Multiplex bead assays enable integrated serological surveillance and reveal cross-pathogen vulnerabilities in Zambezia Province, Mozambique

Abstract

Multiplex serological assays simultaneously measure antibodies to multiple antigens, furnishing insights into exposure and susceptibility to several pathogens and cross-pathogen vulnerabilities. Our serosurvey tests dried blood spots from 1292 individuals for IgG antibodies to 35 antigens from 18 pathogens using a multiplex bead assay for vaccine preventable diseases, malaria, SARS-CoV-2, neglected tropical diseases, and enteric pathogens in Mozambique. We produce pathogen-specific seroprevalence estimates and age-seroprevalence curves and identify spatial differences in seroprevalence. Rural clusters have higher odds of seropositivity to most NTDs neglected tropical diseases, Plasmodium falciparum malaria, and enteric pathogens, but lower odds of seropositivity to SARS-CoV-2 and vaccine preventable diseases compared to urban clusters. This co-occurrence identifies clusters with high vulnerability to multiple pathogens. We identify a candidate group of antigens that are correlated with high overall vulnerability. Our results demonstrate a role for multiplex serology in integrated disease surveillance to guide control strategies for individual and co-endemic pathogens.

Fig. 1. Participant enrollment cascade.

Participants were selected at the individual level irrespective of household. However, during enrollment, some heads of households refused to have anyone in their household participate. Therefore, 22% of participants were lost due to household level refusals. The rest were due to individual level reasons. Quality control issues with the specimens included data collection and laboratory issues such as contamination or low bead counts. Percentages are calculated based on the previous step in the participant cascade.

Fig. 2. Weighted provincial seroprevalence.

Pathogen with antigen name in parentheses are listed on the y-axis; P. is short for plasmodium. Weighted provincial seroprevalence is on the x-axis. The dot represents mean seroprevalence values and 95% confidence intervals are represented by the lines for each antigen. A Black dot represents the provincial estimate (n = 1292). B Green represents 6month to 4-year-olds (n = 432), orange represents 5- to 17-year-olds (439), and purple represents 18- to 49-year-olds (n = 421). C Red represents seroprevalence for females (n = 680), and blue represents males (n = 612). D Red represents seroprevalence among participants who live in rural clusters (n = 1115), and blue represents those in urban clusters (n = 177). Source data are in Supplemental Table 4.

Fig. 3. Age seroprevalence curves.

Seroprevalence curves by age for each antigen are represented here. X-axis is age in years, and y-axis is seroprevalence estimate. Black dots represent observed mean seroprevalence estimates by age. Red lines represent seroprevalence curves by age using semi-parametric cubic regression splines in a generalized additive model (GAM). Red shading represent the 95% confidence intervals for the GAM using bootstrapping. P. vivax (pvrbp2b) failed to converge in calculating confidence intervals for the GAM because there was only 1 seropositive in the dataset. This is based off the seroprevalence responses for all 1292 participants. P is short for plasmodium, NTD is short for neglected tropical disease, and VPD is vaccine-preventable disease.

Fig. 4. Odds of seropositivity to one antigen compared to another.

Adjusted log odds ratios of seropositivity to outcome antigens on y-axis as compared to regressor antigens on x-axis among 1292 participants. All odds ratios were calculated using logistic regression and are adjusted for seropositivity to all other antigens, age in years, GST reactivity, and uninfected Vero cell lysate reactivity and urban or rural environment. P. is short for plasmodium. Statistically significant (p < 0.05 for 2-sided hypothesis test, values adjusted for multiple hypothesis testing) adjusted odds ratios are highlighted in black. Source data are in Supplemental Table 6.

Fig. 5. Overall vulnerability score by cluster.

A Vulnerability scores for each cluster (n = 30), urban clusters in orange and rural clusters in green. B Vulnerability scores for each dried blood spot (DBS) cluster (n = 30) ordered from highest (value of 30) to lowest (value of 1) plotted on map of Zambezia province, urban clusters in orange and rural clusters in green Zambezia province within Mozambique shown in yellow inset. This map uses the shape files by United Nations Office for the Coordination of Humanitarian Affairs (OCHA) licensed under CC BY-IGO and available at https://data.humdata.org/dataset/cod-ab-moz. Shape files were not modified. Source data are in Supplemental Table 4. C Spearman’s rank correlation and corresponding 95% confidence intervals between rank of odds of seropositivity for each antigen across all clusters (n = 30) and rank of overall vulnerability score across all clusters. All p-values were adjusted for multiple hypothesis testing using a Benjamini–Hochberg correction. P-values that were smaller than 0.05 for a 2-sided t-test after adjustment are colored in blue other correlations are colored in red. Size of dots are proportional to the estimated seroprevalence for each antigen. P is short for plasmodium.