View All Abstracts for Session 8
36 Statistical Methods for Evaluating ELISPOT Assays in HIV Vaccine Trials
S. Self*, M. Hudgens, Y. Chiu, and A. J. Rossini
Fred Hutchinson Cancer Res. Ctr., Seattle, WA, USA
Background: The ELISPOT assay will likely be the primary means of assessing cellular responses to vaccine candidates in large-scale HIV vaccine efficacy trials for the purpose of identifying correlates of vaccine-induced protection. These responses will be measured both for bulk PBMC and for subsets of T cells. The identification of an objective method to distinguish positive and negative responses that also controls the rate of false positives is critical.
Methods: We use a simple Poisson model for the numbers of spot-forming units per well and propose criteria for a positive assay based on a likelihood ratio statistic. A correction for testing multiple peptide pools per assay is made to control the false positive rate at the level of study subject. The methods are used to compare different assay formats including number of peptide pools, number of wells per pool, and number of negative control wells per plate. The methods are also used to compare 2 different strategies for evaluating CD8-specific responses. Computer simulations and real data examples are used to evaluate properties of the proposed methods.
Results: The proposed method is shown to maintain nominal false positive rates in both the bulk PBMC and the CD8-specific assays. The currently used ad hoc method is shown to be conservative in some instances and liberal in others. The method also suggests that a reasonable balance between assay sensitivity and operational economies is obtained with a format involving duplicate wells for 10 to 25 peptide pools and 6 negative control wells per plate. The subtraction strategy for assessing CD8-specific responses (bulk PBMC result minus CD8-depleted PBMC result) is shown to be 2 to 3 times less sensitive as the direct strategy (direct assay on CD4-depleted PBMC).
Conclusions: Formal statistical criteria provide an objective method for distinguishing positive responses in the ELISPOT assay that appropriately control the rate of false positives. Such methods are also useful in the larger context of assay development.
Contact Author about this Abstract
