Guillain-Barré syndrome related to Zika virus infection: A systematic review and meta-analysis of the clinical and electrophysiological phenotype

Background The Zika virus (ZIKV) has been associated with Guillain-Barré syndrome (GBS) in epidemiological studies. Whether ZIKV-associated GBS is related to a specific clinical or electrophysiological phenotype has not been established. To this end, we performed a systematic review and meta-analysis of all published studies on ZIKV-related GBS. Methods We searched Pubmed, EMBASE and LILACS, and included all papers, reports or bulletins with full text in English, Spanish or Portuguese, reporting original data of patients with GBS and a suspected, probable or confirmed recent ZIKV infection. Data were extracted according to a predefined protocol, and pooled proportions were calculated. Results Thirty-five studies were included (13 single case reports and 22 case series, case-control or cohort studies), reporting on a total of 601 GBS patients with a suspected, probable or confirmed ZIKV infection. Data from 21 studies and 587 cases were available to be summarized. ZIKV infection was confirmed in 21%, probable in 22% and suspected in 57% of cases. ZIKV PCR was positive in 30% (95%CI 15–47) of tested patients. The most common clinical features were: limb weakness 97% (95%CI 93–99), diminished/absent reflexes 96% (95%CI 88–100), sensory symptoms 82% (95%CI 76–88), and facial palsy 51% (95%CI 44–58). Median time between infectious and neurological symptoms was 5–12 days. Most cases had a demyelinating electrophysiological subtype and half of cases were admitted to the Intensive Care Unit (ICU). Heterogeneity between studies was moderate to substantial for most variables. Conclusions The clinical phenotype of GBS associated with ZIKV infection reported in literature is generally a sensorimotor demyelinating GBS with frequent facial palsy and a severe disease course often necessitating ICU admittance. Time between infectious and neurological symptoms and negative PCR in most cases suggests a post-infectious disease mechanism. Heterogeneity between studies was considerable and results may be subject to reporting bias. This study was registered on the international Prospective Register of Systematic Reviews (CRD42018081959).


Introduction
Guillain-Barré syndrome (GBS) is the most common cause of acute flaccid paralysis worldwide, with an incidence rate of approximately 1 per 100,000 person-years. [1] GBS is an acute immune-mediated polyradiculoneuropathy, and is presumed to be triggered by preceding infections with specific pathogens, such as Campylobacter jejuni, cytomegalovirus (CMV), and Epstein-Barr virus (EBV). [2] Recently, the incidence of GBS increased during Zika virus (ZIKV) epidemics in French Polynesia (2013) and Latin America (2015-2016) and an association between GBS and ZIKV was established through epidemiological studies. [3,4] The classic form of GBS is characterized by a rapidly progressive and symmetrical weakness of the limbs, with sensory symptoms and reduced or absent tendon reflexes. [4] Cranial nerve involvement is frequent, with facial and bulbar muscles most often affected. [5] Electrophysiological studies help to confirm the diagnosis of GBS, and can indicate different subtypes, including acute inflammatory demyelinating polyradiculoneuropathy (AIDP), acute motor axonal neuropathy (AMAN), and acute motor and sensory axonal neuropathy (AMSAN). [4] The majority of patients will lose the ability to walk during the acute phase of the disease and about 25% of patients need to be mechanically ventilated at the Intensive Care Unit (ICU). [6] Clinical presentation and severity of GBS can vary extensively between patients. This variability is thought to be, in part, caused by differences in the type of preceding infections. For instance, C. jejuni has been associated with a pure motor axonal form of GBS with a severe disease course, while CMV has been linked to a sensorimotor GBS with pronounced respiratory insufficiency. [6][7][8] Since the ZIKV epidemics, numerous studies have been published on ZIKV-related GBS, but it has not been established if there is a specific clinical and electrophysiological phenotype of GBS after ZIKV, and whether this differs from GBS triggered by other pathogens. [3,4] Therefore, we have performed a systematic review and meta-analysis of all published studies on ZIKV-related GBS, and give a comprehensive overview of demographic characteristics, clinical features, diagnostic investigations, and outcome of ZIKV-related GBS patients.

