Introduction
Herpes simplex encephalitis (HSE) is a rare but severe condition. In the absence of treatment, prognosis is extremely poor, with a mortality rate of about 70 % [
1]. Improvements in diagnostic techniques (based on use of the polymerase chain reaction, PCR) and the advent of acyclovir treatment decreased mortality rate to 15–20 % [
2,
3]. However, recent epidemiological data are lacking. For example, incidence was evaluated at about 1 case per 1,000,000 inhabitants per year during the 1990s, but PCR was not routinely used at the time [
2]. Moreover, although deaths occurring in the long term due to complications have been well described [
2,
3], deaths occurring in the acute phase have been little analyzed. The studies published to date do not focus on the most severe cases. Few data are available for the most severe cases of HSE, those admitted to intensive care units (ICUs), particularly as concerns their epidemiology, initial management, acute-phase mortality, and long-term outcome. The notion that high intracranial pressure (ICP) might be associated with a poor acute-phase outcome of severe cases of encephalitis arose from the historical series of Barnett et al. [
4]. High ICP is a well-known complication of acute bacterial meningitis [
5,
6] and has also been reported in cases of viral meningitis [
7]. Uncontrolled high ICP may lead to a decrease in cerebral blood flow, potentially resulting in fatal brain herniation. Medical strategies should be considered for the management of patients with high ICP due to HSE, and such approaches are now well established, based on many years of experience with patients with traumatic brain injury [
8]. However, drug-based strategies targeting ICP may be ineffective, and salvation decompressive hemicraniectomy has been proposed and cases reported [
9‐
18]. There are currently no recommendations supporting this strategy, and obtaining evidence to support its use would be difficult, given the rarity of this clinical condition and potential ethical problems. Nevertheless, care providers should be aware of this possible lifesaving procedure and the circumstances in which it might be appropriate. More detailed data are thus required, with better characterization and description of the management of the most severe cases of HSE.
In this study, we aimed to assess the epidemiological and clinical features of severe HSE, and to describe its management and long-term neurological outcome.
Discussion
This study is the first, to our knowledge, to report the incidence and long-term follow-up results for patients with severe HSE since the introduction of routine PCR testing. The incidence of HSE due to confirmed HSV-1 and HSV-2 infections was 1.2 cases per 100,000 inhabitants per year, about ten times higher than that reported at the end of the 20th century [
2]. Presumably, the observed difference reflects improvements in diagnostic performance (routine use of PCR) over time rather than an absolute increase in the rate of infection.
HSE accounted for 8.2 % of encephalitis hospitalization, which is slightly lower than the rate reported in a recent study [
24] using administrative database from the United States, in which HSE accounted for 13.8 % of encephalitis hospitalization.
The high frequency of admission of HSE patients to ICU (32 %) and of the need for mechanical ventilation (17 %) highlight the need to get intensive care specialists more involved in the management of these patients. Furthermore, the overall in-hospital mortality was 5.5 % (at a level similar to the crude mortality observed in the United States: 8.9 % [
24]), but increased up to 12 % for patients admitted in the ICU. This specific data has rarely been reported until now. For patients with all-cause encephalitis admitted to the ICU, mortality was 17 to 18 % in two recent studies [
25,
26] and no clear differences were observed in short-term outcome between infectious and noninfectious (autoimmune) etiologies.
Regarding the 1-year follow up of the university hospital ICU cohort, all deaths occurred in the ICU, highlighting again the importance of initial management for the most severe cases. Hjalmarsson et al
. [
2] reported a similar 12 % 1-year mortality, but in a non-ICU-focused cohort. In the retrospective cohort studied by Raschilas et al
. [
3], in which 71 % of the patients were admitted to the ICU, 1-year mortality was about double that reported here (28 %).
The incidence of disability after HSE was high, with only 50 % of the patients displaying complete neurologic recovery at 1 year. This finding is similar to other observations [
2,
3,
27] and highlights the burden of this infection, long after discharge.
We also investigated the use of decompressive hemicraniectomy to treat life-threatening HSE and the 1-year outcome for patients undergoing this intervention. The use of decompressive hemicraniectomy remains exceptional in HSE management, and there is currently no evidence of its efficacy. Only case reports have been published to date ([
9‐
18], Additional file
4). The overall success rate for the published cases is surprisingly high: 10 of 13 cases in adults resulted in complete neurologic recovery. However, this undoubtedly reflects a nonpublication bias: no case report of severe HSE with high ICP and brain herniation treated medically has been published, and there have also been no reports of decompressive hemicraniectomy failure. We analyzed all patients admitted for severe HSE in the university hospital and we identified no case of high ICP successfully treated with medical therapies only. Furthermore, those undergoing decompressive hemicraniectomy were the patients with the most severe symptoms 24 h after admission to the ICU. These findings validate
, a posteriori, the decision to carry out decompressive hemicraniectomy as a salvation therapy in life-threatening situations, in patients with intractable high ICP and brain herniation. However, the long-term outcome of patients undergoing decompressive craniectomy was not markedly different from that of other patients, despite the initially life-threatening situation due to high ICP and brain herniation. One patient died, but one young patient displayed a spectacular reversal of symptoms after ICP was brought under control by decompressive craniectomy, with the unexpected outcome of a total absence of long-term neurologic complications. No post-surgery complications were observed, not even during intensive care nursing.
