Descriptive analysis of surveillance data for Zika virus disease and Zika virus-associated neurological complications in Colombia, 2015–2017

Zika virus (ZIKV) is a mosquito-borne pathogen that recently caused a major epidemic in the Americas. Although the majority of ZIKV infections are asymptomatic, the virus has been associated with birth defects in fetuses and newborns of infected mothers as well as neurological complications in adults. We performed a descriptive analysis on approximately 106,000 suspected and laboratory-confirmed cases of Zika virus disease (ZVD) that were reported during the 2015–2017 epidemic in Colombia. We also analyzed a dataset containing patients with neurological complications and recent febrile illness compatible with ZVD. Females had higher cumulative incidence of ZVD than males. Compared to the general population, cases were more likely to be reported in young adults (20 to 39 years of age). We estimated the cumulative incidence of ZVD in pregnant females at 3,120 reported cases per 100,000 population (95% CI: 3,077–3,164), which was considerably higher than the incidence in both males and non-pregnant females. ZVD cases were reported in all 32 departments. Four-hundred and eighteen patients suffered from ZIKV-associated neurological complications, of which 85% were diagnosed with Guillain-Barré syndrome. The median age of ZIKV cases with neurological complications was 12 years older than that of ZVD cases. ZIKV-associated neurological complications increased with age, and the highest incidence was reported among individuals aged 75 and older. Even though neurological complications and deaths due to ZIKV were rare in this epidemic, better risk communication is needed for people living in or traveling to ZIKV-affected areas.


Introduction
Zika virus disease (ZVD) is an emerging infectious disease caused by Zika virus (ZIKV), a single-stranded RNA virus that belongs to the genus Flavivirus [1]. In 1952, the first cases of ZVD in humans were reported in present-day Tanzania patients. Although the data are anonymized, identification is a risk given the high geographic resolution and large combination of predictor variables. This determination was made by Comité de Ética y Metodologías de Investigación (CEMIN).
incidence, risk ratios, and tests for statistical significance were calculated for high-risk groups. Understanding risk factors for neurological complications could inform prevention efforts and improve interpretation of ZIKV surveillance data.

Ethics statement
The technical and ethical endorsement of this study was provided by the Comité de Ética y de Metodologías de Investigación (CEMIN) of the Instituto Nacional de Salud (INS) of Colombia (project number .

Data
We analyzed two anonymized line lists on ZVD cases and ZIKV-associated neurological complications. The location of cases used in this study refers to location of likely infection, which is determined by the clinician who reported the case. ZIKV. The ZIKV dataset consists of 106,033 suspected and laboratory-confirmed cases of ZVD reported to Sivigila, Colombia's national public health surveillance system, between 2015 and 2017. We included cases with missing information on municipality (administrative level 2) location. Information related to public health events in the national territory is generated from the local levels by local health service providers. There are roughly 14,000 institutional, municipal, departmental, or national reporting bodies in Colombia. ZVD was added to the list of notifiable conditions in 2015. Each week these data are aggregated and published [26].
The line list data were aggregated by municipality. Data were further aggregated by week based on either date of symptom onset or date of notification. For ZIKV, this resulted in a time series spanning 97-weeks, from the week ending August 15, 2015 to that ending June 17, 2017, or epidemiological week 32 of 2015 to epidemiological week 24 of 2017.
Neurological complications. The neurological complications dataset contains information on 418 patients with neurological complications and recent history of febrile illness compatible with ZVD. It includes those with GBS as well as other conditions such as myelitis and meningoencephalitis but excludes cases with microcephaly and other congenital defects. Four cases in this dataset were laboratory-confirmed for ZIKV by reverse-transcriptase PCR.
The number of cases with neurological complications here is a few hundred smaller than previously published data from Colombia [27,28]. Throughout the epidemic, cases of neurological complications associated with prior ZIKV infection were reported in the INS Weekly Epidemiological Bulletin. This information was made publicly available with the caveat that cumulative case numbers could change following a verification process [28].
Medical records of patients with neurological complications were reviewed using case definitions from the Brighton Collaboration Working Group for GBS, myelitis, encephalitis, and acute disseminated encephalomyelitis (ADEM) [29,30]. With the goal of improving comparability of vaccine safety data, the Brighton Collaboration developed standard case definitions and guidelines for neurologic adverse events following immunization. The criteria could be applied to a range of settings, including across geographical regions as well as different levels of healthcare quality and access. Case definitions are organized according to three levels of diagnostic certainty, from Level 1 (most certain) to Level 3 (least certain) [31]. Patients that did not meet Brighton case definition criteria 1-3 were removed from the dataset.
Both date of symptom onset of neurological complications and date of notification were available for all cases in the neurological complications' dataset. Dates corresponding to symptom onset spanned 122 weeks, from the week ending July 4, 2015

