|Home | About | Journals | Submit | Contact Us | Français|
To analyze the risk factors, presentation, etiologies, and outcomes of adult cancer patients with intracranial hemorrhage (IH).
We analyzed 208 patients retrospectively with the diagnosis of IH from the Memorial Sloan-Kettering neurology database from January 2000 through December 2007. Charts were examined for clinical and radiographic data. Survival was calculated using the Kaplan-Meier method. Survival between groups was compared via the log-rank test. Logistic regression models were used to assess for prognostic indicators of 30- and 90-day mortality.
There were 181 intracerebral and 46 subarachnoid hemorrhages. Sixty-eight percent of patients had solid tumors, 16% had primary brain tumors, and 16% had hematopoietic tumors. Hemiparesis and headache were the most common symptoms. Intratumoral hemorrhage (61%) and coagulopathy (46%) accounted for the majority of hemorrhages, whereas hypertension (5%) was rare. Median survival was 3 months (95% confidence interval [CI] 2-4), and 30-day mortality was 31%. However, nearly one-half of patients were completely or partially independent at the time of discharge. Patients with primary brain tumors had the longest median survival (5.9 months, 95% CI 2.9-11.8, p = 0.05). Independent predictors of 30-day mortality were not having a primary brain tumor, impaired consciousness, multiple foci of hemorrhage, hydrocephalus, no ventriculostomy, and treatment of increased intracranial pressure.
Intracranial hemorrhage in patients with cancer is often due to unique mechanisms. Prognosis is poor, but comparable to intracranial hemorrhage in the general population. Aggressive care is recommended despite high mortality, because many patients have good functional outcomes.
Cerebrovascular disease is the second most common CNS complication in patients with cancer.1 Intracranial hemorrhage (IH) accounts for nearly one-half of cerebrovascular events and is more likely to be symptomatic than ischemic lesions.1 The causes of IH in the cancer population differ from the general population and vary based on the type of malignancy. Solid tumors typically cause IH from intratumoral hemorrhage (ITH), representing an average of 3.1% of spontaneous IH in large autopsy series, whereas hematologic malignancies may cause IH from coagulopathy or leukostasis.1–3 IH generally occurs late in the course of malignancy, although it may be the first manifestation of cancer.4–10 Most hemorrhages are intraparenchymal and can vary in presentation.1,11–13
IH in patients with cancer is often viewed as a catastrophic and terminal event, although limited data on outcome exist.14,15 Most existing literature is autopsy based or precedes the modern era of neuroimaging. Moreover, advances in oncologic therapies and neurointensive care have significantly changed the demographics, pathophysiology, and prognosis of these patients. The objective of this retrospective study was to analyze the clinical manifestations, causes, and treatments of IH in adult patients with cancer.
We reviewed the records of all cancer patients with a radiologically confirmed diagnosis of intracerebral hemorrhage (ICH) or subarachnoid hemorrhage (SAH) who were evaluated by a neurologist between January 2000 and December 2007; patients were identified from the Department of Neurology database. Subdural and epidural hemorrhages were included only if they coexisted with an ICH or SAH.
We excluded patients who had postoperative hemorrhages. Patients with incomplete records and those aged 17 years or younger were also excluded. Patients with recurrent hemorrhages were included only once at the time of first hemorrhage. Most patients had excellent follow-up, and only 6 were not seen by a Memorial Sloan-Kettering Cancer Center (MSKCC) provider after October 2007. We had functional and vital status for all patients; disposition after ITH was known for all but 4. The number of patients with missing data for each clinical and laboratory variable is shown in table 1.
