Ethics statement

All patients were managed in accordance with the amended Declaration of Helsinki (https://www.wma.net/what-we-do/medical-ethics/declaration-of-helsinki) and Good Clinical Practice guidelines. Each child’s legal representative was informed about the study orally and by personal correspondence, giving them the opportunity to explicitly refuse the use of their child’s medical data for research purposes. Oral formal consent was provided by parents or legal guardians of study participants on enrollment. The study was approved by the French National data privacy commission (CNIL, Commission Nationale Informatique et Liberté) under the number N°1897956.

Study setting

Martinique is a French overseas territory belonging to the Lesser Antilles in the Eastern Caribbean Sea, with a population of approximately 384,000 inhabitants. On 18^th December 2013, health authorities reported the first confirmed indigenous CHIKV cases in Martinique, and the outbreak ended in January 2015.

Study design and patients

We conducted a single-center, observational, longitudinal, retrospective study, investigating the neurological manifestations of CHIKF in children admitted to the University Hospital of Martinique, between 1^stJanuary 2014 and 31^st January 2015. It is the sole hospital for the whole island providing inpatient hospitalization services for sick children.

At the start of the epidemic in Martinique (December 2013), the pediatric team, alerted by the description of neurological involvement during the epidemic in the island of Réunion, issued recommendations for the systematic investigation by lumbar puncture (LP) of all suspected cases of neurological involvement.

For the present study, we included all patients aged more than 10 days and less than 18 years, who consulted the pediatric emergency department or were hospitalized at the University Hospital of Martinique, and who presented with CHIKV infection requiring a lumbar puncture (LP).

In line with guidelines for pediatric management, the indications for lumbar puncture were as follows:

• Before 3 months, mainly microbial testing for suspected sepsis, in line with French guidelines. At this age, neurological manifestations usually cannot be distinguished from sepsis.
• Between 3 and 12 months: suspected sepsis or neurological impairment.
• Between 1 and 5 years of age: neurological impairment, except for isolated seizure
  • For children over 5 years’: neurological impairment including isolated seizure

Newborns less than 10 days old were excluded from the analysis to avoid any confusion with maternal-fetal transmission (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow diagram of the study population.

https://doi.org/10.1371/journal.pntd.0014112.g001

Children with bacterial or fungal meningitis were also excluded.

Patient follow-up

Children aged >10 days and < 3 months, as well as those who presented with severe manifestations during hospitalization for CHIKV infection, were followed up between 1 and 6 months after discharge. Furthermore, all study patients were invited to a follow up between 3.5 and 4.5 years after the initial hospital stay. During the follow-up visit, their development quotients were determined as follows: the youngest children (< 2.5 years of age) were tested with the complementary Brunet Lézine tests and the Child Development Inventory [34]. For older children, the Raven’s progressive matrices, the Bell test, the memory of numbers and the Rey–Osterrieth complex figure (ROCF) tests were used. For all children, an interview was conducted with one or both of the child’s parents. The psychomotor development was recorded based on the parents’ statements and the child’s health records.

Study data

For each child, we recorded the following variables: date of diagnosis, age, sex, length of hospital stay, body temperature at admission, dermatological manifestations, neurological manifestations, need for admission to the pediatric intensive care unit (PICU), history of neurological disorders, as well as variables related to CHIKV infection, namely pain level as reported using the French EVENDOL scale or self-evaluated (by visual analog scale, VAS) depending on the child’s age, arthralgia, laboratory parameters and follow-up data (head circumference, progress of child’s schooling). All data were retrospectively retrieved from the medical records.

Criteria for defining and grading severity

The severity of the initial clinical neurological manifestations was classified into three categories, namely severe, intermediate or non-severe according to the definitions of the International Encephalitis Consortium [35].The major criterion was the presence of consciousness disorders, while minor criteria were as follows: fever > 38°C within 72 hours before or after consciousness disturbances, generalized or partial convulsive seizures not related to pre-existing epilepsy, new onset of focal neurological signs, presence of ≥ 5 leukocytes/mm³ in the CSF, imaging or electroencephalogram (EEG) abnormality related to encephalitis.

Manifestations were deemed to be severe in the event of: (1) confirmed or probable encephalitis (major criterion and ≥ 3 minor criteria); (2) possible encephalitis (major criterion + 2 minor criteria); (3) other systematized neurological impairment. Severity was deemed to be intermediate in case of non-encephalitic-CHIKV-associated central nervous system disease (NECACD), defined by the presence of either (1) major criterion + 1 minor criterion, or (2) major criterion alone, or (3) 2 minor criteria except fever. In the absence of these parameters, the presentation was qualified as non-severe.

