Introduction

Encephalitis is associated with high neurological morbidity and mortality and occurs with an estimated incidence of 3.5–7.5 cases/100,000 person-years [1,2,3,4,5,6]. Identifying the cause of encephalitis is challenging with the etiology frequently remaining unknown after thorough investigation [7,8,9]. Establishing the etiology of encephalitis in patients living with human immunodeficiency virus (HIV) is especially difficult given the patients’ susceptibility to a wide variety of opportunistic infections as well as the direct neurological effects of the HIV virus itself in the form of HIV encephalitis [10,11,12] or CD8 + encephalitis [13,14,15].

HIV-infected patients are at risk for all causes of encephalitis seen in persons without HIV, as well as those that are specific to a person with decreased cellular immunity. Although Cryptococcus neoformans, cerebral toxoplasmosis, herpes viruses, and JC-associated progressive multifocal leukoencephalopathy causing encephalitis in HIV patients are well-described in case reports and case series [16,17,18,19,20,21], a large study with an analysis of the etiology of encephalitis in this patient population is lacking. To further clarify the diagnostic and therapeutic approaches to these patients, we conducted a study describing the clinical manifestations, microbiology, etiology, and outcomes of acute encephalitis in adults with and without HIV.

Methods

We performed a retrospective cohort study of HIV-infected and HIV-negative adults admitted to 16 hospitals in Houston, Texas area between January 2005 and July 2020. Clinical information was obtained by a retrospective chart review of electronic medical records.

Case definition

Cases were initially screened by electronic search of ICD-9/10 billing codes. Due to low accuracy of encephalitis-related ICD codes [22], inclusion of each case was determined by manual evaluation by two physicians using the 2013 International Encephalitis Consortium (IEC) criteria. Thus, encephalitis was defined as new-onset altered mental status without alternative cause lasting > 24 h with at least two of the following: (A) fever 72 h before or after presentation, (B) new-onset seizures, (C) new-onset focal neurological findings, (D) white blood cell count > 5 cells/μL in the cerebrospinal fluid (CSF), (E) new neuroimaging findings, or (F) abnormal electroencephalogram consistent with encephalitis [23]. We excluded patients < 18 years of age, those that did not meet criteria for encephalitis based on the IEC guidelines, or if records were considered insufficient for inclusion. Patients without an HIV test performed were not included.

Data collection

Baseline characteristics were obtained, including demographic data (age, sex, and race), comorbidities, and signs and symptoms upon clinical presentation. Charlson comorbidity index was calculated and recorded based on the presence or absence of nineteen comorbid disease processes [24, 25]. Laboratory data, CSF white blood cell count (WBC) and differential, CSF glucose, CSF protein, serum WBC, and serological studies were recorded. Microbiological studies including blood and CSF cultures and molecular diagnostics were recorded. A summary of the laboratory and microbiological tests available to the ordering physician is included as a supplemental file.

The etiology of encephalitis was classified into seven categories: viral, bacterial, fungal, parasitic, prion disease, immune-mediated, or unknown. Some patients were included in more than one category, representing multiple disease processes at the time of presentation. Classification of neurosyphilis was based on the algorithm outlined by Marra et al. with an HIV-negative patient or HIV-infected patient with > 200 CD4 + cells/μL considered to have neurosyphilis if either (1) VDRL is positive in the CSF or; (2) if a serological test from the peripheral blood is positive, CSF WBC is > 20 cells/μL, and the patient has symptoms consistent with neurosyphilis. In HIV-infected patients with a positive serological test from peripheral blood, symptoms of neurosyphilis, and CD4 < 200 cells/μL, a lower limit of 6 WBC/μL was used [26]. A diagnosis of presumed cerebral toxoplasmosis was made based on clinical presentation, appearance of lesions on magnetic resonance imaging (MRI), and response to toxoplasmosis-directed therapy with decrease in size of lesions on follow-up MRI. Confirmed toxoplasmosis was made based on CSF Toxoplasma gondii PCR or Toxoplasma gondii identified on histopathology from autopsy.