Finally, like published case reports, the results for our series support the notion that salvation decompressive hemicraniectomy should be offered to patients with HSE and high ICP not responding to medical treatments. Rapidly growing HSE lesions (due to local bleeding or edema) may act as space-occupying lesions, generating a high ICP and leading to death from temporal swelling with brainstem compression. From this pathophysiological point of view, the brain lesions observed in severe HSE closely resemble the intracranial hypertension observed in malignant middle cerebral artery (MCA) infarction but differ from that observed in cases of traumatic brain injury. Given the demonstrated benefit of decompressive hemicraniectomy [
28,
29] in malignant MCA infarction, this suggests that decompressive craniectomy should be considered in cases of severe HSE with intractable high ICP due to temporal herniation of the brain. However, in cases of malignant MCA infarction, decompressive hemicraniectomy is generally proposed before the occurrence of brain herniation in patients selected according to strict criteria based on clinical evaluation and assessments of the volume of ischemic lesions on brain scans. These conditions cannot be extrapolated to HSE. The clinical symptoms of HSE are often nonspecific, delaying medical consultation. In our study, the median time from symptom onset to the first brain scan was 5 days. By contrast, the corresponding interval for patients with acute stroke is only about 10 h [
30]. Moreover, the potential for predicting brain herniation from the first brain scan is low (50 % sensitivity). Clinicians cannot therefore reliably predict the occurrence of brain herniation and therefore anticipate the potential toned for decompressive craniectomy.
This study has several limitations. There is a risk of coding misuse and errors in administrative databases. However, we focused on major codes, which are less likely to be forgotten or misinterpreted (
herpes encephalitis,
death, admission to ICU, mechanical ventilation). We validated the performance of our algorithm and identified no cases of false-negative or false-positive HSE diagnosis. However, as HSE is a rare condition, the probability of finding a false negative (i.e., diagnosis of HSE recorded in the patient’s medical records but not picked up in the PMSI search) is low. This may have resulted in an overestimation of the negative predictive value and, thus, an underestimation of the true hospital incidence. The validation algorithm was performed on medical files of patients admitted to the university hospital, and not in the other general or private hospitals of the study. However, we previously showed that coding process for infectious diseases was homogenous and showed no difference between the different French hospitals included in this study [
21,
23,
31,
32]. Moreover, a study in France showed that for encephalitis with a precise diagnosis code, concordance between coding approach and prospective data collection was acceptable [
33]. The retrospective design and the small number of patients are also limitations of the monocentric part of the study, but these are key aspects of PMSI use [
21‐
23]. These concerns are intrinsic to the very low incidence of this disease. Furthermore, there is currently no evidence to support the use of decompressive hemicraniectomy, and the use of this procedure differs considerably between centers, as reported in the case reports (Additional file
4). We therefore preferred to carry out a single-center study to avoid the potential problems of a center effect bias. Like any therapeutic procedure, hemicraniectomy for HSE should be evaluated in accordance with the current standards for clinical trials (randomized controlled trials). However, this would be very difficult due to the very low incidence of high ICP in HSE and the ethical and technical problems raised by such a trial. Finally, our study adds new evidence to the debate concerning the relevance of decompressive hemicraniectomy as a salvage therapy for HSE with intractable high ICP and life-threatening brain herniation. The use of the Glasgow Outcome Scale to assess recovery 1 year after the initial admission for HSE may also be questionable. It is a validated tool to evaluate outcome after neurological insult [
34], but has some limitations, and notably, mild neuropsychological impairment might be underestimated [
35]. However, as follow-up has not been standardized, we preferred to use a simple tool, feasible even for patients whom follow-up was provided only out of the university hospital.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
YJ, LGG and AG conceived and designed the study, and were involved in drafting of the manuscript. LGG designed and performed the algorithm for data retrieval from the hospital discharge database and wrote the corresponding sections in the manuscript. YJ, AG and LGG performed the validation of the case definition for the algorithm and performed the statistical analysis. FE and PF made substantial contributions to the conception of the study, were involved in drafting of the manuscript and made critical revisions for the discussion section. XC contributed to design and performed the brain imaging analysis and was involved in drafting the manuscript for the corresponding sections. All authors read and approved the version to be published.