Assessing the effects of limited vaccine supply on risk-based vaccination strategies in regions of endemic foot-and-mouth disease

Assessing the effects of limited vaccine supply on risk-based vaccination strategies in regions of endemic foot-and-mouth disease | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Assessing the effects of limited vaccine supply on risk-based vaccination strategies in regions of endemic foot-and-mouth disease Glen Guyver-Fletcher, Mike Tildesley This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8778801/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Foot-and-mouth disease is an important livestock disease which is endemic in many regions of the world, many of which struggle to supply enough vaccine doses to cover the entire population of animals at risk. We use a previously fitted stochastic metapopulation model of FMD spread to simulate four plausible vaccine allocation strategies in the context of limited vaccine supply and endemic disease. We find that random allocation outperforms other strategies when the objective is to minimise average prevalence, and random allocation also achieves local eradication at lower doses per capita. However, all strategies perform similarly when minimising annual incidence, with the ‘optimal’ allocation strategy varying by doses available per capita. These results suggest that policymakers in regions of endemic FMD should focus more on sourcing high-quality vaccines and achieving high coverage, with risk-based vaccination being a secondary concern. FMD Epidemiology Vaccines Control Optimisation Simulation Policy Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Foot and Mouth disease (FMD) is one of the most infectious livestock diseases known, infecting a large variety of cloven-hoofed animal species, including major livestock animals such as cattle, buffalo, sheep, goats, and pigs [ 1 ]. The etiological agent is a virus, Foot-and-Mouth Disease Virus (FMDV), of the Apthovirus genus and Picornaviridae family. Symptoms vary by species, but common symptoms include lesions around the feet and mouth, fever, and lameness – in cattle reductions in milk production are also common [ 2 ]. The disease is endemic in most countries in Africa and Asia, and it has been estimated that visible production losses due to FMD and vaccination in endemic areas alone amount to between $ 6.5 and $ 21 billion USD annually (Knight-Jones & Rushton, 2013). Efforts to limit or eradicate the spread of FMD are common; historically control efforts have been successful in Europe and South America; however, such efforts relied on mass vaccination campaigns [ 4 , 5 ]. For many countries where the disease is currently endemic, procurement of efficacious vaccines matched to locally circulating variants remains problematic. Even if such vaccines are available, their expense may be prohibitive, leading to smaller numbers of doses being procured than are necessary to vaccinate all animals at risk (Hammond et al., 2021). Given that vaccine supply is limited in many settings, it is important to ensure that doses are allocated in the most effective way. This scenario - where the supply of efficacious vaccine doses is insufficient to meet demand – may have knock-on effects on the allocation of those doses; the optimal allocation when many doses are available may be different from the optimal allocation when few doses are available. Epidemiological modelling allows for the exploration of optimal vaccine dose allocation strategies. Previous work has generally focused on optimising vaccination policies against epidemics of FMD [ 7 – 9 ] although there has been limited work on simulating optimal control policies in endemic areas [ 10 , 11 ]. We therefore aim to use a previously developed and validated simulation model of FMD spread in endemic areas to investigate how the ‘optimal’ allocation of vaccine doses changes as the availability of doses changes. We focus on Turkey due to its combination of endemic FMD and high-quality data necessary for detailed simulation. Specifically, we wish to know whether: 1) The ‘optimal’ strategy changes at different dose availabilities; and 2) which strategies are most optimal, for each criterion investigated. Methods & Materials Data Data are available from Turkey 2001–2012, comprised of: i) Village point locations and cattle headcounts; ii) Village outbreak records, 2001–2012; iii) Village-to-village cattle movement records, 2007–2012.Owing to the detailed data and model and to keep simulation times reasonable, we limit our analysis to a single province, Erzurum. Model The model we use has been developed in the context of previous work on FMD in Turkey; for a more in-depth account of the model and parameters used please see our previously published article (Guyver-Fletcher et al., 2025). We use the joint posterior parameter distributions from that work to simulate our model here. In summary, the model is a stochastic metapopulation model which simulates intra- and inter-village FMDV transmission. A schematic representation of the main model aspects is illustrated in Fig. 1 . Cattle within a village can be in one of eight (8) disease-relevant states: Maternally immune (M), Susceptible (S), Exposed (E), Infectious (I), Recovered (R), Carrier (C), Vaccinated-Susceptible (V S ), and Vaccinated-Recovered (V R ). Rates of progression between these states is determined by Ordinary Differential Equations and implemented stochastically using the tau-leap algorithm. It is assumed that carrier animals do not transmit due to the lack of evidence for this in the field [ 12 ]. Disease transmission between villages is simulated through two routes: 1) direct shipments of cattle, with probability of transmission equalling the probability one of the animals moved is infected; 2) Local transmission, aggregating aerosol transmission and other unseen routes of transmission – this is estimated through a distance-dependent kernel function that scales the probability of transmission with the number of infectious animals in the source village and the number of susceptible animals in the “target” village. Vaccine Allocation Strategies We define δ v as doses per capita – the ratio of available vaccine doses to the total cattle population at each vaccination campaign. For example, a value of δ v = 0.5 indicates a ratio of 1 available vaccine dose per 2 animals. For simplicity, we consider only application of a single dose at a time. We simulate mass vaccination campaigns every 182 days (6 months). No other control policies are simulated. At campaign commencement, the number of available doses is calculated as \(\:{\delta\:}_{v}\times\:{N}_{t}\) , where N t is the total population of cattle in existence at time t . When simulating a vaccination campaign, the relevant epidemiological unit is the village to retain operational realism. We order each village by the relevant criteria defined by the allocation strategy outlined in Table 1 and then vaccinate in order, subtracting the number of used doses, until there are either no doses left or no villages remaining to vaccinate. “Random” allocation, the comparison allocation strategy, consists of vaccinating villages with no targeting by size, connectedness, or local density. We do not consider randomisation at the individual-animal level, as partial vaccination of herds is not operationally realistic for mass FMD vaccination campaigns. Table 1 Vaccine allocation strategies investigated. Allocation Strategy Ordering Criteria Random None – random Population Villages with the greatest cattle headcount are vaccinated first Degree Villages with the greatest number of recorded inward or outward cattle movements are vaccinated first Density Villages with the greatest number of other villages within 5km are vaccinated first Simulation of strategies To