Zika virus disease, microcephaly and Guillain-Barré syndrome in Colombia: epidemiological situation during 21 months of the Zika virus outbreak, 2015–2017

In recent decades, Colombia has been an exceptional witness to an avalanche of emerging arboviruses. Outbreaks of Dengue, Venezuelan equine encephalitis, Chikungunya virus -associated arthralgia- and most recently Zika virus disease (ZVD) have caused significant health burden in Colombia. Other emerging viruses have been detected with unknown disease burden, such as West Nile virus, St. Louis encephalitis virus and hantavirus [1]. Guillain-Barre Syndrome (GBS), microcephaly and encephalitis due to arboviruses are novel and emerging public health issues in Colombia [2].

By March 10, 2017, autochthonous cases and vector-borne transmission of ZVD have been confirmed in 35 countries and 13 regions of the Americas [3]. In Latin America, the countries with higher ZVD cases rate (i.e., [autochthonous suspected + autochthonous confirmed]/100,000 pop) are Honduras (396), Belize (239.62), Colombia (221.29), Venezuela (197.72), El Salvador (187.44) and Brazil (168.09) [3, 4]. In Colombia and Latin American, during the last months, the infection with ZVD has had a huge impact on public health and its effects continue to be felt [2].

In October 2015, the first nine cases of ZVD in Colombia were reported in Turbaco (Bolivar State), in the Caribbean area [5]. So far, June 24th, 2017 and nearly 21 months later the disease has spread to the entire country where Aedes aegypti is established. Similar to other countries affected by Zika virus, the increases of congenital syndromes including microcephaly cases and GBS appeared abruptly coincident with the ZVD outbreak [6‐12].

In Colombia, the National Health Institute (INS) conducts intensified surveillance for microcephaly and GBS cases associated with ZVD [13]. As a result, the INS has published preliminary reports about ZVD cases [8], GBS [8] and microcephaly cases associated with ZVD [8], other reports about transmission and epidemiology of Zika virus in Colombia have also been published [9‐11]. In this study, we analyze the epidemiological situation of ZVD and its association with microcephaly and GBS in Colombia during a 21-month period, from October 2015 to June 2017.

This study provides the first general analysis of ZVD in Colombia in the period from October 2015 to June 2017. The paper shows that Zika virus disease has had a huge impact on public health for the numbers of people affected and the presentation of Guillain-Barre syndrome and microcephaly cases associated to Zika virus disease. This study updates the epidemiological analysis of the Zika virus disease situation, it describes the geographical distribution of Zika virus disease and shows the numerical relationship between Zika virus disease with microcephaly cases and Guillain-Barre syndrome.

According to the retrieved data Zika’s attack rate among the total population of the country was 222 ZVD cases per 100,000 populations. This attack rate is similar to that observed in other Latin American countries such as Brazil, Venezuela, Costa Rica and Salvador [20]. Among the pregnant population, higher attack rates compared with total population were observed, but this situation has already been described previously in Colombia [21] and may be associated with an enhanced surveillance in this population sub-group due to congenital malformations that produces the diseases, such as microcephaly and other congenital malformations of the central nervous system.

Similar to other countries (e.g., Brazil) [22, 23], the most affected population in Colombia were women (AR, 286 per 100,000 population) and people between 20 and 39 years old (AR, 305 to 375 per 100,000 population). Pregnant women (18.5%), and people older than 65 years (3.18%) are considered as groups at risk by INS [14], because their post-infection condition might be more complicated than other groups. On the other hand, the rest of the affected population includes the economically active population, and thus it might be costly for the nation [24]. We recommend to carry out another study including the disability adjusted life years (DALYs) needs to be calculated in the near future.

The increasing number of both GBS and microcephaly cases due to ZKD has been described [7, 8]. By 24th June 2017, 0.22% (AR, 222 ZVD cases per 100,000 population) of Colombian population has been affected by ZVD [16], including almost 19,963 cases of infected pregnant women. This is an alarming situation because congenital defects increase during ZVD transmission, as has been observed in Colombia, Brazil and other regions of America [21]. The number of cases confirmed in laboratories is lower (9.10%) than reported in other countries [3]. Most cases confirmed in laboratories correspond to those notified in pregnant women (64.94%). During pregnancy the infection with ZIKV is associated to microcephaly and other abnormalities of central nervous system [25].

In March 2016, five months after the Zika virus outbreak, microcephaly cases were 1156 increased, which means that cases of microcephaly may correspond to congenital 1157 infections of women infected by ZIKV. Currently (June 2017, epiweek 25) Colombia has 710 ZVD cases associated with microcephaly; 174 were confirmed by laboratory. The increase of microcephaly cases during ZIKV transmission represents new information for clinicians because this active surveillance just started to be carried out after the first confirmed cases in Brazil. Prior to 2016, In Colombia, surveillance of microcephaly cases and other abnormalities of central nervous system, included evaluation for pathogens of the STORCH complex (syphilis, toxoplasmosis, other agents, rubella, cytomegalovirus, and herpes) [26]. ZIKV antibody testing was not done in Colombia due to high cross-reactivity with other endemic flavivirus [26].