Laboratory procedures

At admission, prospective blood CHIKV RT-PCR was performed for all patients between day 1 and day 7 after disease onset. Measurement of CHIK viral load was performed retrospectively on a frozen aliquot.

As the dengue epidemic was winding down when CHIKV emerged, patients were also tested for dengue RNA, at the start of the epidemic and subsequently only in severe cases. Nucleic acid extraction was performed on 200 µL plasma or CSF with an elution volume of 50 µl using EasyMag Nuclisens from BioMérieux (Meylan, France). For detection of viral RNA, we used the RealStar Chikungunya RT-PCR kit 1.0 (Altona Diagnostics, Hamburg, Germany) with amplification on an Applied Biosystems7500 Real-Time PCR System (Thermo Fisher Scientific, USA) and Simplexa Dengue RT-PCR assay (Focus Diagnostics, Cypress, CA) on a 3M Focus system.

Lumbar puncture was performed on admission or at the onset of neurological symptoms. The same technical procedures for CHIKV RT-PCR and dengue RT-PCR were applied to the CSF, with additional detection of herpes simplex virus and varicella zoster virus, using RealStar alpha Herpesvirus PCR 1.0 kit from Altona Diagnostics (Hamburg, Germany) and enterovirus, using ENTEROVIRUS R-gene from Argene (France).

For extraction/amplification of RNA, negative and positive controls (strain dilutions) were included in every round. The University Hospital virology laboratory participates every year in the annual program for External Quality Assessment (EQA)/ Proficiency Testing (PT) from Quality Control for Molecular Diagnostics (QCMD) (Scotland, UK), for all the viruses considered in the present study.

For measurement of viral load, RT-PCR was performed using serially diluted RNA standards with RealStar Chikungunya RT-PCR kit. The cycle threshold (Ct) values obtained were plotted against the log dilution of the plasmid to generate the standard curve. VL in the blood and CSF are presented on a log scale. The ratio was defined as. log CSF VL (copies/mL)/ log blood VL (copies/mL).

Statistical analysis

For all descriptive and inferential analyses, the normality of the data distribution was tested. Means and standard deviations are reported for normally distributed variables. For non-normally distributed variables data are reported as median (minimum (min)-maximum (max) or interquartile range). Categorical variables are reported as number and percentage. The following tests were used for group comparisons when appropriate: Student t test, Wilcoxon-Mann-Whitney test, Chi square test and Fisher’s exact test. A logistic regression model was used to assess the association between neurological severity (dependent variable) and viral load in CSF (independent variable), adjusted for age and time between onset of clinical symptoms and lumbar puncture. Only patients with non-hemorrhagic CSF were included. The level of statistical significance was set at p < 0.05. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

Clinical features at admission

This single-center, observational, longitudinal, retrospective study reports the neurological manifestations observed in children hospitalized during the 2014–2015 CHIKV epidemic in Martinique, and analyzes the relation between CHIKF disease severity and the viral load in the blood and CSF.

All children had fever and were admitted to hospital at the acute phase of infection. Neurological manifestations were the main cause of hospitalization in 40.7% of the children, mainly those older than one year. The onset of neurological signs was concomitant with the onset of fever in 84.9% of the children in our study, which is similar to the 73% reported in the study from Reunion Island by Robin et al. [8,13]. Seizures are common in children, varying from 12% [14] to 56.3% [19]. In the present study, seizures occurred in all age ranges, and was the main reason for lumbar puncture in children aged over 1 year (22 out of 28).

The leading causes of PICU admission in our cohort were severe neurological manifestations (26.7%), very young age and/or poorly tolerated fever or pain, in line with the 24% and 37.9% respectively described elsewhere [8,36]. A septic aspect was the leading reason for lumbar puncture in children younger than 3 months.

Biological parameters, EEG and neuroimaging

The lymphopenia, normal or slightly elevated proteinorachia and low or absent pleocytosis observed in our study are common in CHIKV-infected infants [8,11,13,37]. Slowing waves and paroxysmal peaks, observed on the 21 abnormal EEG of our study patients, and the unremarkable results of the neuroimaging were similar to observations in other studies [8,16].

Viral loads

During the first days of illness, the VL of CHIKV in the blood is known to be very high [26,38,39]. It depends on age and is higher in newborns and older adults [16,40]. Our results (mean VL in the blood > 9 log) are not unexpected, since all the children were hospitalized early after the onset of clinical signs. Moreover, blood VL varied significantly across age groups, with infants younger than 3 months presenting with the highest VL (p < 0.0001).

The possible association between blood VL and clinical manifestations and/or disease severity remains contradictory [41,42]. We found no clear relationship between blood VL and the clinical presentation of children, similar to the findings from an Indian study [26]. This suggests that the blood viral load may not contribute significantly to the severity of CHIKV infection.

