The World Health Organization 2030 goals for onchocerciasis: Insights and perspectives from mathematical modelling

Abbreviations

ABR, annual biting rate; APOC, African Programme for Onchocerciasis Control; ATS, alternative treatment strategies; DALY, disability-adjusted life-year; EMC, Erasmus Medical Center; EOT, elimination of transmission; EPHP, elimination as a public health problem; ICL, Imperial College London; IU, implementation unit; MAP, maximum a posteriori (parameter set); MDA, mass drug administration; NTD, neglected tropical disease; OCP, Onchocerciasis Control Programme in West Africa; PCR, polymerase chain reaction; pOTTIS, provisional operational thresholds for treatment interruption and initiation of surveillance; PPV, positive predictive value; SSA, sub-Saharan Africa; Ta(N)T, test-and-not-treat; TaT, test-and-treat; WHO, World Health Organization; YLD, years of life lost to disability.

Disclaimer

The views and opinions expressed in this article are those of the authors and do not necessarily reflect those of the World Health Organization. Publication in Gates Open Research does not imply endorsement by the Gates Foundation.

Background

Onchocerciasis is a filarial disease caused by infection with Onchocerca volvulus, a vector-borne parasitic nematode transmitted via the bites of several Simulium blackfly species. In sub-Saharan Africa (SSA), where 99% of the cases occur, the main vectors belong to the Simulium damnosum sensu lato species complex. Modelling of data from the 2013 Global Burden of Disease Study suggests that at least 17 million people are currently infected with O. volvulus¹. An estimated 198 million people live in areas where there is potential for transmission of the parasite, although this number may increase as the mapping of areas of low transmission is finalized². The disease is also known as river blindness because blackflies breed in and bite near fast-flowing rivers, and because the most devastating sequela is irreversible loss of vision. High infection load, measured by the density of microfilariae (the larval progeny of adult worms) in the skin, is associated both with blindness incidence³ and excess human mortality, the latter over and above that due to blindness^4,5. In addition to ocular sequelae, onchocerciasis is also responsible for skin disease and troublesome itching (which disturbs sleep and work patterns). The Global Burden of Disease Study 2015 estimated the disability-adjusted life years (DALYs) due to onchocerciasis as 1,136,000 in 2015 (a 21% decrease since 2005)⁶. More recently, the term river epilepsy has been used to highlight the association between onchocerciasis and nodding syndrome (a condition in the epilepsy spectrum)⁷. The burden of onchocerciasis-associated epilepsy for 2015 was estimated as circa 13% of the total years lost due to disability (YLDs) attributable to onchocerciasis and 10% of those attributable to epilepsy⁸.

Currently, the mainstay of onchocerciasis control is through mass drug administration (MDA) of the microfilaricidal drug ivermectin (Mectizan®), donated by Merck & Co. Inc. to endemic countries through the Mectizan Donation Program⁹. Ivermectin MDA (mostly of annual frequency in SSA) started in the late 1980’s, and great strides in reducing morbidity have been made¹⁰. Although questions were raised concerning the feasibility of eradicability based on ivermectin MDA alone^11–13, studies in some foci of Mali and Senegal (in the western extension of the Onchocerciasis Control Programme in West Africa, OCP), indicated that local elimination may be achieved after 15–17 years of ivermectin MDA^14,15. Based on this, in 2010, the African Programme for Onchocerciasis Control (APOC) shifted its goals from elimination of the disease as a public health problem (EPHP), to elimination of transmission (EOT)¹⁶. This shift necessitates a drastic geographical extension of treatment. Whilst for EPHP, mesoendemic and hyperendemic areas had been prioritized for ivermectin MDA, EOT at a Pan-African scale requires that treatment be distributed also in hypoendemic areas. This poses challenges, particularly where loiasis (a filariasis caused by Loa loa) is co-endemic due to severe adverse events, including fatalities, that may result from microfilaricidal treatment of individuals heavily infected with L. loa¹⁷. In onchocerciasis, the original definition of endemicity levels comprises three categories: i) hypoendemic (microfilarial prevalence <30–35%), ii) mesoendemic (prevalence between 30–35% and 60%), and iii) hyperendemic (≥60%)¹⁸; EOT should be achieved in all three. A microfilarial prevalence ≥80% has also been used to indicate holoendemicity¹⁹.