The cases of ZVD in Colombia are decreasing, nevertheless, it is possible that new cases of microcephaly associated to ZVD will be present in the next few months when currently infected pregnant women deliver their new babies. Currently, CDC and INS are collaborating to understand long-term effects of Zika virus infection during pregnancy in Colombia, this surveillance will track for 5 years 5.000 pregnant women, as well as their spouses and children [27]. On the other hand, Zika virus also affects the hearing, visual and cognitive capacities of new-borns [28] and in Colombia this level of impact remains unknown.

Currently with the emergence of Chikungunya, Dengue and Zika a new challenge of differential diagnosis must be faced in the country, however, this emergence will present a high cost for the public health system of Colombia [1]. Geographically, the most affected states correspond to the Caribbean, Andean and Orinoquía regions of Colombia.

The Zika virus seems to have spread from the Caribbean area towards the Andean region in Colombia. It is interesting to note that there are hotspots in Norte de Santander, reporting area that are close to the Venezuelan border, where commercial activity and border crossing is intense. We believe that the high impact of ZIKV in Colombia was influenced by people moving across borders, because it is possible that people moving throughout the borders increased the mobility of the virus. We should remember that Colombia after Honduras y Belize is the country with the third highest density of ZIKV-infected people [3]. With the exception of areas of the country at 2000 m above sea level (i.e. Bogota, DC), all the 32 Colombian states were affected, the spread of the virus affected small, medium and large provinces; some cases were located at or more than 2000 m above sea level, despite the absence of known vectors for ZIKV at high elevation, these high altitude infections may be explained by travel to endemic areas or sexual transmission.

Several studies report that Guillain-Barre is associated with several etiologies, including ZIKV [1]. In this study, we have adopted kernel ridge regression to predict the number of Guillain-Barre cases given the number of ZVD cases. This model is appropriate where functional dependence is not linear (i.e., it is not proportional), in fact, Fig. 3a and b, shows there is no proportional relationship between the number of GBS and ZVD cases. Instances, like this one, evidence that the increasing of ZVD cases does not imply more GBS cases. It is important to emphasize that the linear kernel correspond to classical linear regression model, and there are two kernel which correspond with the more accurate models than the linear one. These outcomes are coherent with the distribution of the data points which cannot be aligned in a straight line in Fig. 3, which also means that there is no proportional relationship between ZVD and GBS cases, as it was above mentioned. As a consequence, the Gaussian kernel is the best choice for the prediction because with this kernel, KRR reported the smallest RMSE during cross validation, i.e., this is the most accurate kernel for this domain. We have performed 5-fold cross validation for tuning parameters (i.e. the radius and regularization parameters for the Gaussian kernel) and choosing the most appropriated kernel between sigmoid, polynomial, and Gaussian kernel. The cross validation results reveal that the regression model with the latter kernel outperforms the others in terms of accuracy.

The above mentioned regression model accurately predicted the number of Guillain-Barre cases. Nevertheless, there is an over fitting risk due to small sample size compared with those available in other domains. However, we lack another independent dataset to carry out the test, so the best setting achieved through cross validation was tested over the only dataset available to us. So in this study we provide evidence which might confirm the link between both diseases, notwithstanding that there are several reports where GBS has multiple associated etiologies.

As this incidence of ZVD decreases, GBS cases are expected to decrease as well. However, because of multiple etiologies, we expect GBS to persist even when ZVD disappears. In fact, we showed that the functional relationship between both diseases is not proportional, so it is not accurate to state the more Zika cases appear, the more Guillain Barre cases shall take place. Instead, as it is shown in Fig. 3a and b, the more Zika cases happen, the Guillain Barre cases will either increase or decrease, otherwise the linear kernel (i.e., the classical linear regression model) would have outperformed other kernels. Nonetheless, the prediction model adopted in this work is able to capture the regular patterns accurately, in spite of there are other known and latent (hidden) factors which explain the relationship between both illnesses. The drawback of KRR is that this model does not allows us interpret the prediction, however, it is the most accurate in this context.

This work contributes with an update on epidemiological analysis of the situation of ZVD in Colombia. The study describes the geographical distribution of ZIKV infection in the country and shows the association with microcephaly cases and GBS. A kernel ridge regression may be important to predict the number of Guillain-Barre cases given the number of ZVD cases. This study provides evidence that confirms the relationship between both diseases. Zika virus may have some influence of the new presentation of GBS in Colombia. There is a public health concern about the future of microcephaly cases in Colombia. Data essential in ZIKA control such as disaggregated data by age-group and by altitude are currently missing.