HBV vaccination and PMTCT as elimination tools in the presence of HIV: insights from a clinical cohort and dynamic model

Ethics approval was obtained from the Ethics Committee of the Faculty of Health Science, University of the Free State, Bloemfontein, South Africa (HIV Study Ref: ETOVS Nr 08/09 and COSAC Study Ref: ECUFS NR 80/2014), and from the Oxfordshire Research Ethics Committee A, ref 06/Q1604/12. Written consent for enrollment into the study was obtained from the child’s parent/guardian.

Recruitment was undertaken in Kimberley, South Africa. In this setting, a standard three-dose HBV immunisation schedule is deployed in infants, with the first dose at 6 weeks. A previous study of HBV serology in adults in the same setting found HBsAg prevalence of 9.5% (55/579) [6]. Children were recruited as part of the Co-infection in South-African Children (‘COSAC’) study as previously described [20, 21]. The lower age limit of recruitment was 6 months in order to limit the detection of maternal anti-HBs.

Children were recruited as follows:

At the time of undertaking this study, children were immunised with three doses of a monovalent HBV vaccine (Biovac Paed). Where possible, we recorded the number of HBV vaccine doses received based on the Road to Health Book (RTHB). The characteristics of the cohorts are summarised in Table 1 and all metadata can be found in Additional file 1.

Testing for hepatitis B serum markers and DNA was performed as previously described, and in keeping with recent implementation of HBV screening in Kimberley [21]. Briefly, HBsAg testing was carried out in Kimberley Hospital, South Africa, using the magnetic parcel chemiluminometric immunoassay (MPCI; Advia Centaur platform). Confirmatory HBsAg testing was carried out by the clinical microbiology laboratory at Oxford University Hospitals (OUH) NHS Foundation Trust, Oxford, UK (Architect i2000). For all samples, anti-HBs and anti-HBc testing were carried out by the OUH laboratory (Architect i2000). Limit of detection of the anti-HBs assay was 10 mIU/ml.

Studies variably quote anti-HBs titres of ≥ 10 mIU/ml or ≥ 100 mIU/ml as a correlate of protection; UK recommendations for testing HBV immunity advocate the more stringent criterion of an anti-HBs titre of ≥ 100 mIU/ml [12], while early vaccine studies suggest a titre of ≥ 10 mIU/ml as a clinically relevant threshold for protection [13, 22]. We have presented our results pertaining to both thresholds.

Here, we summarise the modelling framework, but include a detailed description of the ODE system, model parameters, and Bayesian data fitting approach in Additional file 2. We developed a dynamic model based on ordinary differential equations (ODE), for which parameterisation of HBV transmission and prevention was based both on our Kimberley paediatric cohort and current literature estimates.

In summary, the model takes into consideration the proportion of the population susceptible to HBV infection (S), those with chronic infection (C) and acute infection (I), those who are immune as a result of recovery from prior infection (R) and those who are immune as a result of vaccination (V) (Fig. 1). For simplicity, and assuming vaccination takes place early in life, all individuals are assumed to be born either susceptible (Z) or vaccinated (Z’). Chronic carriers (C) are divided into HBeAg-positive (C+) and HBeAg-negative (C−) to further allow for different parameterisation (e.g. transmission potential) between these two epidemiologically distinct states. To be able to parameterise epidemiological traits by age, (e.g. probability of chronicity, or decay of vaccine-induced protection) susceptible (S) and vaccinated (V) individuals are divided into three subgroups representing infants (i, < 1 years of age), children (c, 1–6 years of age) and older individuals (whom we abbreviate to ‘adults’, a, but comprise children age > 6 years, adolescents and adults). The probability of developing chronicity decreases with age, with (1 − ψ) for infants, (1 − ε) for children and (1 − γ) for older individuals. Vertical transmission takes place from mothers with chronic infection and is dependent on their HBeAg serostatus (not shown on diagram). HBeAg-positive chronic carriers (C+) may become HBeAg-negative at a rate θ. HBeAg-negative chronic carriers (C−) can clear infection spontaneously at a rate ρ, entering the anti-HBc-positive, HBsAg-negative state (R). Acute infections (I) are cleared at a rate σ, also entering the recovered class (R).

