Clinical characteristics and outcomes of patients admitted to hospitals for posterior reversible encephalopathy syndrome: a retrospective cohort study

Background Posterior reversible encephalopathy syndrome (PRES) is usually a benign, yet underdiagnosed clinical condition associated with subacute to acute neurological manifestations primarily affecting white matter. PRES is reversible when recognized promptly and treated early by removal of the insulting factor; however, can lead to irreversible and life-threatening complications such as cerebral hemorrhage, cerebellar herniation, and refractory status epilepticus. Methods We utilized the National Inpatient Sample database provided by the Healthcare Cost and Utilization Project (HCUP-NIS) 2017 to investigate the demographic variables (age, sex, and race) for patients with PRES, concomitant comorbidities and conditions, inpatient complications, inpatient mortality, length of stay (LOS), and disposition. Results A total of 635 admissions for patients aged 18 years or older with PRES were identified. The mean age was 57.2 ± 0.6 years old with most encounters for female patients (71.7%, n = 455) and white as the most prevalent race. Half the patients in our study presented with seizures (50.1%, n = 318), sixty-three patients (9.9%) presented with vision loss, and sixty-four patients (10.1%) had speech difficulty. In addition, 45.5% of patients had hypertensive crisis (n = 289). 2.2% of hospitalizations had death as the outcome (n = 14). The mean LOS was 8.2 (±0.3) days, and the mean total charges were $92,503 (±$5758). Inpatient mortality differed between males and females (1.7% vs. 2.4%) and by race (3.6% in black vs. 1.8% in white) but was ultimately determined to be not statistically significant. Most patients who present with vision disturbance have a high risk of intracranial hemorrhage. Furthermore, end-stage renal disease, atrial fibrillation, and malignancy seemed to be linked with a very high risk of mortality. Conclusion PRES, formerly known as reversible posterior leukoencephalopathy, is a neurological disorder with variable presenting symptoms. Although it is generally a reversible condition, some patients suffer significant morbidity and even mortality. To the best of our knowledge, this is the largest retrospective cohort of PRES admissions that raises clinician awareness of clinical characteristics and outcomes of this syndrome.


Introduction
Posterior reversible encephalopathy syndrome (PRES) is a clinical, radiographic syndrome ranging from subacute to acute neurological manifestations primarily affecting the white matter [1,2]. PRES has been associated with hypertension, eclampsia, chemotherapeutic and immunosuppression medications, and transplant-recipient patients [1,2]. Generally, it is considered a benign disease if recognized early and promptly treated with the removal of causative factors [3,4]. However, there is a myriad of clinical conditions that when manifest with PRES symptoms carry a poor prognosis, such as lifethreatening cerebral hemorrhage, cerebellar herniation, and refractory seizures.
PRES is still an underdiagnosed neurological syndrome with a constellation of symptoms that share characteristic neuroimaging findings [2,3]. Demographically, it commonly affects middle-aged females with no statistical significance in mortality, gender, or race [5]. The most common symptoms include headache, confusion, visual disturbance, altered mental status, and seizures [2,3,6]. Neuroimaging shows a distinctive parieto-occipital finding in a bi-hemispheric distribution that reflects vasogenic edema [2,3].
PRES is reversible when recognized early and treated promptly by removal of the insulting factor; nonetheless, it can lead to irreversible and life-threatening complications which include cerebral hemorrhage, cerebellar herniation, and refractory status epilepticus [7]. In this analysis, we sought to investigate the clinical characteristics and outcomes of patients admitted to hospitals for PRES. Additionally, we aimed to identify predictors of unfavorable outcomes in relation to this syndrome.

Study design/settings
We utilized the National Inpatient Sample database provided by the Healthcare Cost and Utilization Project (HCUP-NIS) 2017. This database is a weighted 20% sample of all-payer hospitalizations in the United States. The selection of these hospitalizations was made systematically by the Agency for Healthcare Resources and Quality (AHRQ) to represent all the hospitalizations in the United States in 2017. The data collected includes demographic variables, admission diagnoses, procedures, type of insurance, geographical location, length of stay, inpatient mortality, and disposition, among others. These encounters represent hospitalizations, not patients, so multiple potential hospitalizations for the same patient are considered as multiple encounters.

Participants
We included all hospitalizations for patients primarily admitted for PRES. Encounters for patients aged < 18 years old were excluded from this study.

Variables
We aimed to investigate the demographic variables (age, sex, and race) for these patients, concomitant comorbidities and conditions, inpatient complications, inpatient mortality, length of stay (LOS), and disposition.

Data measurement
PRES and concomitant comorbidities and complications were identified through their international classification of diseases, 10th revision (ICD-10 codes) recorded in the discharge record for each hospitalization.