Severity of disease course was indicated by admission to the intensive care unit (ICU), length of stay in the hospital (LOS), need for mechanical ventilation, renal failure, Sequential Organ Failure Assessment (SOFA) score > 3 [27], and Assessment of Physiology And Chronic Health Evaluation II (APACHE II) score > 10 [28]. Outcomes were assessed at time of discharge with the Glasgow Outcome Scale (GOS) [29]. In-hospital mortality was recorded for all patients during their hospitalization and one-year follow-up was recorded for those with available data in the electronic medical records of any of the 17 participating hospitals and affiliated clinics.

Ethical consideration

The study was approved by the UT Health Committee for the Protection of Human Subjects, by the Memorial Hermann Hospital Research Review Committee and the Harris Health Research Committee.

Statistical analysis

Categorical outcomes given HIV status were assessed for significant associations with Pearson’s χ2 test or Fisher’s exact test when appropriate. Continuous data were compared using Welch’s t test. Statistical analyses were performed with either IBM SPPS version 26 or R [30].

Results

Following initial screening using ICD-9/ICD-10 billing codes, 1241 adults admitted to the hospitals with a putative diagnosis of encephalitis were identified. Among these, 278 were included in the analysis based on IEC criteria. Eighteen patients (6.4%) did not have an HIV test performed and were excluded. Our analysis focused on the 260 (93.5%) with known HIV status. All patients with HIV were infected with HIV-1; none were infected with HIV-2.

The demographic, clinical, and laboratory characteristics are shown in Table 1. Patients infected with HIV were younger than HIV-negative patients (median 43 years vs 49 years, p < 0.001). HIV-infected patients were more likely to be Black and had a higher Charlson comorbidity index (p < 0.001). There were no other significant differences regarding other demographics or presenting symptoms.

Table 1 Baseline characteristics and clinical presentation of HIV-infected and HIV-negative adults with encephalitis
Full size table

As shown in Table 2, the most common etiologies of encephalitis in HIV-infected patients were viral (45.0%), unknown (27.5%), neurosyphilis (20.0%), parasitic (12.5%), fungal (7.5%), and immune-mediated (5.0%). HIV-infected patients were less likely to have unknown etiologies and more likely to have neurosyphilis [p < 0.001, OR 55 (6.6–450)], parasitic (p < 0.001, OR unreportable) and fungal etiologies [p = 0.005, OR 8.8 (1.4–54.5)]. HIV-infected were more likely to have more than one etiology identified [p < 0.001, OR 13 (4.7–35.9].

Table 2 Etiology of encephalitis of HIV-infected and HIV-negative adults admitted to the hospital
Full size table

As shown in Table 3, the median CSF glucose was lower in HIV-infected patients (p = 0.002), but there was no difference in CSF WBC or CSF protein levels. Serum WBC level was lower in the HIV-infected cohort (p < 0.001). HIV-infected persons were more likely to have a positive VDRL [p < 0.001, OR 18.3 (1.8–183)], positive cytomegalovirus (CMV) PCR [p = 0.022, OR 11.2 (1.18–105)], or varicella zoster virus (VZV) PCR [p = 0.003, OR 4.82 (1.62–14.6)] compared to the HIV-negative patients. HIV-infected patients were more likely to have an abnormal MRI [p < 0.001, OR 15 (2.0–112)]. Of the 30 HIV-infected patients with abnormal MRI, 20 (66.7%) had reports of T2/FLAIR hyperintensity, 11 (36.7%) with intra-parenchymal lesions, 11 (36.7%) with white matter changes, 8 (26.7%) had report of restricted diffusion; five (16.7%) with leptomeningeal enhancement, 2 (6.7%) with ependymitis, and 3 (10.0%) with reports of acute ischemia. There were no significant differences in the results of other diagnostic tests between the two groups. The majority of HIV-infected patients had a CD4 < 200 cells/μL with a median level of 55 cells/μL. Three of the HIV-infected patients had a brain biopsy performed during the studied hospitalization. Two of these were diagnosed with primary CNS lymphoma. The third reported CD8 + predominant inflammation diagnostic for CD8 + -encephalitis. One HIV-infected patient had toxoplasmosis detected on brain lesions at autopsy.