Vaccinated individuals (Vi, Vc, Va) are under the same HBV acute and chronic infection rules as susceptible individuals (Si, Sc, Sa), but are further assumed to have vaccine-induced age-dependent protection against infection (Δi, Δc, Δa). For simplicity, we assume that vaccine-induced protection was equivalent to reducing susceptibility to infection potential (λ), e.g. Δi = 1 would be 100% reduction in susceptibility, or 100% vaccine efficacy against infection. Interventions include routine vaccination at birth (affecting Z’) and other ages (affecting ωa, ωc), as well as PMTCT at birth (affecting Z, Z’) and catch-up events (not shown in diagram).

We used a Bayesian Markov-chain Monte Carlo (bMCMC) approach to fit the dynamic model to the local demographic and epidemiological setting of Kimberley before projecting the impact of interventions (Additional file 2: Figure S1). The bMCMC used informative priors for ODE model parameters for which robust literature support exists. Two parameters (ρ, θ) were left with uninformed priors (uniform, from 0 to 1), for which we later checked if the fitted bMCMC solution recovered posteriors of these parameters compatible with current literature knowledge (as partial validation of the fitted solution). Informed by the clinical cohort data described above, natural decay (age-effects) and the effects of HIV sero-status on vaccine-induced protection (Δi, Δc, Δa) are taken into account (Additional file 2: Figure S2).

SDGs for the year 2030 have been set out in the WHO Global Health Sector Strategy on Viral Hepatitis (GHSSVH) [2]. Given the public health relevance of chronic infections, in particular of HBeAg-positive infections, we measured impact of interventions based on two targets:

From our cohort of 402 children in Kimberley, South Africa, three were HBsAg-positive (0.7%; Table 2), indicating active infection. This HBsAg prevalence is significantly lower than in adults in a comparable study population (e.g. 11.1% in a previous study [6]; p < 0.0001), but is typical of a population with widespread vaccine coverage [10], highlighting the positive impact of the EPI programme to date. Exposure to HBV infection was measured using anti-HBc antibody; this was detected in three children (0.7%), one of whom was also HBsAg-positive. The other two were HBsAg-negative, indicating previous HBV exposure and clearance.

We collected written evidence of immunisation from the RTHB in 90.8% HIV-negative (KReC) subjects and 6.3% of HIV-positive subjects (total 41.3% of the entire study cohort). None of the HBsAg-positive children attended with a written vaccination record. Among those with a RTHB record, 81.3% of HIV-negative and 100% of HIV-positive children were recorded as having received three HBV vaccine doses. Among all children age ≤ 60 months, 238/310 (77%) had an anti-HBs titre ≥ 10 mIU/ml suggesting some degree of vaccine-mediated immunity (Fig. 2A). The median anti-HBs titre in HIV-negative children was significantly higher than among the HIV-positive group (196 mIU/ml, vs. 11 mIU/ml, respectively, p < 0.0001) (Fig. 2B). There was no detectable anti-HBs antibody in 3.4% of HIV-negative vs. 47.8% of HIV-positive children (p < 0.0001). Irrespective of the antibody titre used as a threshold for immunity, anti-HBs was higher in HIV-negative compared to HIV-positive children (Fig. 2C). There was no significant difference in anti-HBs titres between male and female participants, either with or without HIV infection (p = 0.49 and 0.31 respectively, data not shown).

