Introduction

A remote cerebellar hemorrhage (RCH) is a postoperative complication characterized by spontaneous bleeding in the posterior fossa after supratentorial procedures. It is a relatively rare occurrence with an incidence of approximately 0.08–0.6 % [2].

RCH has been described after different supratentorial craniotomies [11, 45], burr holes for hematomas’ evacuation, or trans-sphenoidal procedures [2].

The typical bleeding pattern is defined as “the zebra sign” which is characterized by blood in the sulci of cerebellar hemispheres and vermis facing the tentorium. It has been reported as an isolated pattern or in association with a intracerebellar hemorrhage (ICH). Since RCH was first described by Yasargil and Yonekawa in 1977 [52], it has intrigued the neurosurgical community due to the uncertainty regarding the pathophysiology, risk factors, best treatment, and outcome.

This paper is the first systematic review of the English literature reporting cases of RCH after supratentorial craniotomies. The aim of this study is to identify which procedures are more frequently associated with RCH and discuss the pathophysiological mechanisms and the possible risk factors.

Materials and methods

Study selection

A comprehensive review of the literature was performed using the keywords “remote AND cerebellar AND (hemorrhage OR stroke)” to search in PubMed and Scopus databases. Search terms were identified by consulting MeSH terms and according to the most common definition adopted by the authors. All studies in the English language reporting on cases of RCH after supratentorial procedures were selected. Review papers, letters, and cases of RCH after infratentorial or spinal procedures were excluded. Two couples of reviewers (CLS/MR and FV/VB) independently abstracted data from the included papers. Any differences were resolved by consensus and discussion with the senior author (DdA). The last search was launched in September 2013. The search strategy followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement where applicable, and this checklist was used in designing and reporting our review.

Variables and risk factors assessment

We evaluated the following variables: demographics, primary disease (intracranial aneurysm (IA), tumor, focal epilepsy, vascular malformation, or other), and surgical procedure (clipping, tumor/malformation resection, hematoma evacuation, lobectomy, bypass, or other). In regard to RCH characteristics, we assessed the following variables: pattern (zebra sign, ICH, or mixed), clinical onset (asymptomatic, cerebellar signs, seizures, impairment of consciousness, or coma), timing of bleeding (immediately after surgery, within 24, 48, 72 h or later than 72 h), and the relationship between side of craniotomy and side of RCH (omolateral, contralateral, bilateral, or midline).

Moreover, we considered the different management strategies adopted by the authors (clinical observation, external ventricular drainage (EVD), sub-occipital decompression (SD)), the clinical status at hospital discharge and at follow-up (assessed by Glasgow Outcome Scale (GOS)). For studies reporting a different outcome scale, a correspondent GOS score was assigned. According to the GOS, patient outcome was dichotomized as follows: “good” in those with good recovery or moderate disability, and “poor” in those with severe disability, vegetative state, or dead.

Finally, we explored the frequency of a number of potential risk factors distinguishing them in: preexisting pathological conditions, intraoperative complications, and surgical maneuvers potentially at risk for RCH; we also analyzed their relationship with outcome.

According to the literature, a zebra sign was defined as alternating hyperdense/hypodense curvilinear, slightly irregular stripes along the cerebellar folia, imitating a zebra’s coat, indicating subarachnoid bleeding [7], whereas an ICH was defined as a hemorrhage within the cerebellar parenchyma.

Statistical analysis

Data were individually extracted for each patient. Percentage of incidence and 95 % of confidence intervals (CI) were calculated for all considered variables and outcomes. For those risk factors with the similar clinical significance, we also assessed their cumulative incidence.

Since we collected data from a large time period and pooled together case reports and case series, we stratified the outcomes according to the year of publication (before and after 2001 at intervals of 4 years), and the sample size (less vs more than ten cases), in order to evaluate selection or publication bias.

Quantitative variables were expressed as mean ± standard deviation and Student’s t test was used to compare their means. Fisher’s exact test (two-sided) was instead used to compare categorical variables.

