A Linear Model-Based Approach for Inferences About the Area Under the Curve with Repeated Measures and Values Below the Limit of Detection

• Activity Number: 511
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Rameela Chandrasekhar*+ and Alan D Hutson and Gregory E Wilding
• Companies: Vanderbilt University School of Medicine and SUNY at Buffalo and SUNY at Buffalo
• Keywords: Limit of detection ; censoring ; Area under the curve ; trapezoidal rule ; longitudinal data

Abstract

Outcome versus time data is commonly encountered in biomedical and clinical research. A common strategy adopted in analyzing such longitudinal data is to condense the repeated measurements on each individual into a parameter of interest such as the area under the response versus time curve (AUC). Standard parametric or nonparametric methods are then applied to perform inferences on the conditional AUC distribution. However, outcomes may be subject to censoring due to the limit of detection and specific methods that take this into account need to be applied. Furthermore, condensing repeated measures with a single metric leads to loss of information due to the fact that within-subject variability fails to be captured. These factors motivate us to propose a general linear model-based approach for estimation and hypothesis tests about the mean areas that encapsulates the within-subject correlation and also accounts for censored data. Though the actual estimate of the AUC has the same interpretation as before, we obtain a variance estimate using all the available information. Thus yielding more efficient inference procedures based on the same metric. Inferential properties of our method are investigated using Monte Carlo simulation studies.