EOT is formally defined as a reduction to zero in the incidence of infection (rate at which new cases arise)²⁰. However, since in macroparasite epidemiology the number of parasites in the population is more relevant than the number of cases, the definition of EOT used here is the absence of adult worms and larval stages (in humans and vectors) after cessation of interventions²¹. Onchocerciasis is in the fortunate position that ivermectin reduces: i) morbidity (mainly caused by microfilariae), ii) microfilarial prevalent cases and microfilarial load, and iii) transmission to vectors (microfilariae are the stages infective to blackflies). This is achieved by a combination of ivermectin’s microfilaricidal effect, temporary embryostatic effect on adult female worms (macrofilariae)²², and irreversible effects on female worm fecundity and/or survival^23,24. Given the disease- and transmission-curtailing benefits of ivermectin MDA, the 2012 World Health Organization (WHO) roadmap to accelerate progress in combating neglected tropical diseases (NTDs) proposed that onchocerciasis be eliminated in selected African countries by 2020²⁵. The Joint Action Forum of APOC reformulated this goal to EOT in 80% of endemic African countries by 2025²⁶.

To help evaluate progress towards these goals, the developers of two mathematical models of onchocerciasis transmission dynamics and control (EPIONCHO and ONCHOSIM) have joined forces under the umbrella of the Bill & Melinda Gates Foundation-funded NTD Modelling Consortium. The modelling groups, based at Imperial College London (ICL), UK, and Erasmus Medical Center (EMC), The Netherlands, along with multi-disciplinary collaborators have led the recent onchocerciasis work. Firstly, a comparison of modelling assumptions and resulting outputs was conducted²⁷, highlighting the need to present and discuss the models jointly²⁸, and modify some key assumptions crucial to fitting the microfilarial prevalence trends observed in the elimination studies of Mali and Senegal²⁹. A policy paper followed, discussing the role of intervention strategies other than annual ivermectin MDA to achieve elimination in SSA³⁰. A further understanding of critical uncertainties led to a paper discussing these in light of data needs³¹. Due to knowledge gaps surrounding the processes regulating parasite establishment and fecundity of O. volvulus^32–34, the magnitude of exposure heterogeneity and its interaction with such regulatory processes³¹, and the impact of treatment on population biology parameters, including the development of acquired immunity³⁵, the models have contrasting underlying assumptions leading to differences in model predictions^29,30. Despite these differences, the models generally agree on the treatment strategies required to achieve EOT.

Moving towards the post-2020 goals, new NTD goals have been proposed by the WHO for 2030 within the framework of a consultation process. Table 1 summarizes the current (2020) and proposed (2030) goals for onchocerciasis, the scenarios in which both models agree that EOT is technically feasible with current interventions, the alternative treatment strategies (ATS^36,37) required when current ones are not sufficient, the suitability of available tools for evaluating EOT, the principal uncertainties driving differences in model outputs, and the biggest risks associated with the proposed 2030 goals.

Insights gained from quantitative and mathematical modelling analyses

1) Initial model comparison: Predictions from EPIONCHO and ONCHOSIM were compared over a range of baseline endemicities and treatment scenarios. In particular, comparisons were made on: 1) microfilarial prevalence and intensity during 25 years of (annual or biannual) MDA with ivermectin; 2) required duration of treatment to bring microfilarial prevalence below a provisional operational threshold for treatment interruption (pOTTIS) of 1.4%^16,38; and 3) required duration to drive the parasite population to local elimination (defined by stochastic fade-out in ONCHOSIM and passing the transmission breakpoint in EPIONCHO). In mesoendemic areas, the models predicted that the provisional operational prevalence threshold could be reached with annual MDA. In highly hyperendemic areas, the models indicated that annual MDA would be insufficient. In lower endemicity settings, ONCHOSIM predicted that the time needed to reach 1.4% microfilarial prevalence would be longer than that required to reach local elimination; the opposite was true for higher endemicity settings. In EPIONCHO, the pOTTIS was reached consistently sooner than the breakpoint²⁷.