Two logistic regression models were built in order to assess the independent contribution of those variables significantly associated with outcome at univariate analysis, alternatively introducing the time as variable. Positive (PPV) and negative predictive value (NPV) of the main predictors were also measured, and the receiver operating characteristics curve (ROC) analysis was performed to determine sensitivity and specificity. The associations were considered significant when p < 0.05. The statistical package used was SPSS 13.0 (SPSS Inc., Chicago, IL).

Results

Study selection

The electronic literature search yielded 154 results; 49 articles (from 1977 to 2013) met our inclusion criteria and were analyzed in detail [1, 3–31, 34–40, 42–54]. Eighteen papers were case reports, and 31 were series including from 2 to 42 patients. Overall, 209 patients were included in the study.

Demographic characteristics

Age was reported in 194 out of 209 patients (92.82 %), and the mean age at diagnosis was 49.09 ± 17.07 years (range 12–80). Sex was reported in 207 out of 209 patients (99.04 %) and the male/female ratio was 130/77 (62.8 vs 37.2 %).

Associated procedures

RCH was a complication more frequently observed after supratentorial craniotomies for treatment of IA (39.2 %), tumors (26.3 %), focal epilepsy (20 %), hematomas (7.1 %), vascular malformations (3.3 %), ischemia (1.4 %), and other diseases (2.3 %). All reported IA (26.8 % ruptured) underwent clipping; tumors were treated with debulking in 92.7 % and lobectomy in 7.3 % of cases; patients with focal epilepsy underwent frontal or temporal lobectomy in 14.3 and 85.7 %, respectively; similarly, vascular malformations underwent selective resection in 85.7 % and lobectomy in 14.3 %; all cerebral hematomas were evacuated through external drainage; three patients affected by cerebral ischemia underwent STA-MCA bypass; different procedures were instead performed in the remaining five patients. Aneurysm clipping (39.2 %; 95 % CI, 32.8–45.9), parenchymal lesion debulking (27.2 %; 95 % CI, 21.6–33.6), and lobectomies (22.4 %; 95 % CI, 17.3–28.6) were the procedures most frequently associated with RCH.

About 6.2 % of RCH (13/209; 95 % CI, 3.6–10.3) occurred, instead, after supratentorial craniectomy or burr holes.

Clinical onset and timing

The clinical status after RCH was described in 197 out of 209 patients (94.5 %). In 57 cases (28.9 %; 95 % CI, 23.0–35.6) RCH was incidentally discovered on postoperative imaging with no associated symptoms; in 88 cases (44.6 %; 95 % CI, 37.9–51.6), it was associated with impairment of consciousness, in 25 (12.6 %; 95 % CI, 8.7–18.0) with cerebellar signs, in 17 (8.6 %; 95 % CI, 5.4–13.3) with seizures, and in 10 (5 %; 95 % CI, 2.7–9.0) evolved in a comatose state.

Timing of the clinical onset was reported in 152 out of 209 patients (72.7 %): 48 (31.5 %; 95 % CI, 24.7–39.3) showed symptoms immediately after surgery, 72 (47.3 %; 95 % CI, 39.5–55.2) within 24 h, 22 (14.4 %; 95 % CI, 9.7–20.9) between 24 and 72 h, and 10 (6.5 %; 95 % CI, 3.6–11.6) later than 72 h from surgery. Overall, 80.4 % (95 % CI, 73.1–86.0) of patients had a clinical onset during the first postoperative day.

RCH characteristics

RCH pattern was reported in 194 out of 209 patients (92.8 %): 126 showed a zebra sign (64.9 %; 95 % CI, 58.0–71.3), 36 an ICH (18.5 %; 95 % CI, 13.7–24.6), and 32 (16.4 %; 95 % CI, 11.9–22.3) a mixed pattern (Fig. 1).

Fig. 1
figure 1

Relationship between pattern and side of the RCH. Both pattern and side of the RCH were available in 190 patients

Full size image

The side of the RCH was reported in 205 out of 209 patients (98 %): 113 were bilateral (55.1 %; 95 % CI, 48.2–61.7), and 92 unilateral (44.9 %; 95 % CI, 38.2–51.7). Among them, 42 (20.4 %; 95 % CI, 15.5–26.5) were omolateral to the supratentorial craniotomy, 45 (21.9 %; 95 % CI, 16.8–28.1) were contralateral, and 5 were located at the midline (2.4 %; 95 % CI, 1.0–5.5). Bilateral localization was about four times more frequent among patients with zebra pattern (Fig. 1).

