Keywords
Mathematical, Statistical, Targets, Public Health, Elimination, Transmission, Cross-cutting, NTD
This article is included in the 2030 goals for neglected tropical diseases collection.
Mathematical, Statistical, Targets, Public Health, Elimination, Transmission, Cross-cutting, NTD
The views expressed in this article are those of the authors. Publication in Gates Open Research does not imply endorsement by the Gates Foundation.
The World Health Organization’s (WHO) 2021-2030 Neglected Tropical Disease (NTD) Roadmap was launched on January 28th, 2021, renewing the commitment of the global NTD community to end the suffering caused by these diseases1. The development of the roadmap was guided by extensive global stakeholder consultation, including consultation with mathematical and statistical modellers. Modellers were asked to assess the technical feasibility of proposed goals, to identify major challenges for achieving the new goals from a transmission dynamics perspective, possible acceleration strategies, and key outstanding research questions2. Technical commentaries have been published as a collection in Gates Open Research3–15, which detail these insights for 13 NTDs: Chagas disease, dengue, gambiense human African trypanosomiasis (gHAT), lymphatic filariasis (LF), onchocerciasis, rabies, scabies, schistosomiasis, soil transmitted helminthiases (STH), Taenia solium taeniasis/ cysticercosis, trachoma, visceral leishmaniasis (VL) and yaws.
Neglected tropical diseases continue to affect over one billion people16 as the result of the considerable inequalities in global healthcare systems that fail to support those most in need17. The burden of NTDs falls largely on the poorest communities, resulting in an unrelenting cycle of poverty that is driven by negative social, health and economic impacts of infection on individuals and families, augmenting existing social divides. For infections with a substantial zoonotic component, morbidity and mortality among livestock also affect people’s livelihood with economic impacts that transcend medical implications. Notable progress to reduce the burden of NTDs has been made as a result of the commitments made in 2012 through the WHO 2020 NTD Roadmap18 and the London Declaration on NTDs19. As a result, 500 million people no longer require interventions against several NTDs and 40 countries, territories and areas have eliminated at least one disease1. These wins are the outcome of concerted and consolidated efforts from endemic communities and invaluable volunteers, governments, donor agencies and the pharmaceutical industry. Despite such early gains, reaching the endgame presents some of the greatest challenges – namely sustaining those early gains whilst identifying and averting small numbers of sparsely distributed cases. The 2030 roadmap is shaped around three pillars that aim to support global efforts to maintain the gains, address the challenges and ultimately combat NTDs1: 1. Accelerating programmatic action. 2. Intensifying cross-cutting approaches and 3. Shifting operating models and culture to facilitate in country ownership.
The use of mathematical and statistical modelling in NTD research and policy has until recently, and with a few exceptions (e.g., onchocerciasis20), lagged behind other groups of infectious diseases that receive more focus and funding (often, diseases that impact wealthier individuals and nations, or those perceived to potentially impact these). However, this is changing with the advent of groups like the NTD Modelling Consortium21, who have developed the Policy-Relevant Items for Reporting Models in Epidemiology of Neglected Tropical Diseases (PRIME-NTD) principles, as a guide to communicate the quality and relevance of modelling to stakeholders20. This has added clout to the call for modelling in the policy arena as well as setting a high bar of best practice for the wider modelling community. Having now gained significant traction, the use of modelling in NTD policy has contributed to new intervention tools22, vector control strategies23–26, shaped policy responding to COVID-19-related programme disruptions27–35 and has aided in the development of WHO guidelines36,37. For this positive relationship to continue, it is imperative to invest in a mutual understanding through ongoing conversation between policy-makers and modellers, to determine what kind of questions are the “right” questions, how to interpret uncertainty and what the models can and cannot be used for.