2) Modelling elimination studies: Based on the initial model comparison, technical refinements were implemented (e.g. age-dependent adult worm mortality in EPIONCHO). EPIONCHO and ONCHOSIM projections were tested against microfilarial prevalence data from the two foci in Mali and Senegal where the observed prevalence had been brought to zero in 2007–2009 after 15–17 years of ivermectin MDA only^14,15. Model projections were trained using longitudinal (microfilarial prevalence) data from 27 communities in two transmission foci, incorporating programmatic information (treatment frequency, duration and coverage), and evaluating whether the projected outcome was elimination (local parasite extinction) or resurgence. The epidemiological trends during MDA were captured by both models but resurgence was predicted by EPIONCHO in some communities of the River Gambia focus in Senegal with the highest (inferred) annual biting rates (no. vector bites/person/year) and associated pre-intervention endemicities (Figure 1 and Figure 2)²⁹. Resurgence was never predicted by ONCHOSIM.

Figure 1. Modelled relationship between annual biting rate (ABR) and endemic microfilarial prevalence.

In EPIONCHO (a), the coupled ABR-prevalence data are from nine communities in northern Cameroon, with each ABR measured as an average from multiple years and locations within and around each community, weighted by the proportion of time community residents spent at these locations⁴⁰. Each thin line corresponds to an EPIONCHO parameter set identified by a sampling importance resampling procedure to account for parametric uncertainty. These are colored sequentially from yellow (hypoendemic) to red (hyperendemic). The thick black line corresponds to the parameter set that achieved the highest likelihood. In ONCHOSIM (b), the thin lines correspond to stochastic realizations using the default parameter set²¹, colored sequentially according to endemicity category; the thick black line is the median of 500 simulations. The coupled ABR-prevalence data are not shown in (b) because ONCHOSIM has not been re-fitted to these data. ABR = No. blackfly bites/person/year. This figure has been reproduced from 29 under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.

Figure 2. Observed and modelled microfilarial prevalence dynamics in 14 communities from the River Gambia focus, Senegal.

Left (a) and right (b) panels show, respectively, EPIONCHO and ONCHOSIM projections. The thin lines correspond to community-specific simulations using maximum likelihood estimates of the community-specific ABRs and either the maximum a posteriori (MAP) parameter set (EPIONCHO) or the default parameter set (ONCHOSIM). The estimated ABRs and MAP parameter sets were derived using the complete longitudinal sequence of microfilarial prevalence for each community. For ONCHOSIM there are many stochastic projections for each community; for EPIONCHO there is a single deterministic projection for each community, corresponding to the MAP parameter set. The thick solid lines show the median dynamics by endemicity category (yellow: hypoendemic; orange: mesoendemic; red: hyperendemic). In the River Bakoye focus of Mali, no resurgence was predicted by either model using the entire longitudinal microfilarial prevalence set. Panel insets show the period between 2000 and 2020 using a transformed y-axis for a better visual appraisal of the model projections compared to the data close to zero. This figure has been reproduced from 29 under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.

3) Alternative treatment strategies: ONCHOSIM and the refined version of EPIONCHO²⁹ were used to simulate trends in microfilarial prevalence for 80 different settings. These settings were defined by their baseline endemicity and past programmatic scenario (frequency and coverage of MDA) and future treatment scenarios defined by the frequency (annual, biannual, or quarterly), with or without vector control. Each strategy was assessed whether it eventually led to elimination³⁰. Both models predicted that in areas with 40%–50% pre-control microfilarial prevalence and ≥10 years of annual MDA, elimination may be achieved within a further 7 years without changing strategy. For areas with 70%–80% pre-control microfilarial prevalence, both annual and biannual MDA were predicted by ONCHOSIM (but not by EPIONCHO) to be sufficient strategies to reach elimination by 2025. The likelihood that elimination will be reached thus depends on pre-control endemicity (i.e. transmission intensity), duration and frequency of past MDA, and the strategy implemented from now until 2025. Biannual or quarterly MDA will accelerate progress toward elimination but it cannot be guaranteed by 2025 in high-endemicity areas. In such areas, MDA plus concomitant vector control would be useful³⁰.

