Mathematical modelling and phylodynamics for the study of dog rabies dynamics and control: A scoping review

Background Rabies is a fatal yet vaccine-preventable disease. In the last two decades, domestic dog populations have been shown to constitute the predominant reservoir of rabies in developing countries, causing 99% of human rabies cases. Despite substantial control efforts, dog rabies is still widely endemic and is spreading across previously rabies-free areas. Developing a detailed understanding of dog rabies dynamics and the impact of vaccination is essential to optimize existing control strategies and developing new ones. In this scoping review, we aimed at disentangling the respective contributions of mathematical models and phylodynamic approaches to advancing the understanding of rabies dynamics and control in domestic dog populations. We also addressed the methodological limitations of both approaches and the remaining issues related to studying rabies spread and how this could be applied to rabies control. Methodology/principal findings We reviewed how mathematical modelling of disease dynamics and phylodynamics have been developed and used to characterize dog rabies dynamics and control. Through a detailed search of the PubMed, Web of Science, and Scopus databases, we identified a total of n = 59 relevant studies using mathematical models (n = 30), phylodynamic inference (n = 22) and interdisciplinary approaches (n = 7). We found that despite often relying on scarce rabies epidemiological data, mathematical models investigated multiple aspects of rabies dynamics and control. These models confirmed the overwhelming efficacy of massive dog vaccination campaigns in all settings and unraveled the role of dog population structure and frequent introductions in dog rabies maintenance. Phylodynamic approaches successfully disentangled the evolutionary and environmental determinants of rabies dispersal and consistently reported support for the role of reintroduction events and human-mediated transportation over long distances in the maintenance of rabies in endemic areas. Potential biases in data collection still need to be properly accounted for in most of these analyses. Finally, interdisciplinary studies were determined to provide the most comprehensive assessments through hypothesis generation and testing. They also represent new avenues, especially concerning the reconstruction of local transmission chains or clusters through data integration. Conclusions/significance Despite advances in rabies knowledge, substantial uncertainty remains regarding the mechanisms of local spread, the role of wildlife in dog rabies maintenance, and the impact of community behavior on the efficacy of control strategies including vaccination of dogs. Future integrative approaches that use phylodynamic analyses and mechanistic models within a single framework could take full advantage of not only viral sequences but also additional epidemiological information as well as dog ecology data to refine our understanding of rabies spread and control. This would represent a significant improvement on past studies and a promising opportunity for canine rabies research in the frame of the One Health concept that aims to achieve better public health outcomes through cross-sector collaboration.

The dog vaccination coverage recommended by the WHO (70%) has been generally shown to be sufficient to reach dog rabies elimination [7,12,13], except in Ethiopia where a 90% vaccination coverage was recommended [4]. Vaccination strategies targeting at-risk dog populations are more effective [10]. The role of underreporting is not clear [2,21,25] but heterogeneous vaccination coverage is shown to disrupt vaccination [26].