This piece introduces a collection of papers borne of a meeting in Geneva, in April 2019 attended, among others, by the NTD Modelling Consortium and convened by the WHO: Achieving NTD control, Elimination and Eradication Targets Post-2020; Modelling Perspectives & Priorities2. As new management targets and strategies took shape, the meeting provided policy makers and modellers the space to ask and answer specific questions regarding the proposed 2030 goals and the intended strategies to achieve them. Although the roadmap covers a range of diseases with diverse epidemiologies and differing management recommendations, the priority questions identified by modelers and stakeholders during the 2019 meeting and echoed by the authors of the technical commentaries shared three similar themes that should be considered in NTD modelling moving forward: timelines, programme design, and clinical study design.
Goals are only worth setting in the context of time. It is therefore not surprising that many of the technical commentaries in this collection identified timelines as a priority issue. The public health and economic benefits of reaching goals are innumerable but can only be achieved by the target year through appropriate mobilisation of diverse resources. Modelling in the forms of past inference and forward projections can align many moving parts (for example epidemiological, demographic, and social considerations) to inform our understanding of the reasons why programmes succeed or fail38,39. Forecasts have played a crucial role in understanding whether the 202040 and associated collection41,42, 202543 and 20303–15,34 goals can be reached under current strategies with the caveat that long-term predictions naturally become more uncertain.
In some instances, whether a goal can or will be met on time is relatively easy to ascertain – for example it is a resounding no for leprosy and rabies, which are hindered by passive case control, long quiescent incubation periods, and inadequate investment in interventions15,44. Alternatively, the goals for schistosomiasis11, STH8, and onchocerciasis13 seem achievable in some or most settings, depending on localised parameters like baseline prevalence, and already experienced duration of and adherence to mass drug administration (MDA) programmes. In the case of T. solium, a lack of internationally agreed goals for elimination or control curtails the ability to effectively model timelines; for example, the 2021-2030 NTD roadmap proposes the overall milestone of achieving “intensified control in hyperendemic areas”, without agreeing on technical definitions for T. solium endemicity levels, or defining measurable criteria for attaining “intensified” control14.
The diseases considered by the London Declaration and WHO roadmaps are at differing stages in their trajectories. Whilst some are on the cusp of achieving their goals, others face political and epidemiological barriers to progress. Both scenarios raise several priority questions regarding programme design, where ‘programme’ can mean intervention or surveillance. In addition to determining success or failure within the defined intervention time frames, modelling has provided insights into key factors of operational design like the treatment coverage necessary to reach goals in a given setting. Where it may not be possible, models can be used to test the efficacy of separate and combined chemotherapeutic37 and non-pharmaceutical interventions23,45,46, including combined interventions that target multi-host systems for zoonotic NTDs14. Additionally, deciding the optimal timing47 or frequency48,49 of treatment, and knowing who to treat50,51 are essential to the success of all interventions. Of course, the intervention strategies most likely to lead to achievement of the goals may not be sustainable in terms of cost to individuals, governments, or donors. By partnering highly detailed transmission models with cost-effectiveness analysis, modelling can also contribute to tailored insights regarding the affordability and benefits versus costs of interventions52–62. Models can also be used to explore integration between NTD programmes, or to understand the potential cross-utility of existing NTD programmes on other helminth species, such as exploring the additional benefit of national schistosomiasis control programmes using praziquantel on T. solium prevalence in co-endemic areas14. Understanding this cross-utility is vital to intensifying cross-cutting approaches – one of the three core pillars of the roadmap, that differentiates the framework from its predecessor.
These are all very practical features of intervention programmes that can in principle be planned for, but underlying features of target populations and human nature can undermine these plans. Survey data in recent years have made it evident that whilst the aim may be to deliver treatment at a certain geographical and therapeutic coverage, it is not analogous with consumption, as treatment is systematically not ingested by some63,64, or is not disseminated to the full intended group, reducing the true coverage. There are a variety of reasons for this65,66, but it is likely that similar mechanisms impact participation in surveillance, therefore biasing the estimates of prevalence, particularly when treatment and surveillance are co-occurring (e.g., gHAT9,67, rabies15,68). Modelling shows that the impact of this variable true coverage depends on the pathogen in question and transmission intensity64,69–71 but it undoubtedly has an impact on reaching public health goals72,73.