HIV-positive children with anti-HBs titres ≥ 100 mIU/ml were significantly younger than those with lower antibody titres (median age 17 months vs. 31 months, p = 0.0008), while no such difference was observed within the HIV-negative group (Fig. 3A). Using the lower threshold of ≥ 10 mIU/ml, we found no significant difference by age in either the HIV-positive or the HIV-negative groups (p = 0.17 and 4.48 respectively, data not shown). To expand our view of the HIV-positive group, we also added analysis of an older cohort (92 children aged > 60 months) and demonstrated that anti-HBs titres were significantly lower in this older group (p < 0.0001), with only 2/92 subjects (2.2%) achieving a detectable anti-HBs titre (Fig. 3B). Anti-HBs titres waned significantly with age up to age 60 months in HIV-positive children (Fig. 3C; p = 0.004). We observed a similar trend in the HIV-negative cohort, but this did not reach a statistical significance (Fig. 3C; p = 0.07). The proportion of HIV-positive subjects with a detectable anti-HBs titre declined steadily with age in the cohort, contrasting to the trend in HIV-negative subjects, where individuals maintained protective anti-HBs titres despite a trend towards decreasing mean titres (Fig. 3C).

For HIV-positive children aged ≤ 60 months, ART treatment data were available for 79% of subjects. Within this group, 71% were receiving ART at the time we tested for anti-HBs, and had received a median of 20 months of treatment (IQR 6–33 months). Comparing anti-HBs titres between ART-treated vs. untreated children, we found no significant difference (p = 0.72; 76 ART-treated, median anti-HBs 13.3 mIU/ml and 31 untreated children, median anti-HBs 14.1 mIU/ml, data not shown). There was also no difference between anti-HBs titres of children treated for ≤ 12 months vs. > 12 months (p = 0.50, data not shown). We did not examine the effect of ART on anti-HBs titres in children > 60 months due to the low numbers of subjects with a detectable anti-HBs titre (n = 2).

We used an odds ratio (OR) analysis to identify factors associated with vaccine-mediated protection (Fig. 3D). HIV-positive status was associated with lack of protection, for antibody titres of both < 10 mIU/ml (OR 26.2, 95% CI 11.2–58.6) and < 100 mIU/ml (OR 11.6, 95% CI 6.7–20.4). In contrast, younger age (< 24 months) was protective in HIV-positive subjects, (for anti-HBs < 10 mIU/ml, OR 0.3, 95% CI 0.2–0.5, and for anti-HBs < 100 mIU/ml, OR 0.3, 95% CI 0.2–0.4). Gender, ART, CD4+ count, CD4+ ratio, and HIV viral load were not found to be significantly predictive of anti-HBs titres at either threshold.

The mathematical model was used in three steps, as follows. Step 1: calibration to a demographic background by estimation of rates of movement between the age-structured susceptible classes, with the aim of deriving similar age-proportions to Kimberley (in the absence of transmission). Step 2: calibration to the transmission background without interventions, in which local HBV prevalence (HBsAg, or sum of infected), prevalence of HBV exposure (recovered class R) and relative proportion of HBeAg-negative and HBeAg-positive among chronic carriers were fitted. Step 3: the model was run until it reached equilibrium, effectively reproducing the (mean) desired proportions of step 2, at which point interventions were started and the model was tracked for 1000 years. Details on all steps, the methods and parameters involved can be found in Additional file 2.

In step 2, when calibrating to the transmission background, the model closely reproduced the desired variables (fitted distributions and estimated parameter posteriors in Additional file 2: Figure S1 A1–2). For parameters for which insufficient support was found in the literature, the resulting posteriors were well behaved (Additional file 2: Figure S1 B), while for parameters using informative priors the posteriors matched the priors well (Additional file 2: Figure S1 C-D). Overall, the obtained bell-shaped posteriors highlighted a good data fit and no identifiability issues with the bMCMC approach. Based on these results, we next simulated stochastic projections of intervention impact.

