Serological evidence of Flavivirus circulation in human populations in Northern Kenya: an assessment of disease risk 2016–2017

Yellow fever virus (YFV), Dengue virus (DENV), West Nile virus (WNV) and Zika virus (ZIKV) are important mosquito-borne Flaviviruses that have a potential to cause severe disease and mortalities in humans with economic and ecological consequence [1, 2]. Approximately 831 million people in Africa are at risk of infection with at least one of these viruses [3]. Epidemiologic studies on these mosquito-borne viruses remain a priority given their risk of global spread and high epidemic potential.

There have been no cases of Yellow Fever (YF) in Kenya since the first and last ever documented outbreak in 1992–95 [4, 5]. However, recent re-emergence of the virus with major outbreaks in border countries of Uganda, Ethiopia and South Sudan [6‐9] and regionally in Angola and Democratic Republic of Congo has become a major public health concern [10]. Dengue is currently endemic in parts of Kenya, with outbreaks reported in some specific geographic zones in the coast and northern frontier of Kenya associated with dengue Serotypes 1–3, and neighboring countries like South Sudan, Somalia and Tanzania [11, 12]. West Nile on the other hand, was first isolated in Uganda in 1937 and is now one of the re-emerging zoonotic mosquito-borne pathogens whose occurrence and geographical range continues to spread [2].

Zika virus was first discovered in Uganda in 1947 [13], and since then, there has been limited data available on its circulation in the region. Previously, virological and immunological evidence suggested that although ZIKV was distributed widely in Africa and Asia, Zika fever was not a disease of substantial concern to human beings because only 14 cases had been documented worldwide [14, 15]. In 2016 outbreaks of the virus were reported in the Americas [16] that led to the World Health Organization (WHO) declaration of Zika virus as a public health emergency of international concern. The increased detection of Zika virus worldwide and its association with increasingly large outbreaks of disease has heightened awareness of this emerging mosquito-transmitted pathogen [17].

There is limited epidemiologic knowledge about the presence and spread of arboviral diseases in northern Kenya. Surveillance capacities are lacking, with most resources for study and control of these viruses being focused on epidemic periods. As such, the magnitude of human exposure and the burden of these important Flavivirus diseases in this region remain poorly understood. West Pokot and Turkana Counties are located in areas that border countries that have had and reported outbreaks of these Flaviviruses in recent times and therefore the risk of virus spread from neighboring countries remains high. This study sought to determine the seroprevalence of Yellow fever, Dengue, West Nile and Zika viruses among the human populations in the border locations of West Pokot County bordering Uganda and Turkana County bordering South Sudan and Ethiopia to the north.

The present study reports on the serological evidence of human exposure to Flaviviruses in West Pokot and Turkana counties of northern Kenya. Overall, we found that 14.5% (n = 127, 95% CI 12.3–17.0%) of the samples were positive for at least one flavivirus (ZIKV, YFV, DENV-2, WNV). However, there were variations in the rates of human exposure for the individual viruses between the counties and across sites. Such variation in prevalence by site or even region is not uncommon in Kenya [27‐29]. At the county level, seroprevalence ranged from 0.22% for DENV in West Pokot to 10.7% for YFV in Turkana. Much higher exposure of specific flavivirus has been observed, of up to 60% for DENV in coastal Kenya [27, 30]. Such variability in specific flavivirus as observed in our study and elsewhere in Kenya suggests exposure is likely conditioned by a myriad of ecological and human factors such as vector species and abundance, climate and weather patterns, environmental features, occupational activities, degree of mobility among others.

West Nile virus was detected in Turkana (42/413, 10.2%), but not in West Pokot. This difference could relate partly to ecological variation between the two areas. West Nile virus is associated with Culex mosquitoes, the virus having been previously isolated in some Culex species in Kenya [31‐33]. It is possible the mosquito fauna especially the Culex spp. abundance and diversity varies between the areas. Grossi-Soyster [29] found high risk of both alphaviruses and flaviviruses primarily associated with households near the Lake area in Western Kenya. In this case, the presence of a lake (Lake Turkana) may encourage high diversity of bird species and breeding of Culex species, which are important determinants of the epidemiology of WNV [2]. This may possibly account for higher exposure rates in humans in Turkana than West Pokot. Proximity to Lake Turkana may also influence the economic activities of the people living in the area, such as fishing or grazing livestock. Older men spend more time away from home fishing, taking care of the livestock and looking for pasture in areas around the lake. Such outdoor activities could predispose them to more mosquito infectious bites. This could explain the obvious disparity in exposure rates among males compared to females, even though more samples were analyzed from the latter relative to the former. In fact, males were about three times more likely to be infected with WNV than females (aOR = 2.55, 95% CI 1.22–5.34). Additionally, adult males aged 50+ had more exposure, consistent with previous findings [30]. This alludes to higher risk due to advanced age; however, exposure in the 13–19 age group is indicative of ongoing transmission of this virus. Evidence of isolation of the virus from Culex spp. in this ecology [33] lends support for active virus transmission.

