Statistics, Department of

 

First Advisor

Dr. Jennifer Clarke

Date of this Version

Summer 7-28-2020

Citation

Dutta, E., Statistical Methodology to Establish a Benchmark for Evaluating Antimicrobial Resistance Genes through Real Time PCR assay, University of Nebraska, Lincoln, 2020

Comments

A THESIS Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfillment of Requirements For the Degree of Master of Science, Major: Statistics, Under the Supervision of Professor Jennifer Clarke. Lincoln, Nebraska: July 2020

Copyright 2020 Enakshy Dutta

Abstract

Novel diagnostic tests are usually compared with gold standard tests for evaluating diagnostic accuracy. For assessing antimicrobial resistance (AMR) to bovine respiratory disease (BRD) pathogens, phenotypic broth microdilution method is used as gold standard (GS). The objective of the thesis is to evaluate the optimal cycle threshold (Ct) generated by real-time polymerase chain reaction (rtPCR) to genes that confer resistance that will translate to the phenotypic classification of AMR. Data from two different methodologies are assessed to identify Ct that will discriminate between resistance (R) and susceptibility (S). First, the receiver operating characteristic (ROC) curve was used to determine the optimal Ct by optimizing the area under the curve (AUC), which was further validated by assessing the sufficiency of sample sizes involved in this study and by 5-fold cross-validation. AUC is a straightforward method, using a default probability threshold (Pt) of 0.5 and independent of misclassification cost to discriminate between the classes. An alternative methodology - H measure - is proposed, which selects the Pt based on minimum error rate. The H measure is quite flexible, and the threshold can be selected according to researchers’ interest by minimizing the false positive or negative rate.

A total of 297 lung and 111 nasal swabs from bovine were tested for AMR using the gold standard and rtPCR for three specific drugs. The level of agreement between the two tests were measured using Cohen’s Kappa (. Using the first approach, the optimal Ct for lung tissue samples was between 32.6 and 35.7, with a good level of agreement between the two tests. For the nasal tissue, the rtPCR results were only validated for one drug with a Ct of 33.3 with a moderate level of agreement. For the second approach, the lungs and nasal tissues are combined, and the optimal Ct is evaluated by taking the average from AUC and H measure and lies between 32.0 and 32.9 with a moderate level of agreement.

Adviser: Jennifer Clarke

Share

COinS