Predictors of neurological deterioration during admission for patients with cerebellar strokes

Background: Despite less common, cerebellar stroke frequently results in unfavorable outcomes, especially after deterioration. Therefore, this study was aimed to identify the signicant predictors of neurological deterioration during admission (NDDA) in ischemic and hemorrhagic cerebellar strokes. Materials and methods: We retrospectively reviewed all medical records of the patients diagnosed with ischemic and hemorrhagic cerebellar strokes during 2002-2018 in Songklanagarind hospital. Comparison of patients’ demographic data, initial clinical presentations, neuroradiological results, timing and signs of NDDA, and outcomes between cerebellar strokes were descriptively analyzed. Logistic regression model was applied for determining the signicant predictors of NDDA from initial clinical presentations. Results: 74 of 100 patients were eligible. They comprised of 42 (57%) cerebellar ischemia (CI) and 32 (43%) cerebellar hemorrhage (CH). Elevated diastolic blood pressure (DBP) and the neuro-radiological evidences suggesting increased posterior cranial fossa pressure were signicantly prevalent in neurological deterioration patients. NDDA was found in 31 patients (42%) without signicant difference between CI and CH. 42 (56.8%) patients had poor neurological outcomes. The independent predictors for NDDA were DBP B 120 mmHg (adjusted odds ratio [adj. OR] 15.39, 95% CI 1.58-149.59; p = 0.004), time from onset to arrival (adj. OR 0.98, 95% CI 0.97-1.00; p = 0.044), and hemispheric cerebellar signs (adj. OR 0.22, 95% CI 0.06-0.75; p = 0.012). Conclusions: CH was not an independent predictor of NDDA in overall cerebellar strokes. Only high DBP predicted NDDA, whereas time to arrival and hemispheric signs showed protective impact.


Background
Although cerebellar stroke is less common (1-3% of all strokes), it frequently results in high morbidity and mortality rates (25-100%). [1,2] Because of the limited space of posterior cranial fossa where cerebellum locates and the closed contact of cerebellum, particularly cerebellar vermis, with brain stem, ischemic cerebellar edema or cerebellar hematoma poses high risk of tonsillar herniation as well as direct brain stem compression. These are key mechanisms of secondary neurological deterioration during admission (NDDA) from cerebellar strokes. Generally, NDDA occurs a few days after the onset of cerebellar infarction (CI), while it commonly takes a few hours after cerebellar hemorrhage (CH). [3] Therefore, close observation for early detection of neurological deterioration, and timely initiation of neurosurgical interventions are necessary for favorable outcomes after a cerebellar stroke.
Previous series reported about natural course and outcomes of cerebellar strokes. The predictors of poor outcome in cerebellar strokes from several studies included high blood pressure, high blood glucose level, hydrocephalus and obliterated fourth ventricle from radiologic ndings. There have been limited studies reporting the prognosticators of NDDA that required neurosurgical interventions in initially non-surgically indicated patients. [1][2][3][4][5][6][7][8][9][10] Thence, this study was aimed to determine the prognostic factors of surgically-indicated NDDA for both CI and CH from the patient's demographic and presentation characteristics. Our ndings might help identify patients at risk who need intensive monitoring or early intervention.

Objectives
Our primary objective was to identify predictors of secondary neurological deterioration after admission.
The secondary objective was to know neurological outcomes of patients admitted with cerebellar strokes at discharge and at 90 days after onset.

Study Designs And Setting
This retrospective cohort study was conducted in Songklanagarind Hospital, an 800-bed tertiary and medical teaching university hospital in southern Thailand. The medical records between 1 January 2002 and 31 December 2018 were reviewed.

Eligible Criteria
Medical records of admitted patients aged ≥ 18 years diagnosed with cerebellar strokes during the study period were reviewed. We excluded the patients who had Glasgow coma score (GCS) ≤ 3 or underwent emergency neurosurgical interventions at the time of presentations.