Table 3 Diagnostic studies of HIV-infected and HIV-negative adults with encephalitis
Full size table

As shown in Table 4, HIV-infected persons were more likely to have APACHE II score > 10 (p = 0.005) although there was no difference between the groups based on the SOFA score > 3 (p = 0.7). There were no significant differences between HIV-infected and HIV-negative patients in the need for intensive care unit admission, length of stay, need for mechanical ventilation, renal failure, or adverse clinical outcomes as defined by the Glasgow outcome scale. Most patients with encephalitis—whether HIV-infected (31/37, 83.7%) or HIV-negative (159/187, 85.0%)—were started on an antiviral medications during their hospitalization, although only 11/37 (29.7%) of the HIV-infected patients and 81/151 (58.6%) of the HIV-negative patients were initiated within 24 h of admission to the hospital. Inpatient mortality was similar in the HIV-infected and HIV-negative patients, 15.0% vs 9.5% [p = 0.4, OR 1.67 (0.63–4.44)], but one-year mortality was higher for the HIV-infected patients, 31.3% vs 16.0% (p = 0.04, OR 2.40 (1.02–5.55)].

Table 4 Indication of severity and outcomes of HIV-infected and HIV-negative patients with encephalitis
Full size table

Eight of the 22 HIV-infected patients that were known to survive > 1 year were on ART (36%) in comparison to 1/10 (10%) of the HIV-infected patients who died reported taking ART with an odds ratio of survival 5.14 (0.55, 48.97). The HIV RNA viral load of the one patient that died that reported taking ART was 11,200 copies/mL, concerning for non-adherence. The median CD4 + count of those that died was 33 cells/μL (IQR 16–136). Three (30.0%) had more than one diagnosis identified. Three (30.0%) had an unclear etiology of encephalitis at the time of death. Of the four that died after discharge, one was discharged home, one to acute rehabilitation center, and two to skilled nursing facilities. Table 5 provides a comprehensive summary of the pertinent diagnostics, diagnoses, and outcomes of each of the 40 patients with HIV admitted to the hospital with encephalitis.  

Table 5 Pertinent diagnostics, diagnoses, and outcomes of HIV-infected patients admitted to the hospital with encephalitis
Full size table

Discussion

This large, multicenter cohort study adds valuable data to what is currently available on HIV-infected persons presenting to the hospital with encephalitis. The demographics of HIV-infected patients with encephalitis reflect the demographics of the HIV epidemic in Houston, Texas, with a higher proportion of HIV-infected patients being Black and younger age compared to HIV-negative patients.

The most notable result in our report is the one-year mortality, which is almost twice as high in the HIV-infected group compared to the HIV-negative cohort (31.3% vs 16.0%) with an odds ratio 2.38 (1.02–5.55). Our effect measure of increased odds of death among HIV-infected patients during hospitalization [OR 1.67 (0.63–3.65)] is similar to those previously reported by another large report of patients with encephalitis admitted to the hospital in the US between 2000 and 2010 [OR 1.70 (1.30–2.22)]; but, our analysis fails to reach statistical significance1. HIV-infected patients taking ART had an odds ratio of 5.14 (0.55, 48.97) of surviving one year after hospitalization compared to HIV-infected patients not taking ART. It is well-established that effective ART decreases mortality, but further studies focused on barriers and facilitators to medication adherence could be beneficial to for this vulnerable patient population.

Another important issue to highlight is the number of patients without an etiology of encephalitis. This could be due to an incomplete differential diagnosis constructed by the clinical team, imperfect diagnostic tests, limited CSF available for diagnostic tests, or a combination of these. The number with an unidentified etiology in our HIV-negative cohort is similar to previous studies at 44.5% [3, 5, 7,8,9], but our HIV-infected cohort is lower than these reports at 27.5%. The difference between the HIV-infected and HIV-negative patients in the number that achieved a final diagnosis (72.5% vs 55.5%) may in part be due to more extensive evaluation in persons with HIV, however it is also more likely that HIV-infected persons have more than one disease process simultaneously [p < 0.001, OR 13 (4.7–35.9]. Eight of our HIV-infected patients were given a diagnosis of VZV encephalitis; in four of these (50.0%), VZV was not the only identified disease process. It is important for clinicians to be aware that one confirmed diagnosis may not indicate a complete work-up in patients with HIV.

