Online Program Home
My Program

Abstract Details

Activity Number: 359 - Contributed Poster Presentations: Biopharmacutical Section
Type: Contributed
Date/Time: Tuesday, July 31, 2018 : 10:30 AM to 12:20 PM
Sponsor: Biopharmaceutical Section
Abstract #330566
Title: Statistical Methods to Estimate Diagnostic Testing Accuracy of ULTE4 to Determine Aspirin Intolerance in Asthma Using Meta-Analysis Data
Author(s): Nan Zhang* and John Hagan and Matthew Rank and Yu-Hui Chang and Rohit Divekar and Erin O'Brien and Gerald Volcheck and Devyani Lal and Patricia Erwin and Harry Teaford and Hirohito Kita and Tanya Laidlaw and Christina Hagan
Companies: Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic and Mayo Clinic and Harvard Medical school and Baylor College of Medicine
Keywords: ROC analysis; cluster data; meta-analysis
Abstract:

Receiver operating characteristics (ROC) analysis is commonly used in diagnostic medicine to evaluate the accuracy of a continuous diagnostic test result versus a gold standard and further determine the optimal cut-off value of sensitivity and specificity for the test. However, when cluster data is present, the possible correlation within cluster needs to be taken into account when doing ROC analysis. The purpose of this study is to compare three statistical methods to do ROC analysis with cluster data. The three methods are: (1) the original method without the within cluster correlation; (2) logistic regression with random effects across clusters; (3) logistic regression with sandwich estimator (GEE method). The proposed methods will be applied to a systematic review and meta-analysis data that defines the diagnostic testing accuracy of ULTE4 in determining aspirin intolerance in asthma patients and the comparison results will be presented. Ten previous published studies that analyzed ULTE4 in asthma patients characterized as having or not having aspirin intolerance were included and individual-level data points from the 10 studies were extracted.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program