Worldwide dynamic biogeography of zoonotic and anthroponotic dengue

Abstract

Dengue is a viral disease transmitted by mosquitoes. The rapid spread of dengue could lead to a global pandemic, and so the geographical extent of this spread needs to be assessed and predicted. There are also reasons to suggest that transmission of dengue from non-human primates in tropical forest cycles is being underestimated. We investigate the fine-scale geographic changes in transmission risk since the late 20th century, and take into account for the first time the potential role that primate biogeography and sylvatic vectors play in increasing the disease transmission risk. We apply a biogeographic framework to the most recent global dataset of dengue cases. Temporally stratified models describing favorable areas for vector presence and for disease transmission are combined. Our models were validated for predictive capacity, and point to a significant broadening of vector presence in tropical and non-tropical areas globally. We show that dengue transmission is likely to spread to affected areas in China, Papua New Guinea, Australia, USA, Colombia, Venezuela, Madagascar, as well as to cities in Europe and Japan. These models also suggest that dengue transmission is likely to spread to regions where there are presently no or very few reports of occurrence. According to our results, sylvatic dengue cycles account for a small percentage of the global extent of the human case record, but could be increasing in relevance in Asia, Africa, and South America. The spatial distribution of factors favoring transmission risk in different regions of the world allows for distinct management strategies to be prepared.

Fig 1. Methodological framework for dengue transmission risk modelling.

Vector models result from combining, through the fuzzy union (U), favorable areas for the presence of urban and sylvatic vectors, thus denoting that the presence of one vector species already implies some potential for disease transmission to humans if the pathogen is present. For a given time period and vector species, a vector model is built using mosquito occurrences as dependent variables, and spatial/environmental descriptors as independent predictor variables. Disease models describe the areas favorable to the occurrence of dengue cases, using the presence/absence of dengue-case records as dependent variables, and spatial/environmental/zoogeographic descriptors as independent predictor variables. A temporal stratification differentiating between the late 20^th century and the early 21^st century was applied when the modelled item was subject to a temporally changing dynamic, i.e. to the global distribution of Ae. aegypti, Ae. albopictus, and dengue cases. 20^th-century models were updated by complementing their equations with new variables capable of accounting for the observed changes of distribution. Finally, transmission-risk models quantify the level of dengue-transmission risk, according to the fuzzy intersection (∩) between vector and disease models. The intersection reflects that, for dengue to be transmitted in a given location, two elements, acting as limiting factors, must coincide in the area: 1) suitable environmental conditions for disease cases to occur; and 2) suitable conditions for the presence of vectors. Complete methodological descriptions are provided in the main text.

Fig 2. Global disease, vector and transmission-risk models.

A: maps for the late 20^th century. B: maps for the early 21^st century. The risk of transmission is estimated as the intersection (∩) between favorable conditions for the occurrence of dengue cases and favorable conditions for the presence of vector species. The spatial resolution is based on 7,774-km² hexagons. Recorded occurrences of dengue cases and vector presences are also mapped. Coast lines source: https://developers.google.com/earth-engine/datasets/catalog/FAO_GAUL_2015_level0.

Fig 3. Refined global disease, vector and transmission-risk models for the early 21^st century.

The risk of transmission is estimated as the intersection (∩) between favorable conditions for the occurrence of dengue cases and favorable conditions for the presence of vector species. Compared to the models in Fig 2, additional predictor variables only available for the 21^st century were considered, and the spatial resolution was based on 2,591-km² hexagons. Recorded occurrences of dengue cases and of vector presences are also mapped. See pre-downscaled versions of these models in S4 Fig. Coast lines source: https://developers.google.com/earth-engine/datasets/catalog/FAO_GAUL_2015_level0.

Fig 4. Areas of potential influence of sylvatic cycles on the presence of dengue in humans.

(A) Late 20^th century; (B) early 21^st century. Green: >0.1 increase of favorability values attributed to primate chorotypes; yellow: ≤0.1 increase of favorability values attributed to primate chorotypes; grey: area with low risk of dengue transmission. Venn diagrams: The numbers are percentages of contribution to the distribution of favorability in the disease models (Z: Zoogeographic factor; S/E: Spatial/Environmental factor). Coast lines source: https://developers.google.com/earth-engine/datasets/catalog/FAO_GAUL_2015_level0.

Fig 5

Late 20^th century disease and transmission-risk models in the Indian peninsula (A) and South America (B). These models were calibrated according to human-dengue cases from the late 20^th century (Fig 2A). The locations of dengue cases recorded in the late 20^th and the early 21^st centuries are shown in order to illustrate the predictive capacity of these models (see explanations and implications in the main text). See early 21^st-century models and data for these areas in S8 Fig. Coast lines source: https://developers.google.com/earth-engine/datasets/catalog/FAO_GAUL_2015_level0.