account for parameter uncertainty, we sample 100 particles from the posterior distribution of our model fit to the observed history of outbreaks in Erzurum province [ 11 ]. We simulate a no-control scenario with each particle, and 10 infections seeded at the start of the timeline, for 5 years to achieve a realistic endemic state. For each endemic scenario, we then simulate a 5-year period for each δ v in the sequence {0.0, 0.01, 0.02, …, 0.99}. Each combination of particle and δ v is simulated 10 times. Each strategy is therefore simulated 1,000 times per value of δ v (100 particles, simulated 10 times per value of δ v ). We simulate vaccine efficacy (VE) value drawn from a normal distribution centred on 71% [ 13 ], or centred on 90%; in both cases the standard deviation used was 5%. For each combination of δ v and allocation strategy we calculate the average annual village-level incidence and average prevalence over the simulated period, as well as the animal-level prevalence. We also calculate the probability of eradication as the proportion of simulations where eradication is achieved, and the time to eradication when or if eradication occurs is the first day when eradication is achieved, averaged over all simulations with the same δ v and strategy. Sensitivity Analysis To investigate the sensitivity of our outputs to certain control parameters, we conducted a global sensitivity analysis using LHS-PRCC. For each of the 4 strategies, we took 30 random samples from our posterior distribution and 80 samples from a Latin Hypercube generated with the joint control parameter distribution outlined in Table 2 . We then simulated with a dose ratio δ v of either 0.3 (low availability), 0.6 (medium availability), or 0.9 (high availability) to investigate how sensitivity to control parameters might differ in different dose availability regimes. Table 2 Marginal distributions of control parameters explored for the sensitivity analysis. Control Parameter Description Marginal Distribution Vaccine Time to Effect The delay after vaccination before an animal is protected Uniform (3–14) days Vaccine Efficacy What proportion of animals generate protective immunity after vaccination Uniform (1–100) percent Vaccine Duration The average duration of protective immunity generated by the vaccine Uniform (150–210) days Mass Vaccination Coverage The proportion of villages that are covered by the mass vaccination campaign Uniform (0–100) percent Mass Vaccination Interval The number of days between mass vaccination campaigns Uniform (120–240) days Once we had simulated each particle, we calculated our outputs of interest and calculated the Partial Rank Correlation coefficients (PRCC) using the ` epiR` package version 2.0.80 [ 14 ] in R 4.4.3 (R Core Team, 2025). Results Mean simulated prevalence declines with increasing availability of vaccine doses, as expected (Fig. 2 ). At the majority of simulated δ v values, the random allocation strategy leads to a lower mean simulated prevalence than using any other strategy, with all strategies converging as availability approaches 100%. Increased vaccine efficacy leads to a faster and larger decline. There is a small difference in observed trends between village- and animal-level prevalence (Fig. 3 ) – the magnitude of performance differences between strategies (in terms of animal prevalence) is much smaller than their performance differences in terms of village prevalence; however, the overall ranking of strategies remains similar across the range of simulated doses per capita. Figure 4 demonstrates a different ranking, with the population-based allocation strategy outperforming all others when δ v < 0.6, and random allocation performing best above δ v = 0.6. Mean Annual Incidence with all strategies is relatively flat or increasing until dose availability exceeds 0.75. Ranking of strategies is very similar between different vaccine efficacies. As dose availability approaches 1, mean annual incidence for population, degree, and random allocation strategies declines towards (without reaching) 0 when VE = 90; however, when VE = 71 the decline is much smaller. No strategy, with vaccine efficacy averaged at 71%, exhibited regular eradication of disease circulation, hence, probability of eradication is 0 and TTE is not applicable. The exception is a single simulation with the population-based strategy at δ v = 0.98, leading to a probability of eradication of 0.01. With VE at 90%, three strategies resulted in eradication at high dose availability ratios (Fig. 5 ) – Random, Population, and Degree. Random sometimes led to eradication at the lowest dose availability (0.88), and all strategies peaked at a probability of eradication of approximately 0.29. Between δ v = 0.95 and 1, population-based allocation led to a significantly higher probability of eradication than random allocation. Density-based allocation did not lead to eradication in any simulation. Figure 6 shows the PRC coefficients of the sensitivity analysis against mean annual incidence. For vaccine time to effect and vaccine duration, no coefficient was significantly different from 0 for any strategy or dose availability ratio. Vaccine efficacy was positively associated with incidence in every case, although the coefficient declined as dose availability increased. Coefficients for the mass vaccination interval were not significantly different from 0 at low dose availability, but positively associated with incidence at higher dose availability, although there appeared no difference by strategy. Mass vaccination coverage coefficients differed by dose availability and strategy: at low dose availability it was negatively associated with incidence for the population and degree-based strategies, but not significantly different from 0 for density-based allocation and slightly positively associated for random allocation; at medium dose availability no coefficients were significantly different from 0; at high dose availability all strategies exhibited a strong negative association with incidence. For probability of eradication, no eradication was seen at dose availability ratios below 0.88, so all coefficients for the low and medium availability scenarios were all 0 (Fig. 7 ). For the high availability scenario and for all strategies, vaccine efficacy and mass vaccination coverage were positively associated with eradication, mass vaccination interval negatively associated, and vaccine time to effect and duration not significantly different from 0. Discussion In summary, the `optimal` allocation strategy depends on the outcome being optimised for, the availability of vaccine doses, and the efficacy of the vaccine being used. Random allocation at the herd level outperforms all other strategies at all dose availability ratios when optimising for average prevalence, however, if optimising for minimising incidence at low dose availability a population-based strategy performs best. Eradication only occurs at very high dose availability with a highly effective vaccine, and the population-based allocation strategy maximises the probability of eradication. However, even with full coverage and a highly effective vaccine, accounting for parameter uncertainty the probability of eradication is still less than 0.3. When assessed against mean annual incidence, all strategies are sensitive to vaccine efficacy and mass vaccination coverage, and insensitive (within explored ranges) to vaccine time to protection and immunity duration. The results indicate that random allocation of doses is the most effective strategy for most metrics investigated. It is strictly better than any other explored allocation strategy in reducing average disease prevalence, is somewhat better at reducing incidence at medium-high dose availability and was able to achieve eradication in some simulations at dose