Comparing and combining data from immune assays based on left-censored multivariate normal model assuming common...

Area of Science:

• Biostatistics
• Immunology
• Vaccinology

Background:

• Immune response biomarkers are crucial for vaccine development, aiding in the comparison of vaccine candidates.
• Variability across different assays and laboratories complicates the analysis of immune response data.
• Standardization is needed to accurately compare immunogenicity and protective effects of vaccines.

Purpose of the Study:

• To develop statistical methods for adjusting assay differences in vaccine research.
• To enable accurate comparison of vaccine immunogenicity and evaluation of correlates of risk using data from multiple assays.
• To address challenges posed by measurement error and lower limits of detection in neutralization assays.

Main Methods:

• Proposed methods based on a left-censored multivariate normal model.
• Integration of external paired-sample data with bridging assumptions.
• Adjustment for common assay differences, measurement error, and lower limit of detection.

Main Results:

• The proposed methods provide unbiased calibrated assay means and valid tests for immunogenicity comparison.
• Alternative methods ignoring assay differences lead to biased estimates and inflated type-I errors.
• The new methods significantly outperform existing approaches in reducing bias and improving precision when handling left-censored data.

Conclusions:

• The developed statistical methods effectively adjust for assay variability in vaccine research.
• These methods enhance the reliability of comparing vaccine immunogenicity and evaluating correlates of risk.
• The approach is applicable to real-world data, such as SARS-CoV-2 neutralization assay results from multiple laboratories.