Introduction

In May 2015, a Zika disease outbreak originated in Brazil, and by October 5, 2016, local transmission of the Zika virus had been reported in over 60 countries and territories, with over 750 thousand estimated cases [1]. The World Health Organization (WHO) previously predicted that the virus would establish itself in all countries in the Americas except Canada and Chile [2], and with few exceptions this scenario has proved true. Travel-imported cases have also been increasingly reported throughout the United States, as well as in Australia, New Zealand, Canada, Western Europe, and China [3], representing the first time Zika has been reported in many of these western countries.

The Zika virus was first isolated in 1947 in a sentinel monkey in the Zika forest of Uganda [4] which gave the virus its name. It was first found in humans in 1952 [5]. However, only 14 human cases were documented prior to 2007 [6], and these were limited to small isolated epidemics in equatorial Africa and tropical Asia [6]. Since the 1950s the virus has spread eastwards from Africa through Asia and the Pacific, culminating with the 2015-16 outbreak in Latin America [7, 8].

The first documented recent outbreak of Zika disease occurred on Yap Island in the Federated States of Micronesia in the North Pacific in 2007 with less than 200 acknowledged cases [6]. In 2013 another outbreak occurred in French Polynesia, with around 28,000 suspected cases, at which point Zika began to be generally recognized as a re-emerging infectious disease [9]. The outbreak subsequently spread from French Polynesia to other Pacific Islands including New Caledonia, Cook Island, and Easter Island, where autochthonous transmission cycles were established [10]. Travel–imported cases were also documented in Japan [11], Germany [12] and Norway [10], among other regions.

The recent outbreak in Latin America began in Brazil, with the first documented Zika case reported in May, 2015 [13], although phylogenetic analyses of virus RNA sequences suggest that the virus was introduced into the Americas between May and December 2013 [14]. The virus quickly spread from Brazil throughout Latin America; by February 2016, an estimated 31,555 cases were identified in Colombia alone [2].

Historically, Zika infection has been associated with mild symptoms typically resembling and milder than those of related arboviruses such as dengue and chikungunya; many cases of infection show no symptoms at all. However, the recent outbreaks in French Polynesia and Latin America have been associated with much more serious clinical manifestations of the virus. In Brazil and French Polynesia, a link between Zika and a rare congenital disease, microcephaly, has been identified [15–22]. Additionally, Guillain-Barre Syndrome (GBS) has been reported in patients infected with Zika, firstly in the 2013 French Polynesia outbreak [23], and since in greater numbers in Brazil, El Salvador, Venezuela, Colombia, and Suriname [22]. The unprecedented size of the outbreak, rate of spread, and potential links with microcephaly and GBS prompted the WHO to declare the current Zika virus outbreak a public health emergency of international concern [24].

Zika now joins a list of arboviral diseases such as dengue and chikungunya that are being increasingly reported in new parts of the world, all likely introduced through global transport systems such as passenger air travel and maritime freight [14, 25]. Geographic spread of the virus occurs when infected travelers travel from affected regions to ones without local establishment of the disease, but in which the known and suspected vector species have established populations. Like dengue and chikungunya, Zika is known to be spread by Aedes aegypti; it is also strongly suspected to be spread by Aedes albopictus. While vectorial competence of Ae. aegypti is well established [26–29], and it is now confirmed to be the primary vector in spreading Zika [30–32], the capacity of Ae. albopictus as a secondary vector for spreading Zika is still unclear. There is evidence of the potential role of Ae. albopictus [33, 34], however, there is limited and conflicting quantitative estimate of its efficiency [35, 36]. Jupille et. al. [35] found that both Ae. aegypti from Madeira Island and Ae. albopictus from France were able to transmit the Zika virus, however Ae. albopictus from France was found to be less suitable to sustain local transmission. Chouin-Carneiro et. al. [36] observed high infection but low transmission rates for both Ae. aegypti and Ae. albopictus, while WHO [31] notes the vector competence for both species is similar, but Ae. albopictus is considered to have lower vector capacity than Ae. aegypti. The outcomes from these studies suggests both species are capable of Zika transmission, while also highlighting the uncertainty in the role that Ae. albopictus may play in transmitting, spreading, and helping to maintain the virus in many areas of the world. Further, potential virus adaptation to new vectors, as demonstrated in the case of Chikungunya in La Reunion [37, 38], introduces additional uncertainties.

