambulatory settings and does not capture adherence in the placebo group
- An artificial intelligence (AI) platform that uses computer vision and neural networks to visually confirm ingestion and capture ingestion-related behaviors was evaluated for accuracy in predicting PK outcomes
- The AI platform is the only known modality to be able to distinguish between non-adherence and fraud
- A preliminary analysis using 6 CNS studies and 278 subjects with available PK data was conducted to assess the ability of adherence to predict PK levels
- A subsequent analysis on 256 of those subjects with available PK data was conducted to assess the ability of adherence and fraud to predict PK levels