How does the proportion of never treatment influence the success of mass drug administration programmes for the elimination of lymphatic filariasis?

Introduction

Lymphatic filariasis (LF) is a mosquito-borne neglected tropical disease (NTD) caused by three parasites, namely, Wuchereria bancrofti, Brugia malayi and Brugia timori [1]. LF can cause chronic morbidity, such as hydrocele or lymphedema which are associated with disability, pain, mental health problems, reduced productivity, and social stigmatisation [2–4]. In 2000, the World Health Organization (WHO) established the Global Programme to Eliminate Lymphatic Filariasis (GPELF) with the target of eliminating the disease as a public health problem (EPHP) [5]. The two key goals of the programme are (i) interruption of transmission by using community-wide mass drug administration (MDA) for at least 5 years using a two-drug combination (ivermectin + albendazole [IA] in onchocerciasis co-endemic areas in Africa, and diethylcarbamazine citrate + albendazole [DA] elsewhere), and (ii) to reduce the suffering of patients by managing morbidity and preventing disability. In areas where W. bancrofti is endemic and Anopheles and/or Culex are the principal vectors, the first goal is considered to be met when the level of infection is reduced to less than 1% microfilaraemia (mf) prevalence in the population aged 5 years and above. The achievement of this goal is measured through a series of transmission assessment surveys (TAS) [6].

Great progress has been made towards the WHO target. By 2019, more than 8.6 billion treatments had been successfully distributed resulting in a 74% reduction in the number of individuals infected with LF [7]. WHO aims to validate elimination as a public health problem in 81% of endemic countries by 2030 [1]. To accelerate progress towards this goal in areas lagging behind, the WHO has recommended to use a combination of all three drugs, known as the triple drug (IDA) therapy in eligible settings [8,9]. Due to the risk of adverse events, this combination is not recommended in those African countries with LF-onchocerciasis or LF-loiasis co-endemic areas [8,9].

LF elimination programmes have been successful in some, but not all areas. An important determinant of reaching the 1% mf prevalence threshold is the population coverage of MDA programmes, which needs to be sufficiently high and is recommended to be at least 65% of the total population [5]. However, in addition to population coverage, prospects of achieving <1% mf prevalence have been recognised to depend on patterns regarding who does/does not take treatment in MDA programmes [10]. Especially in settings with persistent transmission after many rounds of MDA, the question arises as to whether there are groups of individuals who are sustaining transmission due to repeatedly missing treatment [15,16]. There are many reasons why someone may never take treatment, including intentional factors (i.e., refusal, non-attendance, failure to ingest, fear of side-effects) and unintentional factors (i.e., out of the village at the time of MDA, treatment not offered, not eligible for treatment) [11–14]. The term ‘never treated’ (NT) has been proposed to capture all causes for never treatment, irrespective of the reason or intentionality and refers to individuals who have never been treated across consecutive treatment rounds [15,16]. Other terms related to people not being treated (e.g., systematic non-compliance, non-participation, non-attendance) may refer to specific causes for non-treatment and those terms are not used in this work.

The achievement of the WHO 2030 goals may be hampered if too many people remain never treated [10,21,26,27]. Therefore, quantifying NT levels (and ultimately, understanding the reasons behind them so they can be minimised) is critically important. The programmatic implications of a particular level of NT are currently unclear. In this work, we provide modelling insight into the impact of NT on the likelihood that LF programmes achieve the 1% mf prevalence threshold in epidemiological settings where Anopheles or Culex mosquitoes are the main vectors.

Methods

We use two individual-based stochastic models, namely, TRANSFIL and LYMFASIM to simulate the impact of NT on the probability of reaching the <1% mf prevalence threshold. Details of the two transmission dynamics models and their parameterization have been published previously [17–26,28].

Models for never treatment patterns in MDA

There are many approaches to modelling patterns regarding the proportion of the population never treated [10,27]. In LYMFASIM, NT is the result of an input parameter for the proportion of people who will ultimately never be treated (dashed horizontal line in Figure 1) and the overall population coverage of MDA. Patterns in repeated (non-)treatment of eligible individuals are assumed to be the result of a mix of systematic and random factors, which is achieved by assigning simulated individuals a trait for their inclination to participate in MDA. This means that the proportion of eligible people that has never been treated at some time point is higher than expected under random treatment. However, the level of NT among eligibles is directly tied to the overall coverage level (such that for increasing MDA coverage there is a limit to the NT values that can be simulated). Therefore, in LYMFASIM, we simulate different levels of NT for a given MDA coverage by changing the parameter for the proportion of people who will never be treated.

