A game-theoretic model of Monkeypox to assess vaccination strategies

Introduction

Monkeypox (MPX) is a zoonotic disease that has the potential to develop into one of the most threatening human Orthopoxvirus infections since the eradication of smallpox (Durski et al., 2018). The causative agent of MPX is monkeypox virus (MPXV), found in the same genus as the variola virus (smallpox), vaccinia virus, and cowpox virus (Shchelkunov, Marennikova & Moyer, 2006; Sklenovská & Van Ranst, 2018). Common symptoms of MPX, though relatively milder than smallpox, include fever, severe headaches, skin lesions, and myalgia (CDC, 2003). Prevention of the disease has remained a challenge for poverty-stricken rural areas with poor infrastructure that lack necessary sanitary supervision (Sklenovská & Van Ranst, 2018).

MPX is endemic to Central Africa and West Africa (Weinstein et al., 2005; Yinka-Ogunleye et al., 2018). West African and Central African strains of MPXV exist, the latter of which is more virulent and symptomatically severe (Likos et al., 2005; Mwamba et al., 2014). In the Democratic Republic of the Congo (DRC) the mortality rate of the Central African strain is 11% (Ježek et al., 1987). Since the first case of human infection in 1970, there have been numerous outbreaks in the DRC (Eteng et al., 2018). Annually, the DRC reports over 2,000 cases of suspected infections and is the only country in an endemic state (Mwamba et al., 2014). This estimate may be modest, as MPX is often misdiagnosed as chickenpox or other diseases that cause rashes (Ježek et al., 1988). Additionally, modern and robust surveillance of MPX is neglected as a consequence of limited funds and resources (Rimoin et al., 2010), and countries other than the DRC are not required to report all cases of MPX (Durski et al., 2018). Thus, the disease may be more severe than previously estimated.

In 2003, 47 cases of MPX were reported across five states in the U.S., originating from a shipment of animals from Ghana imported to Texas (CDC, 2003). In 2018, 3 cases of MPX were reported in the United Kingdom, making it the first time since the 2003 United States outbreak that the disease had reached a country outside of Africa (Eteng et al., 2018).

The clinical presentation of MPX can be found in Di Giulio & Eckburg (2004). The incubation period for the virus ranges from five to 21 days. MPX infection is split into two distinct phases: the invasion period and the skin eruption period. The invasion period starts between 0–5 days and is characterized by fever, lymphadenopathy, intense asthenia, severe headaches, and myalgia. The skin eruption period occurs 1–3 days after the appearance of a fever or lymphadenopathy, and it is characterized by rash formation, which often begins on the face and spreads to the rest of the body. The rash first appears as maculopapules (lesions with flat bases) and progresses to fluid filled blisters called vesicles. The vesicles then burst, forming pustules, and a crust forms over the affected area within 10 days. The number of lesions formed can vary from a few to thousands across the body, with children reportedly experiencing more severe symptoms than adults.

The predominant mode of MPX transmission is through human-animal interaction. Direct contact with an infected animal’s blood, bodily fluids, or lesions can lead to infection. Documented cases of MPX in Central and West Africa show that transmission can occur via the handling of wild animals, predominantly monkeys (Reynolds et al., 2017). Cultural influences, such as consumption of “bush meat,” can be a potential source of transmission. Additionally, direct contact with an infected person’s bodily fluids and skin lesions can lead to the transmission of the disease (McCollum & Damon, 2013).

Despite MPX’s high case fatality rate (Ježek et al., 1987) there are no known cures (Eteng et al., 2018). Until recently, there were no disease-specific preventative measures such as vaccines, though existing smallpox vaccines have historically been around 85% successful (Eteng et al., 2018). However, administration of the smallpox vaccine has ceased since the disease’s eradication in 1980, resulting in lowered immunity against Orthopoxviruses in general. This has led to a supposed increase in population susceptibility to MPXV (Sklenovská & Van Ranst, 2018). In 2019, the vaccine termed Jynneos^® was approved by the US FDA for protection against VARV and MPXV (Meyer, Ehmann & Smith, 2020).