Descriptive analysis
The cumulative incidence (attack rate) of ZVD was estimated using DANE population projections for 2016 as the denominator. Unless otherwise noted, the cumulative incidence of ZIKVassociated neurological complications was estimated using reported cases of ZVD as the denominator. For incidence by geographic location, the total population was included for each reporting area regardless of altitude (except for the capital city of Bogotá). Confidence intervals were calculated using the binomial exact function in the R package epitools (version 0.5-10.1). Risk ratios (RR) and 95% confidence intervals were calculated using the riskratio function in the R package fmsb (version 0.7.0).
In order to estimate the incidence of ZVD by pregnancy status, it was necessary to estimate the number of pregnant females and non-pregnant females in Colombia. In 2017, the Colombian Ministry of Health estimated the annual number of pregnant females for 2017-2019. The mean estimate of 822,396 for 2017 was multiplied by ¾ (females are only pregnant for ¾ of the year on average) to obtain the number of pregnant females in the population at any time (616,797). To obtain the number of females who were not pregnant at any time, this number was subtracted from the projected total number of females in the population in 2016 (24,678,673, from DANE).
In the ZIKV dataset, 1,636 females had missing pregnancy status.

ZIKV
Sex and age trends. More ZVD cases were reported in females than males each year from 2015-2017 (Table 1). Overall, females represented two-thirds (66.2%) of reported cases, which differs significantly from the general population of Colombia (50.6% female [exact binomial test, p < 0.001]). The RR of ZVD in females was nearly two times higher than in males (1.91, 95% CI: 1.88-1.93).
The median age of ZVD cases in Colombia was 29 years (range 0 to >100 years). Nearly half (49.0%) of cases were reported in individuals between the ages of 20 and 39. The number of cases reported in this age group was significantly different than expected compared to the general population of the same age group (31.2% [exact binomial test, p < 0.001]). The cumulative incidence of ZVD was significantly higher for females compared to males across all age groups except 0-4 years and those 80 years and over (Fig 1). Among those 15 to 29 years of age, the risk of ZVD was about three times higher in females compared to males. The highest RR of female to male cases was observed in the 20 to 24-year age group (3.16, 95% CI: 3.04-3.28).
Pregnancy. A total of 19,243 ZVD cases was reported in pregnant females. Pregnant females had much higher risk of being reported as a ZVD case compared to males and nonpregnant females. The cumulative incidence of ZVD in pregnant females was 3,120 reported cases per 100,000 population (95% CI: 3,077-3,164). In contrast, the incidence in non-pregnant females was 205 cases per 100,000 population (95% CI: 203-207), and the incidence in males was 149 cases per 100,000 population (95% CI: 147-151). The differences in incidence of ZVD by sex and pregnancy status were significantly different.
Geographical distribution. Cases of ZVD were reported in all 32 departments of the country. The departments with the highest number of reported ZVD cases included Valle del Cauca (27,712), followed by Santander (10,374) and Norte de Santander (10,361). The highest incidence of ZVD per 100,000 population was reported in San Andrés and Providencia (1,489), followed by Casanare (1,087) and Norte de Santander (758) (Fig 2).
Temporal trends. The peak of the ZIKV epidemic occurred during the week ending on    [32]. The shapefiles were prepared by OCHA and are available under a CC BY-IGO license [33]. No changes were made to the shapefiles. https://doi.org/10.1371/journal.pone.0252236.g002

PLOS ONE
Zika virus disease and Zika virus-associated neurological complications in Colombia have two peaks (Bolívar and Cundinamarca). Chocó, Guainía, Nariño, Vaupés, and Vichada had irregular time series due to small numbers of reported cases.