Clinical features, vascular risk factors, type and stage of cancer, prior and current cancer treatment, presentation, radiographic information, therapy, prognostic factors, and outcome were collected. Cancer diagnoses were confirmed pathologically in all but 2 patients, 1 with a meningioma and 1 with a brainstem glioma. The presence of active disease and systemic or CNS metastases were documented. Prior cranial irradiation or craniotomy and current chemotherapeutic agents were noted. Laboratory values at diagnosis were recorded, including complete blood count, glucose, creatinine, international normalized ratio (INR), partial thromboplastin time (PTT), d-dimer, and fibrinogen. The presence of antiplatelet or anticoagulant agents was documented. All patients had CT, MRI, or both. Hemorrhage location was categorized as supratentorial, infratentorial, lobar, basal ganglia, brainstem, or cerebellum. Hemorrhage size could not be determined because of the diffuse and multifocal nature of many bleeds. Presence and type of herniations and hydrocephalus were also noted. Etiology of hemorrhage was determined by the reviewer and could include multiple diagnoses. ITH consisted of any hemorrhage into an underlying tumor. Coagulopathy was diagnosed if any of the following parameters were fulfilled: platelets <100/mm3, INR >1.5, PTT >45 seconds, disseminated intravascular coagulation (DIC) (fibrinogen <200 mg/dL and d-dimer >290 ng/dL), or full-dose low-molecular-weight heparin within the past 48 hours. Leukostasis was defined as a white blood cell count greater than 100 × 103/mm3.
Treatment of hemorrhage was recorded. Treatment for increased intracranial pressure (ICP) consisted of osmotic diuresis or hyperventilation. The functional status of patients at discharge was determined retrospectively and categorized as completely independent, partially independent, completely dependent, death, and unknown. The disposition and survival of patients were also documented.
Survival was analyzed using the Kaplan-Meier method. Survival time was defined from the date of hemorrhage to the date of death or last follow-up. Survival between groups was compared using the log-rank test. Logistic regression models analyzed 30- and 90-day mortality, and predictors of mortality univariately. Factors univariately significant at the α ≤0.05 level were included in a stepwise logistic regression model with entry and exit levels set at 0.10 to produce the final multivariate analyses. All analyses were performed in SAS 9.2 (SAS Institute Inc., Cary, NC).
This study was approved by the MSKCC Institutional Review Board.
ICH or SAH was diagnosed in 357 patients with cancer at MSKCC between January 2000 and December 2007, but only 208 met our inclusion criteria. There were 116 (56%) men and 92 (44%) women (table 1). The median age was 61 (range 19-91) years, and 77% of patients were Caucasian. One hundred eighty-one patients (87%) had ICH, 46 (22%) had SAH, and 28 (14%) had intraventricular hemorrhage (IVH). Eighteen (9%) had coexistent subdural hemorrhage, and none had epidural hemorrhage. Forty-six patients (22%) had hemorrhage into multiple brain compartments; the most frequent compartmental combinations were ICH plus SAH plus IVH (n = 13) and ICH plus SAH (n = 11). Cerebrovascular risk factors included tobacco use (45%), hypertension (43%), diabetes mellitus (11%), and ethanol abuse (7%). Forty patients (19%) were on therapeutic anticoagulation, but only 6 (3%) were supratherapeutic; 27 (13%) patients were on antiplatelet agents at the time of hemorrhage.
Patients with IH had 42 different cancers; 20 patients had 2 or more cancers. The cancer contributing to hemorrhage was a solid tumor in 141 patients (68%), a primary brain tumor in 34 (16%), and a hematologic tumor in 33 (16%) (table 2). Melanoma (15%), lung (14%), glioma (12%), breast (7%), and leukemia (6%) were the most common primary malignancies. Of the 24 gliomas, there were 14 glioblastomas, 6 oligodendrogliomas, and 4 other astrocytomas. Renal cell (4%), testicular (2%), hepatocellular (1%), and thyroid (1%) cancers were infrequent despite their known propensity to bleed. Most patients had both active disease (93%) and systemic metastases (71%) at the time of hemorrhage. Patients with primary brain tumors were the youngest cohort, with 47% aged 50 years or younger, whereas only 23% of patients with a solid tumor and 7% of those with hematopoietic tumors were in this age range. The median time from cancer diagnosis to hemorrhage was 28 months for solid tumors, 22 months for hematologic tumors, and 6 months for primary brain tumors. Almost half the patients (n = 91, 44%) had a known primary or metastatic brain tumor at the time of hemorrhage. Fifty-four patients (26%) had received prior cranial radiation therapy, and 37 (18%) had received prior craniotomy. Half the patients (n = 104) had received chemotherapy within the past 4 weeks. Five received antiangiogenic therapy with bevacizumab (n = 4) or sunitinib (n = 1); the patient receiving sunitinib had ITH with coagulopathy, but none of the patients receiving bevacizumab had ITH.