All guidelines for the management of encephalitis recommend PCR testing for HSV, VZV, and enterovirus be performed on CSF [23, 31,32,33,34], yet only 12 of our 40 patients with HIV had results for all three of these (30.0%). Our study shows that clinicians are better trained to consider HSV and ordered HSV PCR more often than enterovirus or VZV (90.0% vs 62.5% vs 52.5% in our HIV-infected patients). Only 18 of the 40 HIV-infected patients had results for CMV PCR (45.0%); only 29 had results for VDRL in the CSF (72.5%). CMV may have been low on the clinician’s differential depending on the control of the HIV and CD4 + count of the patient, but guidelines for the management of encephalitis recommend testing for neurosyphilis in all immunosuppressed patients [23, 31,32,33].

HSV encephalitis is reported to be the most common etiology of viral encephalitis reported in the literature in immunocompetent patients [6, 35, 36]. Although surprising that none of our HIV-infected patients were diagnosed with HSV encephalitis, ours is not the only study to report a difference in the prevalence of HSV between the HIV-infected and HIV-negative patients. A multi-center study on the causes of encephalitis in England published in 2010 had similar results with HSV the etiology of 37/172 (22%) of the immunocompetent patients and 1/31 (3%) of the immunocompromised patients [4]. A separate study published in 2000 tested the CSF of 251 samples from 219 patients with HIV and neurologic symptoms for 5 herpes viruses and found only 3% samples positive for HSV (25% positive for CMV, 7% EBV, and 4% VZV) [37]. Since HSV encephalitis is thought most likely due to reactivation of prior infection, it seems immunodeficiency with HIV would increase the risk of HSV encephalitis, and yet the data from these two studies and ours show the opposite to be true. Further studies are needed to try to explain this puzzling result.

There is a notable discrepancy in the number of cases of West Nile Virus (WNV) encephalitis between HIV-infected (1/15, 6%) and HIV-negative patients (32/120, 26.7%), which is difficult to explain. This could in part be due to incomplete work-up since less than half of the HIV-infected patients were evaluated for WNV (15/40). The ordering clinicians may have been more focused on evaluation of diagnoses associated with HIV (neuro-syphilis, tuberculosis, and Cryptococcus) than those that can be seen in both HIV-infected and HIV-negative (WNV, enterovirus, and autoimmune encephalitis). If the remaining 25 HIV-infected patients with encephalitis would have had WNV serology ordered, perhaps additional cases would have been found. A second explanation for the lower number of cases of WNV in HIV-infected patients could be that other viruses seen in patients with HIV (VZV, CMV, and EBV) are most often due to reactivation of prior infection rather than acute infection seen with WNV. It is plausible that the risk of reactivation in the immunocompetent host that holds latent virus from prior infection is higher than risk of exposure to mosquito carrying WNV causing acute infection. A third possibility is a differential environmental exposure. Is it possible that patients with HIV are less likely to be exposed to the WNV vectors (mosquitos) than HIV-negative patients? Could patients with HIV be more concerned about spending extended periods of time outdoors (camping and hiking) knowing that they have a health condition that predisposes them to infection? This seems less likely than the first two reasons, but worth considering.

Although VZV PCR was only ordered in 52.5% of cases, it was a high yield test, with 8 of 21 tests ordered (38.1%) positive in HIV-infected patients. Although it may not be the only active disease process in an HIV-infected patient with encephalitis, it is important that clinicians consider VZV as well as HSV in any patient with encephalitis given the potential benefits of administration of antiviral medication [38].

Some of the tests not ordered may have been considered by the clinician, but clinicians are often faced with the challenge of prioritizing tests in the setting of limited CSF since the differential for encephalitis is broad, and the amount of cerebrospinal fluid available for analysis is limited. Institutions with multiplex PCR testing on CSF samples available to the clinician have the advantage of a more complete work-up using less CSF. This is especially beneficial in persons with HIV since the recommended diagnostic tests in an immunocompromised patient with encephalitis are extensive, and there could be more than one disease process occurring in the same presentation.

Reports in the literature estimate an immune-mediated process the etiology in 21–33% of cases of encephalitis [4, 9]. However, a work-up for immune-mediated encephalitis was performed in only 2 (5.0%) of the HIV-infected patients in our study. As clinicians become more aware of the overlap of these presentations, cases of autoimmune encephalitis will appropriately be differentiated from presumed infectious encephalitis. Although the proportion of infectious vs immune-mediated etiology more often favors an infectious process in HIV-infected patients compared to HIV-negative patients, this may become less pronounced than in years past, as more patients with HIV achieve viral suppression on ART.