Treatment and clinical outcome

Data on treatment were reported in 207 out of 209 patients (99 %). Clinical observation was the strategy adopted in 157 cases (75.8 %; 95 % CI, 69.5–81.1); 27 patients (13 % 95 % CI, 9.1–18.3) who showed a progressive impairment of consciousness due to the appearance of hydrocephalus underwent EVD, whereas 23 (11.1 %; 95 % CI, 7.5–16.1) who showed a significant mass effect with cerebellum and brainstem distortion underwent SD. Among the latter, 16/23 (69.5 %) showed ICH or mixed bleeding.

Outcome at discharge was reported in 179 out of 209 cases (85.6 %). According to the GOS, 134 patients (74.8 %; 95 % CI, 68.0–80.6) had a good outcome, whereas 45 (25.1 %; 95 % CI, 19.3–31.9) had a poor outcome. There was no significant difference in mean GOS at discharge between studies reporting less or more than ten cases. A trend of progressive GOS improvement over time was instead evident stratifying the outcome according to the year of publication, with a significant difference between studies published before and after 2001 (Table 1).

Table 1 Outcome stratification according to year of publication
Full size table

Outcome at follow-up (mean 5.2 ± 10 months) was reported in less than half of patients (91 out of 209; 43.5 %): among them, 88 (96.7 %; 95 % CI, 90.7–98.8) showed a good outcome and 3 (3.3 %; 95 % CI, 1.1–9.2) a poor outcome. Twenty-one patients died during postoperative course (11.7 %; 95 % CI, 7.8–17.2). Regardless of the associated procedure, mortality was mainly (17 out of 21 cases, 80.9 %) due to cerebellar ICH causing severe brainstem compression.

Relationship among demographics, RCH characteristics, treatment, and outcome

Table 2 shows the relationship among sex, mean age, RCH characteristics, clinical onset, treatment, and outcome at hospital discharge.

Table 2 Relationship between demographics, RCH features, risk factors, treatment, and outcome
Full size table

Among the 179 patients whose outcome at discharge was reported, data regarding age were available in 176 (98.3 %), sex in 177 (98.8 %), timing of RCH appearance in 122 (68.1 %), clinical status in 171 (95.5 %), bleeding pattern in 168 (93.8 %), side of hemorrhage in 171 (95.5 %), and mode of treatment in 175 (97.7 %).

Univariate analysis showed that older patients had a significantly worse outcome (p ≤ 0.0001), whereas sex did not seem to affect the prognosis. Ninety-five percent of patients who presented asymptomatic at onset had a good prognosis (p = 0.0006), whereas the timing of RCH appearance did not show association with outcome. Patients showing zebra pattern had a significantly more frequent good outcome (p = 0.0001), inspite of the fact that the majority underwent no surgical treatment. On the other hand, patients showing a pure ICH had a significantly poorer outcome (p = 0.0002). Finally, bilateral localization did not appear to be necessarily related to a worse prognosis.

Risk factors

Percentages of incidence and 95 % CI of potential risk factors are reported in Table 3.

Table 3 Risk factors for RCH
Full size table

A history of hypertension or epilepsy was reported in approximately one fourth of patients (56/209, 26.7 %, and 49/209, 23.4 %, respectively). About 19.1 % of patients (40/209) were taking antiplatelet/anticoagulant drugs approximately within a week before surgery. Nearly 13.4 % (28/209), instead, suffered from pathologies potentially impairing the coagulation (deficit of factors/platelet, hematologic tumors, radiochemotherapy, anemia, alcohol addiction, chronic kidney failure, etc.).

Perioperative seizures and hypertension peaks were the complications most frequently reported in association with RCH (32/209, 15.3 %, and 28/209, 13.3 %, respectively). Also, hydrocephalus was reported in 10.5 % of patients (22/209) with RCH and pneumocephalus in about 8.1 % (17/209). Finally, thrombotic/hemorrhagic complications occurred in less than 5 % of cases (10/209).