The uncertainty surrounding the vectorial competence of different Aedes species in spreading Zika serves as the main motivation for the present analysis. This study explicitly addresses the differences in the potential geographical risk of Zika spread and local disease cycle establishment if Ae. aegypti is the sole competent vector versus if both Ae. aegypti and Ae. albopictus are competent for this purpose. Scenarios which further vary in the relative capacity of Ae. albopictus as a secondary vector are also considered.

As noted earlier, available evidence indicates that the two species differ in their vectorial competence. Moreover, the two species also vary widely in their present geographic distribution: Ae. aegypti is mainly present in wet tropical regions, while Ae. albopictus, a much better disperser, has a wider global presence in temperate regions, including the northern United States and parts of Canada, southern regions of the Americas including Chile, parts of Western Australia and East Asia. The analysis presented here is global and carried out at the finest resolution (1 arc-minute) that was permitted by the available data. Some preliminary findings were reported earlier [39] but the methodology was not described; all the analyses have been expanded and updated here and expectations from those preliminary findings were used to validate conclusions from the analysis using “back-testing”.

Several recent studies have mapped the potential spread of Zika into new regions [40–43]. These studies differ from the present one in either the assumptions made about the competence of potential vector species, in the spatial resolution or geographical extent of the study areas used, or in the methodological tools that were used. Monaghan et. al. [42] simulated Ae. aegypti and Ae. albopictus mosquito abundance based on meteorological models, and overlaid the results with travel and socioeconomic factors to estimate the cities in the United States with the highest expected cases of travel-imported Zika. Nah et. al. [43] presented a global risk model for Zika importation which used survival analysis and publicly available epidemiological and air travel data to predict the risk of importation and local transmission of Zika at the country level. In one study, Bogoch et. al. [40] presented an air travel-based risk map of Zika spread from Brazil into the rest of the Americas, and in another study modeled risk posed to Africa and the Asia Pacific region [41]. Both works [40, 41] implicitly assumed Zika to be equally efficiently spread by both Ae. aegypti and Ae. albopictus, and all studies only considered airline travelers departing the Americas. However, on August 28, 2016 local Zika spread was confirmed in Singapore, and autochthonous Zika transmission has since been reported across multiple community clusters [44]. With Singapore serving as a new potential source of Zika infected travelers, a substantially higher risk is posed to South and South-east Asia, where Aedes mosquito populations are well established, and Zika and dengue are endemic.

The present analysis extends previous work by presenting a global risk analysis based on a new mathematical framework to estimate Zika importation and establishment risk at a city level based on the most recent state of the outbreak, and accounting for uncertainties regarding the vectorial competence of Ae. albopictus. The risk analysis reported in this paper considers six scenarios, A, B, C, D, E and F, respectively, which vary in their assumed relative capacity of Ae. albopictus compared to Ae. aegypti, as a secondary spreading vector of Zika. The scenarios are bounded by two extreme assumptions; in Scenario A spread is assumed to be driven by Ae. aegypti presence alone, while in Scenario F spread is driven by Ae. aegypti and Ae. albopictus presence equally. In Scenarios B through E spread is driven predominately by Ae. aegypti presence with Ae. albopictus presence playing a lesser role. These scenarios are further described in the Materials and Methods section.

Besides air travel data, this work utilizes ecological vector habitat suitability models for Ae. aegyti and Ae. albopictus previously developed to analyze the role of air travel in the risk of geographical spread of dengue [45–47]. Those models are relevant to the risk of Zika spread because the same two vector species are implicated with one difference: while Ae. aegypti is known to be an efficient vector for both diseases, in the case of dengue Ae. albopictus is also known to be a competent but less efficient vector, whereas in the case of Zika it is a likely vector but its relative competence is unknown. Thus, the focus of this analysis will be on four questions:

1. For each scenario, what is the expected relative risk posed to unaffected regions by Zika-infected travelers arriving (by air) from known affected regions?
2. For each scenario, what is the expected relative risk posed by airports in affected regions by Zika-infected travelers departing (by air) to unaffected regions?
3. What are the intercontinental air travel routes that pose the highest risk of spreading Zika into new regions?
4. To what extent did Ae. albopictus play a secondary role in the 2015-2016 Zika pandemic in Latin America?