The two HIV-infected patients in our report classified as immune-mediated etiology were diagnosed with HIV-associated CD8 + encephalitis (HIV-CD8E). HIV-CD8E is an inflammatory disease with infiltration of CD8 + T cells into the cerebral parenchyma and should be considered in HIV-infected patients on ART [13,14,15]. Although biopsy provides definitive diagnosis, CSF HIV viral load is a non-invasive way to test for viral escape, the presence of which supports the diagnosis of HIV-CD8E and provides sufficient evidence to warrant treatment with steroids.

The first patient had a CSF HIV viral load of 15,000 copies/mL despite a peripheral HIV viral load of 630 copies/mL, supporting the diagnosis of HIV-CD8E with viral escape. The second patient had a brain biopsy consistent with HIV-CD8E; no HIV viral load was obtained from the CSF. Both of these had MRI findings classically seen with HIV-CD8E with T2/FLAIR hyperintensities of the periventricular region (both patients), deep white matter (first patient), and brainstem (second patient). A third patient was considered possible HIV-CD8E based on clinical presentation and response to steroids. The patient, a 41-year-old male was brought to the hospital by family for acute altered mental status. Work up during the hospitalization was negative and he was started empirically on corticosteroids for the suspicion of HIV-CD8E. He experienced clinical improvement and was discharged home. An MRI was not performed. The CSF HIV viral load resulted was < 20 copies/mL. Although the authors of this paper agree with the treatment of steroids based on the possibility of the diagnosis of HIV-CD8E, we classified this patient as having an unclear diagnosis given the incomplete work-up and undetectable HIV viral load in the CSF.

A recent case series of 23 patients with confirmed HIV-CD8E provides a thorough summary of what is known about HIV-CD8E [13]. Of the 16 patients with available CSF, 11 (68.8%) had increased viral load in the CSF with HIV viral load suppressed in the peripheral blood (viral escape) [13]. Only two of the nine patients on ART (22.2%) in our study had an HIV viral load performed on the CSF to evaluate for HIV-CD8E. If more of the patients in our study had a HIV viral load from the CSF or brain biopsy, there may have been more diagnosed with HIV-CD8E than the two we identified. With an increased proportion of patients living with HIV achieving viral suppression, HIV-CD8E is a disease process that should be considered in patients with HIV on ART.

It should be noted that we used a broad definition for classification of neurosyphilis. Given the low sensitivity of VDRL in CSF [39, 40], we followed the treatment algorithm outlined by Marra et al. as explained in the methods [13, 14]. This resulted in eight patients classified as neurosyphilis even though CSF VDRL was positive in only three.

Strengths and limitations

Strengths of our study include the large sample size allowing for a comparison of HIV-infected and HIV-negative patients with encephalitis. This was a multicenter study with patients from 17 hospitals reflective of a diverse patient population and socioeconomic class. We included only cases that met the definition of encephalitis by IEC definition and provided a thorough report of the diagnostic tests ordered and interventions.

Weaknesses include the retrospective nature of the study with missing data that could have limited the conclusions of the study, including data related to outcome for patients that received subsequent medical care in a different healthcare system. Although it may have been appropriate in the clinical setting, the gaps in data from unordered diagnostic tests create the potential for selection bias, where only those with higher clinical suspicion have certain diagnostic tests ordered.

Another weakness of our study is that results may not be generalizable to other parts of the country or outside of the United States since the etiology of encephalitis varies by geographical location. In Houston, Texas, it is important to evaluate for West Nile virus during summer months, but our study did not have patients diagnosed with other vector-borne pathogens, such as Japanese encephalitis, Dengue virus, or tick-borne encephalitis virus, which should be considerations in other parts of the world. For example, we identified two cases of coccidioidomycosis in patients that traveled to the western United States, but a clinician practicing in the western United States would likely see a higher proportion of patients with central nervous system (CNS) coccidioidomycosis. Similarly, Mycobacterium tuberculosis would be expected to be higher in many parts of the world than what was reported in our study, especially in the HIV-infected population.

Conclusion

This report demonstrates the severity of illness of patients presenting to the hospital with encephalitis. The high rates of mortality seen in HIV-infected patients presenting to the hospital with encephalitis highlight the importance of appropriate utilization of available diagnostic tests to allow for rapid intervention in this vulnerable population.