Modeling the persistence of Opisthorchis viverrini worm burden after mass-drug administration and education campaigns with systematic adherence

Abstract

Opisthorchis viverrini is a parasitic liver fluke contracted by consumption of raw fish, which affects over 10 million people in Southeast Asia despite sustained control efforts. Chronic infections are a risk factor for the often fatal bile duct cancer, cholangiocarcinoma. Previous modeling predicted rapid elimination of O. viverrini following yearly mass drug administration (MDA) campaigns. However, field data collected in affected populations shows persistence of infection, including heavy worm burden, after many years of repeated interventions. A plausible explanation for this observation is systematic adherence of individuals in health campaigns, such as MDA and education, with some individuals consistently missing treatment. We developed an agent-based model of O. viverrini which allows us to introduce various heterogeneities including systematic adherence to MDA and education campaigns at the individual level. We validate the agent-based model by comparing it to a previously published population-based model. We estimate the degree of systematic adherence to MDA and education campaigns indirectly, using epidemiological data collected in Lao PDR before and after 5 years of repeated MDA, education and sanitation improvement campaigns. We predict the impact of interventions deployed singly and in combination, with and without the estimated systematic adherence. We show how systematic adherence can substantially increase the time required to achieve reductions in worm burden. However, we predict that yearly MDA campaigns alone can result in a strong reduction of moderate and heavy worm burden, even under systematic adherence. We predict latrines and education campaigns to be particularly important for the reduction in overall prevalence, and therefore, ultimately, elimination. Our findings show how systematic adherence can explain the observed persistence of worm burden; while emphasizing the benefit of interventions for the entire population, even under systematic adherence. At the same time, the results highlight the substantial opportunity to further reduce worm burden if patterns of systematic adherence can be overcome.

Fig 1. Distribution of EPG among humans in data collected in 2012 and 2018.

The hatched wide bars on the left show the proportion of individuals with an egg count of zero. The remaining bars show the proportion of individuals with nonzero egg counts, where the bin width equals 1000 EPG. The red lines mark the thresholds of World Health Organization worm burden classification with 1–999 EPG being classified as light infection, 1000–9999 as moderate infection and 10000 and above as heavy infection [43].

Fig 2. Schematic of the the ABM for O. viverrini transmission and control.

The state variables are described in Table 2 and the transmission parameters are described in Table 3. EPG output is calculated from worm counts using a density-dependent function ρ derived from purging studies (Section A.6.2 in S1 Appendix). While ρ is applied to the mean worm burden in cats and dogs and then multiplied with the number of cats and dogs, respectively, it is applied to each individual worm count in humans and subsequently summed up over humans.

Fig 3. Schematic of the PBM for O. viverrini transmission and control.

The state variables are described in Table 4, the transmission parameters roughly correspond to the transmission parameters of the ABM listed in Table 3. As opposed to the ABM, where β_hf,i varies over individuals, there is only a single β_hf in the PBM. Also, the PBM does not translate worm burden in definitive hosts to egg output which results in the transmission parameters β_sh, β_sd and β_sc acting on mean worm burden instead of egg count. Further details on the PBM are provided in Section B in S1 Appendix.

Fig 4. Time series for prevalence, mean worm burden, prevalence of moderate worm burden and prevalence of heavy worm burden as predicted by the models starting in equilibrium followed by five years of interventions without systematic adherence (MDA, education campaign, improved sanitation).

Both models were calibrated to fit the 2012 data in equilibrium but not the 2018 data. Additional variables of the model runs are provided in Fig I in S1 Appendix.

Fig 5. Kernel density estimate of EPG in the 2012 data and in the equilibrium state of the ABM and the PBM.

The x-axis is plotted on a log scale. The dashed vertical red lines indicate the transition from light to moderate and from moderate to heavy worm burden, respectively.

Fig 6. First-order Sobol indices measured in the global sensitivity analysis of fitted intervention parameters (x-axis) on goodness of fit of the model to the objective summary statistics in the 2018 data (y-axis, Table 7).

The indices sum up to one in each row. The parameter controlling the mean effect of the education campaign, μ_e, has mostly a strong impact on the prevalence in humans. Systematic adherence to the education campaign, controlled through the parameter ${\sigma }_e^2$, affects prevalence, prevalence of heavy worm burden, and mean worm burden, though for each one of these outcomes, one of the other two studied parameters is of greater importance. Systematic adherence to the MDA campaign, controlled through the parameter ${\sigma }_m^2$, has a strong effect on the prevalence of moderate and heavy worm worm burden and consequently on the mean worm burden in the population.

Fig 7. Time series for the ABM with fitted systematic adherence where all three interventions are implemented together or in isolation.

Additional variables of the model runs are provided in Fig I in S1 Appendix.