Dengue is the most important arbovirus in global public health [1]. It is spread by the bite of the highly anthropophilic Aedes aegypti mosquito, and to a lesser extent, by Ae. albopictus. Over half of the world’s population inhabit areas at risk of dengue infection [2, 3]. Currently, the WHO reports its presence in more than 125 countries [4] and recent modelling suggest as many as 390 million infections occur annually [5]. Dengue fever results from infection with any of the four closely related dengue serotypes: DENV-1, -2, -3 and -4. In a minority of cases, infection can progress to life-threatening condition such as dengue haemorrhagic fever (DHF). Infection confers protection from subsequent exposure to the same serotype but does not protect against the others [6], and sequential infections from other serotypes increases the risk of DHF [7]. Case fatality rates of dengue infection vary between 0.5 % – 3.5 % [8, 9].

Chikungunya virus (CHIKV) is an alphavirus also transmitted by Aedes spp. mosquitoes. There are three distinct evolutionary clades: West African, Central/East African and Asian CHIKV [10]. Historically, chikungunya was not considered a life-threatening infection but recent epidemiological evidence suggests a case fatality rate of around 0.1 % (mostly affecting the elderly) [11]. A variant of CHIKV first detected in a 2004 Kenyan outbreak spread globally through international travel, leading to autochthonous transmission events in islands of the Indian Ocean in 2005/6, India in 2005/6 and Europe in 2007 [12, 13]. This rapid spread of chikungunya demonstrated for the first time both the devastating magnitude of modern-day outbreaks (India was the worst affected country with over 1.4 million infections) and the ability of transmission in temperate regions [14–16]. More recently, in 2013, the first case of locally transmitted case of CHIKV outside Africa, Asia and Europe was reported in French Guyana; since then, 44 countries in the Americas have reported chikungunya cases in their territories [17].

Both pathogens are transmitted by the same Aedes spp. mosquitoes and so there is a reasonable expectation that the epidemiology of chikungunya and dengue infections is temporally and spatially related. Moreover, because symptoms presented by infected patients are similar and diagnosis of both infections is predominantly symptom-based, there will inevitably be ambiguity in disease recognition in inhabitants of endemic/epidemic regions and returning travellers. Therefore, the aims of this study were to: 1) determine the geographic limits of chikungunya, dengue and the principal mosquito vectors of both viruses, 2) review the available evidence of chikungunya and dengue co-infections, and 3) describe the clinical significance of chikungunya and dengue co-infection.

We are witnessing a rapid expansion in the geographical extent of chikungunya which mirrors that of dengue as described by Gubler in the 1990s [67]. This has come about partly through the increased opportunity for pathogen and vector spread that has resulted from globalisation [68], and the multifaceted effects on infectious diseases of a growing human population with resultant environmental changes [69]. Perhaps equally important, however, is the reporting bias that has obscured the public health impact of this pathogen, from its discovery until quite recently; CHIKV was first isolated in 1953 from the serum of a suspected dengue patient [70] and its conflation with dengue has persisted. Of the 30 studies eligible for inclusion in the current review, only one arose from an investigation of dengue cases, indicating a conspicuous absence of chikungunya diagnoses when dengue is suspected. Synthesising the available literature on chikungunya and dengue co-infection has revealed several limitations in our current understanding of the epidemiology of coinfection with both arboviruses and identified priorities for future research.

Similar to the global compendium of dengue [71], a consolidated, easily updateable and continuously maintained global database of chikungunya case notifications is needed and should be linked with reports of vector species detection. Subsequent to the 2006 chikungunya outbreak in French territory Le Reunion, several European countries (among them, France, Italy and Switzerland) have adopted a linked surveillance system for both arboviruses and vectors, with clear guidelines for curbing spread including educating inhabitants of outbreak foci on personal protection from mosquito bites, and rapid-response integrated vector management control campaigns [72]. Following France’s example, and, particularly in countries at the fringes of transmission and that have the facilities, both arboviruses must be nationally notifiable for this database to be useful in tracking the spread of disease with any fidelity. We note that this is easily implemented for countries that already have national notifiable databases for other diseases, and that are considered at high risk of incursion by these pathogens. One such example is Australia, which lists dengue as nationally notifiable but not chikungunya in all states and territories.

