Clinical Evaluation and Epidemiologic Application of an Adenoviral Species-Typing Model, based on Syndromic PCR Melt-Curve Data.

Abstract

Background: Human adenoviruses (HAdV) elicit diverse infections, most notably within the respiratory tract. While HAdV is a target in clinical-use PCR assays to diagnose respiratory infections, techniques are not widely available to determine individual HAdV species within clinical specimens. An initial model was previously developed to predict HAdV species from BIOFIRE® Respiratory Pathogen Panels (RPP), evaluated in silico and through contrived specimens. This model was based on melt-curve data of 5 individual amplification reactions underlying the adenoviral result.

Methods: In this study, the initial model is updated to better reflect prior knowledge of HAdV respiratory epidemiology and applied to a dataset of clinical HAdV-detected RPP samples independently subtyped via PCR. Revised model performance was further assessed through application to clinical proficiency testing events in the BIOFIRE Syndromic Trends database (Trends), a near-real-time network of clinical-use BIOFIRE testing results. The revised model was applied to >100 000 HAdV-detected results in Trends from the United States since 2019.

Results: Among the independently typed specimens, the revised model accuracy was 95.2% (180/189): 79/82 for HAdV-B, 97/102 for HAdV-C, and 4/5 for HAdV-E. In the Trends dataset, these analyses indicated dynamic epidemiology for HAdV species, including a shift of B-vs-C prevalence at the onset of the COVID-19 pandemic, which has more recently returned to prepandemic ratios, along with low-level prediction of species less typically associated with respiratory infection.

Conclusion: In silico modeling of melt-curve data from the BIOFIRE RPP can enhance HAdV species surveillance efforts and define viral epidemiology at local, regional, and national levels.