availability ratios lower than any other strategy. This surprising result does not appear to be due to the population structure of the simulated province (many smallholdings can lead to a high village coverage with a small animal coverage), as it also holds for prevalence at the level of the animal. According to the results presented here, the second-best strategy is population-based dose allocation, which holds incidence lower than other strategies when few doses are available and maximises the probability of eradication at (very) high dose availability. Here we must discuss the unintuitive pattern seen in Fig. 3 , where mean annual incidence increases before decreasing – this statistic is due to greater variability in observed incidence in the simulations. As more farms are vaccinated with greater dose availability, but not enough to truly eradicate the disease, when immunity wanes the village may be reinfected and recorded as a new infection. Degree-based and density-based allocation strategies are less effective policies – they are neither best at reducing prevalence or incidence at any dose availability ratio. Degree-based (i.e. connectivity-based) allocation does, however, manage to achieve eradication when simulated with high dose availability and vaccine efficacy. The density-based strategy, somewhat surprisingly, never manages to achieve eradication even with enough doses for essentially the entire population. Close investigation revealed that, as there is little correlation between density and population in this landscape, there exist a set of medium-sized villages with approximately 6,000 cattle who would never be vaccinated using this strategy except at 100% coverage– they were too far from every other village and hence were deprioritised in our simulations. This result may not generalise to other landscapes, where density and population may be positively correlated enough to avoid this failure mode. All modelling results should be translated cautiously, and decisions must consider all available evidence and the unique characteristics of each country or region. However, these results suggest that in regions where the disease is currently endemic, and that wish to focus on control or eradication of FMD, it is better to first focus on achieving higher coverage or sourcing higher-quality vaccines than to attempt to ‘optimise’ where vaccines are allocated. If resources are available to optimise, however our results suggest that: If few vaccine doses are available, and the priority is to minimise incidence, vaccines should be allocated to the largest villages first. If the aim is to minimise average prevalence, random allocation should be used. At high numbers of doses, population-based allocation maximises the probability of eradication. Density-based allocation does not appear to offer any benefit over alternatives. One limitation of these results that must be borne in mind is that, due to time and resource constraints, we only consider allocation of doses in a mass vaccination scenario. Such campaigns are generally carried out with other reactive control measures, such as ring vaccination around identified farms or livestock movement restrictions. Previous work of ours has addressed this [ 11 ], however, we have not addressed the optimal allocation of doses between mass vaccination campaigns and reactive ring vaccination – we believe this would be interesting avenue for future research. There is also the question of generality. Although our data from Turkey are high quality, and our model has demonstrated good fit to those data, the applicability of the model and results to other settings is unclear. This should be considered before any policy decisions might be made. We have focused on simple allocation strategies that we believe are straightforward to implement, however, we have not considered all possible allocation strategies. One that we did not consider, which is sometimes used, was risk-based spatial zoning – i.e. vaccinating as many animals as possible in a defined geographic area identified as high risk. It would be interesting to compare that strategy to these results. Our results suggest that attempting to optimize vaccine allocation is, in many cases, counterproductive and leads to greater incidence and prevalence than using a simple random allocation strategy. They also caution against using density-based or degree-based allocation. Whilst our model could be extended to consider additional vaccine targeting strategies, we believe that this work can be a useful reminder that the problem of FMD can mostly be solved with enough high-quality vaccines applied at regular intervals. Declarations Ethics approval and consent to participate Not applicable Consent for publication Not applicable Availability of data and materials The datasets supporting the conclusions of this article are available in the Zenodo repository under the reserved DOI 10.5281/zenodo.18471687 and will be made publicly available upon publication. Competing interests The authors declare that they have no competing interests. Funding The authors would like to acknowledge the funding of this project by the 11 th Technical Call of the European Commission for the Control of Foot-and-Mouth Disease (EuFMD), reference 81579. This work was also supported by an Ecology and Evolution of Infectious Diseases joint National Science Foundation (NSF) and Biotechnology and Biological Sciences Research Council (BBSRC) grant (BB/X005224/1). Author’s contributions GGF made substantial contributions to the conception and design of the work, the analysis and interpretation of the data, the creation of software for the work, and drafted the work. MJT made substantial contributions to the conception and design of the work and substantively revised it. All authors read and approve the final manuscript. Acknowledgements The authors would like to acknowledge the contribution of the SAP Institute of the Republic of Turkey, and EuFMD, for the provision of agricultural and epidemiological data used in this work. References Arzt J, Juleff N, Zhang Z, Rodriguez LL. The Pathogenesis of Foot-and-Mouth Disease I: Viral Pathways in Cattle. Transboundary and Emerging Diseases. 2011;58:291–304. https://doi.org/10.1111/j.1865-1682.2011.01204.x. Grubman MJ, Baxt B. Foot-and-Mouth Disease. Clinical Microbiology Reviews. 2004;17:465–93. https://doi.org/10.1128/CMR.17.2.465-493.2004. Knight-Jones TJD, Rushton J. The economic impacts of foot and mouth disease – What are they, how big are they and where do they occur? Prev Vet Med. 2013;112:161–73. https://doi.org/10.1016/J.PREVETMED.2013.07.013. Leforban Y. Prevention measures against foot-and-mouth disease in Europe in recent years. Vaccine. 1999;17:1755–9. https://doi.org/10.1016/S0264-410X(98)00445-9. Rivera AM, Sanchez-Vazquez MJ, Pituco EM, Buzanovsky LP, Martini M, Cosivi O. Advances in the eradication of foot-and-mouth disease in South America: 2011–2020. Front Vet Sci. 2023;9:1024071. https://doi.org/10.3389/FVETS.2022.1024071/BIBTEX. Hammond JM, Maulidi B, Henning N. Targeted FMD Vaccines for Eastern Africa: The AgResults Foot and Mouth Disease Vaccine Challenge Project. Viruses 2021, Vol 13, Page 1830. 2021;13:1830. https://doi.org/10.3390/V13091830. Keeling MJ, Woolhouse MEJ, May RM, Davies G, Grenfellk BT. Modelling vaccination strategies against foot-and-mouth disease. 2003. Tildesley MJ, Savill NJ, Shaw DJ, Deardon R, Brooks SP, Woolhouse MEJ, et al. Optimal reactive vaccination strategies for a foot-and-mouth outbreak in the UK. Nature. 2006;440:83–6. https://doi.org/10.1038/NATURE04324;KWRD=SCIENCE. Roche SE, Garner MG, Sanson RL, Cook C, Birch C, Backer JA, et al. Evaluating vaccination strategies to control foot-and-mouth disease: a model comparison study. Epidemiol Infect. 