With West Nile virus (WNV) now endemic throughout North America, with upwards of three million infected cases, it is a priority to document the long-term clinical outcomes (1–3). Among a large cohort of participants with a history of WNV infection in Houston, TX, USA, we found that 40% of those who presented with clinical disease continued to experience WNV-related morbidity up to 8 years postinfection (4). This percentage was up to 80% for those who initially presented with encephalitis. Sequelae varied in severity, with a higher than expected proportion developing chronic neurological and renal disease (5–10). Other shorter-term studies report neurocognitive impairment, debilitating fatigue, persistent neuromuscular paralysis, tremors, and depression as outcomes experienced by patients with a history of West Nile Neuroinvasive Disease (WNND) (7, 11–16). There are currently no published studies correlating neurocognitive evaluations to neuroimaging outcomes in patients with WNND.

MRI is the most sensitive neuroimaging technique for intracranial infections because of its superior soft tissue detail, sensitivity to subtle inflammatory processes, vascular imaging techniques, and functional imaging. Limited MRI data are available to evaluate the long-term effects of WNV, and published studies of brief case report describe changes during acute infection, with areas of interest including the basal ganglia, thalamus, pons, and lobar gray and white matter (17–19). The goal of this study was to understand the long-term neurological function impairment, neurocognitive deficits, and/or MRI abnormal findings in patients with a history of WNV infection.

Table 1 shows the demographics of our study population, which consisted of mostly white, non-Hispanic (83%) individuals, and an almost equal proportion of males (53%) and females. Mean age at the time of exam was 57 years (range 18–89 years). Most participants had graduated high school (98%), and almost half (47%) were employed full time at the time of their neurological exam. Almost one-third of participants were hypertensive (32%), and 15% were diabetic. The majority of participants (77%) were four or more years past their acute WNV infection. Participants evaluated had a history of asymptomatic disease (23%), West Nile Fever (WNF) (33%), or WNND (44%).

Across the entire cohort, almost half (49%; 57/117) of participants had some type of abnormal neurological exam finding, with most abnormalities being unilateral in nature (Tables 2 and 3). The most common abnormalities included abnormal (decreased) strength (26%; 30/117), abnormal reflexes (14%; 16/117), and tremors (10%; 12/117). Decreased strength was more common among core muscle groups than in the extremities, with the exception of finger strength, and was associated with decreased reflexes in 16 out of 30 patients consistent with lower motor neuron involvement. Abnormal reflexes were consistent among all five muscle groups. Finally, tremors were more common among the upper body than the lower body, which was distinct from other findings. There was no statistical difference between time from infection (5 years or more vs. less than 5 years) and having an abnormal neurological examination (p = 0.73).

The frequencies of gross neurocognitive and ADL abnormalities were quite low, with fewer than 5% of the cohort producing abnormal scores on the GCS, GOS, GOAT, and Barthel Index (Tables 4 and 5). Similarly, fewer than 5% of the sample had IQ scores in the Borderline or extremely low range (i.e., <80). However, the frequency of milder higher-order neurocognitive and mood disturbances was somewhat higher. Using a 1SD cutpoint derived from published normative standards, we observed a 22% overall rate of impairment on the RBANS, with the domains of immediate (31%) and delayed memory (25%) being the most frequently impaired. Seven percent of participants had RBANS total scores in the moderately to severely impaired range (<2SD). There was no significant difference between timing of the exam (five or more years since infection versus less than 5 years) and low RBANS score (p = 0.16). Consistent with previous findings, approximately one-fifth of the sample reported current depressed mood using a conservative BDI cutoff score of 17, which is commonly recommended in medical and neurological populations (32).

On MRI, WNV participants showed significant cortical thinning as compared to age- and gender-matched controls (Figure 2A; raw data in Tables S1 and S2 in Supplementary Material). Significant cortical thinning in the left hemisphere (LH) included parts of the posterior cingulate cortex (pink), parts of the superior frontal cortex and medial-orbito frontal region (magenta), anterior cingulate cortex and inferior frontal cortex (violet blue), parts of the cuneus (orange) and para-hippocampal region (pale green), and in the right hemisphere (RH) included parts of the middle and inferior temporal cortex, and supramarginal region (light blue), inferior frontal region and insular cortex (bright green and red), parts of the superior frontal cortex (light green, blue, and violet), cingulate cortex (light purple), and inferior frontal region (blue). Lower neurocognitive functioning as determined by total RBANS score was associated with cortical thinning in regions of the LH only: caudal middle frontal gyrus (r = 0.49, p = 0.008), rostral middle frontal gyrus (r = 0.49, p = 0.007), and supramarginal gyrus (r = 0.51, p = 0.004). Scatterplots comparing cortical thickness and total RBAN score by region and hemisphere among WNV participants can be found in Figure 3.