Based on SDGs [2], Fig. 4 shows the projected impact of varying coverage of neonatal vaccination and PMTCT. Both HBsAg incidence (Fig. 4A1) and HBeAg-positive prevalence (Fig. 4B1) reduce faster with increasing neonatal immunisation coverage, resulting in shorter times to reach SDGs (Fig. 4A2, B2). Importantly, even immunisation of 100% of neonates is predicted to take ~ 99 years (95% CI 61–186) for the HBsAg incidence target (Fig. 4A2) and ~ 175 years (95% CI 103–278) for the HBeAg prevalence target (Fig. 4B2).

With PMTCT interventions, both HBsAg incidence (Fig. 4C1) and HBeAg prevalence (Fig. 4D1) reduced faster in time for increasing efforts, resulting in shorter times to reach the elimination targets (Fig. 4C2, D2). However, the impact of PMTCT was smaller than neonatal vaccination, resulting in longer times to reach SDGs. In fact, for the majority of PMTCT effort levels simulated, SDGs could not be reached within 500 years (beige areas in Fig. 4C2, D2). For HBeAg-positive prevalence, only when PMTCT effort was 1 (i.e. no vertical transmission), was the elimination target attainable within 500 years.

For complementary results using neonatal vaccination and PMTCT with impact on total prevalence (acute and chronic), see Additional file 2: Figure S3, and for time to reach SDG targets, see Additional file 2: Figure S4.

We projected impact of combined interventions by the year 2030 (Fig. 5A1, B1), and predicted the year at which SDGs would be reached (Fig. 5A2, B2). Strikingly, HBsAg incidence could already have been reduced by > 90% (Fig. 5A1) if both neonatal vaccination and PMTCT had been deployed at 100% coverage since they became available in 1995 (mean predicted year of elimination 2017; Fig. 5A2). In reality, complete coverage is not possible, and we therefore projected outcomes based on < 100% coverage. For example, combining neonatal vaccination and PMTCT with 90% coverage of each since 1995 would achieve the HBsAg incidence target by 2028; if this is reduced to 80% coverage, then goals would be attained by 2044. To achieve the target reduction in HBeAg prevalence, modelled on 90% coverage and 80% coverage of interventions, the projected years are 2072 and 2096 respectively (Fig. 5B1, B2).

Overall, the highest probability of achieving elimination targets is through a combination of 100% neonatal vaccination coverage and PMTCT (Fig. 6A1, red line). Based on 90% coverage of neonatal vaccine and PMTCT (Fig. 6A1, green line) as proposed in WHO’s GHSSVH [2], there was only 50% probability of reaching the HBsAg incidence target by 2030 and approaching 100% probability only by 2050. For the target based on HBeAg prevalence, the probabilities of achieving the goal were pushed forward by approximately four decades (Fig. 6A2).

For complementary results on impact and time to reach SDGs when considering combinations of PMTCT and routine vaccination at the age of 6, see Additional file 2: Figure S5, and for combinations of PMTCT and neonate routine vaccination plus a complete catch-up campaign, see Additional file 2: Figure S6.

We simulated the impact of combining neonatal vaccination and PMTCT with additional vaccine deployment (Fig. 6A1, A2), through the routine vaccination of older children (≥ 6 years of age), and one-off catch-up vaccination of children (< 6 years) and others (> 6 years). Adding catch-up vaccination campaigns makes no impact on the probability of reaching SDGs (Fig. 6A1, A2, blue and cyan lines). Routine vaccination at 6 years of age, even when delivered at 100% coverage, is markedly less effective than any other projected intervention (Fig. 6A1, A2, magenta line).