4) Resilience to interventions: Both models incorporate microfilarial density-dependent establishment of O. volvulus L1 larvae within savannah species of S. damnosum sensu lato (i.e. S. damnosum sensu stricto / S. sirbanum)²⁸. However, only EPIONCHO assumes transmission intensity-dependent parasite establishment within humans^31,32. An individual-based, stochastic version of EPIONCHO (EPIONCHO-IBM) was developed to explore the interaction between individual heterogeneity in exposure to infection (via vector bites) and parasite establishment within humans (greater at lower transmission intensities and vice versa). Transmission intensity is measured by the annual rate of human exposure to infective (L3) larvae (no. L3 larvae/person/year). EPIONCHO-IBM was fitted (using Latin hypercube sampling) to matched (savannah) pre-intervention microfilarial prevalence, load and annual biting rate (ABR) data (extending Figure 1a range). Density dependence in parasite establishment within humans was estimated for different levels of (fixed) exposure heterogeneity (Figure 3)³¹. The interaction between overdispersion in exposure to blackfly bites (parameter k_E) and density-dependent parasite establishment stabilizes low (hypoendemic) prevalence, but accentuates resilience to MDA, explaining the lower EOT probabilities predicted by EPIONCHO.

Figure 3. Observed and predicted pre-intervention microfilarial prevalence and intensity vs. annual biting rates (ABRs).

EPIONCHO-IBM-predicted (solid lines) (a) microfilarial prevalence (percent) and (b) intensity (mean no. of microfilariae, mf, per mg of skin) for ABRs in the epidemiological dataset (solid color circles), using the estimated parameters. The overdispersion exposure parameter (of a gamma distribution) k_E was varied between 0.2 (stronger aggregation of blackfly bites among humans) and 0.4 (lesser aggregation). A value of k_E = 0.3 provided the best overall fit. Error bars are 95% confidence intervals (bootstrapped for intensity when raw data were available). Fitting data are from Cameroon [1]⁴⁰, [2]⁴¹, Burkina Faso/Côte d’Ivoire⁴²; validation data are from the Onchocerciasis Control Programme in West Africa (OCP)⁴³, Venezuela⁴⁴ and Ecuador⁴⁵ (the latter two for vectors with similar vector competence to S. damnosum sensu stricto). This figure has been reproduced from 31 under a Creative Commons Attribution 4.0 International (CC BY 4.0) license.

In view of (1)–(4) above, Box 1 summarizes the insights gained thus far and their implications for post-2020 goals.

Practical implications of the currently proposed EOT goals

Risks faced by treatment programs that need to be mitigated to achieve EOT

Some programs, particularly those in areas affected by conflict are more likely to under-perform, suffer from interrupted MDA schedules, or be delayed in starting and up-scaling implementation. Internal population displacement as well as mass migration to other countries could jeopardize progress towards EOT. Weak programs/health systems need to be identified, supported and strengthened⁶⁰. Movement of infected individuals (between endemic communities within countries, and between-countries) due to traditional cultural and trade patterns may lead to re-introduction of infection in areas achieving EOT. Risk of declining drug efficacy following multiple rounds of MDA/existence of sub-optimal responses to treatment should be monitored^59,64. Low coverage and systematic non-adherence may lead to persistent transmission, particularly (but not exclusively) in loiasis co-endemic areas⁶⁵. Factors associated with non-compliance need further investigation. Reported coverage levels are generally higher than true coverage, possibly because coverage has been a primary indicator of program effectiveness. Population denominators may not be correct due to lack of updated censuses. Reported coverages are aggregated by district, but disaggregate data are needed to evaluate and correct coverage heterogeneity.

Modelling priorities to support goals in the 2030 horizon and beyond

Table 2 outlines the priority modelling questions for further research that were elaborated in discussion with the WHO.

Data availability

No data are associated with this article.