Improved cartographic refinement to a sub-national level is a logical next step that would build on the current exercise. While this was possible for some countries, data were not available to inform a global, sub-national level map. Differentiating endemic from epidemic regions for both chikungunya and dengue, and introducing an ordinal categorisation of disease level, such as has been developed for malaria [73], would enable tracking changes of the burden of disease and facilitate prioritisation of interventions. Enhanced geographical refinement and improved categorisation of at risk areas would not only enable focused targeting of surveillance and vector control, but also inform the denominator of co-infection prevalence.

In the current study we have identified a wide range of reported coinfection prevalence estimates (from 1.0–36.4 %); a key limitation with interpreting this finding is that it is set against a variable and dynamic background of monotypic infection prevalence. Furthermore, population standardised data is required to estimate the overall or by region DENV/CHIKV co-infection prevalence [74]; currently, it is not possible to compute a pooled estimate using the available data provided in the studies. Importantly, determining whether infection with one of the arboviruses enhances or attenuates host susceptibility to heterologous infection is not possible through indirect inference of relative prevalence levels; and this potential for ecological fallacy has been discussed fully in the context of more classically recognised mixed infections, for example the polyparasitism of soil-transmitted helminths [75]. The limited available information on infectivity of co-infected individuals provided by the 2012 Gabon study of Caron and colleagues suggests that co-infection reduces viral load relative to monotypic infection [30]. Determining how robust this result is across studies is important both immediately in terms of outbreak and control threshold estimation and in the longer term in the co-evolutionary context of these co-circulating pathogens.

Of related epidemiological significance is the determination of vector competence in virus-infected and superinfected mosquitoes [76, 77]. A recent review and modelling analysis by Christofferson et al. (2014) demonstrates the importance of considering the different combinations of pathogen-vector pairs at a finer resolution than serotype-genotype because of the variation in transmission potential found in even closely related strains [78]. Additionally, experiments suggest co-infection with multiple dengue serotypes may interfere with the vector’s ability to transmit virus [79]; whereas transmission enhancement has been demonstrated in the context of some other arboviruses [80]. Whether the chikungunya E1-226 V mutant that significantly increases chikungunya infectivity to Ae. albopictus also affects co-infected mosquitoes in their capacity as dengue vectors is unclear. Identifying any synergistic or antagonistic pathogen interactions within the vector constitutes an important, achievable future milestone in assessing the epidemiological consequences of chikungunya and dengue co-distribution.

The current study emphasises the likelihood of misdiagnosis of chikungunya infections among background dengue transmission (and vice versa). Critically, misdiagnosis not only hampers epidemiological understanding of both diseases but can profoundly affect the clinical picture of, and outcome for, infected patients. For example, misdiagnosis of dengue fever as chikungunya (or missing a dengue infection when coinciding with chikungunya) risks delaying or disrupting dengue-specific intensive supportive treatment [81] which can have a ten-fold impact on likelihood of progression from dengue fever to severe disease [82–85]. It also risks inappropriate prescription of arthralgia-alleviating nonsteroidal anti-inflammatory drugs (often employed in treating chikungunya patients) which could lead to severe bleeding in patients with thrombocytopenia or DHF [35]. The opposite and potentially more likely scenario in which chikungunya infection is misdiagnosed as dengue (or missed in a co-infected individual) masks the true geographical extent of CHIKV and population at risk of infection. It also obscures the likelihood of progression to severe disease in chikungunya patients: did the increased fatality rate reported post 2004 [11] result from a mutated CHIKV or was it simply easier to correctly attribute deaths from dengue-like illness due to increased awareness of chikungunya during the outbreak?

In this study we provide evidence of widespread co-distribution and co-infection with dengue and chikungunya. Our results suggest that clear protocols are urgently needed for realistic and effective control procedures which a) include emergency responses that take advantage of the shared transmission route of these arboviruses, b) are tempered by local transmission settings and informed by linked pathogen-vector databases and c) capitalise upon modern modelling methods for informing both the biology of infection and transmission processes as well as the strategy and tactics of disease control. Quantitative methods have been capitalised upon to great effect in terms of geospatial statistical approaches for generating high-resolution global maps of dengue risk [5]; early warning systems of dengue outbreaks [86]; biologically detailed multi-serotype mathematical models of dengue spread and control [87, 88]; and combinations thereof [89]. The time is ripe to take advantage of these developments to accelerate corresponding developments for chikungunya as well as dengue-chikungunya co-distribution and co-infection, to facilitate a more holistic understanding of the rapidly evolving, global epidemiology of these arboviruses.