Methods
This systematic literature review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and was registered on the international Prospective Register of Systematic Reviews (PROSPERO) with number CRD42018081959. [9] Information sources and search strategy First, by selecting key words from relevant articles, search strategies were constructed for the Pubmed, EMBASE and LILACS databases (Fig 1), which were searched on 24 November 2017 and on 24 January 2019. Second, the titles and abstracts were screened by two researchers (JDLB and SC) to identify the key words ('Guillain-Barre Syndrome', 'viruses', 'virus', 'Zika virus' and 'Zika'), and to exclude in vitro or in animal studies and reports from meetings or congresses. The selected papers were read in full by two independent reviewers (CCBS, MFPMA) and a third reviewer (SEL) was consulted in case of disagreement.
We included all papers, reports or bulletins with available full text in English, Spanish or Portuguese, without restriction in year of publication, reporting original data of patients with GBS and a suspected, probable or confirmed recent ZIKV infection, of any age, gender and in any setting. Predefined exclusion criteria were: GBS within 3 months after a vaccination or other proven triggering infection (e.g. C. jejuni), and studies with no information on age, residence, and at least one clinical variable of interest. When the study population of reported cases overlapped with cases published in other papers, the paper reporting the highest amount of cases was included. When only part of the cases in a study fulfilled our inclusion criteria, only these cases were included, but if separate data of these cases were not available after contacting the corresponding author, the article was excluded.

Data extraction and management
Data were extracted independently by one of three reviewers (CCBS, MFPMA, SEL) according to a predefined protocol. The data extraction was then checked by one of the other two reviewers, and discrepancies were solved by discussion among all of them. Variables of interest comprised demographics, clinical characteristics (symptoms and signs of arbovirus infection and GBS), ancillary diagnostic investigations (electrophysiology and CSF), treatment, clinical course, and outcome of GBS. The corresponding authors were requested to share data on variables of interest that were not reported.
Cases were classified according to the reported diagnostic certainty levels for GBS and ZIKV infection. To classify the diagnosis GBS we employed the Brighton Collaboration

PLOS NEGLECTED TROPICAL DISEASES
Criteria (2011). [10] If the Brighton Criteria were not reported, these were defined based on available reported data, and if the clinical description did not correspond to the reported Brighton level, cases were reclassified after clarification was sought with the corresponding author. The diagnostic certainty of ZIKV infection was classified as confirmed, probable or suspected, according to the Centers for Disease Control and Prevention (CDC) criteria [11] ( Table 1), based on the results of laboratory tests: case-by-case in case reports and series, and all cases combined in larger studies.
Clinical characteristics were retrieved as the number of patients in whom the variable was present in the numerator, and the total number of reported cases in the denominator: n/N (%). For arbovirus symptoms, we assumed symptoms were absent rather than missing if they were not cited in the manuscript, to account for the reporting bias, and therefore described as zero (n) out of the total number of reported cases (N). For the neurologic findings, variables not cited were considered missing data, because a risk of measurement bias was deemed higher than a risk of a reporting bias for these variables. If clinical characteristics were reported at multiple time points, data representing the full disease course were presented. Continuous variables (age, time between infectious and neurologic symptoms, duration of progression and plateau phase of GBS, duration hospital admission) were extracted as medians and or means, depending on how they were presented in the original article.

Statistical analysis
First, we calculated the proportions per study of each variable of interest, and then the pooled proportions with data from all included studies reporting more than one GBS case. We were unable to summarize continuous variables, as in most studies these were reported as medians without availability of individual data or means. To address the possibility of an ascertainment bias of ZIKV infection among study populations, we then performed a subgroup sensitivity analysis, repeating the pooled analysis with grouped data of only probable or confirmed ZIKV cases (overall study populations comprising only probable/confirmed ZIKV cases, and subsamples of probable/confirmed cases from studies that also included suspected ZIKV cases, when available). We also performed sensitivity analyses by excluding papers that recruited only ICU patients, to account for selection bias in the pooled proportion of mechanical ventilation and ICU assistance.

Study selection
We identified 1716 articles in the databases researched, of which 35 studies were included in our systematic review. The 35 selected studies reported on a total of 601 GBS cases with a suspected, probable or confirmed ZIKV infection with data of at least one variable of interest, and consisted of 13 single case reports and one cohort in which only one case fulfilled our inclusion criteria (n = 14, Table 2), and 14 case series and seven case-control studies (n = 587, Table 3).
For the pooled analysis of the studies, we were only able to use the studies that reported on more than one case. ( Table 3). For the subgroup meta-analysis of probable/confirmed ZIKV cases, data of 165 GBS cases with probable or confirmed ZIKV infection, from 14 studies, could be pooled (Fig 1).

