Characterizing hepatitis B virus infection in children in the Democratic Republic of Congo to inform elimination efforts

Study population and design

We characterized HBV epidemiology among children 6-59 months through three primary objectives: estimation of childhood HBsAg-positivity prevalence nationally, provincially, and by subgroup, including by age group and sex; assessment of correlates of HBsAg-positivity prevalence among children; and estimation of odds of infection among children given exposure to an HBV-infected adult household member through a nested case-control study. We used survey data and DBS specimens from the 2013-14 DRC DHS, a population-representative survey that included DBS collection in a 50% random sample of households (Supplementary Material). Detailed survey methodology is described in the 2013-2014 DRC DHS final report.(20)

For the first and second study objectives, all children with sufficient DBS sample available were included. For the nested case-control study in the third objective, case households were defined as those with exactly one child aged 6-59 months who was HBsAg-positive in our biospecimen analysis (described below); these children (no more than one per household) formed the case population. Control households were sampled in a 6:1 ratio (control:case households) from households in which all children tested were HBsAg-negative; these children (possibly more than one per household) were the controls. Within these case and control households, adult household members with DBS collected were evaluated for HBV infection as a potential correlate of infection for the case vs. control children.

Biospecimen analysis

We determined HBV status by eluting DBS samples and analyzing HBsAg presence on the ARCHITECT platform (Abbott Laboratories, Abbot Park, IL, USA)(21) in a CLIA-approved laboratory (Supplementary Material). The ARCHITECT instrument defines a positive HBsAg result using a ratio of “signal-to-cutoff” (S/CO). To reduce the risk of false-positive results, we employed a two-step approach in interpreting a positive HBsAg result. First, any sample with S/CO >1 was retested automatically using the same eluant (technical replicate). Second, if there was sufficient DBS sample remaining, the assay was repeated using a fresh 6mm DBS punch (biological replicate) for any positive sample with a S/CO between 1-100 or with an adjacent positive on the elution plate. Samples with insufficient material for repeat testing were considered positive in the primary analysis if the average S/CO of both technical replicates was ≥5, and negative otherwise. All other samples were considered positive if both positive biological replicates (four total technical replicates) were positive (S/CO ≥5), and negative otherwise. We conducted sensitivity analyses with different positivity cut-offs (S/CO of 1, 2, and 100).

HBsAg prevalence and correlates analysis

For representative estimates of HBsAg positivity prevalence, we applied two types of weights. To account for the probability of selection into the survey, we used DHS survey weights available through the DHS Program. To account for sample missingness or exhaustion among those selected to provide biospecimens, we followed an established approach to calculate propensity score weights using inverse probability of sampling (elsewhere called “treatment”) weighting.(22,23) Briefly, we identified attributes that could be associated with the outcome (HBsAg status) or sample missingness/exhaustion (Supplementary Material). These variables were included as independent predictors in a binomial logistic regression model in which having an HBsAg result (yes/no) was the dependent variable. We calculated predicted probabilities from this model to obtain propensity scores, the inverse of which formed the weights. We stabilized the inverse propensity score weights by dividing each weight by the sum of weights in the group with samples. The stabilized weights were multiplied by the DHS survey weights to create final survey weights in the analysis.

We identified variables available from the DHS survey to assess as potential correlates of HBV infection. Details on variable coding are available in Supplementary Material, but briefly: variables included were age, sex, province, rurality, location (provincial capital, small city, town, countryside), household wealth, anemia, P. falciparum malaria infection, tetanus serology results (a proxy for pentavalent vaccination, which includes diphtheria-pertussis-tetanus [DPT] and HepB3), reported number of DPT vaccine doses, growth stunting, injections received in the last 12 months (such as antibiotics and antimalarial medications), use of a new and unopened needle/syringe in the last injection, and reporting justification of violence toward women in the household. P. falciparum infection was determined using quantitative real-time PCR targeting the P. falciparum lactate dehydrogenase (pfldh) gene, conducted prior to this study and previously described.(24)Plasmodium and HBV both infect hepatocytes, but the biological mechanisms of coinfection are poorly characterized. Malaria is known to dampen immunologic responses to polysaccharide and possibly protein-based vaccines,(25) and frequent malaria infections, common in high-prevalence areas like the DRC, result in a higher frequency of blood transfusions, a known risk factor for HBV infection.(26–28) Tetanus antibody results are available publicly through the DHS Program and were produced using an ELISA-based assay.(29)