The analysis reported here only considers potential vectorial transmission of Zika. It ignores other modes of transmission that have been reported including sexual transmission [22] and congenital transmission [22].

Materials and methods

Throughout this analysis, the expected relative risk was computed as a function of three main components, the volume of travelers moving between Zika-affected regions and unaffected regions, the probabilistic expectation of established vector populations at the origin, and the probabilistic expectation of established vector populations at the destination. Because of the role of air travel, airports are central to this analysis as both origins and destinations for the spread of Zika. The vector suitability models and travel statistics used to compute passenger travel volumes are explained in further detail in the Data Section.

Results

The destination risk model results are illustrated in Fig 1. The top 100 cities to which Zika may be imported from affected regions for scenarios A, C, D, E and F are shown. The results for Scenario B are too hard to distinguish from A and C in the figure, so it is left out. The size of the circle represents the estimated expected relative risk posed to each city, with the color indicating the scenario. For those cities which are served by more than one international airport, the relative risk for all airports which serve the given city are aggregated. Solid dark red indicates the risk from Ae. aegypti alone, i.e., Scenario A, and the color of the circles lightens progressively from Scenario A to Scenario F. All risk values are computed using eq (3), for their respective scenarios.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Destination Risk Map for Spread of Zika.

The circles depict the top 100 cities to which Zika may be imported from source regions, resulting in local outbreaks. The countries shaded in gray are those with reported autochthonous Zika transmission as of October 5, 2016. The size of the circle is the estimated relative risk, with the color indicative of the scenario. The darkest shade indicates the risk from Ae. aegypti alone. Lighter color circles indicate total risk from both species, weighted according to scenarios C, D, E and F.

https://doi.org/10.1371/journal.pntd.0005487.g001

S1 Table contains the list of the top 100 at risk destination cities included in the map for all six scenarios, including their corresponding rank, relative risk, and designated country.

To gain a better understanding of the risk posed by outgoing travelers, the risk posed by each city in a known affected region for exporting infected travelers is also assessed. The top 100 origin cities in the affected regions likely to export Zika to new regions are listed in S2 Table, including their corresponding rank, relative risk, and designated country. Similarly to the destination risk, the relative risk at the city level is aggregated over all airports which serve a given city.

S3 and S4 Tables further breaks down the previous results to identify those routes which carry the most risk into and out of cities, and include the top 100 highest risk origin-destination city travel pairs for Scenario A and F, respectively.

Finally, in regard to validation, the model was run for each scenario using the set of sources as of February 15, 2016 along with travel data for February 2015. The destination risk results were aggregated to the country level, ranked and compared with the actual set of 29 counties/territories that were added to the CDC list of confirmed affected regions between February 15th, 2016 and October 5, 2016. Results from the back-testing analysis for each scenario are presented in S5 and S6 Tables. S5 Table lists the set of 29 Countries with local Zika transmission confirmed between February 15 and October 5 and the relative ranking for each of those countries computed for each of the six scenarios. S6 Table lists the top 29 countries at risk for the six scenarios. The results reveal Scenario A identified more of the 29 countries in it’s top 29 ranking than the other three scenarios, however all scenarios identified at least 15 of the 29, in their top 29. A more detailed discussion of these results will be presented as part of the Discussion below.

Discussion

By October 5, 2016 Zika had already become established throughout most of the Americas, which was also predicted by our earlier analysis [39]. Based on the status of the outbreak in early October 2015, the results reveal the United States and South and Southeast Asia to be at greatest risk of Zika becoming locally established. This is true for all scenarios considered, although the top 100 set of cities and corresponding rank does vary among the six scenarios. The main differences being if Ae. aegypti is the only competent Zika vector, then risk of local establishment in north America is concentrated in Texas and Florida, with South and East Asia and Australia following closely. However, if Ae. albopictus is a competent vector of Zika, then there is increased risk of local establishment in much of Western Europe and the U.S., Canada, Chile, Australia, New Zealand, as well as South and East Asia. In total, the risk posed to each country is increased by at least a factor of three, when both species of mosquitoes are considered.

