He was distressed, however, at having to part with his sheep, which he gave to the Bordeaux Academy of the Sciences. The Academy proposed as the theme for its yearly prize competition an investigation into the reason for the redness of the sheep's wool. The prize was won by a scholar from the north, who demonstrated by A+B−C/Z that the sheep must of necessity be red, and furthermore that it would die of the rot.
From Voltaire's Candide (1759)Introduction
This quotation describes one of the earliest attempts to use mathematical modeling to solve disease problems, and must surely be the first such example concerning diseases of sheep. It is personally disappointing that Voltaire chose the color red as, for many years, I have used mathematical modeling to address a disease that turns sheep – well, their tongues at least – blue. This disease is ‘bluetongue’, an insect-borne viral disease of ruminants, which causes fever, swelling, inflammation, and hemorrhage, leading sometimes to death.
Bluetongue has emerged dramatically in Europe in the last 15 years, causing the deaths of millions of sheep and many cattle (Mellor et al., Reference Mellor, Carpenter, Harrup, Baylis and Mertens2008). Before this outbreak, the disease had never occurred in Europe north of Madrid but, in 2006, there were cases in Northern Europe, in 2007 it occurred for the first time ever in the UK and in 2008 it reached Southern Scandinavia. This remarkable spread has led to questions about the role of climate change in the emergence of bluetongue.
Bluetongue is caused by an arthropod-borne virus (or arbovirus) called bluetongue virus (BTV) (Mellor et al., Reference Mellor, Boorman and Baylis2000). It is transmitted to its ruminant hosts by Culicoides spp. biting midges, tiny flies, which are often hugely abundant. Other important vectors of arboviruses are mosquitoes, which transmit Japanese encephalitis virus, West Nile virus, dengue virus, Yellow fever virus and many more, and ticks, which transmit tick-borne encephalitis virus, loping ill virus, Crimean-Congo hemorrhagic fever virus and others. Arthropods are usually extremely sensitive to climate and to the disease agents they spread, including arboviruses, and have therefore long been identified as those most likely to respond to climate change. Given this threat, it is sensible to ask: as our climate changes, how will vector-borne diseases respond? Which ones will spread, and how fast? Where will they spread? Answering such questions will help with climate change mitigation or adaptation.
Climate change and the emergence of bluetongue in Europe
Despite widespread concern about climate change, and recognition that one of its impacts could be the spread of vector-borne diseases (IPCC, 2007), until recently there had been relatively little effort expended on exploring what those effects might be. Indeed, only 12 years ago researchers concluded that there was a lack of strong evidence for an effect of climate change on vector-borne diseases (Kovats et al., Reference Kovats, Campbell-Lendrum, McMichael, Woodward and Cox2001).
The recent bluetongue outbreak was first linked to climate change by demonstrating that its early emergence in Southern–Central Europe coincided approximately in space and time with the area of the continent that warmed the most between the 1990s and 2000s (Purse et al., Reference Purse, Mellor, Rogers, Samuel, Mertens and Baylis2005). The strength of evidence was limited, however, because it did not explain the spread to Northern Europe which followed; it was only a correlation in space and time, with no mechanistic processes demonstrated.
Stronger evidence for the role of climate change would be a mechanistic model for the risk of bluetongue, which successfully predicts the disease's emergence when run at warmer, observed, temperatures. This rationale led us recently to build a climate-driven model for the disease (Guis et al., Reference Guis, Caminade, Calvete, Morse, Tran and Baylis2012). Our aim was a model with which we could explore how predicted changes in climate would affect the risk of bluetongue, expressed through the epidemiological concept of the basic reproduction number or R 0. The model demonstrates that the effects of the increase in temperature over the last two decades can explain many aspects of bluetongue's recent emergence. The model was also driven with future climate scenarios, leading to predictions of further increases in the risk of bluetongue in Europe up to 2050. While broadly successful, this exercise led us to discover gaps in arboviral (and other pathogen) vector research that need to be closed to improve the accuracy of predictive models for vector-borne diseases under scenarios of climate change.
