Mixture modelling of Bordetella pertussis serology samples to evaluate anti-pertussis toxin immunoglobulin G titre thresholds for positivity: England 2008–2022

Abstract

Introduction. Antibody testing for evidence of a recent Bordetella pertussis infection by estimating anti-pertussis toxin immunoglobulin G (anti-PT-IgG) titres by enzyme-linked immunosorbent assays is often recommended for those with a cough lasting more than 14 days. Interpreting results varies, with studies recommending different anti-PT-IgG titre thresholds for assigning positivity. In England, early work looking at antibody titre distributions for samples submitted from April 2010 to July 2012 found an optimal threshold of greater than 70 IU ml−1 for good sensitivity, specificity and positive predictive value. Aim. The aim of this study is to use the same mixture modelling technique to determine if the 70 IU ml−1 threshold remains appropriate when assessing data before, during and after the outbreak of pertussis in 2011–2012. Methods. We reviewed titres for all serology-tested samples in England between 1 July 2008 to 30 June 2022. IgG titres were used to calculate the positivity based on the current threshold of 70 IU ml−1, the median duration of cough for individuals who tested positive and, through mixture modelling, the sensitivity, specificity, positive and negative predictive values (PPV and NPV) of assay thresholds. Results. Positivity rates increased from 21.7 % prior to the outbreak to 30.3 % during the outbreak and dropped to 25.1 % post-outbreak; similar to estimates from the mixture model of 20.5, 33.3 and 28.7 %, respectively. Although the estimated sensitivity dropped during and after the outbreak when applying the 70 IU ml−1 threshold, the PPV remained high and therefore no change to this threshold is warranted. Conclusion. Mixture modelling is a useful tool to establish thresholds, but reassessment should also be done when there have been changes to prevalence and/or testing regimes to determine whether there have been any changes in sensitivity, specificity, PPV, and NPV and whether the threshold should be revised.