Neutralizing antibodies against YFV were detected in samples from both counties, although significantly higher in Turkana than West Pokot (P < 0.0001). Site specific variation in exposure rates was also evident ranging from 7% in Kare Edome to about 20% in Elelea in Turkana County. It has been about three decades since the last documented outbreak of YF in Kenya in the neighboring Rift Valley [4, 5]. The exposure in the 13–19 age group, although higher in 30–39 age category, could indicate low-level ongoing transmission and persistent exposure in this region. The participants’ responses to their YF vaccination status made it difficult to determine the definite source of infection for YF reactive samples. Record of vaccination status was based on self-reporting which could not be verified, as no vaccination cards were provided during survey. However, the presence of YF neutralizing antibodies in these two counties may be attributable in part to the vaccinations that were rolled out in 2003 in response to the outbreak that was reported in the neighboring South Sudan [6]. Additionally, anecdotal information from the Kenyan Ministry of Health and Sanitation suggests that, some of the population received YF vaccination in Uganda in response to the 2011 YF outbreak [7], while they had crossed the borders to Uganda for pasture and trade. However, the level of immunity seems too low for a population that was vaccinated so recently to forestall an outbreak from a neighboring country. If vaccination was carried out to prevent an outbreak, then a seroprevalence of 10% is a far cry from the desired protective herd immunity level (80%) and should be enhanced. We also cannot rule out the circulation of YF in low levels in these regions, because of cross border travel by this population to outbreak areas in the neighboring countries in search of pasture and trade contributing to this high seroprevalence. Our findings somehow point to active circulation of YFV in low levels in these regions; similar exposure rates have been observed in other regions of Kenya [30].

Only one sample in West Pokot was reactive to DENV-2 and none from Turkana. This in part may suggest low human exposure to this serotype in this region. Previous studies have highlighted the prevalence of DEN-2 serotype in Kenya through serological studies [12, 28, 34]. Nonetheless, human exposure to the other serotypes (1, 3–4) has been reported [30]. While we employed PRNT which is the gold standard for determining serological specific among viruses, cross reactivity of human sera, however, occurs with all four DENV serotypes [35] and even with other flaviviruses. The limitation of our analysis to only a single serotype (DEN-2), suggests more studies are needed to determine the exact burden of DEN in this region and possibly establish the serotype distribution.

The results in the current study suggest circulation of Zika in both counties, yet there have never been reported or confirmed cases of ZIKV in Kenya. The prevalence of Zika was significantly higher in West Pokot than in Turkana County. Coincidentally, the highest seroprevalence was recorded in Kanyerus (Fig. 1), proximal to the Uganda where the virus was first detected in 1947 [13]. Exposure could be influenced by the degree of human mobility between this site and across border into Uganda. As nomadic pastoralists, long distance movement of livestock in search of pasture or water is common practice especially during drought. Few studies have reported data on human exposure to ZIKV in Kenya, the last dating > 40 years [36]. The high prevalence among the age group 13–19 years possibly indicates evidence of active circulation of this virus in this ecology, which may have gone unnoticed due to lack of detection facilities. This calls for focused arboviral surveillance to forestall severe consequences of the virus emergence and incriminate the important vectors to enhance preparedness in dealing with potential outbreaks.

Our sampling approach focused only on asymptomatic population and could not tell whether the exposure represented here did manifest clinically. It is well known that clinical presentations of arboviral diseases are often highly non-specific, commonly indistinguishable from each other and from other common tropical diseases like malaria [37]. The lack of diagnostic capability for arboviruses being routinely implemented in any health facility in the study area makes tracking of clinical cases very difficult, outside the confines of research studies. The need for continuous monitoring remains a priority given the unanticipated emergence or reemergence of arboviral diseases in recent years mainly in the form of outbreaks. Active surveillance including virus activity in vectors would help to shed light on the specific vectors involved and risk trajectory. Such studies could employ emerging technologies such as Next Generation Sequencing (NGS) to identify and characterize even unknown pathogens that could potentially be of human health impact.

This study has some limitations that should be considered. Our sampling approach targeted asymptomatic population in a cross-sectional study design. There was no clinical information related to these infections in the study population. Further studies should include clinical disease monitoring and should be extended to include young children and pregnant women, as in the case of ZIKV which has been found to be associated with congenital birth defects, including microcephaly in the developing fetus [38]. In PRNT analysis, we employed only DEN-2 serotype which affects definite conclusions that can be drawn regarding DEN exposure and other serotypes in the study areas. Also, while we used the Firth’s logistic regression model to reduce bias in estimates and to overcome the computational problems experienced with the conventional logistic regression model, the low frequencies of the viruses – as has been reported in the literature – does make it difficult to associate the prevalence with the risk factors.