Ethical considerations
The NIS databases are considered "limited data sets" by the HIPAA privacy regulations; thus, do not require revision by an institutional review board (IRB) (http:// privacyruleandresearch.nih.gov). The study was conducted in agreement with the principles of the Declaration of Helsinki.

Statistical methods
Continuous and categorical variables were described as mean with standard deviation and frequencies, as appropriate. Chi-square was used to compare categorical variables, and an independent sample t-test was used to compare continuous variables. Multivariate logistic regression models (Enter method) were used to predict outcomes, and Hosmer and Lemeshow (HL) test were used to assess goodness of fit. Variables that predicted outcomes in univariate analyses (p < 0.1) were included in the multivariate models. Correlation matrices were created and evaluated to rule out any potential multicollinearities. All analyses were done using IBM SPSS statistics version 26.0 (IBM Corporation, Artmonk, NY). An alpha value (p) of 0.05 was used to ascertain statistical significance.

Results
A total of 635 admissions for PRES were identified. Missing data were only 26 values in race variable and three values in primary expected payer. All analyses were done using pairwise deletion for missing data, as missingness of less than 5% of data is unlikely to introduce bias [8].

Demographic variables
The mean age was 57.2 ± 0.6 years old. Most encounters were for female patients (71.7%, n = 455). White was the most prevalent race. Medicare, followed by Medicaid, were the most common primary payers (Table 1).
Outcomes 2.2% of hospitalizations had death as the outcome (n = 14). Mean LOS was 8.2 (±0.3) days, and the mean total charges were $92,503 (±$5758). Inpatient mortality differed between males and females (1.7% vs. 2.4%); however, the difference was not statistically significant (p = 0.561). Mortality also differed by race (3.6% in black vs. 1.8% in white), but none of the differences were statistically significant (p = 0.786). Similarly, there was no statistical difference in length of

Predictors of unfavorable outcomes
The predictability of different variables for unfavorable outcomes (mortality, cerebral infarction, intracranial hemorrhage) was evaluated using univariate analyses ( Table 2). The presence of anemia, infection, malignancy, atrial fibrillation, difficulty speaking, end stage renal disease (ESRD), malnutrition, and coagulopathy predicted mortality (p < 0.1) in univariate analyses. In a multivariate regression model only malignancy, atrial fibrillation, and ESRD predicted mortality (Table 3) and the HL test showed a good fit (p = 0.88). Vision loss and hypertension predicted intracranial hemorrhage in univariate analyses (p < 0.1). However, only vision loss predicted intracranial hemorrhage (OR 3.76 (95% CI 1.59-8.93, p = 0.003), with HL showing a good fit (p = 0.97). ESRD and dehydration predicted cerebral infarction in univariate analyses (p < 0.1) although none of them predicted it in a multivariate logistic regression model.

Discussion
At present, the exact pathophysiology of PRES unknown; however, multiple theories have centered around the disruption of the blood-brain barrier (BBB) by an insulting factor leading to extravasation of intravascular fluid and cerebral edema [9]. Rapid elevation of blood pressure is well known to disrupt the BBB and has been noted in most reported cases of PRES [6,10]. Other causes of BBB disruption include autoimmune diseases, which were also noted in patients with PRES, especially those with normal blood pressure [11]. Endothelial damage can also be cytokine-mediated [11]. Cytokines stimulate the secretion of vasoactive factors such as vascular endothelial growth factor (VEGF), which increases permeability. The following discusses a few important prognostic indicators and factors that impact patient outcomes.

Demographics
PRES can present at any age and can range widely. Our average age of presentation for our patients was 57 years old; however, cases have been reported as old as 90 years old [3]. PRES is most commonly seen in women than in men. Our study demonstrated similar results, with 72%  of PRES patients being females. It has been proposed previously that women may be a higher risk and have more severe symptoms which may be due to women having fewer inter-neuronal connections and more diffusivity in parieto-occipital regions [12,13].

Clinical presentation
Presenting symptoms of our patients were consistent with those observed in prior studies. The main symptoms included seizures and hypertension [2,3,6]. Our study validates that hypertension may play a key role in the development of PRES as about 87% of admitted patients were hypertensive. This was explained in previous studies by the direct effect of blood pressure on the blood-brain barrier function and the resultant brain edema [14]. In an animal model, a rapid elevation of blood pressure resulted in leakage or exudation of plasma, macromolecules, or red blood cells [15]. Vision disturbance was also reported in 63 patients (9.9%). This relates to the occipital lobe involvement that gives various presentations such as vision loss and visual hallucinations [16,17]. Our study revealed a strong correlation between vision loss, among all other presentations, with intracranial hemorrhage (OR 3.76, p = 0.003). The pathogenesis of parenchymal or sulcal subarachnoid hemorrhage in PRES could be due to either postischemic reperfusion damage or impaired autoregulation [18,19]. Speech difficulty was reported in 64 patients (10.1%).