The majority of patients were symptomatic (94%) from the hemorrhage (table 3). Patients with primary brain tumors (18%) were more likely to be asymptomatic than patients with solid (4%) or hematopoietic tumors (3%). Hemiparesis (48%), headache (41%), and impaired consciousness (34%) were the most common symptoms or signs; few patients had seizures (17%) or coma (6%). Only 5% of patients were acutely hypertensive with a systolic blood pressure greater than 180 mm Hg or a diastolic blood pressure greater than 120 mm Hg at time of ICH. The mean white blood cell count was 7.8 × 103 cells/mm3, and the mean creatinine was 1.1 mg/dL. Only 7% of patients had a serum glucose greater than 200 mg/dL. Nineteen patients had mild thrombocytopenia (50-99 × 103 platelets/mm3), 15 had moderate thrombocytopenia (20-49 × 103/mm3), and 25 had severe thrombocytopenia (<20 × 103/mm3) (table 1). Twenty-eight patients (14%) had INR values greater than 1.5, and 26 (13%) had PTT values greater than 40 seconds. DIC was documented in 4 patients, 2 of whom had acute promyelocytic leukemia.
Radiographic diagnosis was established in all patients by CT (24%), MRI (12%), or both (64%). Most patients bled into the intracerebral compartment (87%), and nearly half (n = 91, 44%) had multiple foci of hemorrhage. IH was usually supratentorial and equally dispersed between hemispheres of the brain (left 52%, right 48%). Infratentorial hemorrhages occurred in 37 patients (18%) and were most common in the cerebellum (n = 29). Herniation was seen radiographically in 50 patients (24%), 38 subfalcine, 19 uncal, 14 transtentorial, and 6 tonsillar. Hydrocephalus of any form was seen in 37 patients (18%).
ITH (61%) and coagulopathy (46%) were the most common causes of IH in our population (table 3). Less common causes were head trauma (6%), hypertension (5%), hemorrhagic conversion of an ischemic stroke (4%), and venous thrombosis (2%). Only 1 patient with acute myelocytic leukemia had leukostasis-associated hemorrhage (white blood cells 150.1 × 103 cells/mm3), and this patient was also moderately thrombocytopenic. Etiology was multifactorial in 33% of patients and most commonly was due to the combination of ITH and coagulopathy (n = 44, 21%). ITH occurred in isolation in 44% of patients with solid tumors and 47% of those with primary brain tumors. Conversely, patients with hematopoietic tumors were more apt to hemorrhage from coagulopathy in isolation (46%) or coagulopathy in conjunction with other, less frequent diagnoses (33%), such as trauma. SAH was multifactorial in 30% of cases and was most commonly due to coagulopathy (n = 30), trauma (n = 10), or ITH (n = 6); aneurysmal SAH was rare (n = 4).
Steroids were administered to 156 patients (75%). Reversal of coagulopathy and prevention of hematoma expansion were attempted with platelet transfusion (29%), fresh-frozen plasma or cryoprecipitate infusion (21%), or vitamin K administration (19%). Craniotomy for ventriculostomy or resection was performed in 55 patients (26%), whereas emergency therapy for increased ICP was administered to 26 (13%). Forty-six percent of patients were treated with anticonvulsants, though many were already taking seizure medications before their hemorrhage. Forty-eight patients (23%) were ultimately treated with cranial irradiation.
At the time of statistical analysis, only 11% of patients were alive. At discharge, 15% of patients were completely independent, 33% were partially independent, 30% were completely dependent, and 22% were dead. Patients with solid tumors had the best functional outcome at discharge, with 53% being completely or partially independent, compared with 44% of patients with primary brain tumors and 30% of patients with hematologic tumors. Patients with hematopoietic tumors had the highest rate of death during hospitalization (36%), whereas patients with primary brain tumors (8%) had the lowest. Thirty-day mortality was 31% and 1-year mortality was 78% for the entire cohort. Median survival for all patients was 3 months (95% confidence interval 2-4). A difference (p = 0.05) in survival existed among tumor subtypes, with median survival being 5.9 months for patients with primary brain tumors, 2.1 months for patients with solid tumors, and 1.5 months for patients with hematologic tumors (figure). Median survival also varied based on etiology of hemorrhage and was 3.7 months for patients with ITH, 0.3 months for patients with coagulopathy, 1.8 months for patients with both ITH and coagulopathy, and 3.9 months for patients with other diagnoses (p = 0.009). The most common dispositions excluding death were home (50%), hospice (10%), acute rehabilitation (6%), subacute rehabilitation (5%), and transfer to another hospital (3%).