• Download figure
• Open in new tab

Figure 1:

Proportion of the all-time eligible population (y-axis) that has been treated in N MDA rounds (grey colour) by the X-th round (x-axis) as predicted by the model for MDA participation implemented in LYMFASIM, assuming an annual coverage of 65% of the eligible population, and assuming that 5% will never participate in the long run (dashed horizontal line). The never treated proportion (black) is highest in the first round and declines with the number of treatment rounds, approaching the proportion that will never participate. The all-time eligible population is the subgroup of the population who were eligible for treatment during all X treatment rounds.

In the version of TRANSFIL used here, patterns of repeated (non-)treatment are modelled based on Griffin et al. [29] (as reformulated by Dyson et al. [10]). The model contains a parameter, p, which controls the correlation of individuals attending treatment in different rounds. This parameter governs the relative contribution of random and systematic factors to the probability of an individual being treated across consecutive treatment rounds (p = 0 corresponds to completely random and independent probability of attendance in each round; p = 1 corresponds to completely systematic). For TRANSFIL, we simulate different levels of NT by varying the parameter p. In addition to the above, both models consider age-dependent eligibility for treatment.

NT is dynamic over time as it depends on how many rounds have been administered (illustrated in Figure 1). In the current study, we define NT as the proportion of people who are never treated after 5 rounds of MDA among individuals who are eligible (based on their age and health status) for treatment during each of those MDA rounds. We quantify NT after 5 rounds of MDA, as this is the time-point at which most LF programmes evaluate infection prevalence and programme success, and coverage surveys are likely to be implemented. This definition excludes individuals that were or became eligible for treatment during those MDA rounds. The proportion never treated will be higher in young children who were ineligible during all or some of the 5 MDA rounds.

Other than age-dependent eligibility for treatment, no age/sex-specific variation in treatment probabilities (e.g., related to work and mobility or pregnancy status) is considered in the current study. We also assume that ‘drugs received’ are ‘drugs swallowed’ (e.g., there is no coverage-compliance gap). Both models assume no association between treatment and exposure to infection.

Results

In Table 2, and Supplementary Material Tables S1⍰S4 we present the number of annual MDA rounds, as a range over both models, required to achieve the 1% mf prevalence threshold with 90% probability for different treatment regimens, MDA coverage, NT, and baseline endemicity levels in treatment-naïve settings. For each setting, we find that: (i) the higher the value of NT (going down the rows in each panel), the higher the number of MDA rounds required for the programme to achieve the target; (ii) the higher the coverage, the fewer the number of rounds required to achieve the target (going across the columns in each panel), and (iii) the higher the baseline mf prevalence (going down the panels), the lower the value of NT tolerable by the programme to achieve the target.

Discussion

Our findings indicate that the level of NT above which the 1% mf prevalence threshold cannot be achieved depends on the baseline endemicity, the employed drug combination, and the MDA coverage levels. The MDA coverage needed to achieve <1% mf prevalence depends on transmission setting, baseline endemicity level, drug combination and the level of NT. In Africa, after five rounds of MDA, the proportion of NT among all-time eligible individuals should not exceed 10% in low endemic settings or 1% in high-endemic settings. The use of more effective drug combinations (DA and IDA compared to IA) can bring these NT thresholds up to some extent, although these cannot be used everywhere (e.g., in areas co-endemic for onchocerciasis or loiasis). In India and other culicine transmission settings, after five rounds of annual MDA at 65% coverage, the proportion of NT should not exceed 20% in low endemic areas. A higher coverage level leads to more optimistic outcomes, in which any NT level can be permitted to achieve <1% mf prevalence. For higher endemic areas, we obtain similar NT thresholds as for the Africa-like settings with anopheline transmission when using the same treatment regimen, coverage level and for a baseline mf prevalence of 10%.

It should be noted that these outcomes are highly dependent on the assumptions that (i) NT occurs completely at random (not clustered in households or sub-communities) and (ii) NT is independent of exposure/infection status. In reality, the first assumption may not necessarily apply: never treated individuals may be clustered geographically, leading to hotspots of ongoing transmission and a larger negative impact of never treatment on required treatment duration.

