Censoring Point in Logistic ELISA Standard Curves

• Activity Number: 250
• Type: Contributed
• Date/Time: Tuesday, August 8, 2006 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ryan E. Wiegand*+ and Elizabeth H. Slate and Elizabeth G. Hill and Jyotika K. Fernandes
• Companies: Medical University of South Carolina and Medical University of South Carolina and Medical University of South Carolina and Medical University of South Carolina
• Keywords: limit of detection ; censoring ; bias ; clustering

Abstract

We address the analysis of concentrations of various cytokines in gingival crevicular fluid (GCF) arising from a multiplex ELISA study where GCF was collected at multiple sites from each patient. Since a sample's cytokine concentration is determined by back-solving from an optical density via the standard curve, any sample with an optical density within the curve's asymptotes will map to a reliable concentration. Values close to the low asymptote, however, are unreliable because they may be indistinguishable from zero and, similarly, values at the high asymptote represent saturated signals. Hence, these samples must be censored. Using an analysis method that simultaneously accounts for left- and right- censoring and clustering, we explore various methods of determining the censoring points and evaluate the benefits of each method. This work was funded by NIH grant P20 RR-017696.