It is possible that this lack of preventative measures is partially explained by a matching lack of literature on the potential dangers of inter-human transmission of MPX (Rimoin et al., 2010; Mwamba et al., 2014; Doshi et al., 2018; Sklenovská & Van Ranst, 2018). The urgency for better research on MPX is exacerbated considering vaccination cessation and immunocompromised populations in Central Africa, so a need for comprehensive preventative strategies is apparent (WHO, 2017). The factors such as (1) a lack of an effective vaccination strategy from fixed bases, (2) a shortfall in the vaccine supply and (3) logistical and security problems associated with the distance from the health centers, all contribute to the challenges of vaccinating the whole population in Central and Western Africa (Herp et al., 2003).

The identity of MPXV reservoir host(s) remains unknown (Di Giulio & Eckburg, 2004; Falendysz et al., 2017). The seroprevalence of MPXV was found highest in a population of moribund rope squirrels (Funisciurus anerythrus) in Zaire (now DRC), and the virus was also found in sun squirrels (Heliosciurus rufobrachium) and non-human primates in DRC (Khodakevich et al., 1987; Khodakevich, Ježek & Messinger, 1988) as well as in Gambian pouched rats (Cricetomys gambianus) (Doty et al., 2017; Doshi et al., 2019). In West Africa, African dormice (Graphiurus sp.) and ground squirrels (Xerus sp.) were identified as additional hosts (Reynolds et al., 2010). The majority of reported human cases originate from an interaction with an infected animal (Arita et al., 1985). The transmission of MPX among animals can be affected by environmental conditions (Brown & Leggat, 2016). Deforestation and flooding could potentially increase or decrease the MPX reservoirs, depending on how the animal population is affected by these conditions (Brown & Leggat, 2016). Long-distance transportation of potential MPX carriers may result in the expansion of the geographical range of the MPX reservoir, as exemplified by the 2003 US outbreak.

Currently, MPXV likely needs the animal reservoir as the human-to-human transmission chains of MPX are relatively short; the maximum number of generations reported in literature is seven (Learned et al., 2005). Nevertheless, as demonstrated by the case of H1N1 influenza (swine flu), some virus mutations can increase viral fitness in humans (Elderfield et al., 2014). We note that poxviruses have linear, double-stranded DNA genomes that vary from 130 to 230 kbp (Moss, 2013) and as such are evolving much slower than H1N1. Nevertheless, they can still adapt rapidly (Elde et al., 2012) and genetic engineering and modern molecular biology already turned a mousepox virus into an unusually lethal strain (Jackson et al., 2001; Di Giulio & Eckburg, 2004).

Epidemiologic compartmental models have been used to better understand the potential implications of disease transmission and infection (Blackwood & Childs, 2018; Bidari & Goldwyn, 2019). For MPX, the framework for a mathematical model has been tentatively set, but existing iterations have had shortcomings, failing to address some of the aforementioned aspects of the disease in their entirety. Bhunu & Mushayabasa (2011) introduced a basic SIR vector-borne compartmental model between humans and primates, yet deem an endemic state solely in humans as trivial. Usman & Adamu (2017) build upon this framework by introducing an SVEIR compartmental model to account for the disease’s incubation period and potential vaccine.

Game theoretical models attempt to study complex scenarios in which self-interested individuals will take an action based on the decisions of the rest of the population (Bauch & Earn, 2004). The model is a predictive tool in populations for extracting an optimum decision-making strategy (Chang et al., 2020). Game theory has been applied to protection strategies to control diseases such as smallpox (Bauch, Galvani & Earn, 2003), toxoplasmosis (Sykes & Rychtář, 2015), cholera (Kobe et al., 2018), measles (Shim et al., 2012), rubella (Shim, Kochin & Galvani, 2009), influenza (Galvani, Reluga & Chapman, 2007), African sleeping sickness (Crawford et al., 2015), malaria (Orwa, Mbogo & Luboobi, 2018; Broom, Rychtář & Spears-Gill, 2016), (Zika Padmanabhan, Seshaiyer & Castillo-Chavez, 2017; Banuelos et al., 2019) (Polio Cheng et al., 2020), Ebola (Brettin et al., 2018), chikungunya (SRM Klein, AO Foster, DA Feagins, JT Rowell, IV Erovenko, 2019, unpublished data), meningitis (A Martinez, J Machado, E Sanchez, I Erovenko, 2019, unpublished data), typhoid (C Acosta-Alonzo, IV Erovenko, A Lancaster, H Oh, J Rychtář, D Taylor, 2020, unpublished data), Hepatitis C (Scheckelhoff, Ejaz & Erovenko, 2019) and Hepatitis B (Chouhan et al., in press) among others. In this paper, we apply a similar approach to MPX to investigate a scenario in which individuals have the option of vaccinating to reduce the chance of contracting the virus. We further evaluate vaccination strategies on an individual and population-wide level by discussing the vaccination rates required to achieve herd immunity and Nash equilibrium.