The destination cities at highest risk are similar across all scenarios, with Orlando, Fort Lauderdale, Houston, Bangkok and Hong Kong topping all lists. As the assumed role of Ae. albopictus increases across the scenarios, the risk posed to major cities in the U.S. such as New York City and Los Angeles increases, as does the risk to major cities in Western Europe, specifically Rome, London, Paris and Amsterdam. When Ae. albopictus is considered an equal spreading vector, New York City becomes one of the top five at risk destinations, outranking Houston. The high risk posed to cities in Florida, Houston and New York is predominately due to the recent establishment of Zika in Miami. In fact, the Miami-to-Houston travel route is ranked in the top ten of all travel routes for Scenario A; however, it is replaced by the Miami-to-New York travel route for Scenario F. Within the U.S. our results are also consistent with those from [42]; both studies conclude Texas and Florida are at highest risk, specifically Miami, Orlando and Houston. Increasingly, travel cases are being reported in almost all U.S. States, and other countries including Argentina, Chile, France, Italy, New Zealand [22], which are all identified to be at high risk by our model.

The more recent increased risk posed to cities in Asia is predominately due to the local establishment of Zika in Singapore. Singapore is a global travel hub, with especially high connectivity to major cities throughout the Asia Pacific region, which also have established populations of Aedes mosquitoes, and are prone to outbreaks of dengue and even some small outbreaks of Zika. For this reason, cities in Thailand, Hong Kong, Indonesia, the Philippines, China, India, Taiwan, Vietnam and Malaysia are at high risk of importation resulting in local outbreaks. In fact, in September 2016 the CDC issued a Zika travel notice to 11 Zika endemic countries in Southeast Asia after confirmed local outbreaks in the region, however, the actual risk posed to locals and travelers remains unclear. Cities in Australia and New Zealand are similarly at increased risk due to the high travel connectivity with the Asia-Pacific region. For the scenarios which consider additional risk posed by Ae. albopictus, Tokyo, Japan as well as multiple European cities including Rome, London, Paris, Amsterdam, and Lisbon, and even San Francisco are identified to be at risk of local establishment. Of the 20 highest ranked cities in Scenario F, four are served by at least two major airports, including Tokyo, London, Paris, and New York, which partially contributes to their increased risk of Zika importation and establishment, as illustrated in Fig 1. The destinations at highest risk of Zika importation should prepare and implement mosquito surveillance and control efforts, in order to reduce the likelihood of local establishment, or, in those Southeast Asian cities where Zika is already endemic, the likelihood of new local outbreaks.

An additional means to control the spread of Zika is to conduct surveillance at the sources of travel to limit the number of outgoing cases. However, it is unrealistic to implement control at every airport in a region affected by Zika. By having a better understanding of the risk posed by individual airports and cities the available resources in a country can be optimally allocated. For all scenarios Singapore is identified as the highest risk origin. This is again due to its highly connected and globally central location. Miami, Florida, San Juan, Puerto Rico, and Cancun, Mexico are ranked second, third, and fourth in all scenarios, with Nadi, Fiji and Nassau, Bahamas ranking 5th and 6th in scenarios A, B and C, and Sao Paulo, Brazil ranking 5th in Scenario D, E and F. This outcome is of particular concern because many of these cities are highly popular tourist destinations for many U.S. and European residents. In order to prevent further global spread and establishment of Zika, it is imperative that these major travel destinations get the local outbreaks under control.

In efforts to better understand the destination and origin rankings discussed previously, the city-level analysis was further disaggregated to explicitly identify the travel routes carrying the highest risk of spreading Zika into new regions. This type of analysis allows for the specific incoming and outgoing flights which pose the highest risk to be identified and targeted for mosquito surveillance and targeted passenger screening. As noted above, these route-level results are included in S3 and S4 Tables, for Scenario A and F only. For these two scenarios, six and eight of the highest risk routes originate in Singapore. When only Ae. aegypti is considered, six of the top ten highest risk routes depart Singapore for cities in south and east Asia, with the top one arriving in Bangkok, followed by Hong Kong. Three of the top ten routes arrive in Florida, from either San Juan or Nassau, and the travel route between Miami and Houston is listed 10th. For Scenario F, eight of the top ten highest risk routes depart Singapore to cities in Asia. The only two routes in the top ten list not departing Singapore are from San Juan to Orlando and Miami to New York. These routes align closely with the top origins and destinations discussed previously.

Supporting information