Surgical maneuvers responsible for prolonged CSF leakage, such as the placement of epidural drainages, were performed in approximately one fourth of patients (52/209); a spinal drainage was placed in almost 14 % (29/209); a ventricular/subdural catheter in 11 % (23/209). Overall, at least one drainage was present in about 34 % of patients (71/209).

The presence of at least one risk factor (cumulative risk) for impaired coagulation, hypertension, or seizures was present in about one third of patients (70/209, 33.4 %, and 72/209, 34.3 %), respectively. In particular, assumption of antiplatelet/anticoagulant drugs and factors impairing coagulation appeared significantly associated with poor outcome (Table 2).

Multivariate analysis

A logistic regression model, considering poor outcome at discharge as dependent variable and zebra pattern, symptomatic onset, antiplatelet/anticoagulant intake, and age as covariates, confirmed that symptomatic onset, intake of antiplatelet/anticoagulant drugs, and older age were independently associated with poor outcome (Table 4). Factors potentially impairing coagulation other than antiplatelet/anticoagulant drugs were excluded from the multivariate analysis, as it was postulated that they may have contributed to altering the coagulation properties. Likewise, the surgical treatment was excluded as the majority of patients were managed conservatively and surgery was reserved for only those who showed a more severe clinical picture. Contrariwise, the zebra sign appeared as a more benign hemorrhagic pattern (Table 4). However, when the year of publication was added to the statistical model as an additional covariate, the zebra pattern no more appeared as a significant variable.

Table 4 Logistic regression model assuming poor outcome at discharge as dependent variable
Full size table

Since all significant predictors of good/poor outcome on univariate analysis showed a decreasing trend in their association strength over time, we also built a second model of multivariate analysis only including studies published after 2001. In this new model, only symptomatic onset and advanced age appeared as significant predictors of poor outcome.

Sensitivity for predicting poor outcome was 36 % and specificity was 94.4 %. The PPV was 60.0 % and the NPV of 86.4 %. The ROC curve analysis calculated for these two variables presented an area under the curve of 0.845 (95 % CI, 0.77–0.92).

Discussion

This review shows that RCH after supratentorial craniotomies is a complication more frequently reported for IA clipping, tumors’ resection, and lobectomies for focal epilepsy. Overall, these results reflect those observed in the largest series published in literature [11, 16, 20] and support the hypothesis that procedures with greater CSF leakage have higher risk of postoperative RCH. The main RCH pattern has been in fact described as a predominantly subarachnoid bleeding, which probably originates from ruptured veins as they course through the cerebellar fissure or where they enter the cerebellar parenchyma. An excessive CSF leakage, in fact, may cause cerebellum shift, thus resulting in stretching and tearing of these bridging veins [32, 33]. In contrast with this hypothesis, the majority of patients who have a pattern of zebra hemorrhage do not show associated parenchymal infarction, which would be expected after interruption of these veins [11].

RCH was an incidental finding on postoperative imaging in more than one third of cases (37 %), while in the remaining two thirds, it was associated with progressive impairment of consciousness or cerebellar signs; rapid progression toward coma was found in only a minority of patients.

Hence, RCH represents a fairly benign condition, with good outcome in almost three fourths of all patients (74.8 %, 95 % CI, 68.0–80.6) and in more than 85 % of those with the zebra pattern (85.2 %; 95 % CI, 77.1–90.8).

However, the overall mortality rate reported after RCH is not negligible (11.7 %), but if we consider only that observed in patients with the zebra pattern, the rate is significantly lower (about 1.5 %). Accordingly, zebra hemorrhage and ICH could represent two different nosological entities which do not share the same pathophysiology. Compared to pure ICH, in fact, the zebra pattern appears three times more frequent, usually bilateral, and characterized by a significantly better outcome.

This study also reveals that the influence of bleeding pattern on outcome appears progressively reduced over time, probably due to advances in neurosurgical and intensivistic management.