2015;143:1256–75. https://doi.org/10.1017/S0950268814001927. Ringa N, Bauch CT. Impacts of constrained culling and vaccination on control of foot and mouth disease in near-endemic settings: A pair approximation model. Epidemics. 2014;9:18–30. https://doi.org/10.1016/J.EPIDEM.2014.09.008. Guyver-Fletcher G, Gorsich EE, Jewell C, Tildesley MJ. Controlling endemic foot-and-mouth disease: Vaccination is more important than movement bans. A simulation study in the Republic of Turkey. Infect Dis Model. 2025;10:702–15. https://doi.org/10.1016/J.IDM.2025.02.006. Stenfeldt C, Arzt J. The carrier conundrum; a review of recent advances and persistent gaps regarding the carrier state of foot-and-mouth disease virus. Pathogens. 2020;9. https://doi.org/10.3390/PATHOGENS9030167. Knight-Jones TJD, Bulut AN, Gubbins S, Stärk KDC, Pfeiffer DU, Sumption KJ, et al. Retrospective evaluation of foot-and-mouth disease vaccine effectiveness in Turkey. Vaccine. 2014;32:1848–55. https://doi.org/10.1016/J.VACCINE.2014.01.071. Stevenson M, Sergeant E. epiR: Tools for the Analysis of Epidemiological Data. 2025. Additional Declarations No competing interests reported. Supplementary Files tc11supplementary.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8778801","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":606112763,"identity":"79f4ffd4-507d-4ddc-bb57-8b9f3ef6cfbc","order_by":0,"name":"Glen Guyver-Fletcher","email":"data:image/png;base64,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","orcid":"","institution":"University of Warwick","correspondingAuthor":true,"prefix":"","firstName":"Glen","middleName":"","lastName":"Guyver-Fletcher","suffix":""},{"id":606112764,"identity":"1774b2ae-9a5a-4f84-a254-6b7b079d5a76","order_by":1,"name":"Mike Tildesley","email":"","orcid":"","institution":"University of Warwick","correspondingAuthor":false,"prefix":"","firstName":"Mike","middleName":"","lastName":"Tildesley","suffix":""}],"badges":[],"createdAt":"2026-02-03 17:23:28","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8778801/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8778801/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104783098,"identity":"b7bcc458-e6fd-4d50-8109-3289eacbbbbd","added_by":"auto","created_at":"2026-03-17 07:58:13","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":252482,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSchematic representation of intra- and inter-village transmission. \u003c/strong\u003eCattle proceed through disease states within a farm/village at certain rates, defined by Ordinary Differential Equations (ODEs). Transmission between villages is simulated through two different routes: local transmission in the form of a distance-dependent kernel, and explicit simulation of cattle shipments.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8778801/v1/b3d49c023a774199160eed1f.png"},{"id":104754561,"identity":"a1af426e-a6cb-48b9-90df-776216e9e04b","added_by":"auto","created_at":"2026-03-16 21:29:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":411858,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAverage village-level prevalence by doses, strategy, and vaccine efficacy (panels).\u003c/strong\u003e Minimum and maximum simulated values are indicated by the shaded area around each line. Prevalence with the random allocation strategy declines linearly as more doses become available, at both values of VE. Allocation by any other strategy exhibits a non-linear curve in predicted prevalence, initially flatter below dv = 0.5, before declining much faster.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8778801/v1/19f35b306dfc5c2fb1a63832.png"},{"id":104754563,"identity":"0692b667-36f5-4565-9ed8-70b301a6186c","added_by":"auto","created_at":"2026-03-16 21:29:44","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":306749,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAverage animal-level prevalence by doses, strategy, and vaccine efficacy (panels). \u003c/strong\u003eMinimum and maximums of the mean across simulation are indicated by the shaded areas. Prevalence with the random allocation strategy declines\\n linearly as more doses become available, at both values of VE. Allocation by any other strategy exhibits a non-linear curve in predicted prevalence.\"\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8778801/v1/69f2dc240f2fe4e64ff4ab47.png"},{"id":104783025,"identity":"9f94beb7-0480-4dec-9aae-0cc8e1d058f7","added_by":"auto","created_at":"2026-03-17 07:58:07","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":325379,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMean Annual Incidence by dose availability ratio, allocation strategy, and vaccine efficacy (panels). \u003c/strong\u003eMinimum and maximum predicted values are indicated by the shaded areas. Mean annual incidence increases with greater dose availability for all allocation strategies, before declining, with the inflection point varying by allocation strategy and VE.\u003c/p\u003e","description":"","filename":"floatimage41.png","url":"https://assets-eu.researchsquare.com/files/rs-8778801/v1/2e53044abb816b534b1d3ec3.png"},{"id":104754564,"identity":"3e822426-1a0c-4a04-b9d2-51f93e28bde2","added_by":"auto","created_at":"2026-03-16 21:29:45","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":247910,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSimulated probability of eradication for all strategies where eradication occurred, with vaccine efficacy at 90%. \u003c/strong\u003eEradication is first seen with the random strategy at d\u003csub\u003ev\u003c/sub\u003e = 0.88 and increases to 0.29 as d\u003csub\u003ev\u003c/sub\u003e approaches 1. Population- and degree-based strategies also exhibit eradication at high dose availability (and efficacy), however, density-based allocation did not exhibit eradication at all.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8778801/v1/71880f26b6a0fd62ab596e49.png"},{"id":104754567,"identity":"74ebb65c-3c0c-487b-8866-0d4295921738","added_by":"auto","created_at":"2026-03-16 21:29:45","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":409618,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePRC coefficients against Mean Annual Incidence.\u003c/strong\u003e Each row is a different strategy (right), each column a different dose availability ratio.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8778801/v1/21288bca3d918898bada4f32.png"},{"id":104754566,"identity":"17d75f40-fbb0-4d2a-8717-f2b660f76a4a","added_by":"auto","created_at":"2026-03-16 21:29:45","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":197922,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePRC coefficients against probability of eradication. \u003c/strong\u003eEach row is a different strategy (right), each column a different dose availability ratio. No eradication was seen at d = 0.3 or 0.6, so coefficients are 0.\u003c/p\u003e","description":"","filename":"floatimage71.png","url":"https://assets-eu.researchsquare.com/files/rs-8778801/v1/d0b87e4cf5ceb8d33ec4706b.png"},{"id":104784921,"identity":"ec4425fa-0f31-4e96-b70e-2ce58eb850ab","added_by":"auto","created_at":"2026-03-17 08:09:08","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2767753,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8778801/v1/1c71dbeb-0f8c-4290-abee-647a58932465.pdf"},{"id":104783031,"identity":"ee618665-caf2-4095-b5d6-18c47afe4859","added_by":"auto","created_at":"2026-03-17 07:58:07","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":475452,"visible":true,"origin":"","legend":"","description":"","filename":"tc11supplementary.docx","url":"https://assets-eu.researchsquare.com/files/rs-8778801/v1/21d3832b2d117fd34c862972.