A semiparametric method for comparing the discriminatory ability of biomarkers subject to limit of detection

Receiver operating characteristic curves and the area under the curves (AUC) are often used to compare the discriminatory ability of potentially correlated biomarkers. Many biomarkers are subject to limit of detection due to the instrumental limitation in measurements and may not be normally distrib...

Full description

Saved in:
Bibliographic Details
Published inStatistics in medicine Vol. 36; no. 26; pp. 4141 - 4152
Main Authors Yin, Lixuan, Diao, Guoqing, Liu, Aiyi
Format Journal Article
LanguageEnglish
Published England Wiley Subscription Services, Inc 20.11.2017
Subjects
Area Under Curve
AUC
Autism
Autistic Disorder
Biomarkers
Biomarkers - analysis
Computer Simulation
Humans
Likelihood Functions
Limit of Detection
Medical statistics
Models, Statistical
nonparametric maximum likelihood estimator
ROC Curve
semiparametric transformation model
Statistics, Nonparametric
semiparametric transformation model
ROC curve
limit of detection
nonparametric maximum likelihood estimator
AUC
Online AccessGet full text

Cover

More Information
Summary:Receiver operating characteristic curves and the area under the curves (AUC) are often used to compare the discriminatory ability of potentially correlated biomarkers. Many biomarkers are subject to limit of detection due to the instrumental limitation in measurements and may not be normally distributed. Standard parametric methods assuming normality can lead to biased results when the normality assumption is violated. We propose new estimation and inference procedures for the AUCs of biomarkers subject to limit of detection by using the semiparametric transformation model allowing for heteroscedasticity. We obtain the nonparametric maximum likelihood estimators by maximizing the likelihood for the observed data with limit of detection. The proposed estimators are shown to be consistent, asymptotically normal, and asymptotically efficient. Additionally, we propose a Wald type test statistic to compare the AUCs of 2 potentially correlated biomarkers with limit of detection. Extensive simulation studies demonstrate that the proposed method is robust to nonnormality while performing as well as its parametric counterpart when the normality assumption is true. An application to an autism study is provided.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:0277-6715
1097-0258
1097-0258
DOI:10.1002/sim.7415