Once a strategy has been deemed effective and prevalence targets are attained, it is likely that these interventions either transition, such as going from MDA to identified case management, or they stop all together. Establishing robust surveillance strategies at this point is vital, but obviously not everyone can be regularly sampled and not every incident infection case will be detected. Stochastic events like reinfection and reintroduction are risks that can drive resurgence. Modelling can support the identification of the optimal surveillance strategy and determine which prevalence or intensity indicators need to be monitored to ensure the desired public health goal74–76, although challenges remain in developing long-term strategies77. Modelling can make useful contributions in developing sustainable, effective interventions and surveillance strategies and should therefore be included in any programmatic design from the start. As embodied by the 2021-2030 NTD roadmap, impactful interventions cannot be achieved by working in silos, but instead require continuous communication between all parties of an interdisciplinary team.
Though modelling is increasingly used in public health decision making, the use of modelling to direct clinical trial design and drug development is not so common, and even less so for NTDs. To reach goals like elimination as a public health programme (trachoma, STH, schistosomiasis and LF) and elimination of transmission (onchocerciasis) novel drug development will be critical1,4,5,11,13,78. However, financial returns on investments into NTDs are limited and therefore largely unappealing, particularly because of the heavy reliance by endemic nations, on donations from pharmaceutical producers. Increased use of mathematical modelling could reduce the financial waste associated with the drug-development-to-distribution-pipeline78. If we consider this pipeline in three parts; pre-clinical, clinical trial and distribution, it is clear that modelling can provide valuable insight at each stage. Onchocerciasis and LF have recently benefited from pharmacokinetic-pharmacodynamics modelling, translating pre-clinical non-human experimental results into quantitative insights relevant to human treatment79. Clinical trial simulations are designed to include all aspects of a clinical trial protocol including (but not limited to) recruitment criteria, drug properties/ effectiveness and follow-up times80, providing valuable guidance that translates into more effective, efficient, cost-efficient and robust clinical trials. In addition to providing insight into the optimal distribution of new drugs81, rethinking the distribution of existing drugs to achieve public health targets can also be guided by modelling37,48.
Modelling has certainly addressed many of the key questions asked of modellers at the 2019 meeting2. However, cross-disease challenges remain82. The most common of these, highlighted by all groups involved in the meeting report2 and this collection, is undoubtedly a lack of data or poor data quality. This could be because certain parameters simply cannot be measured; because of vast heterogeneity or because they have yet to be collected83. For example, VL has a highly variable incubation period, unknown duration of asymptomatic infection and estimates for the duration of lasting immunity are ill-defined6,84,85, introducing uncertainty into the temporal dynamics underlying any projections. Chagas disease, gHAT and leprosy also suffer from indeterminate incubation periods9,12,21 impacting case detection and adding greater uncertainty in epidemiological estimates fitted to by models85,86. Asymptomatic or pre-symptomatic infection is common of many NTDs and presents a significant challenge to their management. For example, asymptomatic VL infections cannot be treated, whereas it is possible to teat asymptomatic gHAT but only if it is able to be detected. Identifying their respective proportions in an infected population, particularly in the absence of high surveillance coverage, means accounting for this group using roundabout methods and proxy diagnostics6,9.
Many diagnostics are indirect, proxy measures of case detection, often with less than perfect sensitivity or specificity87,88, and have a direct effect on perceived prevalence and individual burdens of infection89,90. Given that models are only as good as the data to which they are fitted, this has a significant impact on the utility of model results. For example, in the instances of STH and intestinal schistosomiasis (Schistosoma mansoni), WHO targets are given in terms of eggs per gram of faecal matter as detected with the Kato-Katz method, which notoriously suffers from poor sensitivity, particularly for low intensity infections91, invariably underestimating prevalence. Where a multi-host system is present for zoonotic NTDs, though it is possible to measure infection through direct observation of parasite stages in the animal host(s)14, via necropsy or other methods92, it is likely that this approach is inappropriate for monitoring and evaluating the likes of T. solium control programmes, due to the large animal sample sizes required to detect a statistically meaningful impact on transmission, especially in low prevalence settings14. Molecular xenomonitoring (testing vectors for the parasite instead of human hosts) for LF and onchocerciasis has shown promise93 but operational research gaps remain, impacting large-scale utilisation94. Reconciling these different streams of imperfect diagnostic data will be key to their utility in modelling and indeed to reaching and sustaining public health goals.