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eAssessing the effects of limited vaccine supply on risk-based vaccination strategies in regions of endemic foot-and-mouth disease\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eFoot and Mouth disease (FMD) is one of the most infectious livestock diseases known, infecting a large variety of cloven-hoofed animal species, including major livestock animals such as cattle, buffalo, sheep, goats, and pigs [\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e]. The etiological agent is a virus, Foot-and-Mouth Disease Virus (FMDV), of the \u003cem\u003eApthovirus\u003c/em\u003e genus and \u003cem\u003ePicornaviridae\u003c/em\u003e family. Symptoms vary by species, but common symptoms include lesions around the feet and mouth, fever, and lameness – in cattle reductions in milk production are also common [\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e]. The disease is endemic in most countries in Africa and Asia, and it has been estimated that visible production losses due to FMD and vaccination in endemic areas alone amount to between \u003cspan\u003e$\u003c/span\u003e6.5 and \u003cspan\u003e$\u003c/span\u003e21\u0026nbsp;billion USD annually (Knight-Jones \u0026amp; Rushton, 2013).\u003c/p\u003e \u003cp\u003eEfforts to limit or eradicate the spread of FMD are common; historically control efforts have been successful in Europe and South America; however, such efforts relied on mass vaccination campaigns [\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e5\u003c/span\u003e]. For many countries where the disease is currently endemic, procurement of efficacious vaccines matched to locally circulating variants remains problematic. Even if such vaccines are available, their expense may be prohibitive, leading to smaller numbers of doses being procured than are necessary to vaccinate all animals at risk (Hammond et al., 2021).\u003c/p\u003e \u003cp\u003eGiven that vaccine supply is limited in many settings, it is important to ensure that doses are allocated in the most effective way. This scenario - where the supply of efficacious vaccine doses is insufficient to meet demand – may have knock-on effects on the allocation of those doses; the optimal allocation when many doses are available may be different from the optimal allocation when few doses are available.\u003c/p\u003e \u003cp\u003eEpidemiological modelling allows for the exploration of optimal vaccine dose allocation strategies. Previous work has generally focused on optimising vaccination policies against epidemics of FMD [\u003cspan class=\"CitationRef\"\u003e7\u003c/span\u003e–\u003cspan class=\"CitationRef\"\u003e9\u003c/span\u003e] although there has been limited work on simulating optimal control policies in endemic areas [\u003cspan class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWe therefore aim to use a previously developed and validated simulation model of FMD spread in endemic areas to investigate how the ‘optimal’ allocation of vaccine doses changes as the availability of doses changes. We focus on Turkey due to its combination of endemic FMD and high-quality data necessary for detailed simulation. Specifically, we wish to know whether: 1) The ‘optimal’ strategy changes at different dose availabilities; and 2) which strategies are most optimal, for each criterion investigated.\u003c/p\u003e "},{"header":"Methods \u0026 Materials","content":"\u003cp\u003eData\u003c/p\u003e\u003cp\u003eData are available from Turkey 2001–2012, comprised of: i) Village point locations and cattle headcounts; ii) Village outbreak records, 2001–2012; iii) Village-to-village cattle movement records, 2007–2012.Owing to the detailed data and model and to keep simulation times reasonable, we limit our analysis to a single province, Erzurum.\u003c/p\u003e\u003cp\u003eModel\u003c/p\u003e\u003cp\u003eThe model we use has been developed in the context of previous work on FMD in Turkey; for a more in-depth account of the model and parameters used please see our previously published article (Guyver-Fletcher et al., 2025). We use the joint posterior parameter distributions from that work to simulate our model here.\u003c/p\u003e\u003cp\u003eIn summary, the model is a stochastic metapopulation model which simulates intra- and inter-village FMDV transmission. A schematic representation of the main model aspects is illustrated in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e. Cattle within a village can be in one of eight (8) disease-relevant states: Maternally immune (M), Susceptible (S), Exposed (E), Infectious (I), Recovered (R), Carrier (C), Vaccinated-Susceptible (V\u003csub\u003eS\u003c/sub\u003e), and Vaccinated-Recovered (V\u003csub\u003eR\u003c/sub\u003e). Rates of progression between these states is determined by Ordinary Differential Equations and implemented stochastically using the tau-leap algorithm. It is assumed that carrier animals do not transmit due to the lack of evidence for this in the field [\u003cspan class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eDisease transmission between villages is simulated through two routes: 1) direct shipments of cattle, with probability of transmission equalling the probability one of the animals moved is infected; 2) Local transmission, aggregating aerosol transmission and other unseen routes of transmission – this is estimated through a distance-dependent kernel function that scales the probability of transmission with the number of infectious animals in the source village and the number of susceptible animals in the “target” village.\u003c/p\u003e\u003cp\u003eVaccine Allocation Strategies\u003c/p\u003e\u003cp\u003eWe define δ\u003csub\u003ev\u003c/sub\u003e as doses per capita – the ratio of available vaccine doses to the total cattle population at each vaccination campaign. For example, a value of δ\u003csub\u003ev\u003c/sub\u003e = 0.5 indicates a ratio of 1 available vaccine dose per 2 animals. For simplicity, we consider only application of a single dose at a time.\u003c/p\u003e\u003cp\u003eWe simulate mass vaccination campaigns every 182 days (6 months). No other control policies are simulated. At campaign commencement, the number of available doses is calculated as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\delta\\:}_{v}\\times\\:{N}_{t}\\)\u003c/span\u003e\u003c/span\u003e, where \u003cem\u003eN\u003c/em\u003e\u003csub\u003e\u003cem\u003et\u003c/em\u003e\u003c/sub\u003e is the total population of cattle in existence at time \u003cem\u003et\u003c/em\u003e.\u003c/p\u003e\u003cp\u003eWhen simulating a vaccination campaign, the relevant epidemiological unit is the village to retain operational realism. We order each village by the relevant criteria defined by the allocation strategy outlined in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e and then vaccinate in order, subtracting the number of used doses, until there are either no doses left or no villages remaining to vaccinate. “Random” allocation, the comparison allocation strategy, consists of vaccinating villages with no targeting by size, connectedness, or local density. We do not consider randomisation at the individual-animal level, as partial vaccination of herds is not operationally realistic for mass FMD vaccination campaigns.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\"\u003e\u003c/div\u003e\u003ctable id=\"Tab1\" border=\"1\"\u003e \u003ccaption\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eVaccine allocation strategies investigated.