Asia
Current situation: Disease-free in Japan. Recent introductions in Indonesia and Philippines. Endemic in China and continental South-East Asia.
Data: Human rabies cases from passive surveillance, dog rabies cases from active (China) or passive surveillance, contact tracing, dog vaccination data, dog density from surveys, dog movements from household surveys, historical records of dog rabies epidemics in Osaka and of dog and human censuses, RABV genetic sequences from dogs, wildlife and humans.
Main findings: Rabies introductions in the disease-free islands of the Philippines result from single introductions from neighboring rabies-endemic islands followed by local transmission [39,44]. At the continental scale, RABV lineages are spatially clustered [37,38] but transboundary movements markedly influence rabies spread [37]. China is endemic for rabies and multiple RABV lineages co-circulate across the country, notably Asian, Arctic-like and Cosmopolitan lineages [36,42,43]. It is thought to be one of the main sources of RABV lineages in Asia [36,37]. A decade after achieving rabies elimination, it resurged in Yunnan and is currently circulating uncontrolled. This Chinese province corresponds to a crossroads area where multiple RABV lineages circulate, probably resulting from multiple transboundary movements [35,45]. Moreover, rabies dispersal velocity is weakly associated with forest coverage, croplands and accessible areas [45]. Whereas human-mediated movement is not statistically associated with rabies velocity in the Yunnan province [45], it is suspected to have played a role in rabies dispersal in the Shaanxi province [41]. More studies are needed to unravel the interactions between RABV, reservoir ecology and humans in Asia. In general, rabies is estimated to spread at low grade with an R lower than two [26][27][28][29][30][31][32]. Occasional long distance migrations which were documented in the Philippines [44], Indonesia [40] and China [30,43,46] might contribute to disease persistence. The role of wildlife has been poorly studied and remains unclear in endemic areas [31,42]. Dog vaccination is the most effective strategy [27,28,31,32] and may be improved by complementary measures such as domestic and stray dog management [27][28][29], dog confinement [34] or increasing public awareness [29,31,33]. Homogeneous vaccination coverage was shown to yield better elimination prospects [8,30,34] which might be due to its robustness to human-mediated movements [8]. In Japan, Kadowaki et al. [33] showed that the current contingency plan is adapted to the rapid detection, control and elimination of rabies after an introduction. The authors emphasized the benefits of dog owner awareness and the control of stray dogs in the improvement of the plan [33]. The time to detection is also a crucial factor in the success of rabies elimination after introduction. The faster the disease is detected, the higher the odds of eradicating it [8]. For example, it's estimated that the surveillance system detected rabies circulation one year after its introduction in the Luzon island group in the Philippines [44]. This delay would have been greater with a lower reporting capacity [44].

Middle East
Current situation: Endemic.
Data: RABV genetic sequences from dogs, wildlife, and humans.
Modelling aims: Spatiotemporal dynamics and interactions of canine and wildlife RABV lineages [47,48]. Identification of circulating lineages and environmental factors impacting rabies spread [47].

Main findings:
Many lineages circulate that are phylogenetically related to Asian, Arctic/Artic-like or Cosmopolitan lineages resulting from sustained circulation in dogs and wildlife after introduction [47,48]. There is a strong spatial segregation of RABV lineages circulating in Iran. Overall, their spread is not driven by road connectivity, but humans presumably play a role since lineages tend to disperse towards and remain in highly populated areas. Lineages were less likely to spread towards grasslands and to occur in areas with barren vegetation. These results may be influenced by biased sampling towards populated areas however [47]. Wildlife seems to play a role in rabies maintenance in dog populations [47,48] but data are not sufficiently available to study host shift and dynamics between reservoirs.

South America
Current situation: Endemic for bat rabies and localized resurgences of rabies in dogs.
Data: Serological data and RABV genetic sequences from dogs and wildlife.
Main findings: Despite extensive dog vaccination campaigns, multiple dog-related RABV lineages circulate in Brazil with a relatively recent common ancestor estimated in the 1950s [50,51]. Dog lineages are generally spatially clustered [50,51] and lineages circulating in wild foxes and dogs are phylogenetically and dynamically independent [50].
Data: Dog population structure, dog roaming behavior (GPS data, questionnaires/interviews of dog owners), dog contacts, census data.
Main findings: Australian studies focused on rabies spread in remote rural and peri-urban locations where rabies is expected to be introduced and where surveillance systems might be weakened by the remoteness. Rabies dynamics are expected to differ between dog categories, such as explorer dogs, roaming dogs or domestic dogs [54][55][56], and consequently between rural and peri-urban areas [56]. Reactive vaccination after the detection of rabies introduction is the only beneficial strategy [52,53,55,56]. A 90% dog vaccination coverage is recommended to break down rabies spread [55,56] and targeting at-risk dogs should enhance vaccination campaigns efficacy [53,56].