ISCTM 16th Annual Scientific Meeting
Washington DC
February 19-21, 2020
1) Digital Biomarkers
AiCure was honored to participate and present at this year’s ISCTM 16th Annual Scientific Meeting. Isaac Galatzer-Levy, Ph.D., VP of Clinical and Computational Neuroscience, a global expert in PNS/CNS & Digital Biomarker development as well as physiological and behavioral clinical phenotype development, is leading AiCure’s research efforts.
Isaac Galatzer-Levy, co-presented a novel study design using AiCure’s platform to capture and calculate visual and auditory biomarkers of Parkinson’s disease (PD) to enhance understanding of symptom severity in response to treatment. The natural fluctuating course of PD and the lack of scalability in traditional clinical measures of PD, present significant obstacles in the development of new therapies.
2) Predictive Analytics
The use of AiCure’s platform to virtually capture dose adherence data is helping sponsors better answer the key clinical and regulatory questions asked during the drug development process. As partner in the process, AiCure is leveraging rapid advances in ML/AI that make it possible to collect large datasets previously unattainable within the confines of standard clinical practice. These new datasets and their operational implementation are enabling sponsors to turn vast amounts of unstructured data into actionable insights, allowing for a full 360 view of trial performance. Sponsors are using AiCure’s unique content to plan, conduct, and analyze trials from the portfolio level down to the individual dose.
Adherence Prediction Model
Isaac Galatzer-Levy, Ph.D. and Anzar Abbas, Ph.D. presented the development of an Adherence Prediction Model. Based on current measure of dose adherence, they found that future medication adherence could be predicted with high accuracy, allowing for proactive clinician interventions. This injects considerable risk into the trial, both statistical and operational. The ability to identify participants at risk of partial or full non-adherence on an ongoing basis allows clinical site staff to engage in focused proactive interventions, preventing dropout before it is too late. Different predictive models were trained and tested across multiple indications.
Download: “Prediction of medication adherence in clinical trials using machine learning”
Intentional Non-Adherence Scope and Prediction: Collaboration with Tufts’ CSDD
In collaboration with Ken Getz from the Tufts Center for the Study of Drug Development (CSDD), Laura Shafner, Co-Founder and VP of Value, presented landmark data on the use of AI/ML and human raters to account for and manage investigational dose nonadherence in clinical trials. AiCure is unique in its ability to distinguish ‘vanilla’ non-adherence from intentional non-adherence. Based on 257,672 dosing administrations across 23 trials, the current study characterizes the scope of the problem as well as factors associated with, and predictive of, intentional non-adherence. More precise estimates of dose adherence are being used by sponsors to enhance study designs and reduce operational risk.
Composite Site Performance Score – optimizing site selection and reducing operational risk
Site performance traditionally has focused on operational metrics, including speed and volume of enrollment. The emphasis on enrollment, has, in part, derived from the need to increase patient enrollment in order to account for a possible reduction in treatment effect due to non-adherence. Laura Shafner presented a Composite Site Performance Score prototype, developed to help sponsors quantify and normalize site performance. The model used historical data from 701 sites, across 52 trials. Two factors were identified that accounted for 75% of the total variance measured – study completion rate and intentional non-adherence. Ranking, prioritization, or exclusion of sites based on pre-defined percentile-based scores may be used by sponsors during trial planning for site selection and/or during the conduct of the study to optimize enrollment targets.