In the multivariate models, impaired consciousness, not having a primary brain tumor, multiple foci of hemorrhage, hydrocephalus, increased ICP treatment, and not receiving ventriculostomy were significant predictors of 30-day mortality (table 4). All of these variables, except not having a primary brain tumor, were also independently predictive of 90-day mortality. Additionally, hemiparesis and current chemotherapy were significant predictors of mortality at 90 but not 30 days via multivariate analysis. Neither anticoagulant nor antiplatelet use was a significant predictor of mortality.
Cancer is an important and common cause of IH. Clinical and autopsy studies have found that brain tumors represent 0.9% to 11% of spontaneous IH.2,7,11,16–22 The largest autopsy study of IH in patients with cancer was performed at MSKCC from 1970 to 1981 and identified 244 patients, 57% of whom were symptomatic.1 Coagulopathy occurred in 57% of those with ICH, and 89% of these patients had hematologic malignancies, particularly leukemia (78%). ITH occurred in 38% of patients and was the second most common cause of ICH; all but 2 of these patients had a solid tumor, with melanoma and germ cell tumors being the most frequent. Hypertension was responsible for only 8% of parenchymal hemorrhages in this cohort.
Our report represents the largest clinical series of IH in patients with cancer. Systemic solid tumors were the most common underlying cancer, which differs from many previous studies where primary brain tumors accounted for the majority of malignancies.13,21,23,24 In one series of 110 patients with bleeding cerebral neoplasms, 77% had primary brain tumors and 23% had metastatic solid tumors.21 Similarly, 62% of patients with ITH in a different cohort of 58 cases had primary brain tumors.13 This changing demographic may represent longer survival of patients with solid tumors or the higher likelihood of primary brain tumors with IH to be clinically silent and thus not lead to acute neuroimaging.
The frequency of hemorrhage into intracranial neoplasms ranges from 1.3% to 14.6% and varies markedly depending on the pathology of the underlying cancer.4,7,11,12,19,21,25–27 Malignant and hypervascular neoplasms have the highest predilection for hemorrhage.12,25,26,28,29 Factors favoring hemorrhage include overexpression of vascular endothelial growth factor and matrix metalloproteinases, imbalances in the fibrinolytic cascade, retiform type capillaries, rapid tumor growth, vascular invasion, tumor necrosis, and neovascularization.30–33 Melanoma, lung, glioma, breast, and leukemia were the most common tumor types in our cohort. The high rates of hemorrhage in patients with breast and prostate cancer were unexpected. Alternatively, we saw low rates of hemorrhage in patients with renal, germ cell, thyroid, and hepatocellular carcinomas previously thought to be highly associated with ITH.
ITH, coagulopathy, or the combination of the 2 accounted for the majority of IH. Hypertension, amyloid angiopathy, and other common causes of IH in the general population were rare, highlighting the unique pathophysiology of IH in patients with cancer. Patients with solid or primary brain tumors most often bled from ITH, whereas patients with hematopoietic tumors most often bled from coagulopathy. Patients with hematologic malignancies were also most likely to bleed from the rarer etiologies in our cohort, such as trauma or hemorrhagic conversion of stroke. Previous studies had found hyperleukocytosis and DIC to be common causes of IH in acute leukemia,1,3,34,35 but both syndromes were rare in our series, reflecting the advances in therapy and changing natural history of acute leukemia.