The operational units over which epidemiological data are collected, and projections made are also often over somewhat arbitrary administrative borders that infectious diseases do not adhere to. For rabies, non-spatial models are inadequate for capturing the low-endemicity incidence rates15 such that more data-intensive modelling approaches are required. In addition to questionable detection success, VL surveillance has operated over geographical units that are too large to evaluate the success of control methods6, despite modelling showing that transmission is highly localised over smaller spatial scales (i.e. 85% of inferred transmission distances ≤300m)95. Similarly for onchocerciasis, modelling shows that the rate at which interventions can be scaled down depend strongly on the spatial units of assessment13,96. Clustering of T. solium porcine cysticercosis around human taeniasis carriers, particularly evident in South American communities, demonstrates the need for spatially explicit models in certain settings14,97, such as the recently developed CystiAgent model for Peru98, capable of testing spatially structured interventions. From this it is evident that whilst spatial heterogeneity requires nuanced model structure, the leading challenge here is the paucity of data at the spatial level necessary to parameterise the models for spatially relevant insights. This will become ever more important as all NTDs move towards low-prevalence and spatially-heterogenous incidence patterns.
The assumptions made to overcome these uncertainties often differ across models – which then produce differing results. Whilst this is somewhat overcome by the practice of model comparison99,100, which highlights important biological and population processes that impact epidemiological trajectories, these unknowns wave a clear flag for collaborative opportunities between modellers, field epidemiologists and clinicians. Indeed, the optimal working relationship is a synergistic pathway, where the model’s needs drive data collection, the data shapes further model iterations, and these then inform policy and the outcomes at the programmatic and clinical level51,82,83,85,101–109. Improving communication between these groups is critical to achieving the desired public health gains20.
The increased use of mathematical and statistical modelling over the last decade has helped move the field of NTDs into a more quantitative space, providing the link between epidemiological concepts and observed reality. For modelling to continue to fill this role and influence decision-making, ongoing conversations and engagement between all parties will be paramount. These will, in turn, overcome the continuous challenges of data quality and access, and the consequent model assumptions required. As programme and disease management move towards a country-ownership framework under the new roadmap, it will be key that modelling follows suit, overcoming systematic notions of knowledge ownership and challenging associated power dynamics110–112. In this way, future modelling will work to support this new NTD landscape.
No data are associated with this article.