In the present study, we build on the work of Usman & Adamu (2017), see also Lauko, Pinter & TeWinkel (2018) for a simplified SIR version of the model. The mathematical model of the MPX dynamics is shown in the next section. Then, we provide closed-form formulas for equilibrium states of MPX dynamics; the formulas provided in Usman & Adamu (2017) do not allow for direct calculations of the equilibria. We also show the existence of a “semi-endemic” equilibrium. This was not previously discussed in Usman & Adamu (2017), although it appears in Lauko, Pinter & TeWinkel (2018). We apply a game-theoretic approach to evaluate individual and population-wide vaccination strategies on the basis of cost and probabilistic disease acquisition and then we perform sensitivity analysis. We conclude the study by a discussion.

Mathematical Model

We adopt the compartmental epidemiological model introduced in Usman & Adamu (2017) and shown in Fig. 1. We consider squirrels to be the primary reservoir hosts. The population is divided into squirrels and humans, denoted by s and h subscripts, respectively. Individuals are born as Susceptible (S) at rate Λ. Susceptible humans vaccinate (move to V_h) at rate α_h. Vaccinated humans are assumed to never contract the disease in the remainder of their lifetime. Susceptible squirrels become Exposed (E_s) by coming into contact with infected squirrels with effective transmission rate β_ss. Susceptible humans become exposed by coming into contact with either infected squirrels (with effective transmission rate β_sh) or infected humans (with effective transmission rate β_hh). After an incubation period ν⁻¹, the exposed individuals become Infected (I). Infected individuals are infectious. They Recover (R) at rate ρ. Any individual may die due to natural causes at rate µ. Infected individuals can also die from the disease at rate d. The notation is summarized in Table 1. The model yields the following differential equations, see for example Blackwood & Childs (2018).

(1)${\displaystyle \frac{d{S}_s}{dt}={\Lambda }_s-\left({\mu }_s+{\beta }_{ss}\frac{I_s}{N_s}\right)S_s}$ (2)${\displaystyle \frac{d{E}_s}{dt}={\beta }_{ss}\frac{I_s}{N_s}{S}_s-\left({\mu }_s+{\nu }_s\right)E_s}$ (3)${\displaystyle \frac{d{I}_s}{dt}={\nu }_s{E}_s-\left({\mu }_s+d_s+{\rho }_s\right)I_s}$ (4)${\displaystyle \frac{d{R}_s}{dt}={\rho }_s{I}_s-{\mu }_s{R}_s}$ (5)${\displaystyle \frac{d{S}_h}{dt}={\Lambda }_h-\left({\mu }_h+\left({\beta }_{sh}\frac{I_s}{N_s}+{\beta }_{hh}\frac{I_h}{N_h}\right)+{\alpha }_h\right)S_h}$ (6)${\displaystyle \frac{d{V}_h}{dt}={\alpha }_h{S}_h-{\mu }_h{V}_h}$ (7)${\displaystyle \frac{d{E}_h}{dt}=\left({\beta }_{sh}\frac{I_s}{N_s}+{\beta }_{hh}\frac{I_h}{N_h}\right)S_h-\left({\mu }_h+{\nu }_h\right)E_h}$ (8)${\displaystyle \frac{d{I}_h}{dt}={\nu }_h{E}_h-\left({\mu }_h+d_h+{\rho }_h\right)I_h}$ (9)${\displaystyle \frac{d{R}_h}{dt}={\rho }_h{I}_h-{\mu }_h{R}_h}$