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003c/colgroup\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\"\u003e \u003cp\u003eAllocation Strategy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\"\u003e \u003cp\u003eOrdering Criteria\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eRandom\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eNone – random\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\"\u003e \u003cp\u003ePopulation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eVillages with the greatest cattle headcount are vaccinated first\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eDegree\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eVillages with the greatest number of recorded inward or outward cattle movements are vaccinated first\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eDensity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eVillages with the greatest number of other villages within 5km are vaccinated first\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/table\u003e\u003c/div\u003e\u003cp\u003eSimulation of strategies\u003c/p\u003e\u003cp\u003eTo account for parameter uncertainty, we sample 100 particles from the posterior distribution of our model fit to the observed history of outbreaks in Erzurum province [\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e]. We simulate a no-control scenario with each particle, and 10 infections seeded at the start of the timeline, for 5 years to achieve a realistic endemic state.\u003c/p\u003e\u003cp\u003eFor each endemic scenario, we then simulate a 5-year period for each δ\u003csub\u003ev\u003c/sub\u003e in the sequence {0.0, 0.01, 0.02, …, 0.99}. Each combination of particle and δ\u003csub\u003ev\u003c/sub\u003e is simulated 10 times. Each strategy is therefore simulated 1,000 times per value of δ\u003csub\u003ev\u003c/sub\u003e (100 particles, simulated 10 times per value of δ\u003csub\u003ev\u003c/sub\u003e).\u003c/p\u003e\u003cp\u003eWe simulate vaccine efficacy (VE) value drawn from a normal distribution centred on 71% [\u003cspan class=\"CitationRef\"\u003e13\u003c/span\u003e], or centred on 90%; in both cases the standard deviation used was 5%.\u003c/p\u003e\u003cp\u003eFor each combination of δ\u003csub\u003ev\u003c/sub\u003e and allocation strategy we calculate the average annual village-level incidence and average prevalence over the simulated period, as well as the animal-level prevalence. We also calculate the probability of eradication as the proportion of simulations where eradication is achieved, and the time to eradication when or if eradication occurs is the first day when eradication is achieved, averaged over all simulations with the same δ\u003csub\u003ev\u003c/sub\u003e and strategy.\u003c/p\u003e\u003cp\u003eSensitivity Analysis\u003c/p\u003e\u003cp\u003eTo investigate the sensitivity of our outputs to certain control parameters, we conducted a global sensitivity analysis using LHS-PRCC. For each of the 4 strategies, we took 30 random samples from our posterior distribution and 80 samples from a Latin Hypercube generated with the joint control parameter distribution outlined in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. We then simulated with a dose ratio δ\u003csub\u003ev\u003c/sub\u003e of either 0.3 (low availability), 0.6 (medium availability), or 0.9 (high availability) to investigate how sensitivity to control parameters might differ in different dose availability regimes.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\"\u003e\u003c/div\u003e\u003ctable id=\"Tab2\" border=\"1\"\u003e \u003ccaption\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMarginal distributions of control parameters explored for the sensitivity analysis.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003c/colgroup\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\"\u003e \u003cp\u003eControl Parameter\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\"\u003e \u003cp\u003eDescription\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\"\u003e \u003cp\u003eMarginal Distribution\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eVaccine Time to Effect\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eThe delay after vaccination before an animal is protected\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eUniform (3–14) days\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eVaccine Efficacy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eWhat proportion of animals generate protective immunity after vaccination\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eUniform (1–100) percent\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eVaccine Duration\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eThe average duration of protective immunity generated by the vaccine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eUniform (150–210) days\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eMass Vaccination Coverage\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eThe proportion of villages that are covered by the mass vaccination campaign\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eUniform (0–100) percent\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eMass Vaccination Interval\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eThe number of days between mass vaccination campaigns\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\"\u003e \u003cp\u003eUniform (120–240) days\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/table\u003e\u003c/div\u003e\u003cp\u003eOnce we had simulated each particle, we calculated our outputs of interest and calculated the Partial Rank Correlation coefficients (PRCC) using the `\u003cem\u003eepiR`\u003c/em\u003e package version 2.0.80 [\u003cspan class=\"CitationRef\"\u003e14\u003c/span\u003e] in R 4.4.3 (R Core Team, 2025).\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eMean simulated prevalence declines with increasing availability of vaccine doses, as expected (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). At the majority of simulated δ\u003csub\u003ev\u003c/sub\u003e values, the random allocation strategy leads to a lower mean simulated prevalence than using any other strategy, with all strategies converging as availability approaches 100%. Increased vaccine efficacy leads to a faster and larger decline. There is a small difference in observed trends between village- and animal-level prevalence (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) \u0026ndash; the magnitude of performance differences between strategies (in terms of animal prevalence) is much smaller than their performance differences in terms of village prevalence; however, the overall ranking of strategies remains similar across the range of simulated doses per capita.\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e demonstrates a different ranking, with the population-based allocation strategy outperforming all others when δ\u003csub\u003ev\u003c/sub\u003e \u0026lt; 0.6, and random allocation performing best above δ\u003csub\u003ev\u003c/sub\u003e = 0.6. Mean Annual Incidence with all strategies is relatively flat or increasing until dose availability exceeds 0.75. Ranking of strategies is very similar between different vaccine efficacies. As dose availability approaches 1, mean annual incidence for population, degree, and random allocation strategies declines towards (without reaching) 0 when VE\u0026thinsp;=\u0026thinsp;90; however, when VE\u0026thinsp;=\u0026thinsp;71 the decline is much smaller.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eNo strategy, with vaccine efficacy averaged at 71%, exhibited regular eradication of disease circulation, hence, probability of eradication is 0 and TTE is not applicable. The exception is a single simulation with the population-based strategy at δ\u003csub\u003ev\u003c/sub\u003e = 0.98, leading to a probability of eradication of 0.01.\u003c/p\u003e \u003cp\u003eWith VE at 90%, three strategies resulted in eradication at high dose availability ratios (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e) \u0026ndash; Random, Population, and Degree. Random sometimes led to eradication at the lowest dose availability (0.88), and all strategies peaked at a probability of eradication of approximately 0.29. Between δ\u003csub\u003ev\u003c/sub\u003e = 0.95 and 1, population-based allocation led to a significantly higher probability of eradication than random allocation. Density-based allocation did not lead to eradication in any simulation.