Gaps in our understanding of the mechanism(s) underlying bluetongue emergence
In the case of BTV, there is one particularly obvious gap: what are the vectors? The occurrence of the disease outside of areas where the Afrotropical species, Culicoides imicola, was found, first in the Balkans from 2000 (Purse et al., Reference Purse, Nedelchev, Georgiev, Veleva, Boorman, Denison, Veronesi, Carpenter, Baylis and Mellor2006), and then in Northern Europe from 2006 (Mellor et al., Reference Mellor, Carpenter, Harrup, Baylis and Mertens2008), indicated that there must be other vectors in Europe. Indigenous Culicoides Obsoletus and Pulicaris species group midges were soon implicated (Caracappa et al., Reference Caracappa, Torina, Guercio, Vitale, Calabro, Purpari, Ferrantelli, Vitale and Mellor2003; Mehlhorn et al., Reference Mehlhorn, Walldorf, Klimpel, Jahn, Jaeger, Eschweiler, Hoffmann and Beer2007; Meiswinkel et al., Reference Meiswinkel, van Rijn, Leijs and Goffredo2007; Balenghien, Reference Balenghien2008; Dijkstra et al., Reference Dijkstra, van der Ven, Meiswinkel, Holzel and Van Rijn2008), but there is still debate about which of the different species play the most important roles. In our model, we focused on C. imicola and the Obsoletus group only.
Second, there is poor information about where these vectors occur. Prior to the European bluetongue outbreak, the only published distributions in Europe were from Spain and Portugal, as surveys were undertaken following an earlier African horse sickness (AHS) outbreak. AHS is also caused by an arbovirus that is transmitted by Culicoides. Since the start of the European bluetongue outbreak, several countries have undertaken Culicoides surveys but often with different trap types or protocols. Even today the UK has not published any national distributions; and there are no Europe-wide distribution maps in existence. And yet, bluetongue readily crosses frontiers and Europe-wide information is required for an international approach to prevention and control.
Light traps are used for Culicoides surveillance and we used light trap data from Spain to model vector distributions across Western Europe. However, this led us to a major gap in understanding, which is currently pervasive in vector-borne disease studies. Models of vector-borne diseases utilize the number of vectors per host (often calling this ratio m), but how does the number of vectors caught in a trap relate to the number per host? One approach is to use arbitrary multipliers of trap catches (such as 100× (Hartemink et al., Reference Hartemink, Purse, Meiswinkel, Brown, de Koeijer, Elbers, Boender, Rogers and Heesterbeek2009)) without evidence, to estimate the number of vectors in a region, which can then be divided by the host density to estimate m. This causes various difficulties however: estimates of R 0 then scale with the arbitrary multiplier, introducing huge subjectivity; and very high estimates of R 0 arise where host density is very low or zero (in large cities, for example) and where the risk of an outbreak is, intuitively, low. In our work, we made the assumption that trap catches are estimates of m itself, and do not need to be multiplied and divided by host number. After all, a light trap may be a proxy for a host; and as the number of vectors that come to a host should, by definition, be proportional to m, so the number coming to a trap may also be proportional to m. While the argument is interesting, the real point is that while we use insect traps for surveillance, we have poor understanding of what the contents of the trap mean in terms of the risk of disease.
Three additional variables required for estimating R 0 are the mean lifespan of vectors, their feeding interval, and the extrinsic incubation period (EIP): the length of time between a vector taking an infected blood meal and becoming infectious. These variables combine to tell us how many times, on average, a vector will feed on a suitable host in its remaining life after becoming infectious. While these variables are very important for understanding transmission risk, we have poor knowledge of them in Europe. First, there are no colonies of European indigenous Culicoides vectors or of C. imicola, and so data on the EIP (which must be obtained in high containment facilities) are misappropriated from other species. Mortality rates and feeding intervals can be obtained for the correct species in the field, but remarkably little research has been undertaken on the species present in Europe.
The above discussion makes two important points regarding our understanding of the epidemiology of bluetongue in Europe. Most of the required information, to understand bluetongue epidemiology, model disease spread or to predict the effect of climate change, pertains to the vectors; and much of it does not exist or is inadequate.