PRES and ESRD
In our study, ESRD was a high predictor of mortality (n = 108 [17%]) in patients with PRES, with the likelihood of dying about seven times higher than those without ESRD (OR 7.27, 95% CI 2.06-25.63, p = 0.002). Many studies revealed a heterogeneous association of renal diseases with PRES, ranging from no association in one series to approximately 30% in other large series [6,20]. Several possible etiological factors play a role in the pathogenesis including hypertension, drugs, and fluid/ electrolyte disturbances. ESRD patients have a higher likelihood of hypertension and fluid overload, which supports the hyperperfusion theory of PRES in ESRD patients [21,22]. In these patients, toxic metabolic effects of urea may lead to cytokine-mediated damage to the blood-brain barrier. Another possible mechanism is vasoconstriction from erythropoietin infusion [23]. There have also been reports of PRES in patients who are started on hemodialysis [24]. It is hypothesized that initiation of dialysis results in a difference in pretreatment and post-treatment urea levels and resultant acidosis. This promotes the shift of sodium, potassium, and other electrolytes which contributes to endothelial dysfunction and resultant vasogenic edema [21,25]. Our study did not review patients who were initiated on dialysis; however, it still does show a strong correlation between ESRD and mortality in PRES patients. Patients with ESRD and PRES should be monitored closely due to the higher chance of mortality.

Malignancies
Our study found malignancy to be a strong predictor of mortality in PRES (OR 4.18, 95% CI 1.07-16.32, pvalue = 0.039), which was consistent with prior studies [25,26]. Cancer patients are more likely to have other comorbidities than non-cancer patients, such as thrombocytopenia, renal failure, or medication-induced hypertension [26]. Immunosuppressive and chemotherapeutic agents have another potential etiology for the increased incidence of PRES in these patients. The neurotoxicity of these agents is well known in the literature but still not yet fully understood [10]. Toxic levels in the blood are not required to develop PRES and can develop at any time during therapy [10]. Few studies have reported complete resolution of the clinical and radiological symptoms despite the continuation of medications [27]. Because of the markedly increased mortality in cancer patients who develop PRES, we stress early measures to diagnose and treat these patients when there is a slight suspicion of PRES. Early lowering of the drug dose or complete discontinuation of the cytotoxic or immunosuppressive drug is recommended to prevent deterioration of the clinical condition, avoiding permeant deficit and poor outcomes [26].

Atrial fibrillation
Atrial fibrillation was documented in approximately 10% of our patients with a odds ratio suggesting a four time higher risk of mortality (OR 3.842, 95% CI 1.01-14.55, p-value = 0.048). A prior study by Hinduja et al. examining patients with PRES and predictability of intensive care unit admissions showed that all patients with atrial fibrillation required ICU admission [28]. This implied the correlation that patients with PRES and atrial fibrillation may be more critically ill.

Sepsis and severe infections
In our study, 198 patients (31.2%) had an associated severe infection or sepsis. These patients had a high odds ratio or mortality; however, this did not reach statistical significance (OR 3.1, 95%CI 0.95-10.23, p = 0.061). Severe infection or sepsis association with PRES was previously reported in the literature; however, was not well established as other comorbidities (e.g., hypertension, pre-eclampsia/eclampsia, cytotoxic/immunosuppressive, and autoimmune disease) [29]. Many theories have been put forward to explain the pathophysiology related to this association. As an example, Marra et al. described the mechanism of endothelial damage of PRES in sepsis patients by the primary inflammatory involvement with T cell activation and cytokine release [9]. Separately, Gao et al. in their neuropeptide theory suggested that endothelial dysfunction is induced by the release of endothelin-1, prostacyclin, and thromboxane A2 [30]. Powerful stimulation of the systemic inflammatory response may be related to the development of PRES in severe infection [31,32]. Further investigations are needed accurately to determine a possible correlation between PRES and infections and its impact on mortality.

Conclusion
Posterior reversible encephalopathy syndrome (PRES), formerly known as reversible posterior leukoencephalopathy, is a neurological disorder with variable presenting symptoms. Although it is generally a reversible condition, some patients suffer significant morbidity and even mortality. Upon reviewing the literature to date, this is the most extensive retrospective study of PRES admissions. Most patients who present with vision disturbance have a high risk of intracranial hemorrhage. Furthermore, ESRD, atrial fibrillation, and malignancy seemed to be linked with a very high risk of mortality. Ours is the largest study that raises awareness of PRES among the internists, neurologists, nephrologists, and hematooncologist and calls for a collaborative approach to mitigate the devastating consequences of this syndrome. Based upon the findings, we suggest prompt and meticulous fluid/electrolyte balance, blood pressure control, and avoidance of rapid changes from baseline, and chemotherapeutic agents in the management and reversal of this condition. Since kidney injury can be an important determinant of outcomes, further studies are needed to determine the impact and degree of renal insufficiency on PRES.