In addition to blood screening, the presence of CHIKV RNA in the CSF has also been studied as a means to explore the virological basis of neurological manifestations” [8,13], In our study, all CSF samples were positive except for two cases, probably due to delayed hospitalization (>3 days after symptom onset) in one case; the cause remains unexplained for the other.

CSF VL in CHIKF patients has been reported to be related to plasma VL and is generally lower [16,28]. The results of this study are in agreement with those found in a cohort of encephalitis and NECACD in Reunion Island [15]. As with plasma VL, CSF VL was significantly higher in patients aged less than 3 months (p < 0.001) in our study cohort. These findings further report that blood VL was on average 4.5x10⁵ higher than CSF VL. The ratio (CSF VL/Blood VL) varied across age groups. Since the time between lumbar puncture and blood sampling was ≤ 24h, and hemorrhagic CSF samples were excluded from calculation, these findings suggest that there might be passive diffusion of viral RNA from the blood to the CSF, rather than intrathecal replication. This hypothesis should have been further explored by assessing cytokines and intrathecal antibody response, but this could not be done given the small quantity of CSF available and the retrospective study design.

Severity of the neurological manifestations and outcome

Seven of the 86 (8.1%) children included in this study developed encephalitis; when compared with the 651 positive cases diagnosed in hospital, this rate fell to 0.1%.

These results seem less worrying than those reported from Reunion Island, [8,16]. Among children hospitalized for CHIKF in Honduras in 2015, it was reported that 6% had meningoencephalitis [18]. CHIKV strains are different, with an Asian lineage in Martinique and the IOL strain in Reunion Island, but likely, the same strain in Martinique and Honduras. However, a possible difference in neurovirulence between ECSA and Asian strains cannot be ruled out. This question remains to be studied by comparing the clinical and virological data of patients with severe neurological signs associated with CHIKV. Patient recruitment was also different, since newborns were excluded from our study, and all the children had CHIKV infection proven by positive RT-PCR in the blood and concomitant screening in the CSF at the acute phase of infection.

Severity does not seem to be related to neurological history, since no encephalitis occurred in the 18 children who had previous neurological manifestations.

VL and ratios were not significantly different in the 3 groups (severe, intermediate and non-severe), as in the study of Gerardin et al. [16], where they studied a cohort including infants and adults. In our study, children aged over one year accounted for up to 75% of severe cases, whereas they represented only 32.6% of the cohort. The differences in severity according to age are biased in our study since the reason for undergoing LP, and therefore for inclusion in the cohort, was different for infants and for children over one year of age. However, at any age, the severity of neurological manifestations did not differ according to the VL or the ratio.

Our study showed an almost constant spread of CHIKV in the CSF, raising the question of possible cognitive sequelae in children. Severe cognitive disorders were observed in newborns infected during the epidemic on the Island of Réunion [22].

Our preliminary study of the children’s psychomotor development is worrying, although the low number of patients, and the absence of control group precludes drawing any firm conclusions.

None of our patients died, whereas other studies reported mortality of up to 6.6% [20,22]. During the same epidemic, the crude mortality rate was 26% among 65 adults (10 of whom had acute neurologic manifestations) admitted to intensive care units in Martinique and Guadeloupe (a neighboring French island) [9]. These data suggest that the fatality rate of CHIKV infection seems to be less related to the viral genotype, and more to other factors, including older age and comorbidities.

The linearity of CSF VL and ratios according to age is all the more remarkable since our study includes pediatric populations from different age groups. This observation argues in favor of the hypothesis of passive diffusion of CHIKV in the CSF, and no role of intrathecal replication in the pathophysiology of neurological impairment. The latter therefore remains to be explored. The spread of the virus in the CSF has been demonstrated in a mouse model. After peripheral inoculation followed by viremia, the CHIKV can spread and invade the CNS via the choroid plexus, then spread in the CSF [32].

The great variability in the incidence of this disease during the various epidemics around the world raises two hypotheses: the role of viral variants and that of the immune specificities (genetics or cross-protection) of each population. The early onset of neurological manifestations in our study suggests, as with dengue encephalopathy, that there is a vascular mechanism secondary to primary cytokine response. The only three autopsies we found in the literature are in favor of this hypothesis [30,31].

Recently, in a study screening many cytokines, Santos Almeida and coworkers [43]. proposed the IL-22/IL-17A ratio as the best marker of neurological damage in adults infected by Arbovirus. These authors suggested that this cytokine profile promotes the breakdown of the blood-brain-barrier. A study of cytokines in the CSF could provide a better understanding of the mechanisms and treatment of neurological damage. However, since our study has shown that LP does not yield any benefit in the immediate management of children, in the future, LP will only be ethically justified for differential diagnosis and severe forms.