Tensor-based morphometry analysis of participants’ scans vs. control scans revealed significant regional atrophy in the overall cohort of 30 WNV participants that underwent MRI (Figure 2B). Significant regional atrophy was identified in the cerebellum, brain stem, thalamus, putamen, and globus pallidus. No significant correlation between a low RBANS score and these regions was identified.

This study contributes one of the largest cross-sectional examinations of the neurological and neurocognitive outcomes post-WNV infection, and the only known study to link these quantitative measures to MRI findings, allowing for valuable insight into the pathology of infection. These findings can be useful in the clinical management of patients with a history of WNV, particularly those with WNND.

Results of the neurological exam showed weakness and decreased reflexes consistent with lower motor neuron damage in a substantial proportion of patients (Tables 2 and 3) and identified WNND participants to be at an increased risk for these outcomes, findings consistent with current literature. Weatherhead et al. described abnormal reflexes and muscle weakness as common abnormalities in a prospective study that evaluated WNND and WNF participants in the Houston WNV cohort 1–3 and 8–11 years after their initial WNV infection (10), with results indicating that participants with a history of WNND with encephalitis were more likely to have persistent and progressive abnormalities.

When comparing the results of the neurocognitive exams between neuroinvasive and non-neuroinvasive cases, we see a significant difference in only a few of the measures (GOAT, RBANS total score, RBANS immediate memory, RBANS visuospatial, and Barthel ADL Index) (Table 4). A larger than expected number of the WNF cases (20%) had observable deficits. In order to meet the case definition of WNND, it was required to have a lumbar puncture procedure with evidence of pleocytosis; therefore, it is possible that these cases were misclassified as WNF, when indeed, they had WNND. Newer guidelines now permit the diagnosis of encephalitis without lumbar puncture if patients present with altered mental status for greater than 24 h with no alternative explanation and meet at least three of the following criteria: fever 38 C or above, generalized or partial seizures unrelated to a preexisting seizure disorder, new onset of focal neurologic findings, abnormality of brain parenchyma on neuroimaging, abnormality in electroencephalography (33). Since we do not have this information for all participants, we cannot reevaluate their classification in this study. We acknowledge this limitation and that the significance of WNND as a risk factor for sequelae may be underrepresented in this study population. Also, none of our patients evaluated had WNV-associated acute flaccid paralysis, which is known to have poorer clinical outcomes than other presentations of WNND (5, 6). It would be valuable in future studies to further distinguish neurological and neurocognitive outcomes among the different clinical presentations of WNND (meningitis vs. encephalitis vs. AFP), since these different presentations have varying severity of morbidity (4).

MRIs to evaluate cortical thinning and atrophy were performed only on participants with ongoing neurological complaints and/or a low RBANS score. When compared to age- and gender-matched controls from the OASIS database, cortical thinning was identified in the frontal and limbic lobes of both the left and RHs, with small regions of the temporal and parietal lobes also indicated in the RH. Interestingly, participants with a low RBANS score also had significant thinning within regions of the LH. Significant regions were within the middle frontal gyrus and supramarginal gyrus, both within the parietal lobe. The parietal lobe is responsible for sensation and perception and integrating sensory input, areas that were identified to be weakened in some participants. The lack of overlap between regions affected by WNV compared to controls and cortical thinning correlates of the RBANS analysis of WNV participants may be a simple artifact of measurement. Namely, the RBANS is a brief neurocognitive screening measure and does not include tasks that assay the frontal regions identified in the cortical thinning analysis. This lack of overlap warrants additional neurocognitive measures in future studies with larger participant groups.