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e shows the PRC coefficients of the sensitivity analysis against mean annual incidence. For vaccine time to effect and vaccine duration, no coefficient was significantly different from 0 for any strategy or dose availability ratio. Vaccine efficacy was positively associated with incidence in every case, although the coefficient declined as dose availability increased. Coefficients for the mass vaccination interval were not significantly different from 0 at low dose availability, but positively associated with incidence at higher dose availability, although there appeared no difference by strategy. Mass vaccination coverage coefficients differed by dose availability and strategy: at low dose availability it was negatively associated with incidence for the population and degree-based strategies, but not significantly different from 0 for density-based allocation and slightly positively associated for random allocation; at medium dose availability no coefficients were significantly different from 0; at high dose availability all strategies exhibited a strong negative association with incidence.\u003c/p\u003e \u003cp\u003eFor probability of eradication, no eradication was seen at dose availability ratios below 0.88, so all coefficients for the low and medium availability scenarios were all 0 (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). For the high availability scenario and for all strategies, vaccine efficacy and mass vaccination coverage were positively associated with eradication, mass vaccination interval negatively associated, and vaccine time to effect and duration not significantly different from 0.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn summary, the `optimal` allocation strategy depends on the outcome being optimised for, the availability of vaccine doses, and the efficacy of the vaccine being used. Random allocation at the herd level outperforms all other strategies at all dose availability ratios when optimising for average prevalence, however, if optimising for minimising incidence at low dose availability a population-based strategy performs best. Eradication only occurs at very high dose availability with a highly effective vaccine, and the population-based allocation strategy maximises the probability of eradication. However, even with full coverage and a highly effective vaccine, accounting for parameter uncertainty the probability of eradication is still less than 0.3. When assessed against mean annual incidence, all strategies are sensitive to vaccine efficacy and mass vaccination coverage, and insensitive (within explored ranges) to vaccine time to protection and immunity duration.\u003c/p\u003e \u003cp\u003eThe results indicate that random allocation of doses is the most effective strategy for most metrics investigated. It is strictly better than any other explored allocation strategy in reducing average disease prevalence, is somewhat better at reducing incidence at medium-high dose availability and was able to achieve eradication in some simulations at dose availability ratios lower than any other strategy. This surprising result does not appear to be due to the population structure of the simulated province (many smallholdings can lead to a high village coverage with a small animal coverage), as it also holds for prevalence at the level of the animal.\u003c/p\u003e \u003cp\u003eAccording to the results presented here, the second-best strategy is population-based dose allocation, which holds incidence lower than other strategies when few doses are available and maximises the probability of eradication at (very) high dose availability. Here we must discuss the unintuitive pattern seen in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, where mean annual incidence increases before decreasing \u0026ndash; this statistic is due to greater variability in observed incidence in the simulations. As more farms are vaccinated with greater dose availability, but not enough to truly eradicate the disease, when immunity wanes the village may be reinfected and recorded as a new infection.\u003c/p\u003e \u003cp\u003eDegree-based and density-based allocation strategies are less effective policies \u0026ndash; they are neither best at reducing prevalence or incidence at any dose availability ratio. Degree-based (i.e. connectivity-based) allocation does, however, manage to achieve eradication when simulated with high dose availability and vaccine efficacy. The density-based strategy, somewhat surprisingly, never manages to achieve eradication even with enough doses for essentially the entire population. Close investigation revealed that, as there is little correlation between density and population in this landscape, there exist a set of medium-sized villages with approximately 6,000 cattle who would never be vaccinated using this strategy except at 100% coverage\u0026ndash; they were too far from every other village and hence were deprioritised in our simulations. This result may not generalise to other landscapes, where density and population may be positively correlated enough to avoid this failure mode.\u003c/p\u003e \u003cp\u003eAll modelling results should be translated cautiously, and decisions must consider all available evidence and the unique characteristics of each country or region. However, these results suggest that in regions where the disease is currently endemic, and that wish to focus on control or eradication of FMD, it is better to first focus on achieving higher coverage or sourcing higher-quality vaccines than to attempt to \u0026lsquo;optimise\u0026rsquo; where vaccines are allocated. If resources are available to optimise, however our results suggest that:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eIf few vaccine doses are available, and the priority is to minimise incidence, vaccines should be allocated to the largest villages first. If the aim is to minimise average prevalence, random allocation should be used.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eAt high numbers of doses, population-based allocation maximises the probability of eradication.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eDensity-based allocation does not appear to offer any benefit over alternatives.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eOne limitation of these results that must be borne in mind is that, due to time and resource constraints, we only consider allocation of doses in a mass vaccination scenario. Such campaigns are generally carried out with other reactive control measures, such as ring vaccination around identified farms or livestock movement restrictions. Previous work of ours has addressed this [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e], however, we have not addressed the optimal allocation of doses between mass vaccination campaigns and reactive ring vaccination \u0026ndash; we believe this would be interesting avenue for future research.\u003c/p\u003e \u003cp\u003eThere is also the question of generality. Although our data from Turkey are high quality, and our model has demonstrated good fit to those data, the applicability of the model and results to other settings is unclear. This should be considered before any policy decisions might be made.\u003c/p\u003e \u003cp\u003eWe have focused on simple allocation strategies that we believe are straightforward to implement, however, we have not considered all possible allocation strategies. One that we did not consider, which is sometimes used, was risk-based spatial zoning \u0026ndash; i.e. vaccinating as many animals as possible in a defined geographic area identified as high risk. It would be interesting to compare that strategy to these results.