We described the distribution of the overall population according to each potential correlate of interest and estimated HBsAg-positivity prevalence and Wald-type 95% confidence intervals (CI) within strata of each. We further estimated prevalence differences and associated Wald-type 95% CIs in HBsAg-positivity across correlate strata.

Spatial analysis

We characterized spatial variation in HBsAg-positivity at the DHS cluster level and at the province level. GPS locations were available through the DHS Program for DHS clusters; for clusters with missing GPS location, we used province designation to randomly select latitude and longitude coordinates within the province to impute GPS coordinates. We calculated prevalence of HBsAg-positivity within each DHS cluster. We also analyzed choropleth maps of weighted HBsAg-positivity prevalence at the province level, overall and by age (categorizing age in months according to age in years), sex (boys vs girls), and HepB3 vaccination proxies (tetanus antibody-positive versus antibody-negative and reported receipt of any DPT vaccine dose versus none). To inform focused public health action, we mapped province-level prevalence differences by these same variables, which we chose for province-level analysis on the basis of expected regional differences in vaccine distribution efforts and cultural practices that could confer differential HBV exposure by sex.

Nested case-control analysis

For the nested case-control study, we used logistic regression to calculate weighted, unadjusted odds ratios comparing HBsAg positivity in case vs control children according to two separate exposure variables: 1) exposure to at least one HBsAg-positive adult household member of any type, and 2) exposure to an HBsAg-positive mother, specifically.

All data were imported and analyzed in R (v4.3.3, R Core Team, Vienna, Austria),(30) using the tidyverse (v2.0.0), survey (v4.1-1), and srvr (v1.2.0) packages. Spatial analysis was conducted using the sf (v.1.0-11) package. R code is publicly available at https://github.com/IDEELResearch/dhs_hbv. This study was approved by the Institutional Review Board at UNC and the Ethics Committee at Kinshasa School of Public Health.

Study population

We analyzed 5,679 samples from children 6-59 months of age (Figure 1). The mean age among children was 32 months (SD 0.23), and their households had an average of seven residents (Table 1). Most children (70%) were from rural areas, and 45% resided in households in the lower two wealth quintiles (as defined by the DHS Program for the full DRC population). Similar percentages of boys and girls were studied. Twenty-eight percent had PCR-confirmed P. falciparum infections, and about a third (36%) had detectable tetanus antibodies. Nutritional status was poor overall: almost half (44%) were moderately-to-severely stunted and about a third (35%) had moderate-to-severe anemia. About half of children (53%) had documentation of DPT vaccine completion, and 18% had not initiated the series. Almost one-third of children (31%) had received at least one injection in the last year, and for 7% of children, a reused needle or syringe was applied in their last injection. Approximately two-thirds of children (69%) lived in a household in which physical violence toward a woman was reported as justified. These distributions of demographics of children evaluated for HBV were similar to those of all children selected for DHS biospecimen collection (Supplementary Table 1).