Study characteristics: case selection, case ascertainment and risk of bias
In Table 2, the single case reports are presented alphabetically with a brief clinical description per case. Eleven cases were from ZIKV epidemic or endemic regions and three were travelers returning from epidemic regions. Eight cases were positive for ZIKV PCR, four for IgM and plaque-reduction neutralization test (PRNT), and two were reported to be ZIKV positive with no further information provided. Six of eight cases of whom the Brighton classification was reported, fulfilled level 1. The most frequent clinical phenotype was a demyelinating sensorimotor GBS with facial and/or bulbar palsy.
In Table 3, the 21 studies reporting more than one patient are displayed according to the location and time-period of cases, in line with the global spread of the ZIKV epidemics on the Pacific islands (Oct 2013-Dec 2014) and Latin America (Dec 2014-2017). The first study was from French Polynesia in 2013-2014, [4] and the last was from Mexico in 2016-2017. [14] One study reported cases during and outside of a ZIKV outbreak period in Singapore [15] Inclusion criteria, case selection and setting differed between studies. A diagnosis of GBS was the inclusion criterion in 14 studies, and seven studies also included other acute neurologic illnesses besides GBS. [16][17][18][19][20][21][22] Six studies included all GBS patients in their reference population, [4,15,[23][24][25][26] one study included all GBS patients >12 years old, [27] and one study included all arbovirus-related neurologic manifestations. [19] All other studies included a convenience sample of patients seen at one or more health-care centres. Three studies only included patients admitted to the ICU, [17,22,28] and nine studies only included GBS patients with a clinical suspicion or laboratory evidence of a ZIKV infection. [14,17,[20][21][22][28][29][30][31] Seven studies were set in a specialized hospital (academic or reference centre), [4,16,18,23,26,31,32] and two multi-centre studies were set in both specialized and non-specialized hospitals. [21,28] These differences are potential sources of selection bias within studies and heterogeneity across studies.
Sixteen studies reported the criteria that were applied for diagnostic certainty of GBS, and 13 used the Brighton Criteria. In four studies the Brighton Criteria were prospectively applied by a physician; in seven, retrospectively through records review; two studies gave no information on how the Brighton level was assessed; and three employed other criteria. The risk of ascertainment bias of GBS is likely to be low or very low, as the vast majority of all cases with this data available in this review fulfilled Brighton levels 1-3 (396/407, 97%).

Patient characteristics
Demographics. The median age of the study populations varied between 34 and 61 years, and only 11 pediatric patients were included in four studies. [19,24,27,30] The majority of patients was male (62%) and the male:female ratio of all studies combined was 1.63. In multicenter studies or those including all GBS cases in the reference population, the male:female ratio was 1:1, with the exception of studies from French Polynesia [4] and Martinique [26], which had ratios of 3:1 and 2:1, respectively ( Table 3).
Diagnostic investigations. PCR, principally in serum, was the most frequently performed test for ZIKV diagnosis, although anti-ZIKV IgM was positive twice more often ( Table 5). In the CSF, ZIKV PCR was positive in only 10 of 244 tested cases. Presence of neutralizing antibodies against ZIKV in the serum was tested in eight studies. [4,15,17,23,24,26,27,31] To differentiate ZIKV from DENV, IgM antibodies against DENV were tested in 18 studies (426 cases), and were positive in 70 patients. [4, 14-18, 21, 23-33] Of these patients, 54 were also positive for ZIKV PCR, IgM and/or ZIKV neutralizing antibodies, and in 16 cases no separate information on ZIKV test results was available. Infection with CHIKV was investigated in nine studies and 187 cases, of which 16 were PCR or IgM positive. [14, 17, 19-21, 25, 28, 30] Only five studies tested all ZIKV suspected cases for other infections that have been associated with GBS (C.jejuni, CMV, EBV, Hepatitis E virus, Mycoplasma pneumoniae). [4,17,23,26,31] And all tested cases (80/587; 14%) were negative for recent infection. None of the studies tested for all of these pathogens. Heterogeneity was considerable for all ZIKV laboratory tests (I 2 = 75-100%).
CSF was examined in most studies, and information on protein level and cell count was provided by about half of these. Increased protein level and albuminocytological dissociation were present in the vast majority of cases and results were similar between all studies combined and the probable/confirmed subgroup. Eleven studies reported the CSF cell count, which did not exceed 55 cells/mm 3 , and medians were below 5 cells/mm 3 (Fig 4). [4, 16-18, 20, 21, 26, 27, 29, 31, 32] Heterogeneity was limited for increased protein level and albuminocytological dissociation in all studies combined and the probable/confirmed subgroup.