A baseline scenario was defined by the epidemiological setting fitted by our model in the context of Kimberley, using local HIV prevalence for each of the modelled age groups (Fig. 6B1, B2, solid line). In a sensitivity exercise, alternative scenarios were considered in which baseline HIV prevalence was altered to zero or higher prevalence. When compared with no HIV (Fig. 6B1, B2, dotted line), the presence of HIV at the prevalence seen in Kimberley (Fig. 6B1, B2, solid line) adds an estimated 4 years to achieve a 50% chance of reaching the goals (Fig. 6B1). A higher baseline HIV prevalence (× 2, × 3 and × 4 baseline data for Kimberley) was used to investigate the potential impact of coinfection in high-risk populations. Increasing HIV prevalence has a negative impact on the success of interventions for HBV, but the effects are relatively modest. In particular, doubling HIV prevalence would shift the 50% probability endpoint into the future by ~ 4 years for the HBsAg incidence target and ~ 7 years for the HBeAg prevalence target.

Variation in prevalence of HBV has been reported from different regions of South Africa, ranging from 2.0–35.2%, depending upon the population sampled [26, 27]. The adult HBsAg prevalence of 8–11% that we have recorded in our own previous studies of the region [6, 28] is represented in this current analysis; while recognising that there is considerable heterogeneity between populations, we believe this epidemiology also broadly reflects other sub-Saharan African settings, where an adult prevalence of ≥ 8% would be typical. The much lower prevalence of infection in children demonstrates the success of inclusion of HBV vaccine in the EPI schedule [10, 21, 29].

Our results support previous evidence that HBV vaccine-mediated immunity wanes over time independently of HIV serostatus, but faster for HIV-positive individuals [30]. Impaired vaccine responses have previously been reported in HIV-positive individuals [16, 26, 31‐33], but it is also possible that vaccine coverage is lower in HIV-infected children [34]. However, waning of anti-HBs titres does not necessarily correlate with loss of clinical protection; anamnestic responses are thought to occur in a proportion of those vaccinated [35], although this memory may be attenuated by HIV [36, 37].

ART has previously been associated with improved HBV vaccine responses [38, 39], although we did not replicate this finding in our cohort. This can potentially be explained by a previous study in the same setting, demonstrating that immune reconstitution takes a median of 5 years after ART initiation [40]. Our current study is underpowered to detect any ART effect, given both the relatively short durations of therapy and the small number of untreated children. Interestingly, despite the lack of direct association with ART, children with lower HIV viral loads had significantly higher anti-HBs titres, in keeping with previous studies [15, 38].

Our cohort highlights day-to-day challenges of drug provision and monitoring in this setting: we did not have access to detailed prospective ART treatment data, guidelines have changed numerous times since 2002, and lamivudine (3TC) was intermittently used as a substitute for nevirapine (NVP) due to supply issues. During the period covered by our study, ART was only introduced in children achieving certain immunological criteria, while new guidelines recommend that all HIV-infected children are started on ART [41]. The immune reconstitution of this population over time is likely to reduce differences between HIV-positive and HIV-negative groups. ART treatment is relevant to outcomes in individuals with HIV/HBV coinfection, as first line ART regimens include either 3TC or tenofovir (TDF), both of which have activity against HBV. Alternative approaches for HBV prevention in HIV-positive subjects, such as supplementing the current schedule with booster vaccinations and increased vaccine doses, have been trialled with variable results [15, 16].

Our projections propose that HIV does have a negative effect on HBV interventions, although HIV prevalence only marginally increases time to reach elimination targets, which may not be significant in light of the long overall time-frames projected, even in the absence of HIV. The high HIV prevalences modelled can occur in specific high-risk groups including sex workers and men who have sex with men [42], and it is likely that increased intervention will be required in these groups to minimise HBV transmission.

The model suggests long time-lines, enumerated in centuries rather than decades, before control targets are reached using vaccination or PMTCT alone. Combinations of these interventions show much shorter time scales. Based on currently available interventions, major scaling up of both neonatal vaccination and PMTCT efforts will be required. Importantly, the prevalence of HBeAg-positive carriers, who are at an elevated risk of chronic liver disease and hepatocellular carcinoma, as well as being at higher risk of transmitting their infection, will decline at a slower rate. Setting a control target based on reduction in the number of new HBV cases (i.e. HBsAg incidence) can therefore lead to the most optimistic projections but distract attention from the importance of reducing HBeAg prevalence.

