Clinical predictors of encephalitis in UK adults–A multi-centre prospective observational cohort study

Objectives Encephalitis, brain inflammation and swelling, most often caused by an infection or the body’s immune defences, can have devastating consequences, especially if diagnosed late. We looked for clinical predictors of different types of encephalitis to help clinicians consider earlier treatment. Methods We conducted a multicentre prospective observational cohort study (ENCEPH-UK) of adults (> 16 years) with suspected encephalitis at 31 UK hospitals. We evaluated clinical features and investigated for infectious and autoimmune causes. Results 341 patients were enrolled between December 2012 and December 2015 and followed up for 12 months. 233 had encephalitis, of whom 65 (28%) had HSV, 38 (16%) had confirmed or probable autoimmune encephalitis, and 87 (37%) had no cause found. The median time from admission to 1st dose of aciclovir for those with HSV was 14 hours (IQR 5–50); time to 1st dose of immunosuppressant for the autoimmune group was 125 hours (IQR 45–250). Compared to non-HSV encephalitis, patients with HSV more often had fever, lower serum sodium and lacked a rash. Those with probable or confirmed autoimmune encephalitis were more likely to be female, have abnormal movements, normal serum sodium levels and a cerebrospinal fluid white cell count < 20 cells x106/L, but they were less likely to have a febrile illness. Conclusions Initiation of treatment for autoimmune encephalitis is delayed considerably compared with HSV encephalitis. Clinical features can help identify patients with autoimmune disease and could be used to initiate earlier presumptive therapy.


Introduction
Encephalitis is inflammation and swelling of the brain most often caused by an infection, or by the body's immune defences [1] Patients typically present with altered consciousness, which may range from mild behavioural abnormality to deep coma. Although there is a long and growing list of potential microbial aetiologies, herpes simplex virus (HSV) remains the most common sporadic infectious cause [2][3][4][5]. During the last fifteen years, non-infectious immune-mediated causes of encephalitis have also been recognised with increasing frequency, especially associated with antibodies against the N-methyl-D-aspartate receptor (NMDAR), leucine-rich glioma inactivated 1 (LGI-1) protein which is part of the voltage gated potassium channel complex and glutamic acid decarboxylase (GAD) [6][7][8][9].
The outcome of encephalitis is improved with prompt recognition and initiation of appropriate treatment. Several studies have shown that for HSV encephalitis, starting the antiviral drug aciclovir within 48 hours of hospital admission is associated with a better outcome [10]. The drug is therefore typically started as soon as encephalitis is suspected. In NMDAR encephalitis retrospective studies have shown that better outcomes are also associated with earlier immunomodulatory treatment [11]. However, identifying patients with encephalitis, and with particular aetiologies, can be difficult because of the non-specific nature of the symptoms, and overlap of the clinical features for the different causes [3][4][5][12][13][14]. This leads to delays to delays in accessing the appropriate health care and receiving prompt investigation and management [15,16]. We therefore studied the clinical features of adults presenting with suspected encephalitis and assessed which of the features were associated the different types of encephalitis, in particular HSV and the autoimmune forms, in order to prompt earlier treatment.
Patients were eligible if they were aged 16 years or older with clinically suspected encephalitis defined as an acute or sub-acute alteration in consciousness, cognition, personality or behaviour changes for more than 24 hours, along with any two of fever, prodromal illness, new onset seizures, focal neurological signs, pleocytosis (cerebrospinal fluid [CSF] white cell count of > 4 cells x10 6 /L), neuroimaging or electroencephalogram (EEG) compatible with encephalitis; [5] or any clinical suspicion of encephalitis when these investigations had not been completed at the time of recruitment and with no alternative diagnosis evident (Table 1).
Patients were followed up for 12 months, and the Glasgow Outcome Score (GOS) which ranges from 1 (death) to 5 (full recovery) recorded. The study protocols were approved by participating sites and the National Research Ethics Service (now part of the Health Research Authority) East Midlands Nottingham 1 committee (reference 11/EM/0442). Written consent for entry into the study was obtained from patients or an accompanying relative. Standardised case record forms for clinical, laboratory and radiological data were recorded on a secure online database (Openclinica TM ).