Anticoagulation of patients with primary or metastatic brain tumors and thromboembolism is often avoided for fear of catastrophic IH despite existing data suggesting that the risk is small.36,37 One retrospective study of 42 patients with cerebral metastases on therapeutic anticoagulation found only 3 symptomatic IHs, 2 of which occurred in the setting of supratherapeutic values.36 Nineteen percent of our patients were on therapeutic anticoagulation at the time of hemorrhage, and only 3% were supratherapeutic. There was no significant association between anticoagulation and mortality, supporting the safety of this treatment in hypercoagulable cancer patients. This holds true for patients with primary or metastatic brain tumors, because nearly half of our patients had known brain lesions before IH. Furthermore, antiangiogenic therapies were not associated with a substantial risk of ITH. This was also observed in a recent study of 21 patients with glioma receiving concurrent bevacizumab and anticoagulation, none of whom developed severe IH.37
Patients with cancer have a poor overall prognosis after IH. Median survival was 3 months, and only 22% of patients lived 1 year. Despite short survival, many patients had excellent functional outcomes, and nearly half were completely or partially independent at the time of discharge. Half of our cohort was also discharged home. Mortality at 30 days (31%) was similar to that of ICH in the general population and only began to worsen compared with historic controls as time from hemorrhage advanced.38,39 This trend suggests that many patients died from their systemic malignancy and not from the hemorrhage itself. Patients with primary brain tumors had the best prognosis, with a median survival of 5.9 months. These patients lived a median of 12.3 months from diagnosis, comparable to the typical life span of patients with malignant brain tumors. Patients with hematopoietic tumors had the worst prognosis, with more than a third dying during hospitalization. This can be attributed to their high rate of coagulopathy, which in isolation was associated with a median survival of only 0.3 months. These patients typically had severe coagulopathy and would develop refractory hemorrhage into multiple cerebral compartments.
Poor prognostic indicators were impaired consciousness, hemiparesis, multiple hemorrhagic foci, hydrocephalus, and treatment for increased ICP (table 4). Recent chemotherapy, not having a primary brain tumor, and lack of ventriculostomy also predicted poor outcome. Recent chemotherapy likely indicates patients with more aggressive disease and higher rates of pancytopenia. Not receiving a ventriculostomy may identify patients treated less aggressively. Steroid administration was common in our IH cohort (75%) and correlated with the rate of ITH (61%). Glucocorticoids were used to suppress tumor-associated edema and promote oncolysis in some hematologic malignancies; its use did not affect outcome (30-day mortality odds ratio 0.8, p = 0.45).
There are several limitations to our study, most importantly its retrospective nature. We did not address isolated subdural or epidural hematomas. Subdural hemorrhage in patients with cancer is often associated with bone and dural metastases and has a different course and outcome from IHs.40 Epidural hematomas were not included as most are postoperative at our institution. Our study also excluded outpatient hemorrhages because we had no comprehensive method of ascertainment. Hemorrhages recognized in outpatients are more likely to be asymptomatic or mildly symptomatic and unlikely to affect morbidity or mortality. Our median survival would probably improve if these hemorrhages were included. Last, MSKCC is not a designated stroke center. Some patients were likely taken to nearby stroke centers, potentially decreasing the number of larger and more disabling hemorrhages.
This study describes the unique pathophysiology and outcome of patients with cancer and IH. ITH and coagulopathy cause most IHs in patients with cancer, whereas hypertension and other causes typical in the general community are rare. Changes in etiology of IH in patients with cancer parallel improvements in management of the underlying neoplasm, particularly in the hematologic malignancies. Prognosis is generally poor, although many patients retain independence after IH and ultimately succumb to their underlying malignancy and not the hemorrhage itself.
Statistical analysis was performed by Anne S. Reiner, MPH, and Katherine S. Panageas, DrPH.
Dr. Navi, Dr. Reichman, Dr. Berlin, Ms. Reiner, and Dr. Panageas report no disclosures. Dr. Segal serves on the speakers' bureau for Boehringer Ingelheim. Dr. DeAngelis has served on a scientific advisory board for Genentech Inc.; serves on the editorial board of Neurology; receives publishing royalties for The Neurologic Complications of Cancer (Oxford University Press, 2008); and has received research support from the NIH (UO1 CA-105663-01 [Participating Member in the NABTC]).
Address correspondence and reprint requests to Dr. Lisa M. DeAngelis, Department of Neurology, Memorial Sloan-Kettering Cancer Center, 1275 York Ave., New York, NY 10065 deangell/at/mskcc.org
Disclosure: Author disclosures are provided at the end of the article.
Received August 6, 2009. Accepted in final form November 10, 2009.