PLOS NEGLECTED TROPICAL DISEASES
Electrophysiological studies were done in about half of reported cases. In five studies no information on electrophysiological examination was reported. [16,19,[28][29][30] Criteria used to classify cases into the different electrophysiological subtypes were reported in only five studies, [18,23,26,31,32] and included criteria by Hadden et al, Ho et al, and Rajabally et al. [50][51][52]. The most frequent electrophysiological subtype was AIDP in 62% (95%CI 38-83), followed by AMAN in 16% (95%CI 0-41), with both similar pooled proportions in the probable/confirmed ZIKV subgroup. In most studies, the majority of cases had an AIDP subtype, except for the study from French-Polynesia [4] where all cases were classified as AMAN, three studies with similar percentages of AMAN and AIDP, [20,22,27] a study from Singapore [15] with similar frequencies of AIDP and a normal EMG (in patients with MFS), and a Brazilian case series [21] reporting only a normal EMG and AMAN or AMSAN subtypes.
Treatment and disease progression. All but three studies [15,20,30] provided information on treatment, and in most studies almost all cases were treated with IVIg, except for three large studies, from Colombia [24,32] and Brazil [19], where only 55-70% of patients were treated with immunomodulating therapy. Three studies provided no information on ICU admission, [23,29,30] which was necessary in about 50% of all reported cases, and even more  Eight studies informed about the time between onset and nadir of neurologic deficits (progressive phase), and only three studies reported the duration of the plateau phase (Fig 3). Only one large study from French-Polynesia informed about the functional evaluation of mobility of patients at nadir, showing incapacity to walk in 27/42 and difficulty to walk in 3/42. [4] The mobility of patients at 6 months after onset of disease was described in a study from Brazil [27] (33/50 walking without aid, 17/50 incapacity to walk) and a study from Puerto Rico [25] (48/79 able to walk 10 meters without aid, 39/79 any difficulty walking, and 12/79 incapacity to walk).