With regard to possible etiological risk factors, this review shows that alterations of coagulation, hypertensive status, and perioperative CSF drainages are described in more than one third of patients with postoperative RCH, respectively. However, the frequency of a previous history of hypertension in these patients does not significantly differ from that of the general population with the same mean age. Also, since most of the surgical procedures associated with RCH were elective, we could expect that the majority of patients had suspended the anticoagulant/antiplatelet drugs at least a week before surgery. Thus, the association with this factor could only reflect a more general relationship with cardiovascular risk factors.

Finally, this review does not provide sufficient information to validate other different pathophysiological hypotheses, such as the role of the intraoperative head rotation and its influence on jugular outflow [11, 41, 45], or the Papanastassiou’s hypothesis, which proposes that RCH would preferentially occur contralateral to the supratentorial craniotomy as this part of cerebellum, due to CSF leakage, moves posteriorly to abut the transverse sinus, kinking, and obstructing the veins draining from the superior aspect of the hemisphere [36]. In fact, in patients with unilateral RCH (about 42 %), we observed that the percentages of hemorrhages omolateral and contralateral to the side of the craniotomy were approximately the same.

Also, these data cannot certainly establish if the high percentage of RCH reported in patients with a previous history of epilepsy only reflects an association with the procedures of lobectomy or if the occurrence of perioperative seizures (also undetected) may cause venous congestion and subarachnoid bleeding.

Similarly, the relatively high frequency of RCH among patients who underwent large tumor resections only partially corroborates the hypothesis of Konig et al., who suggested that removal of supratentorial space occupying masses may induce a reduction of intracranial pressure and an associated critical increase of transmural venous pressure causing bleeding [25].

This study presents several limitations. First of all, 18 of the included papers were case reports and 18 were case series reporting less than five patients. Also, the reported evidence is observational and noncomparative; in fact, data regarding the denominator were lacking in the majority of the analyzed studies. Therefore, the proportions reported in this review only reflect the published cases and may be influenced by some form of publication bias. On the other hand, the stratification of outcomes according to the sample size did not show significant differences, therefore limiting the impact of a possible selection bias. However, we cannot exclude that the severity of the underlying disease, in addition to the occurrence of a RCH, has affected the outcomes (including mortality) in some cases. In fact, it is noteworthy that the mean percentage of good outcome among patients with RCH has increased over time maybe due to the introduction of a better management, being limited to about 50 % until 2000 and constantly around 80 % in the papers published since 2001 (Table 1).

Moreover, data regarding the variables and short-term outcomes we investigated were not available for all patients and the reported percentages were assessed in subgroups. However, these data were accessible in approximately 90 % of patients we considered. Instead, data regarding follow-up were available only in 43.5 % out of patients and they did not include a reliable evaluation of long-term prognosis.

Furthermore, in regard to patients with ruptured IA, neither it was possible to differentiate the contribution of a the previous SAH on the postoperative RCH, nor its influence on outcome.

Finally, risk factors analyzed in this study has showed a NPV of only 86.4 %, and a PPV not exceeding 60 %.

However, this study has some strengths. We followed a comprehensive systematic review process and extracted all available demographic, clinical, and neuroradiological data in order to improve the knowledge about the RCH characteristics and the associated risk factors. Moreover, so far, this is the only systematic review on this topic.

Conclusions

RCH after supratentorial craniotomy is a fairly benign and frequently self-limiting condition, particularly when it occurs in younger patients and shows a zebra hemorrhage pattern. The presence of ICH, instead, may result in a more severe clinical picture. All available clinical and neuroradiologial data suggest a venous origin of the bleeding. The role of coagulation disorders, perioperative CSF drainages, and hypertension as possible risk infactors appears limited, since their incidence does not exceed about one third out of the reported cases, respectively.

In the opinion of most authors, RCH usually does not require surgical treatment except for cases associated with hydrocephalus or progressive deterioration of consciousness.

The correct interpretation of this hemorrhagic complication far from the field of surgery is therefore mandatory to plan a prolonged clinical surveillance and to identify those cases which may potentially evolve and result in severe neurological impairment.