Neurological complications
Sex and age trends. The median age of ZVD cases with neurological complications was 41 years (range 0-93). The highest number of neurological complications was reported in middle-aged groups from 35-49, and the cumulative incidence of neurological complications was highest in those at least 75 years of age with a high degree of uncertainty.
For both sexes, the greatest number of complications was reported in middle-aged groups. The cumulative incidence of neurological complications per 1,000 cases of ZVD by age and sex is shown in Fig 5. For both sexes, the incidence of neurological complications increased with age. Males aged 75 and older had the highest incidence but also the most uncertainty, which is reflected in the wide confidence intervals.
Diagnosis and final condition. Three hundred and fifty-two patients (85%) with ZIKVassociated neurological complications were diagnosed with GBS. Encephalitis and myelitis were the second and third most common diagnoses, respectively, with 29 and 17 cases each. Facial paralysis was recorded for 11 patients. Meningitis (N = 2), meningoencephalitis (N = 2), and optic neuritis (N = 1) were the least commonly diagnosed complications.
Thirty-five cases (8.4%) resulted in death. Geographical distribution. Twenty-eight out of 32 departments in Colombia were reported as the location of likely infection for at least one case of ZIKV-associated neurological complications (Fig 6). The highest number of cases was reported in Atlántico with 104. Norte de Santander and Valle del Cauca had the next highest number of cases with 61 and 44, respectively. The highest cumulative incidence of ZIKV-associated neurological complications per 1,000 cases of ZVD was reported in Nariño followed by Chocó and Atlántico (Table 2 and Fig  6). However, 24 departments had 10 cases or fewer, and the estimated incidence should therefore be interpreted with caution.
At the municipality level, Barranquilla was reported as the location of likely infection for the highest number of ZIKV-associated neurological complications with 80, followed by Cúcuta with 44 and Cali with 23. Cases were spread out geographically and tended to cluster in large cities.
Temporal trends. The distributions of ZVD cases and cases of ZIKV-associated neurological complications over time are similar. There are differences, however, in the time series of   [32]. The shapefiles were prepared by OCHA and are available under a CC BY-IGO license [33]. No changes were made to the shapefiles.
https://doi.org/10.1371/journal.pone.0252236.g006  neurological complications when date of symptom onset is considered rather than date of notification. The week with the highest number of neurological complications according to date of symptom onset was the same week in which the most ZVD cases were reported (30 cases during the week ending on February 6, 2016). In contrast, the most neurological complications according to date of notification were reported two weeks earlier, in the week ending on January 23, 2016 (33 cases). Similar to the ZIKV dataset, which did not have date of symptom onset for all cases, date of notification (reporting) of ZIKV-associated neurological complications seemed to decrease during the Christmas/New Year and Easter holidays.
Although the distribution of neurological complications by sex did not vary across time, more cases were consistently reported in males than in females throughout the epidemic.

Discussion
This work builds on a preliminary report of ZVD cases in Colombia that also used data from the national population-based surveillance system [26]. The earlier report was published just after the peak of the epidemic and included 65,726 cases reported between August 9, 2015 and April 2, 2016. This analysis adds an additional 40,307 cases reported until June 17, 2017 and data on severe cases with neurological complications, including GBS.

ZIKV
Compared to the general population of Colombia, ZVD cases were more likely to be reported in individuals in their 20s and 30s. Several factors can affect the age distribution of cases found through epidemic surveillance, including age-related variation in susceptibility, reporting bias, pre-existing immunity, the age distribution of the population, and level of exposure to infection [34].
Seroprevalence studies, which test for antibodies indicative of past infection, can be used to assess age-related variation in susceptibility. Although at least one such study found a positive association between ZIKV infection and age, several studies in different countries have found no significant association [35][36][37][38][39][40].
Based on the timing and origins of ZIKV arriving in the Americas from Southeast Asia and the Pacific [41], no immunological protection for ZIKV was assumed at the population level prior to this epidemic. If there was pre-existing immunity, lower infection rates would have been expected in older age groups that had been exposed in the past assuming long-lasting immunity.
Risk factors for ZIKV infection are poorly understood [37]. On Yap Island, Duffy et al. found no behavioral risk factors for ZIKV infection [35]. In contrast, Lozier et al. found higher prevalence of ZIKV among those who reported being bitten by mosquitoes at home in bivariate analyses, but the association was not statistically significant in multivariable analysis [36].
In this study, females had higher cumulative incidence of reported ZVD than males. Case ascertainment was likely higher in females of child-bearing age due to concerns about birth defects. However, reporting bias does not explain the elevated risk in females versus males between the ages of 45 and 79. This result could be explained by differences in susceptibility or exposure. If females in this age range spend more time at home than their male counterparts, they might experience higher exposure to Ae. aegypti mosquitoes, which tend to live in and around people's homes in urban areas [42,43]. The higher incidence of ZVD in females is consistent with ZIKV epidemics in other locations such as Brazil, Puerto Rico, and Yap [35,36,44]. Spending more time at home might also increase the risk of exposure to mosquito bites and therefore ZIKV infection in young children, which could explain the higher incidence of ZVD in the youngest age group (0-4 years) compared to older children in this study.
Notification of ZVD cases decreased during Christmas/New Year and Easter holidays in 2015-2016. This trend could be seen at the national and subnational level as well as during the CHIKV epidemic. Changes in the number of patient consultations for general practice services on and immediately after public holidays have been observed in several countries, including the U.K. This has been called the "public holiday effect" [45]. In Colombia, where 93% of the population identifies as Christian, surveillance for notifiable diseases is likely impacted by religious holidays [46].