Aetiological testing
All CSF samples had microscopy, culture and polymerase chain reaction (PCR) performed according to national guidelines in the admitting hospitals, regional diagnostic centres or at the University of Liverpool. This included standard first line testing of HSV-1 and 2, varicella zoster virus (VZV), enteroviruses and parechoviruses, and second line testing depending on the clinical presentations [17,18]. Serum, and where available CSF, were tested for a panel of autoantibodies either at the hospital site, the University of Liverpool, or the University of Oxford [6,7,19]. All patients with suspected autoimmune encephalitis had, as a minimum, tests to detect NMDAR, VGKC-complex, glutamic acid decarboxylase (GAD) and

Methods for analysis
Patients were classified into causes of encephalitis with different degrees of certainty using published case definitions [5-7, 20, 21]. They were reclassified when updated autoimmune case definitions became available ( Table 2) [22]. Where the classification was unclear, cases were discussed by a panel of experts and a consensus reached. There were two primary analyses; one looked for an association between baseline variables and HSV encephalitis, the other between baseline variables and autoimmune encephalitis. Univariate analysis was performed initially on each variable to look for associations using unpaired t-tests, Mann-Whitney U test, chi-square test and Fisher's exact test where appropriate. Clinical variables that had potential for significance were selected using the lasso variable selection approach to perform a multiple logistic regression model. The stepBIC approach to multiple logistic regression model was used to identify variables significantly associated with encephalitis. Data were analysed using R [23].

Results
Between 2012 and 2015, 341 patients with suspected encephalitis were enrolled, of whom 233 (68%) met the case definition of clinically diagnosed encephalitis (Fig 1); 108 (32%) did not have encephalitis, but a range of mimicking conditions, most commonly septic encephalopathy 18 (17%). Of the 233 patients with encephalitis, 108 (46%) had an infectious cause including 65 (28%) with HSV type 1 or 2, 32 (14%) with other viruses and 11 (5%) with bacterial or fungi. Autoimmune encephalitis was diagnosed in 38 (16%) patients, of whom 35 (15%) had a confirmed diagnosis with autoantibodies and 3 (1%) had probable; in addition, 23 (10%) had possible autoimmune encephalitis making 61 in total. Sixty-four (27%) patients with encephalitis had no cause identified, despite further testing for infectious and autoimmune causes. Clinical features of the study population are shown in Table 3.

HSV encephalitis
Patients with HSV encephalitis typically presented with a brief febrile illness (  In a multivariate logistic regression analysis having a fever on examination, not having a rash on examination, and lower serum sodium were associated with HSV encephalitis compared to other causes of encephalitis (Table 4). For validity, a sensitivity analysis was performed using imputed data which found minimal difference in the associations found; multivariate analysis demonstrated the same associations. The combined presence of fever, lack of rash, and presence of hyponatremia had a sensitivity of 86% for diagnosis of HSV encephalitis and specificity of 42%, with a positive predictive value (PPV) of 0.36 and negative predictive value of 0.88.

Autoimmune encephalitis
The 38 patients with confirmed or probable autoimmune encephalitis typically presented with a few weeks' history (median 30 days [IQR [25][26][27][28][29][30][31][32][33][34]) of altered personality and/or confusion, often with agitation and seizures; the most commonly identified antibodies were against VGKC complex and NMDAR (Fig 1). Of note 7 of the 10 VGKC complex positive individuals were tested for LGI1 of which 1 was positive in CSF (serum was not available for retrospective testing  (Table 3). In a multiple logistic regression, female sex, abnormal movements, lack of hyponatremia and a CSF white cell count <20 x10 6 /L were associated with confirmed or probable autoimmune encephalitis, compared to other causes ( Table 5). The presence of all these features gave a sensitivity of 97%, specificity 12%, with positive predictive value (PPV) of 0.18 and negative predictive value (NPV) of 0.96.
Because new diagnostic guidelines have been introduced which include case definitions for possible, as well as probable and confirmed autoimmune encephalitis, we repeated the analysis including 23 such patients. This showed that the 61 patients with confirmed, probable or possible autoimmune encephalitis were likely to be younger and have seizures, psychosis, abnormal movements, and a lower CSF white cell count, than those with other forms of encephalitis (Table 3); in addition, they were more likely to have agitation.
In a multivariate logistic regression analysis, a lack of fever on examination, the presence of abnormal movements, and a CSF white cell count of <20 cells x10 6 /L were associated with a diagnosis of confirmed, probable or possible autoimmune encephalitis, compared to other causes (S1 Table). For validity, a sensitivity analysis was performed using imputed data which found minimal difference in the associations found; multivariate analysis demonstrated the same associations. The combined absence of fever, presence of abnormal movements and CSF white cell count <20 cells x10 6 /L had a sensitivity of 95% for diagnosis of confirmed, probable or possible autoimmune encephalitis and specificity of 37%, with a PPV of 0.35 and NPV of 0.95.