Regarding the second assumption, the potential (positive or negative) correlation between an individual risk of infection acquisition (e.g., via exposure to mosquito bites) and the probability that an individual has never been treated can influence the impact of NT on achieving the TAS epidemiological thresholds. A positive correlation, where a higher bite-risk corresponds to a higher NT value, decreases the impact of MDA in achieving <1% mf prevalence and would make our results more pessimistic. This is particularly important in settings for which our models predict that the 1% mf prevalence threshold can be achieved with relatively low MDA coverage and high NT values (i.e., in India settings, treating with DA at 65% coverage and 20%NT), which in case of positive correlation, would allow for relatively large reservoirs of infection in untreated people. A negative correlation, where a higher bite-risk corresponds to a low NT, can increase the probability of elimination, and impact the time in achieving the 1% mf prevalence threshold.

Pragmatically, a coverage-NT combination may be acceptable when such a combination permits reaching the 1% mf prevalence threshold, within 10 years of MDA, even though a lower value of NT could have led to achieving the target faster. For example, we can achieve the 1% mf prevalence threshold within 8 years when the annual MDA coverage is 80% and NT is 20% (Table S3-S4). However, we can achieve this target twice as fast when the annual MDA coverage is 80% and NT is 1%. In both scenarios, we use the same number of MDA tablets per round, but over the total programme duration (to reach the 1% mf prevalence threshold) many more tablets are needed when NT is high. Therefore, in order to optimize/prioritize drug use, given limited resources, we need to minimize NT.

Our results can inform policy makers on optimal treatment strategies and show the importance of quantifying the level of NT in a community/implementation unit (and ultimately understanding the reasons behind NT). However, in practice, it might be challenging to quantify NT levels, as it is difficult to identify individuals who have never been treated without implementing longitudinal surveys that record the treatment-related behaviour of each individual during any round of MDA. Although many LF studies have measured the level of NT [30–32], very few longitudinal studies of never treatment have been conducted [33]. Furthermore, most of these studies have measured the level of NT retrospectively and not at each round of MDA, mostly due to financial and logistical constraints. In order to properly model the impact of NT, longitudinal cohort studies with information on who is treated (disaggregated by age, gender, occupation, education) and when, are essential.

It should be noted that there are other factors, not included in this analysis, which can influence the impact of NT on achieving elimination. One of these factors is age/gender-variation in coverage and NT, which depends on age-related patterns of exposure and contribution to transmission and can increase the number of years required to achieve elimination. Here we have explored the impact of MDA as a standalone intervention upon achieving the 1% mf prevalence threshold for a range of NT values. Consideration of other interventions, such as the addition of vector control could alter the results presented here by reducing the vector/human ratio (a component of the vector biting rate, which is a key determinant of the basic reproduction ratio (R₀) of the infection and hence, of baseline endemicity) and the vector biting rate more generally, helping to decrease the number of years required to achieve the 1% mf prevalence threshold [21].

Human movement (migration) can also play an important role, particularly as elimination is approached, which can either reduce the probability of elimination by adding a source of infective material acting as an external reservoir or increase the probability of elimination by reducing the prevalence of infection in that area [34,35]. This impact depends on human demographic and sociological factors of the area.

Authors’ contributions

K.K.: conceptualization, formal analysis, investigation, methodology, visualization, project administration, writing—original draft, writing—review and editing; W.A.S.: conceptualization, formal analysis, investigation, methodology, visualization, writing—original draft, writing—review and editing; M.G.B.: conceptualization, methodology, writing—original draft, writing—review and editing; B.S.C.: conceptualization, software, writing—review and editing; S.J.D.V., P.D., K.G.: conceptualization, methodology, writing—review and editing; M.G: conceptualization, software, writing—review and editing; T.D.H., J.D.K., A.K.: conceptualization, methodology, writing—review and editing; R.M.A.: conceptualization, methodology, visualization, supervision, writing—review and editing. L.E.C.: conceptualization, formal analysis, investigation, methodology, visualization, writing—original draft, writing—review and editing.

Funding

The authors are grateful for funding by the Bill & Melinda Gates Foundation through the NTD Modelling Consortium (grant INV-030046). This supplement is sponsored by funding of Professor T. Deirdre Hollingsworth’s research by Li Ka Shing Foundation at the Big Data Institute, Li Ka Shung Centre for Health Information and Discovery, University of Oxford and funding of the NTD Modelling Consortium by the Bill & Melinda Gates Foundation (INV-030046). K.K., M.G.B., B.S.C., and R.M.A. also acknowledge funding from the MRC Centre for Global Infectious Disease Analysis (MR/R015600/1), jointly funded by the UK Medical Research Council (MRC) and the UK Foreign, Commonwealth & Development Office (FCDO), under the MRC/FCDO Concordat agreement and is also part of the EDCTP2 programme supported by the European Union.