• Download figure
• Open in new tab

Figure 1. Flowchart of study participants

Prevalence of HBsAg positivity, overall and by correlates of infection

Overall HBsAg-positivity prevalence in children 6-59 months was 1.3% (95% CI: 0.9%, 1.7%) (Table 2), ranging from 1.2% to 1.7% in sensitivity analyses (Supplementary Table 2). HBsAg-positivity prevalence was 1.8% (95% CI: 1.2%, 2.7%) in boys and 0.7% in girls (95% CI: 0.4%, 1.3%), corresponding to a prevalence difference (PD) of 1.1 cases (95% CI: 0.2-2.0) per 100 children (Figure 2). HBsAg-positivity prevalence did not differ appreciably by age, ranging from 1.0% in 12–to-23-month-olds to 1.6% among 36-to-47-month-olds (Table 2). Prevalence among children in rural areas (1.5%, 95% CI: 1.0%, 2.1%) was nearly double that of children from urban areas (0.8%, 95% CI: 0.4%, 1.5%). Prevalence decreased with increasing household wealth, from 1.9% (95% CI: 1.1%, 3.1%) in the poorest wealth quintile to 0.6% (95% CI: 0.2%, 2.1%) in the richest wealth quintile. HBsAg-positivity prevalence among children infected with P. falciparum was 1.5% (95% CI: 1.0%, 2.2%), compared to 1.2% (95% CI: 0.8%, 1.8%) among those uninfected with P. falciparum.

• Download figure
• Open in new tab

Figure 2. HBsAg-positivity prevalence differences for potential correlates of interest among children 6-59 months.

Children with detectable tetanus antibodies had 0.8 (0.1, 1.6) fewer HBsAg-positive cases per 100 children compared with those without detectable antibodies. DPT vaccination reported by caretakers was associated with lower HBsAg-positivity, with children reported to have initiated the series (1-2 doses) having 1.7 (0.3, 3.0) fewer infections per 100 compared with those who had not received any doses (Table 2, Figure 2). Children in households with reported justification of physical violence toward women had 0.6 (−0.1, 1.3) more HBsAg-positive cases per 100 children compared to children in households where violence was not reported as justified. We observed no significant association between HBsAg-positivity and anemia, growth stunting, or receiving injections in the last 12 months.

Spatial variation in HBsAg-positivity prevalence, overall and by covariates

Thirty DHS clusters, representing 750-900 households, had an HBsAg-positivity prevalence greater than 10% (Figure 3). We observed substantial provincial variation, from 0% (95% CI: 0%, 0%) in several provinces to 5.6% (2.6%, 11.8%) in the northwestern Sud-Ubangi province (Figure 3, Supplementary Table 3). Most (21 of 26) provinces had at least one HBsAg-positive child. Twelve provinces had higher HBsAg-positivity prevalence among boys compared with girls, with the largest differences in Sud-Ubangi, Équateur, and Ituri; two provinces (Tshuapa and Kwilu) had higher HBsAg-positivity prevalence among girls compared with boys (Figure 4, Supplementary Figure 2). In 10 provinces, children who were tetanus antibody-negative had higher HBsAg-positivity compared with children who were seropositive, with the largest magnitude in Sud-Ubangi (PD: 5.9 per 100 children, 95% CI: 1.4, 10.5) (Figure 4, Supplementary Figure 3). Reported DPT vaccination resulted in similar directionality of association as that observed between tetanus seropositivity and HBsAg status in most provinces, but two provinces (both in the Katanga region) had an opposite result, with reporting of any DPT doses associated with higher HBsAg-positivity prevalence (Supplementary Figure 4). In no province did HBsAg positivity consistently decrease with decreasing age cohort (Supplementary Figure 5).

• Download figure
• Open in new tab

Figure 3. Geographic distribution of the prevalence of HBsAg-positivity in children 6-59 months in the DRC, by A) DHS cluster location, with major cities and rivers indicated, and B) province.

• Download figure
• Open in new tab

Figure 4. Province-level prevalence differences in HBsAg-positivity in children 6-59 months, by sex (A-B) and by tetanus antibody status (C-D).

Nested case-control study

In our nested case-control study of children who resided with no HBsAg-positive children (or other HBsAg-positive children, for cases), we observed that children living with any HBsAg-positive adult household member had 2.3 (95% CI: 0.7, 7.8) times the odds of HBsAg-positivity compared with children without any HBsAg-positive adult household member (Table 3). In particular, children living with a mother who was HBsAg-positive had 7.2 (95% CI: 1.6, 32.2) times the odds of HBsAg-positivity compared to those living with an HBsAg-negative mother.