Informatics driven real world analysis of SARS-CoV-2 serologic response and in vitro diagnostic accuracy
CERSI Collaborators: Thomas Durant, MD (Yale), David Ferguson (MT [ASCP]), William (Garrett) Jenkinson, PhD (formerly of Mayo Clinic), Benjamin Pollock, PhD, MSPH (Mayo Clinic), Joseph Ross, MD, MHS (Yale), Wade Schulz, MD, PhD (Yale) (PI), Elitza Theel, PhD (Mayo Clinic), Frederick Warner, PhD (Yale), Patrick Young, PhD (Yale)
FDA Collaborators: Sara Brenner, MD (CDRH), Aloka Chakravarty, PhD (Formerly of OC/ODT/ODAR), Mary Jung, PhD, (Formerly of OC/ODT/ODAR) Tamar Lasky, PhD (Formerly of OC/ODT/ODAR), Jacqueline Puigbo, PhD, MS, (Formerly of OC/ODT/ODAR), Veronica Sansing-Foster, PhD, (Formerly of OC/ODT/ODAR), Mary Ritchey, Jill Marion
Project Start Date: October 19, 2020
Project End Date: April 18, 2022
Regulatory Science Challenge
The SARS-CoV-2 pandemic has rapidly and dramatically changed healthcare and daily life. Many questions about the diagnostics for SARS-CoV-2 and the immune response to infection remain unknown, making it difficult to develop definitive plans for a safe reopening for healthcare systems that are resuming routine services and, further, the phased reopening of society. The rapid implementation of serologic assays that have been approved through Emergency Use Authorization (EUA) pathways has further complicated the assessment of SARS-CoV-2 prevalence, as largely unknown accuracy and likely variation between tests make it difficult to interpret population-level results. Correlation of diagnostic test results with clinical disease is critical to guide our interpretation of serologic assays (antibody tests) within the clinical laboratory.
Project Description and Goals
This project will focus on developing resources for studies that are used to assess the performance of artificial intelligence software tools. This includes designing hardware that can be controlled virtually through the web, software, and statistical methods that can be used in all environments, especially low resource. This study will aim to determine the validity of artificial intelligence software tools.
The goal of this project was to gain a better understanding of the performance characteristics of commercial serology assays. Specifically, Yale-Mayo CERSI implemented a real-time, digital approach to identify patients with prior SARS-CoV-2 infection based on a diagnosis and/or test result and assess the performance of various clinical tests. As several tests have been developed that assess different antibodies, with significant uncertainty about the overall accuracy of and correlation between these tests, the data gained from this project has the potential to provide valuable information from a real-world data set to better understand serologic results. This will address several key areas of impact, including the advancement of regulatory science and inform regulatory decision-making for a critical public health need.
Goal 1: Implement real-time categorization to identify specimens for validation of clinical serologic testing.
Goal 2: Characterize the antibody response to SARS-CoV-2 infection and the accuracy of clinical assays over time.
Research Outcomes/Results
Results showed that SARS-CoV-2 molecular and antigen-based testing were identified consistently using our data collection method, with variation corresponding to known waves of the COVID-19 pandemic, suggesting that real-time categorization was successful.
Due to testing quality and availability, multiple different serologic tests were used throughout this study, potentially impacting the likelihood that a given test indicated patients have antibodies. In addition, several patient characteristics varied among the two sites and between those who tested positive by serologic test.
In this multi-site study of SARS-CoV-2 serologic test performance, the researchers identified several potential associations with positive serologic test results, including higher rates among subjects with symptoms at presentation, suggesting that more severe disease may lead to a stronger humoral immune response. The collaborative project demonstrates that while real-world data can provide novel and detailed insights into in vitro diagnostic test results, care is required when interpreting these data given the number of potential biases, and a hybrid approach that leverages real-world and protocol-driven data are likely to give the most robust picture of in vitro diagnostic performance.