We would like to acknowledge all of those who contributed to the pieces within this collection. We list them here in alphabetical order; Fernando Abad-Franch30,42, Bernadette Abela-Ridder23, Emily Adams36, Maryam Aliee38,55, Marina Antillon16,51, Benjamin F. Arnold1, Robin L. Bailey8, Seth Blumberg1, Anna Borlasse2, Uffe C Braae18,43, María Soledad Castaño16,51, Joel Changalucha27, Nakul Chitnis16,51, Sarah Cleaveland33, Ronald E Crump38,55, Derek A.T. Cummings53, Christopher Davis38,55, Emma L. Davis2,55, Michael Deiner1, Brecht Devleesschauwer15,26, Andy P Dobson14, Daniel Engelman40, Neil Ferguson31, Claudio Fronterre7, Sarah Gabriël26, Federica Giardina21, William Godwin1, Sébastien Gourbière50, Jonathan I D Hamley37,39, Wendy E Harrison45, Alex Holmes38, Ching-I Huang38,55, Maria V Johansen24, John Kaldor35, Matt J. Keeling38,46,55, Charles H. King3, Lea Knopf23, Periklis Kontoroupis21, Epke A. Le Rutte21, Justin Lessler34, Michael Z Levy13, Thomas M. Lietman1, Kennedy Lushasi27,33, Mary Veronica Malizia21, Jodie McVernon40,49, Edwin Michael11, Philip Milton37,39, Elizabeth Miranda28, Eric Q. Mooring17, Pierre Nouvellet32, David Pigott54, Travis C. Porco1, Jorge E. Rabinovich9, Sylvia Ramiandrasoa47, Isabel Rodriguez-Barraquer52, Kristyna Rysava55, Veronika Schmidt6,19, Morgan E. Smith11, Andrew Steer40, Michelle Stanton7, Fabrizio Tediosi16,51, Tenzin Tenzin41, S. M. Thumbi4,44, Michael Tildesley55, Panayiota Touloupou22, Chiara Trevisan12, Caroline Trotter25, Inge Van Damme26, Carolin Vegvari37, Juan-Carlos Villar10,29, Sake J. de Vlas21, Martin Walker20,37, Ryan Wallace5 Andrea S. Winkler6,19 & Peter Winskill37.
Views | Downloads | |
---|---|---|
Gates Open Research | - | - |
PubMed Central
Data from PMC are received and updated monthly.
|
- | - |
Is the rationale for the Open Letter provided in sufficient detail?
Yes
Does the article adequately reference differing views and opinions?
Partly
Are all factual statements correct, and are statements and arguments made adequately supported by citations?
Partly
Is the Open Letter written in accessible language?
Yes
Where applicable, are recommendations and next steps explained clearly for others to follow?
Not applicable
References
1. Jones-Hepler B, Moran K, Griffin J, McClure E, et al.: Maternal and Neonatal Directed Assessment of Technologies (MANDATE): Methods and Assumptions for a Predictive Model for Maternal, Fetal, and Neonatal Mortality Interventions. Global Health: Science and Practice. 2017; 5 (4): 571-580 Publisher Full TextCompeting Interests: No competing interests were disclosed.
Reviewer Expertise: PC NTDs, implementation science, design and evaluation of infectious disease control programs
Is the rationale for the Open Letter provided in sufficient detail?
Yes
Does the article adequately reference differing views and opinions?
Partly
Are all factual statements correct, and are statements and arguments made adequately supported by citations?
Partly
Is the Open Letter written in accessible language?
Yes
Where applicable, are recommendations and next steps explained clearly for others to follow?
Yes
References
1. NTD Modelling Consortium Lymphatic Filariasis Group: The roadmap towards elimination of lymphatic filariasis by 2030: insights from quantitative and mathematical modelling.Gates Open Res. 2019; 3: 1538 PubMed Abstract | Publisher Full TextCompeting Interests: No competing interests were disclosed.
Reviewer Expertise: Infectious disease modelling broadly, with some experience with LF but not the other NTDs covered by this letter.
Alongside their report, reviewers assign a status to the article:
Invited Reviewers | ||
---|---|---|
1 | 2 | |
Version 2 (revision) 02 Feb 22 |
read | |
Version 1 27 Jul 21 |
read | read |
Provide sufficient details of any financial or non-financial competing interests to enable users to assess whether your comments might lead a reasonable person to question your impartiality. Consider the following examples, but note that this is not an exhaustive list:
Sign up for content alerts and receive a weekly or monthly email with all newly published articles
Register with Gates Open Research
Already registered? Sign in
If you are a previous or current Gates grant holder, sign up for information about developments, publishing and publications from Gates Open Research.
We'll keep you updated on any major new updates to Gates Open Research
The email address should be the one you originally registered with F1000.
You registered with F1000 via Google, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Google account password, please click here.
You registered with F1000 via Facebook, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Facebook account password, please click here.
If your email address is registered with us, we will email you instructions to reset your password.
If you think you should have received this email but it has not arrived, please check your spam filters and/or contact for further assistance.
Comments on this article Comments (0)