Results

Herd immunity and Nash equilibrium vaccination rates

The average cost of not vaccinating when the population vaccination rate is α_h is denoted C_notV(α_h) and it is given as a product of the cost of the MPX infection (C_MPX) and the probability of moving from the S_h compartment to the I_h compartment, i.e., (12)${\displaystyle C_{notV}\left({\alpha }_h\right)=C_{MPX}\cdot \left(\frac{\left({\beta }_{sh}\frac{I_s}{N_s}+{\beta }_{hh}\frac{I_h}{N_h}\right)}{\left({\beta }_{sh}\frac{I_s}{N_s}+{\beta }_{hh}\frac{I_h}{N_h}\right)+{\mu }_h}\right)\cdot \left(\frac{{\nu }_h}{{\nu }_h+{\mu }_h}\right).}$

As the vaccination rate α_h increases, R_0hh decreases by Eq. (11). Furthermore, the fraction $\frac{I_h}{N_h}$ at the appropriate equilibrium also decreases. Consequently, the cost of not vaccinating decreases. In the semi-endemic equilibrium, the cost eventually becomes 0 when the vaccination rate reaches (13)${\displaystyle {\alpha }_{HI}=max\left\{0,\frac{{\nu }_h{\beta }_{hh}{\mu }_h}{\left({\mu }_h+d_h+{\rho }_h\right)\left({\mu }_h+{\nu }_h\right)}-{\mu }_h\right\}.}$ At that point, herd immunity is achieved and the disease is eradicated from the population. In the fully endemic equilibrium, there is always a reservoir of MPX in the squirrel population. This reservoir causes an influx of MPX infections amongst humans. Therefore the disease can never be fully eradicated and the cost of not vaccinating will never reach 0, see Fig. 2.

When the vaccination rate is such that C_notV(α_h) = C_V, the vaccination rate is at Nash equilibrium, α_NE. When α_h < α_NE, it is beneficial for the individual to vaccinate; when α_h > α_NE, it is beneficial for the individual not to vaccinate.

Figure 3 shows a scenario where β_hh = 60. While this value is unrealistically high, we investigated this hypothetical scenario to see what would happen if MPXV mutates as was the case of H1N1 influenza (swine flu) or is genetically engineered as was the case of mousepox (Jackson et al., 2001; Di Giulio & Eckburg, 2004). When β_hh is large enough, specifically when (14)${\displaystyle {\beta }_{hh}>\frac{{\alpha }_h+{\mu }_h}{{\alpha }_h}\cdot \left({\mu }_h+d_h+{\rho }_h\right)\cdot \frac{{\mu }_h+{\nu }_h}{{\nu }_h}}$ MPX no longer needs squirrels to persist in the human population. In particular, it can become endemic even in countries without natural squirrel population (i.e., even when β_sh = 0). At the same time, in the semi-endemic equilibrium, the disease can be controlled through vaccination. Note that there is almost no difference between the Nash equilibrium rate α_NE (a solution to C_notV(α_h) = C_V) and the rate α_HI needed for herd immunity (a solution to C_notV(α_h) = 0).

Sensitivity analysis

As shown in Fig. 2, as α_h increases, C_notV(α_h) approaches an asymptote. Consequently, the value of α_NE, a solution to C_notV(α_h) = C_V can be very sensitive to C_V when C_V ≈ 3. Any small decrease of C_V can cause a significant increase of α_NE. The same sensitivity is demonstrated in Fig. 3. Figure 4 shows the sensitivity analysis and how α_NE depends on variation of different parameters. We can see the high sensitivity of α_NE on the squirrel-to-human transmission rate, β_sh, and on the cost of vaccination, C_V, for low values of C_V. Moreover, the figure demonstrates that α_NE can be quite sensitive on the effective transmission rate amongst squirrels, β_ss and the squirrels recovery rate, ρ_s. For ${\beta }_{ss}<\left({\mu }_s+d_s+{\rho }_s\right)\cdot \frac{{\mu }_s+{\nu }_s}{{\nu }_s}$, there is a semi-endemic equilibrium and α_NE = 0. However, as β_ss increases above that threshold value, α_NE rapidly increases and, when β_ss > 45, there is no Nash equilibrium vaccination rate. Similarly, when ρ_s is large enough to have a semi-endemic equilibrium, the optimal vaccination rate is 0. However, for small ρ_s, there is no Nash equilibrium and the change is relatively abrupt as in the case of β_ss.