Although most WNV-infected participants were recruited in the Houston area, we were not able to match controls from OASIS based on location for this study. Additionally, the OASIS database contains data from controls scanned with a different machine with regards to both field strength and manufacturer when compared to our scans (1.5T Siemens and 3T Philips, respectively). While these differences could impact our findings and lead to systematic bias, a recent publication from our group (34) found that when examining the effect of field strength between 1.5 and 3T, the cortical thickness showed field dependence only in certain regions that include posterior cingulate and parahippocampal regions in the LH and middle temporal cortex in the RH. In this current study on WNV, we observed significant cortical thinning in the LH that included parts of the posterior cingulate cortex (pink), parts of the superior frontal cortex and medial-orbito frontal region, anterior cingulate cortex and inferior frontal cortex, parts of the cuneus and para-hippocampal region, and in the RH included parts of the middle and inferior temporal cortex, and supramarginal region, inferior frontal region and insular cortex, parts of the superior frontal cortex (light green, blue, and violet), cingulate cortex, and inferior frontal region. Thus, many regions in which cortical thickness differences were observed in WNV patients did not show field dependence. Nonetheless, predicting systematic bias when using different scanners is difficult and this is an important limitation of our study. It will be crucial to confirm these results in future studies by concurrently enrolling and imaging control participants using the same scanners.

Regional atrophy was noted in the cerebellum, brain stem, thalamus, putamen, and globus pallidus of WNND patients who were not apparent in controls. These regions clearly mediate functions that are altered in WNND patients. Cerebellitis has only been described in one case report of a 10-year-old child (35). Infectious cerebellitis is most commonly caused by other etiologies such as JC virus, varicella zoster virus, and Listeria monocytogenes (36). The most common causes of brain stem/rhomboencephalitis are L. monocytogenes and Enterovirus 71 (37). WNV can occasionally have brain stem involvement (38). Encephalitis with thalamic and basal ganglia involvement should raise the possibility of respiratory viruses and WNV infection (39, 40). This thalamic and basal ganglia involvement is responsible for the facial and upper extremity tremors, motor weakness, and abnormal reflexes seen in WNND (41). Although regional atrophy did not statistically correlate with the RBANS total score, atrophy in WNV participants compared to controls still provides pathological insights into the impact of WNND disease. Larger sample sizes within WNND with encephalitis, meningitis, and/or acute flaccid paralysis would allow for further analysis of the effects of WNND and its specific clinical syndromes on regional atrophy. Also, serial MRI scans from acute disease onset to later timepoints would allow for documentation of atrophy and cortical thinning over time. Unfortunately, we were limited in this study to make comparisons to prior MRIs, as we only had MRI result data on six patients from the acute phase of illness, all of whom had a “normal” result provided.

Understanding the long-term complications of WNV infection are important to improving clinical outcomes and decreasing the costs on our health systems. A 2014 study determined that it can cost up to $400,000 per patient to treat WNV sequelae, not including the cost of care during the acute phase of illness (42). This study provides new insights of long-term sequelae by comparing neurological, neuropsychological, and MRI results of affected patients 3–8 years after initial infection. Although some correlations of functional evaluations and MRI were not statistically significant, the biological relevance of the alterations identified via MRI are interesting. For the first time, this study identified brain areas of interest, such as the cerebellum, putamen, thalamus, supramarginal gyrus, and middle frontal gyrus, in patients continuing to experience neurological complications. These data will be crucial for future studies regarding the progression of WNV disease and associated long-term complications. Afflicted regions also suggest the damage pattern could be similar to what is seen in traumatic brain injury. At this time, with no specific preventive or therapeutic options to treat WNV infection other than supportive care, continuing these types of large-scale and long-term prospective studies will allow us to generalize expected consequences of infection. In turn, this allows for proactive treatment of sequelae that would otherwise lower a patients’ quality of life. Ultimately, public health efforts to prevent infection are critical, as many of the issues identified in study participants are not reversible.

This study was approved by Institutional Review Boards at both Baylor College of Medicine (H-30533) and the University of Texas Health Science Center at Houston (HSC-SPH-03-039). All participants provided informed written consent for their involvement in the study. Participants were registered as part of study protocol at the Memorial Hermann’s Clinical Innovation and Research Institute and University of Texas 3T MRI Center in Houston, TX, USA.

Conception and design of study: KM, PN, SW, and RH. Acquisition and analysis of data: KM, MN, SR, SD, KG, LS, SW, and RH. Manuscript/figure drafting: KM, MN, SR, KG, PN, SW, and RH.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer BS and handling Editor declared their shared affiliation.