Leveraging paired serology to estimate the incidence of typhoidal Salmonella infection in the STRATAA study

Abstract

Serologic surveillance of at-risk populations can be used to directly estimate the incidence of typhoidal Salmonella infection across a variety of settings, including those without access to facility-based blood-culture surveillance. We collected paired blood samples approximately three months apart from an age-stratified random sample of healthy children and adults in Bangladesh, Malawi, and Nepal as part of the Strategic Typhoid Alliance Across Asia and Africa (STRATAA) study. We used a multiplex bead assay to measure the concentration of IgG antibodies against seven Salmonella typhi/paratyphi antigens (CdtB, FliC, HlyE, LPSO2, LPSO9, Vi, and YncE) in each sample and identified recently infected participants by fitting a regression mixture model to the change in IgG concentration between participants' samples. We estimated the seroincidence of infection in a Bayesian framework for each study site, age group, and antigen target. Finally, we compared the seroincidence estimates with crude and adjusted estimates of clinical incidence based on blood-culture surveillance. Seroincidence estimates were significantly higher than enteric fever incidence across all study sites, age groups, and antigen targets, even after adjusting for underreporting (median ratio: 24.2, interquartile range: 11.4-58.9). Seroincidence consistently peaked in the 0-4-year age group and declined moderately between children and adults (33% to 58% decline in HlyE seroincidence between the 5-9 and 30 + year old age groups), while enteric fever incidence peaked in older children and fell sharply in adults (71% to 95% decline in adjusted clinical incidence). Seroincidence estimates based on the FliC, YncE, and HlyE antigens individually had the strongest correlation with observed enteric fever incidence across age groups and study sites (r = 0.72, 0.69, and 0.63, respectively). These findings suggest that in endemic settings, both children and adults are frequently infected by typhoidal Salmonella serotypes, although only a fraction of these infections present as clinically identifiable enteric fever cases.

Fig 1. Classification of Anti-HlyE IgG Responses.

IgG measurements refer to standardized log-transformed fluorescence intensity (see methods). Panels in the top, middle, and bottom rows correspond to the Bangladesh, Malawi, and Nepal study sites, respectively. In each panel, each point represents the anti-HlyE IgG at baseline (x-axis) and the change in IgG from baseline to the follow-up visit (y-axis) of a single participant. Left: Most participants are clustered around the horizontal dashed line corresponding to no change in IgG between visits. Middle: Participants are colored by the posterior probability of having a large change (either an increase or decrease) in IgG between visits. This metric is derived from a two-cluster linear regression mixture model. Right: Participants who experienced a large rise in IgG between visits (posterior probability > 0.5) and had a higher IgG level at follow-up than a matched negative control were classified as infected during this period. All other participants were considered to be uninfected.

Fig 2. Seroincidence by Age, Antigen, and Study Site.

Each panel corresponds to the specific antigen target which was used to classify participants’ infection status when calculating seroincidence. Solid lines denote the median seroincidence (y-axis) in each age group (x-axis). Vertical lines represent the 95% credible intervals of the seroincidence estimates, and dashed lines indicate the adjusted incidence of enteric fever. Green, orange, and purple lines correspond to the Bangladesh, Malawi, and Nepal study sites, respectively.

Fig 3. HlyE Seroincidence and Enteric Fever Incidence by Age in Young Children.

Panels in the top and bottom rows display HlyE seroincidence and crude enteric fever incidence, respectively, at the Bangladesh (left), Malawi (middle), and Nepal (right) study sites. Age-specific incidence is shown for each of the first 5 years of life (black), and for 5-9 year-olds overall (red). Vertical lines represent 95% credible and confidence intervals for seroincidence and enteric fever incidence, respectively.

Fig 4. Relative Trend in HlyE Seroincidence and Enteric Fever Incidence by Age and Study Site.

Each column of corresponds to a different study site. Panels in the top row display the median seroincidence based on the HlyE antigen for each age group, while the panels in the bottom row display the adjusted incidence of enteric fever cases for each age group. Seroincidence is highest in the 0-4-year age group and declines relatively gradually with age, while the incidence of enteric fever peaks in older children before rapidly declining to a low level in adults. Vertical lines represent the 95% credible intervals of seroincidence estimates.