H3N2 influenza hemagglutination inhibition method qualification with data driven statistical methods for human clinical trials

Abstract

Introduction: Hemagglutination inhibition (HAI) antibody titers to seasonal influenza strains are important surrogates for vaccine-elicited protection. However, HAI assays can be variable across labs, with low sensitivity across diverse viruses due to lack of standardization. Performing qualification of these assays on a strain specific level enables the precise and accurate quantification of HAI titers. Influenza A (H3N2) continues to be a predominant circulating subtype in most countries in Europe and North America since 1968 and is thus a focus of influenza vaccine research.

Methods: As a part of the National Institutes of Health (NIH)-funded Collaborative Influenza Vaccine Innovation Centers (CIVICs) program, we report on the identification of a robust assay design, rigorous statistical analysis, and complete qualification of an HAI assay using A/Texas/71/2017 as a representative H3N2 strain and guinea pig red blood cells and neuraminidase (NA) inhibitor oseltamivir to prevent NA-mediated agglutination.

Results: This qualified HAI assay is precise (calculated by the geometric coefficient of variation (GCV)) for intermediate precision and intra-operator variability, accurate calculated by relative error, perfectly linear (slope of -1, R-Square 1), robust (<25% GCV) and depicts high specificity and sensitivity. This HAI method was successfully qualified for another H3N2 influenza strain A/Singapore/INFIMH-16-0019/2016, meeting all pre-specified acceptance criteria.

Discussion: These results demonstrate that HAI qualification and data generation for new influenza strains can be achieved efficiently with minimal extra testing and development. We report on a qualified and adaptable influenza serology method and analysis strategy to measure quantifiable HAI titers to define correlates of vaccine mediated protection in human clinical trials.

Keywords: antibody; clinical trials; data pipeline; hemagglutination inhibition (HAI); human; influenza; qualification; statistical analysis.

Figure 1

HAI assay design data pipeline. This figure shows the data flow from raw data generated by CypherOne Software to processed data used for analysis and data sharing. Light grey arrows indicate decision making steps: generated reports and visualization assists with determining the quality of the data and analysis readiness. Dark grey arrow indicates main processing steps performed during data upload. Black arrows indicate data pipeline steps which are scripted and secured.

Figure 2

HAI assay qualification for H3N2 influenza strain A/Texas/71/2017, overview of qualification plan and parameters tested during qualification, recommended acceptance criteria, and results obtained.

Figure 3

HAI assay qualification on H3N2 influenza strain A/Texas/71/2017, system suitability criteria results. Geometric mean titers [GMTs] for positive control Ab2210 IgG1 and negative control 7B2 IgG1 from all experiments. Nineteen data points plotted for both controls. Titers of<20 or<10 are converted to half of the lowest dilution tested in assay, in this case 10 or 5, respectively, to enable statistical analysis. The Ab2210 IgG1 titers are illustrated with red dots, the 7B2 titers are illustrated with orange dots. The dotted line indicates a GMT of 20.

Figure 4

HAI assay qualification on H3N2 influenza strain A/Texas/71/2017, matrix effect testing, linearity analysis results. Scatter plot showing correlation analysis output from SAS CORR procedure, between log 10 transformed GMTs for Ab2210 IgG1 diluted in plasma [plotted on y-axis] and Ab2210 IgG1 diluted in serum [plotted on x-axis]. The 70% and 80% prediction ellipses are plotted in red and blue colors, respectively. Sixteen data points are plotted as circles, some circles overlap due to having same values for both x and y axis. The Pearson correlation coefficient between serum and plasma log10 GMTs was calculated and determined to be 0.98, with a p-value<0.0001.

Figure 5

(A) HAI assay qualification on H3N2 influenza strain A/Texas/71/2017, linearity testing plate visual. A CypherOne instrument graphic demonstrating assay linearity, from a representative assay with the in-house reference standard (IHRS). The HAI titer decreased two-fold for each two-fold decrease in analyte concentration. (B) HAI Assay Qualification on H3N2 influenza strain A/Texas/71/2017, linearity testing results. The plot shows SAS output from REG procedure, showing the linear regression analysis between log10 GMT [plotted on y-axis] and log10 dilution [plotted on x-axis]. The resulting R² and R was 1.

Figure 6

HAI assay qualification on H3N2 influenza strain A/Texas/71/2017, range testing results. The plot shows SAS output from REG procedure, showing the linear regression analysis between the expected [plotted on x-axis] and observed log10 GMTs [plotted on y-axis]. The resulting R² and R was 1.