Cerebellar strokes
The diagnosis of cerebellar strokes must be con rmed by neuroimaging either brain CT or brain MRI interpreted by a certi ed radiologist.
Secondary neurological deterioration during admission (NDDA) Secondary neurological deterioration during admission was de ned by the emergence of any new pyramidal tract signs, new brainstem signs, new cerebellar signs or lowering of GCS ≥ 1 point from the baseline assessment.

Neurological outcome
The neurological outcomes evaluated by modi ed Rankin score (mRs) on the day of hospital discharge as well as 90 days after stroke onset. Stroke outcomes evaluated by mRs were classi ed into 2 groups: favorable outcome (mRs 0-2) and unfavorable outcome (mRs 3-6).

Data Collection
We collected data from the electronic medical record including the patients' demographic data, presenting symptoms and signs, results of routine blood analysis, brain imaging reports, date of NDDA occurrence, neurosurgical interventions performed for treating NDDA.

Statistical analysis
Comparison of patient demographic data, initial clinical characteristics, the presence of NDDA and outcome between CI and CH were descriptively analyzed. The discrete data were analyzed by Chi-square test. The continuous data were analyzed by independent t-test and Mann Whitney U test. The signi cant variables with p < 0.2 in univariate analysis were entered to multivariate logistic regression model. The variables were considered as independent predictors if they meet the statistical signi cance with p < 0.05 in multivariate logistic regression analysis.

Results
Demographic and presenting characteristics One hundred cerebellar stroke patients were initially included. After exclusion of twenty-six patients (11 cases with GCS ≤ 3, and 17 cases with surgical interventions performed at the presentation), 74 initially non-surgically indicated patients admitted for supportive treatments and clinical observation were eligible for nal analysis. (Fig. 1) There were fty-two (70%) male and 22 female patients. Among them, 42 patients were CI and 32 patients were CH. No signi cant difference in median (IQR) age was found between the two groups of patients. (Table 1) The overall median (IQR) time from stroke onset to hospital arrival was 8 (2.12, 27.94) hours with signi cantly shorter time in neurological deterioration patients. Thirteen patients (5 CH and 8 CI) were initially misdiagnosed as peripheral vertigo on their presentations to the emergency department.    In comparison of the number of cases with unfavorable outcomes between patients with and without NDDA, there were 28 (90%) and 24 (77%) cases in NDDA group comparing to 14 (33%) and 11 (26%) cases in no NDDA patients, at the discharge date and 90 days after onset, respectively. (Fig. 1) We also found that NDDA cerebellar stroke patients had signi cant proportion of unfavorable outcomes evaluated by mRs at hospital discharge (p < 0.001) and at 90 days after onset (p < 0.001). (Fig. 2)