Discussion
Our systematic review and meta-analysis show that published studies on ZIKV-related GBS typically report a classic sensorimotor type of GBS often with a facial palsy and a demyelinating electrophysiological subtype. The disease course is frequently severe with high rates of respiratory dysfunction and ICU admission. The time between onset of infectious and neurologic symptoms and negative PCR in most patients suggests a post-infectious rather than a direct infectious disease mechanism. These results should however be interpreted with caution as the studies included in this systematic review are variable in study design and setting, selection criteria, diagnostic ascertainment, and reporting of variables, which are potential sources of bias. The combination of sensorimotor signs with facial palsy and respiratory insufficiency and a demyelinating electrophysiological subtype has previously been described in GBS patients with other preceding virus infections, such as CMV, indicating that such a clinical and electrophysiological profile may be related to preceding virus infections in general, in contrast to a bacterial infection with C.jejuni, that is associated with a pure motor axonal type of GBS. [7,8,53,54] Additionally, although GBS is generally more common in men than in women, we found equal distributions of male and female frequencies in larger studies, similar to previous reports on GBS after other virus infections, suggesting that females may be more prone to virus-related GBS. [7,53] This finding could however also be due to a higher incidence of ZIKV disease in females compared to males as has been shown in some studies. [55,56] Another interesting finding was the high frequency of paraparesis (24%) compared to previous literature on GBS (1-11%), indicating that this may be a GBS variant related to ZIKV, although a lower percentage of paraparesis in the subgroup of patients with probable/confirmed ZIKV makes this feature less specific. [5,57,58] Furthermore, in some studies it is not clear if the paraparesis evolved to tetraparesis at a later time point, and whether myelitis, which has been linked to ZIKV in other studies, was excluded. [5,[57][58][59] Some included studies diverged from the generally reported phenotype. Most importantly, the study from French Polynesia [4], in which all 42 patients had an AMAN electrophysiological subtype, 17 (40%) had a paraparesis and only 26 (62%) had hypo-or areflexia; and the study from Singapore [15], in which 4 out of 12 patients (33%) had MFS and one (8%) had MFS-GBS overlap syndrome. The high percentage of MFS in Singapore is in line with other publications that show high prevalence of MFS in Asian countries, but whether an AMAN subtype is typical for the Pacific region has not been studied. [5] As most of the other studies described cases from Latin America and the Caribbean, these discrepancies may be due to regional differences in host and/or environmental factors, including differences in the ZIKV strains. [5,60] However, some dissimilarities could also be due to differences in diagnostic and electrophysiological accuracy between studies. For instance, the interpretation of electrophysiological data in the study from French Polynesia [4] has previously been questioned, as the prolonged distal motor latencies, found at first examination and persisting after 4 months, would be more consistent with the AIDP subtype. [61] The median time between the onset of infectious symptoms and the start of neurologic symptoms varied between 5 and 12 days, which is similar to other infections preceding GBS. [7,62,63] Considering that the incubation period of ZIKV infection is estimated at 1-2 weeks, the latency between ZIKV infection and GBS was more than a week for most cases, suggesting a post-infectious immunopathogenesis, rather than direct neuronal damage or a para-infectious mechanism, as has been suggested in previous publications. [64,65] A low frequency of ZIKV PCR positivity in blood and CSF, and a low cell count in the CSF in the majority of cases, further argues against a direct infection. These findings are in line with an in vivo study that showed resistance of peripheral nerve cells to infection by ZIKV. [66] Remarkably, half of all cases combined and more than a half of probable/confirmed cases were admitted to the ICU. This proportion is higher than expected based on other literature (15-30%) [67,68], and remained higher (40%) after we excluded papers that only included patients admitted to the ICU. These data may indicate that GBS following ZIKV infection is often severe enough to necessitate ICU admission. However, the percentage of mechanically ventilated patients (20%) is similar to most other publications. [5,58,69,70] It is not clear what causes this discrepancy. A possible explanation is that presence of autonomic symptoms, rapid progression, severe weakness, or respiratory problems that did not evolve into respiratory insufficiency, were reasons to admit to the ICU, especially during the ZIKV epidemic when an increased vigilance for GBS may have lowered the threshold for intensive care monitoring.
Furthermore, many studies were done in specialized centres that may receive more severely affected patients referred from other centres, or may more easily admit patients to the ICU for monitoring compared to non-specialized centres.
The large variability of study designs and settings, selection criteria, diagnostic ascertainment and citation of variables were important sources of bias within studies and heterogeneity across studies, which is a critical limitation of our meta-analysis. Most importantly, diagnostic ascertainment of GBS and ZIKV differed, and electrophysiological criteria were not reported in most studies. Diagnostic certainty of ZIKV infection was limited in most studies, and other preceding infections in GBS were often not excluded. Furthermore, the type of hospital may have biased the inclusion of severe cases, causing heterogeneity in both clinical signs and disease progression. We calculated the I 2 to quantify this heterogeneity between studies, and have performed a sensitivity analysis to estimate the pooled frequencies among a subgroup of cases with only probable/confirmed ZIKV to analyse the clinical picture of GBS among cases with a higher ascertainment of ZIK infection.
The I 2 was considerable for most infectious symptoms, which is likely due to recall and reporting bias, and as we assumed infectious symptoms were absent, rather than missing, if not reported, we may have increased this heterogeneity. Heterogeneity in neurologic symptoms and signs was considerable for some variables, which may be due to differences in study design and methodology and geographical location. Heterogeneity of arboviral test results was also considerable, which may be due to differences between timing of sample collection and variation in incubation and viremia periods. In general, the variables with considerable heterogeneity are difficult to interpret and preclude any firm conclusions to be drawn from these data. However, the I 2 in the probable/confirmed ZIKV subgroup was generally lower than in all cases combined, indicating that the heterogeneity was partly caused by differences in the diagnostic certainty of ZIKV infection, providing more evidence for a specific clinical and electrophysiological phenotype of ZIKV-related GBS.

Conclusion
Published studies on ZIKV-related GBS generally report a sensorimotor demyelinating GBS with a frequent facial palsy and a severe disease course that often necessitates ICU admittance. The paraparetic variant of GBS is also common, which should caution clinicians to exclude myelitis in ZIKV-related cases. The time between onset of infectious and neurologic symptoms and absence of viral genome detected by PCR in most cases suggest a post-infectious, rather than a direct infectious or para-infectious mechanism.