Additional arboviral disease outbreaks
Similar cases can be made for many other arboviral diseases in Europe; and yet, Europe's people and animals have never seemed at higher risk. A second Culicoides-transmitted viral disease of ruminants, Schmallenberg, has followed hot on the heels of bluetongue (Hoffmann et al., Reference Hoffmann, Scheuch, Hoeper, Jungblut, Holsteg, Schirrmeier, Eschbaumer, Goller, Wernike, Fischer, Breithaupt, Mettenleiter and Beer2012). It is believed to have first emerged in mid-2011, was only identified in late 2011 and by late 2012 it had spread over much of Western Europe. Its full impact remains to be seen. Meanwhile, mosquito-borne viruses are affecting people. Nearly 200 people were reported with Chikungunya in Italy in 2007 (Rezza et al., Reference Rezza, Nicoletti, Angelini, Romi, Finarelli, Panning, Cordioli, Fortuna, Boros, Magurano, Silvi, Angelini, Dottori, Ciufolini, Majori, Cassone and Grp2007). West Nile is occurring over a large area of Europe (more than ten countries in 2012). Dengue occurred in France in 2010 (La Ruche et al., Reference La Ruche, Souares, Armengaud, Peloux-Petiot, Delaunay, Despres, Lenglet, Jourdain, Leparc-Goffart, Charlet, Ollier, Mantey, Mollet, Fournier, Torrents, Leitmeyer, Hilairet, Zeller, Van Bortel, Dejour-Salamanca, Grandadam and Gastellu-Etchegorry2010) and there have been more than two thousand cases on the Portuguese island of Madeira since October 2012. A woman died of Crimean–Congo hemorrhagic fever, which is tick-borne, in Greece in 2008 (Papa et al., Reference Papa, Dalla, Papadimitriou, Kartalis and Antoniadis2010).
The recent emergence of these arboviral diseases in Europe, the predominance of vector-related variables in epidemiological models, and the paucity of data regarding vector biology in Europe, points to the need for an overhaul of arboviral research. We need significant investment in the training of a new generation of taxonomists, fluent in both morphological and molecular methods of insect identification. This field is ‘unsexy’ to many, but so important, that it should perhaps be achieved with a dedicated funding stream. We need more basic research on the biology of our vectors or potential vectors. And European researchers need to work more closely together, not just to learn from each other, but to start integrating vector surveillance efforts. We tend to share arboviruses and the vectors that spread them; we now also need to share sampling methods and the data so-obtained.
The relevant authorities should reappraise the value of restricting much arboviral research to high containment facilities. In the UK, access to highly contagious viruses that cause devastating diseases of livestock, such as foot-and-mouth disease and rinderpest, is restricted to a single high-containment facility, the Pirbright Laboratory. Some decades ago virologists at Pirbright added BTV and AHS virus (AHSV) to the portfolio, and restrictions designed for viruses that spread by direct contact or aerosol were applied to arboviruses that require a vector. Elsewhere in the UK, interested scientists were not permitted to work on live BTV or AHSV, even outside of the vectors. The rationale was ‘biosecurity’ but, really, the risk of BTV or AHSV escaping from a laboratory and going on to cause an outbreak was virtually nil, unless Culicoides themselves were infected and released. The end result of the restrictions was that the research base for such diseases was, and has remained, small; and to some extent left the UK poorly prepared for bluetongue when it arrived.
Of course, not all research gaps pertain to vectors: some relate to hosts. Important variables like host mortality and recovery rates can often be estimated from a number of laboratory studies, if not always for the right serotype of arbovirus. In the case of bluetongue, there is good information on the density of farmed ruminants in the UK (and Europe) with which to model spatial spread. However, were the UK or continental Europe to be invaded by AHSV, which could be devastating for our horse populations as AHS has a fatality rate of up to 95%, our ability to model, to explain or to predict the outbreak, and to test the efficacy of interventions, would be limited by the remarkable paucity of information on the distribution of equids. For example, a database of equids in the UK began in 2008, was discontinued 4 years later, and was voluntary and therefore, probably, incomplete.
Conclusion
No one can predict what will be the next arboviral disease to enter the UK, but that does not mean we should not prepare for it. In the event of an arbovirus outbreak, be it AHS, African swine fever, West Nile, dengue, Chikungunya, or any one of many others, we know already what sort of information will be required to deal with it, and also to answer questions posed by ministers, media or the public. We will need to know what and where the vectors are, and how or whether the vectors can be controlled. We will need to know whether hosts, be they human or animal, can be protected from vector bites. We will need to know if and how the arbovirus can survive the vector-free period. Much of this information can be obtained inexpensively because entomology tends to come cheap, like many of its practitioners. However, the work is slow, laborious and often seasonal, and almost certainly cannot be delivered in a useful time frame during an outbreak. With climate change seemingly being behind the emergence of bluetongue, and other arboviral diseases emerging in Europe for this or other causes, we need a significant upturn in arboviral disease research, with a primary focus on the vectors.