Treatment and outcomes
276 patients with suspected encephalitis had aciclovir initiated a median 15 hours (IQR 5-44) after hospital admission; this included the 61 patients with HSV encephalitis whose time to treatment was 14 hours (IQR 5-50). Time to treatment with aciclovir improved during the course of the study from 23 hours (IQR 10-56) in the first year, to 12 hours (IQR 4-28) in the 2 nd and 10 hours (IQR 3-34) in the 3 rd year. Among the 38 patients treated for autoimmune encephalitis the median time to first immunosuppressant drug (corticosteroids in 28, intravenous immunoglobulin in 6) was 125 hours (IQR, 45-250). This did not change significantly across the course of the study. Overall, 87 (37%) of the 233 patients were admitted to intensive care; those with confirmed or probable autoimmune encephalitis were more likely to be admitted than those with other diagnoses (20 [53%] of 38 verses 67 [34%] of 195, p = 0.022). 12 (5%) of 233 patients died in hospital and 86 (37%) made a full recovery; the median Glasgow Outcome Scale score at discharge was 4; these outcomes did not differ significantly between groups (Fig 2). By 12 months the mortality had increased to 26 (11%) and the number with full recovery to 92 (54%) of 171 with the information available, so that the median Glasgow Outcome Scale score had increased to 5.