Discussion
Our study found signi cantly higher proportion of headache as a presenting symptom and elevated DBP in CH patients. Moreover, white blood cells (WBC) ≥ 10,000 cells/mm 3 , obstructive hydrocephalus, and distortion of the fourth ventricle on initial brain scans were signi cantly more prevalent in CH as well.
( Table 2) We propose that an abrupt and rapid increment of posterior cranial fossa pressure from cerebellar hematoma plus its later expansion possibly causes the anatomical distortions. Furthermore, acute physiological reaction to the acutely elevated intracranial pressure (ICP) probably results in leukocytosis. A study by Furlan showed that leukocytosis on admission was associated with poor outcome too. [11] Notably, we found no signi cant difference in NDDA between the two subtypes of cerebellar strokes. In addition, CH was not an independent risk factor of NDDA by logistic regression analysis in the current study. (Table 3) However, the time from onset to NDDA was signi cantly shorter in CH. (Table 2) This is explainable by more rapid rising of ICP in CH than in CI. Unlike a few previous studies, [6,8] which included wider range of severe cerebellar stroke cases, the signi cant differences in characteristics between CH and CI were more obvious than ours. As we aimed to determine the predictors of NDDA in the patients initially without indication for neurosurgical interventions, we excluded all the cases with neurological deterioration at their rst presentations. And, the limited number of cases enrolled in our study was likely to have fewer clinical parameters with statistically signi cant differences reported.
The shorter time from stroke onset to hospital arrival and hemispheric cerebellar signs at presentation were protective factors on multivariate analysis. (Table 3) We considered that the presence of cerebellar hemispheric signs was well realized by most physicians of having hemispheric cerebellar disorder, facilitating immediate neuro-imaging study and therapy. In contrast, in cases of cerebellar vermis stroke mostly are under evaluated, or missed as a peripheral vestibular disorder causing delayed diagnosis and also proper management. As found in our study, the presence of hemispheric cerebellar signs was a signi cant protective factor for NDDA in our study. (Table 3) Positive hemispheric cerebellar signs corresponding with the presence of hemispheric cerebellar lesions seen on the imaging studies were 51/74 (70%) cases in our study. 34/43 of them (79%) acquired favorable outcome eventually. A study by Erik, et al. supported our nding as they reported that cerebellar vermis hemorrhage was associated with higher rates of neurological deterioration. [6] Direct compression of the hematoma against brainstem was attributed.
Our study revealed that DBP 120 mmHg was an independent predictor of NDDA (adj. OR 15.39, 95% CI 1.58-149.59; p = 0.004). (Table 3) Elevation of blood pressure has been considered as a response to elevation of ICP at the stroke onset, however, it probably leads to neurological deterioration because of increased risk of massive cerebral edema and hematoma expansion as well. [12][13][14][15][16][17] Hyperglycemia (Blood sugar 140 mg/dL) was not an independent predictor of NDDA in the current study. (Table 3) Actually, we found that the median (IQR) Blood sugar (BS) level in our cases (117 (100, 133.7) mg/dL) was lower than some previous studies (> 150 mg/dL). [7,8,18] Therefore, lower BS levels would contribute to better cerebellar stroke outcomes in our study. To our knowledge, hyperglycemia worsens the overall stroke outcomes, because high blood sugar level has been known to exert adverse effects on the structures of cerebral vascular endothelial cells, and to induce acute oxidative stress along with vascular endothelial in ammation. [19,20] Koh [1] concluded that hydrocephalus, brain stem deformity and basal cistern compression were associated with NDDA in cerebellar infarction. St. Louis [6] also reported that patients with a cerebellar vermis hematoma and acute hydrocephalus were at high risk for NDDA. Furthermore, Ho [8] reported that obliteration of basal cistern on the initial CT brain scans was associated with NDDA in cerebellar hemorrhages. Based on our available results, we found no neuro-imaging abnormality as a predictor of NDDA by multivariate analysis. (Table 3) Since most of the brain images done in our study were CT scans, demonstration of such mentioned imaging abnormalities in association with NDDA is possibly obscured.
Thirty-one (42%) of all cerebellar stroke patients developed NDDA, and 28 of the 31 (90%) patients acquired unfavorable neurological outcomes at hospital discharge. When compare with the overall cerebellar stroke outcomes, only 37.8% had favorable outcomes at discharge (2.4 folds higher in NDDA cases). Some previous studies reported a slightly higher percentage of unfavorable nal outcomes (50%). [1,6,8] With the available information and based on our current ndings, a worse neurological outcome is undoubtedly higher in NDDA cerebellar stroke patients.
Under limitation of accessibility of MRI brain, particularly under emergency service setting in our center and the similar others, we speculated that some clinical presentation characteristics could be practically useful to predict the occurrence of NDDA among the initially non-surgical-indicated patients. We hope that our ndings could facilitate appropriate monitoring and timely starting of necessary neurosurgical interventions aiming at favorable cerebellar stroke outcomes.

≥ ≥
The limitations of the current study are retrospective design and single-center study with limited sample size. Further prospective and multi-center studies, which include more study samples with variability of cerebellar stroke severity would be useful in providing an appropriate management decision on initiation of early neurosurgical interventions for cerebellar stroke patients.

Conclusions
Cerebellar hemorrhage was not an independent predictor for NDDA for all cerebellar strokes in this study.
Some initial clinical presentations of cerebellar stroke, regardless of types of cerebellar stroke, are potentially applicable to predict NDDA along with favorable short and long-term outcomes. Comparison of modi ed Rankin scores (A) at discharge date and (B) 90 days after onset between cerebellar stroke patients with and without neurological deterioration