Neurological complications
Despite higher cumulative incidence of ZVD cases in females compared to males, higher incidence of neurological complications was observed in males, a finding which is consistent with GBS epidemiology and other studies [47,48].
The median age of ZVD cases with neurological complications was 12 years older than that of ZVD cases. The cumulative incidence of ZIKV-associated neurological complications increased with age, and the highest incidence was reported among individuals aged 75 and older. The positive association between age and neurological complications here is also consistent with GBS epidemiology [47].
Time of symptom onset of ZIKV infection was not available for all cases who went on to develop neurological complications in this dataset. According to a 2016 study of 71 patients in Puerto Rico, the median time between onset of ZIKV infection symptoms and GBS was seven days (range 0-21 days) [13]. Similarly, a 2015-2016 study reported that the median time between onset of ZIKV infection symptoms and GBS symptoms was 7 days (interquartile range, 3-10 days) for 66 patients from six Colombian hospitals [14].
The most common diagnosis among cases in the neurological complications dataset was GBS. However, six other neurological conditions were also documented, including encephalitis, myelitis, facial paralysis, meningitis, meningoencephalitis, and optic neuritis. Although some studies have focused exclusively ZIKV-associated GBS [13,14,49 . Some of these reports involved fatalities, young patients, and previously healthy individuals. In addition to patients with ZIKV-associated neurological complications and CZS, studies using human and animal models have accumulated broader evidence that ZIKV is neurotrophic. The virus targets neuronal cell types in the brain, including neural progenitor cells, mature neurons, and astrocytes [59]. ZIKV infection of the central nervous system has been found in both young and adult animals such as mice and non-human primates [59].
Reports of neurological complications associated with ZVD were reported in nearly every department and tended to cluster in large cities with better access to healthcare. This pattern reflects the widespread dissemination of ZIKV throughout Colombia. The city of Barranquilla had the highest number of reported neurological complications. While there was a large ZIKV epidemic in Barranquilla, the city was also subjected to more intensive surveillance for ZIKVassociated neurological complications compared to other cities [60].
There is some agreement between locations with the highest number of ZVD cases and locations with the highest number of ZIKV-associated neurological complications. However, the locations with the highest cumulative incidence of ZIKV per 100,000 population and locations with the highest incidence of neurological complications per 1,000 ZVD cases are discordant. This mismatch could be due to randomness associated with reporting small numbers of rare events or differences in reporting mild versus severe ZVD cases across the country.

Conclusions
A strength of this analysis is the quality of the datasets. The ZIKV dataset encompasses the entire duration of the epidemic in Colombia, and all patients in the neurological complications dataset were checked against standardized case definitions. Limitations include lack of detailed clinical information and lack of laboratory confirmation for most ZVD cases. Interestingly, although our report does not include any congenital complications, a recently published report found that out of 5,673 pregnancies with laboratory-confirmed ZVD in Colombia, 2% of infants or fetuses had neurological or eye complications [61]. Another recent study from Colombia found that nine out of 60 children (15%) with laboratory-confirmed ZIKV infection at ages 1-12 months had adverse outcomes on neurologic, hearing, or eye examinations at 20-30 months of age. Six of the remaining 47 children (12.8%) had an alert score in the hearinglanguage domain [62].
Neurological complications and deaths due to ZIKV were rare in this epidemic. However, more awareness about these risks is needed for people living in or traveling to ZIKV-affected areas. While GBS is relatively easy for non-neurologists to identify, variants such as Miller-Fisher syndrome may not be [57]. Future research should investigate long-term patient outcomes as well as the pathophysiology of these conditions, which can improve treatment strategies [53]. To fully understand the burden of ZIKV, surveillance should encompass a broader spectrum of neurological symptoms of ZVD beyond GBS and microcephaly. Surveillance should also focus on young children, considering the neurotropism of the virus and its effects on postnatal development.