Discussion
This is one of the largest prospective studies on unselected encephalitis patients from any western industrialised nation, recruiting more patients with HSV encephalitis than any previous prospective report. We found just under half the patients had an infectious cause, approximately one quarter had a confirmed, probable or possible autoimmune encephalitis, and one quarter had no aetiology identified. Over the last 20 years the epidemiology of encephalitis has changed considerably. Whilst HSV continues to be the most important sporadic cause, patients with autoimmune aetiologies are being recognised increasingly. Encephalitis remains relatively rare overall, with an incidence of 2-5 per 100,000 per year; [24] however, it causes a disproportionately high disease burden because of the devastating impact of this acquired brain injury on those affected, particularly if there are delays in recognition, diagnosis and treatment. For HSV encephalitis, the link between worse outcome and delays in treatment, particularly beyond 48 hours after hospital admission, has been recognised for some years [10,25]. National guidelines for viral encephalitis were developed in the UK in 2012 [17]. Their introduction through an NIHR programme grant in applied research was associated with a reduction in the time to aciclovir treatment for those with HSV encephalitis. The programme examined where delays occur, introduced innovative measures such as a lumbar puncture pack to streamline management [15,26], and included intense efforts to raise awareness of encephalitis and the national guidelines, in conjunction with The Encephalitis Society, our partner in patient and public involvement and engagement [27].
Across the three years of our study, we found the median time from admission to treatment reduced from 23 hours to ten hours, which is still longer than the 6 hours recommended by the national guidelines. Our findings that fever on examination, absence of a rash, and hyponatremia were associated with a diagnosis of HSV encephalitis may help improve things. A small retrospective study from Argentina suggested fever, headache, and higher CSF white cell count and protein were more likely in infectious than non-infectious encephalitis [28]; in a study from Turkey hyponatremia was associated with HSV encephalitis [29].
A retrospective comparison of 95 patients from China found involuntary movements and memory deficits were more common in autoimmune than infectious encephalitis patients [30]. A similar retrospective study in Australia evaluated retrospectively 84 patients evaluated over a ten-year period and found headache, fevers, altered consciousness, psychiatric symptoms, and a CSF pleocytosis were more common in infectious encephalitis, whereas seizures were more common in autoimmune [31].
In practice, the typical management for patients with suspected encephalitis is to perform a lumbar puncture and neuroimaging whist starting intravenous aciclovir and antibiotics. If an infectious cause is confirmed by PCR or culture of the cerebrospinal fluid, the antimicrobials are adjusted accordingly. The dilemma clinicians face currently is what to do when the investigations for infectious causes are negative. The results of tests for antibodies associated with autoimmune encephalitis, such as NMDAR, LGI-1, and GAD, often take several days to come back, and may not be helpful in seronegative autoimmune cases, which are being recognised increasingly [22,32]. In our study the median time from hospital admission to immunosuppressive therapy was 5 days.
Our study found female sex, abnormal movements, a lack of hyponatremia and a CSF white cell count <20 x10 6 /L were associated with confirmed or probable autoimmune encephalitis. These features have also been reported in studies of individual causes of autoimmune encephalitis: female sex, abnormal movements and a moderate pleocytosis are reported for NMDAR [33], a moderate pleocytosis for LGI-1 [34], and female sex for GAD-associated encephalitis [9]. Interestingly, other features reported frequently for individual forms of autoimmune encephalitis, such as abnormal mental status and seizures, were not helpful in distinguishing autoimmune from other causes of encephalitis in our study. Although neurologists focus on the clinical features which can distinguish between different causes of autoimmune encephalitis, for the general and infectious disease clinician, the critical thing is to identify patients who may have any form of autoimmune encephalitis, and thus benefit from prompt initiation of presumptive treatment with immunosuppressive drugs. We thus analysed all patients with probable or confirmed autoimmune encephalitis as a single group, whatever the underlying aetiology. There are very few such prospective studies of non-selected patients with encephalitis of all causes. A retrospective comparison of 95 patients from China found involuntary movements and memory deficits were more common in autoimmune than infectious encephalitis patients [30]. A similar retrospective study evaluated 84 encephalitis patients in Austria admitted over a ten-year period, including 34 with confirmed infection and 17 with confirmed autoimmune disease; headache, fevers, altered consciousness, psychiatric symptoms, and a CSF pleocytosis were more common in infectious encephalitis, whereas seizures were more common in autoimmune patients [31].
An expert review concluded that we are too reliant on antibody testing, and need syndrome-based diagnostics of possible, probable and confirmed encephalitis to allow prompt initiation of therapy [22]. Our study preceded these guidelines, but we modified our classification to include patients with possible autoimmune encephalitis, and repeated our analysis including such patients. This confirmed the importance of abnormal movements and a moderate pleocytosis in identifying patients with autoimmune disease. Although there are data to suggest earlier treatment is beneficial in patients with proven autoimmune encephalitis, at what stage to start treatment in patients with suspected autoimmune encephalitis is less clear, and there are no randomised trials [22].
The diagnosis of antibody-associated encephalitis for most of our patients was based on antibody detection in the serum, rather than CSF. For GAD encephalitis in particular this may be important, because GAD antibodies occur at 1% of healthy people and in 80% of those with type 1 diabetes; there is not yet convincing evidence that GAD antibodies are pathogenic, rather than an epiphenomenon [9]. However, none of GAD positive patients in our study had diabetes, and all were felt to need immunosuppressive drugs, as judged by neurologists.
The presence of female sex, abnormal movements, lack of hyponatremia plus a CSF white cell count <20 x10 6 /L had a sensitivity of 95% for diagnosis of autoimmune encephalitis and specificity of 37%, with a PPV of 0.35 and NPV of 0.95. The high sensitivity and high negative predictive value are what is required to minimise the chance of overlooking patients with the condition. Such indicators should raise a strong suspicion, and might be used to guide clinicians to start presumptive corticosteroid treatment for patients, as encouraged by the recent guidelines [22]. A prospective study is needed to evaluate them fully.
With increasing recognition of autoimmune encephalitis, and the wider availability of antibody testing, the time to initiation of treatment is likely to have improved since our study was completed. Corticosteroids are the first line therapy for autoimmune encephalitis, and additional immunosuppressive treatment is recommended for some forms, but there have been few randomised trials. A prospective placebo-controlled trial of intravenous immunoglobulin in autoimmune encephalitis is currently recruiting in the UK.
Our study had several limitations. Firstly, although ours was one of the largest prospective studies of unselected patients with encephalitis, the size was limited, especially in terms of numbers with specific autoimmune diagnoses and also inherently biased towards those autoimmune causes that would present acutely or subacutely whereas some forms are known to take more insidious and longer presentation courses. Secondly, because it was performed some years ago, the number of patients with some of the rarer antibodies, was limited. We reevaluated samples as newer antibody tests became available, but applying tests retrospectively to samples which may have been frozen and thawed many times is never ideal. There may also have been a bias towards identifying well recognised forms of encephalitis, such as HSV, rather than newer emerging forms of autoimmune disease. The time to initiation of treatment for autoimmune encephalitis may have reduced since our study, due to increasing recognition of the condition. We were also reliant on testing of serum for the most part rather than CSF. We now know that antibody testing of CSF has greater specificity. Although our primary purpose was identifying patients with any form of autoimmune encephalitis, there is heterogeneity among different types of autoimmune encephalitis, and grouping them together will have blunted these differences.
In summary, we have shown that among adults admitted to hospital with a diagnosis of encephalitis, clinical features can be used to identify those with HSV and those with autoimmune disease to guide presumptive treatment. Currently, the initiation of treatment for patients with HSV is significantly quicker than for those with autoimmune disease. The presence of female sex, abnormal movements, lack of hyponatremia and a CSF white cell count <20 x10 6 /L predicted autoimmune encephalitis with high sensitivity and high negative predictive value and may be used to guide presumptive treatment, but this requires further evaluation in a prospective study.
Supporting information S1 Table. Multivariate