\u003c/p\u003e \u003cp\u003eOur results suggest that attempting to optimize vaccine allocation is, in many cases, counterproductive and leads to greater incidence and prevalence than using a simple random allocation strategy. They also caution against using density-based or degree-based allocation. Whilst our model could be extended to consider additional vaccine targeting strategies, we believe that this work can be a useful reminder that the problem of FMD can mostly be solved with enough high-quality vaccines applied at regular intervals.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eEthics approval and consent to participate\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003eConsent for publication\u003c/p\u003e\n\u003cp\u003eNot applicable\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials\u003c/p\u003e\n\u003cp\u003eThe datasets supporting the conclusions of this article are available in the Zenodo repository under the reserved DOI \u003cstrong\u003e10.5281/zenodo.18471687\u0026nbsp;\u003c/strong\u003eand will be made publicly available upon publication.\u003c/p\u003e\n\u003cp\u003eCompeting interests\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003eFunding\u003c/p\u003e\n\u003cp\u003eThe authors would like to acknowledge the funding of this project by the 11\u003csup\u003eth\u003c/sup\u003e Technical Call of the European Commission for the Control of Foot-and-Mouth Disease (EuFMD), reference 81579. This work was also supported by an Ecology and Evolution of Infectious Diseases joint National Science Foundation (NSF) and Biotechnology and Biological Sciences Research Council (BBSRC) grant (BB/X005224/1).\u003c/p\u003e\n\u003cp\u003eAuthor\u0026rsquo;s contributions\u003c/p\u003e\n\u003cp\u003eGGF made substantial contributions to the conception and design of the work, the analysis and interpretation of the data, the creation of software for the work, and drafted the work. MJT made substantial contributions to the conception and design of the work and substantively revised it. All authors read and approve the final manuscript.\u003c/p\u003e\n\u003cp\u003eAcknowledgements\u003c/p\u003e\n\u003cp\u003eThe authors would like to acknowledge the contribution of the SAP Institute of the Republic of Turkey, and EuFMD, for the provision of agricultural and epidemiological data used in this work.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eArzt J, Juleff N, Zhang Z, Rodriguez LL. The Pathogenesis of Foot-and-Mouth Disease I: Viral Pathways in Cattle. Transboundary and Emerging Diseases. 2011;58:291\u0026ndash;304. https://doi.org/10.1111/j.1865-1682.2011.01204.x.\u003c/li\u003e\n\u003cli\u003eGrubman MJ, Baxt B. Foot-and-Mouth Disease. Clinical Microbiology Reviews. 2004;17:465\u0026ndash;93. https://doi.org/10.1128/CMR.17.2.465-493.2004.\u003c/li\u003e\n\u003cli\u003eKnight-Jones TJD, Rushton J. The economic impacts of foot and mouth disease \u0026ndash; What are they, how big are they and where do they occur? Prev Vet Med. 2013;112:161\u0026ndash;73. https://doi.org/10.1016/J.PREVETMED.2013.07.013.\u003c/li\u003e\n\u003cli\u003eLeforban Y. Prevention measures against foot-and-mouth disease in Europe in recent years. Vaccine. 1999;17:1755\u0026ndash;9. https://doi.org/10.1016/S0264-410X(98)00445-9.\u003c/li\u003e\n\u003cli\u003eRivera AM, Sanchez-Vazquez MJ, Pituco EM, Buzanovsky LP, Martini M, Cosivi O. Advances in the eradication of foot-and-mouth disease in South America: 2011\u0026ndash;2020. Front Vet Sci. 2023;9:1024071. https://doi.org/10.3389/FVETS.2022.1024071/BIBTEX.\u003c/li\u003e\n\u003cli\u003eHammond JM, Maulidi B, Henning N. Targeted FMD Vaccines for Eastern Africa: The AgResults Foot and Mouth Disease Vaccine Challenge Project. Viruses 2021, Vol 13, Page 1830. 2021;13:1830. https://doi.org/10.3390/V13091830.\u003c/li\u003e\n\u003cli\u003eKeeling MJ, Woolhouse MEJ, May RM, Davies G, Grenfellk BT. Modelling vaccination strategies against foot-and-mouth disease. 2003.\u003c/li\u003e\n\u003cli\u003eTildesley MJ, Savill NJ, Shaw DJ, Deardon R, Brooks SP, Woolhouse MEJ, et al. Optimal reactive vaccination strategies for a foot-and-mouth outbreak in the UK. Nature. 2006;440:83\u0026ndash;6. https://doi.org/10.1038/NATURE04324;KWRD=SCIENCE.\u003c/li\u003e\n\u003cli\u003eRoche SE, Garner MG, Sanson RL, Cook C, Birch C, Backer JA, et al. Evaluating vaccination strategies to control foot-and-mouth disease: a model comparison study. Epidemiol Infect. 2015;143:1256\u0026ndash;75. https://doi.org/10.1017/S0950268814001927.\u003c/li\u003e\n\u003cli\u003eRinga N, Bauch CT. Impacts of constrained culling and vaccination on control of foot and mouth disease in near-endemic settings: A pair approximation model. Epidemics. 2014;9:18\u0026ndash;30. https://doi.org/10.1016/J.EPIDEM.2014.09.008.\u003c/li\u003e\n\u003cli\u003eGuyver-Fletcher G, Gorsich EE, Jewell C, Tildesley MJ. Controlling endemic foot-and-mouth disease: Vaccination is more important than movement bans. A simulation study in the Republic of Turkey. Infect Dis Model. 2025;10:702\u0026ndash;15. https://doi.org/10.1016/J.IDM.2025.02.006.\u003c/li\u003e\n\u003cli\u003eStenfeldt C, Arzt J. The carrier conundrum; a review of recent advances and persistent gaps regarding the carrier state of foot-and-mouth disease virus. Pathogens. 2020;9. https://doi.org/10.3390/PATHOGENS9030167.\u003c/li\u003e\n\u003cli\u003eKnight-Jones TJD, Bulut AN, Gubbins S, St\u0026auml;rk KDC, Pfeiffer DU, Sumption KJ, et al. Retrospective evaluation of foot-and-mouth disease vaccine effectiveness in Turkey. Vaccine. 2014;32:1848\u0026ndash;55. https://doi.org/10.1016/J.VACCINE.2014.01.071.\u003c/li\u003e\n\u003cli\u003eStevenson M, Sergeant E. epiR: Tools for the Analysis of Epidemiological Data. 2025.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"FMD, Epidemiology, Vaccines, Control, Optimisation, Simulation, Policy","lastPublishedDoi":"10.21203/rs.3.rs-8778801/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8778801/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eFoot-and-mouth disease is an important livestock disease which is endemic in many regions of the world, many of which struggle to supply enough vaccine doses to cover the entire population of animals at risk. We use a previously fitted stochastic metapopulation model of FMD spread to simulate four plausible vaccine allocation strategies in the context of limited vaccine supply and endemic disease. We find that random allocation outperforms other strategies when the objective is to minimise average prevalence, and random allocation also achieves local eradication at lower doses per capita. However, all strategies perform similarly when minimising annual incidence, with the \u0026lsquo;optimal\u0026rsquo; allocation strategy varying by doses available per capita. These results suggest that policymakers in regions of endemic FMD should focus more on sourcing high-quality vaccines and achieving high coverage, with risk-based vaccination being a secondary concern.\u003c/p\u003e","manuscriptTitle":"Assessing the effects of limited vaccine supply on risk-based vaccination strategies in regions of endemic foot-and-mouth disease","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-16 21:29:40","doi":"10.21203/rs.3.rs-8778801/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"15bfcbd1-5e5f-409b-8e50-3d004025ba3d","owner":[],"postedDate":"March 16th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-12T18:09:48+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-16 21:29:40","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8778801","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8778801","identity":"rs-8778801","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}