Our results also underscore that a major public health impact is possible, even without achieving elimination. Careful adjusting of expectations and aims, according to the scale on which particular changes occur, may inform the setting of realistic targets (e.g. reduction in the prevalence of HBeAg could be the most informative outcome measure). The wrong choice of either target or timescale may result in unnecessary abandonment of a strategy that could have a major impact in a few decades. In addition to informing rational use of interventions that have a positive population impact, our study is also important in cautioning against the use of strategies that may have little or no lasting population impact. This is illustrated by our results for catch-up HBV vaccination, which adds little in situations where high coverage of both neonatal immunisation and PMTCT can be attained.

Different approaches to recruitment of our HIV-positive and HIV-negative cohorts may have introduced unintentional bias. The KReC children may be less healthy than HIV-negative children in the community, and this approach to recruitment predominantly selected younger children (on average 9.4 months younger than the HIV-positive cohort). We set out to focus on children aged < 60 months in order to collect data from the RTHB. However, in practice, we did not capture good RTHB data and data collection from the RTHB is itself subject to bias, as families who attend with such records may be those who are most likely to have immunised their children. Numerous complex social factors are also relevant in determining immunisation status; babies born to mothers who have HIV and/or HBV are more likely to be in disadvantaged by poverty, and by illness and death in the family, such that they might be less likely to present for (or respond to) vaccination. However, in this setting (and others where antenatal HBV screening is not routinely deployed [8, 43, 44]), we deem it unlikely that there is a significant difference in vaccination rates between infants born to HBV-positive versus HBV-negative mothers.

HBV DNA is a more sensitive screening tool than HBsAg but was not practical due to high cost and lack of availability in this setting. The relatively small numbers in each age group and the lack of longitudinal follow-up for individual children puts limitations on the data showing anti-HBs waning over time, but the trends we observe here are biologically plausible and consistent with existing literature [30, 45].

Although we have estimated and parameterised the impact of HIV status on HBV vaccine-induced protection, we have not modelled other factors related to HIV infection. Namely, we have not included the potential for increased susceptibility to HBV infection or increased risk of vertical transmission. These factors would have required further model classes and specific parameterisation, for which little literature support exists. We have also not considered the influence of population migration on the success of HBV interventions to reach the elimination targets. Migration of non-immune and/or infected individuals into an area would delay the time to achieve the targets estimated by our modelling approach.

We make no age-specific assumptions about force of transmission. Given that reliable data do not exist to inform this parameterisation, we could potentially add a lot of additional uncertainty by trying to estimate age-specific variables (by need of additional age-dependent model parameters). Keeping parameterisation simple was an intended approach, to avoid introducing uncertainty and to make the model applicable as a tool in other settings. We therefore elected to apply and estimate an age-homogeneous force of transmission. Consequently, the model assumes a uniform risk of HBV infection and transmission potential in the > 6 years age group. In reality, particular sub-groups in this category are at increased risk of HBV infection, including sex workers, people who inject drugs and healthcare workers [46]. It is likely that these minority risk groups contribute little to the overall burden of HBV infection and transmission in most settings, but further analysis would be required to quantify the specific impact of targeting these groups for intervention. While our modelled projections suggest that generalised catch-up vaccination of adults is insufficient to eliminate HBV at the population level, targeted vaccination of vulnerable groups remains important to protect individuals, and may contribute to public health efforts.

We used 50 stochastic simulations to model the mean and variation of interventions. Although this is a relatively small number, we found no significant differences between the resulting mean and variation for each new simulation after a total of approximately 30. This was likely a result of the stable posteriors obtained, and then sampled to run these simulations (Additional file 2: Figure S1). On these grounds, it is unlikely that extending the number of simulations would alter the final results.