Discussion

The phylogenetic relatedness between MPXV and variola virus grants the smallpox vaccine an 85% effectiveness in preventing MPX (Reynolds & Damon, 2012). Poxviruses from the Orthopoxvirus genus have cross-reactive antibodies, meaning that vaccinated individuals would have a much lower risk of infection and mortality compared to unvaccinated individuals (Louten, 2016). The imperfect prevention rate of the vaccine was omitted in the design of the mathematical model for the sake of simplicity. As noted in Wu, Fu & Wang (2011), imperfect protection aggravates the dilemma of voluntary protective actions as lower vaccine effectiveness can lead to better vaccine coverage and smaller free-riding effects; however, the impact of the epidemic can be harder to mitigate.

It is of interest to identify and evaluate possible preventative measures in addition to vaccination that would have a measurable effect on the transmission of MPX. For instance, decreasing the animal-to-human contact and launching an education campaign about dangers of eating raw meat which seems to be the main culprit behind squirrel-to-human transmission Khodakevich, Ježek & Messinger (1988) would significantly decrease the animal-to-human transmission rate. It could still come at a considerable individual cost (such as decrease of meat supply) but it would not require a complex or well-developed healthcare infrastructure needed for the vaccination, thus providing the general population with an easily accessible preventative measure. The mathematical model for such a measure would become more complex. The main idea would follow the spirit of Kobe et al. (2018) that investigated a situation for cholera prevention where individuals could either vaccinate or avoid drinking potentially contaminated water.

The reservoir host for monkeypox remains unclear (Di Giulio & Eckburg, 2004; Falendysz et al., 2017). We focused on a moribund rope squirrel, Funisciurus anerythrus, but we note that the disease has also been confirmed in other animals (Arita et al., 1985; Khodakevich, Ježek & Messinger, 1988; Reynolds et al., 2010). As noted in Falendysz et al. (2017), in a recent outbreak of MPX in DRC, no association was found between contact with rope squirrels and human infection (Nolen et al., 2015). Additionally, a recent survey of 34 villages in the Tshuapa region of DRC did not detect contact with a rope squirrel carcass in the previous 30 days, although they reported contacts with red-legged sun squirrel, Heliosciurus rufobrachium, (Monroe et al., 2015) which was identified in Khodakevich, Ježek & Messinger (1988) as another frequent host of monkeypox.

Conclusions

We modeled MPX dynamics using the compartmental model of Usman & Adamu (2017). As one of our major contributions, we provided closed form formulas of the equilibrium states of the dynamics. Moreover, we also showed a potential existence of the semi-endemic equilibrium, in which there is no infection in the squirrel population and the disease still persists in the human population. Currently, MPX does not seem to have the viral fitness to become endemic solely through human transmission. Yet, simple mutations in viral proteins could still occur and increase successive inter-human cases as seen in the H1N1 virus outbreak (Le et al., 2009). Should this mutation occur, a careful understanding of the semi-endemic equilibrium will be needed.

In addition, we applied a game-theoretical approach to assess vaccination decision-making developed by Bauch & Earn (2004). Individuals in any population susceptible to MPX have the choice to vaccinate against the disease or risk the possibility of contracting the disease. Naturally, it is in the individual’s best interest to choose the option with the smaller expected cost. The model quantifies the costs and benefits of getting smallpox vaccine. We found that the optimal vaccination rate is about 0.04, i.e., individuals should vaccinate about once every 25 years.

We must note that the parameter values we used are only estimates based on available literature. In reality, the parameters may be quite different, in large part because the reservoir hosts are different as discussed above. The performed sensitivity analysis allows us to gain insight into how our results depend on the specific parameter values. We observed that the optimal vaccination rate, α_NE, is about 10 times more sensitive to parameters related to animal hosts than to a corresponding parameter related to humans. It is therefore important to establish more accurate parameters. Consequently, greater efforts are needed to track the true prevalence and recurrent cases of MPX in all populations rather than relying on suspected cases.

Though not perfect in practice, mathematical modeling of diseases remains a powerful tool that grants a more profound understanding how MPX operates under certain conditions. The scope of epidemiological modeling and game-theoretic cost analysis is wide. As cases of MPX become increasingly reported among humans (Antwerpen et al., 2019), we hope that the models may serve as a predictive tool to better study the spread of MPX.

Supplemental Information