Recurrent intracerebral hemorrhage

Primary intracranial cerebral hemorrhage (ICH) is the cause of 5 to 16% of all strokes in Europe and United States^1-11 and between 20 to 35% of strokes in Asia.^12-15 ICH is generally a one-time event^16-19 with exceptional recurrences. Recently, three Asian studies^20-22 concluded that recurrences of hypertensive ICH were 1.8 to 5.3%. Two European studies have shown that 6 to 24% of patients with ICH have a recurrence over 6 years of follow-up.^23,24 Many clinicopathologic publications over the last decade have related the tendency to recurrence of ICH to cerebral amyloid angiopathy.^25-31 We systematically reviewed 24 cases of recurrent ICH encountered in a period of 10 years between 1984 and 1994. We compared this population with 81 patients admitted during the same period with ICH without second hemorrhage, followed over a period of more than 24 months and chosen from the 351 patients with a single ICH, to assess the predictors of recurrence of ICH in survivors of primary intracranial hemorrhage.

Patients and Methods. Between January 1984 and December 1994, 375 patients with ICH were admitted to the Department of Neurology at Jean Bernard Hospital in Poitiers, a University Hospital in France. All patients underwent CT and, in selected cases, MRI and cerebral angiography. Twenty-four (6.4%) of these 375 patients had recurrent ICHs that were also proved by CT or MRI.

Known causes of recurrent intracerebral bleeding were excluded: aneurysm, arteriovenous malformation, intracranial tumors, head trauma, primary or secondary disorders of coagulation, Rendu-Osler-Weber disease, moyamoya disease, usage of antithrombotic agents (anti-coagulants, fibrinolytics) or drugs (amphetamines, amphetamine-like substances, and cocaine), and hemorrhagic brain infarct. Simultaneous ICH was also excluded as patients with recurrence of the intracerebral bleeding at the same location in order to exclude angiographically occult vascular malformations.

For each subject, we determined the occurrence of known risk factors of ICH^32,33: arterial hypertension (blood pressure > 160 mm Hg systolic or > 95 mm Hg diastolic on at least two occasions, or previous arterial hypertension with treatment), hypercholesterolemia (previous treatment or special diet, or serum cholesterol over 250 mg/dl on at least two occasions), cigarette smoking(never, previous smoking, or current smoking), heavy drinking (chronic alcoholism was defined by intake of ≥500 g alcohol per week for men and≥250 g per week for women), impaired liver function (increase of GOT, GPT, or alkaline phosphatase at admission).

The location of hematomas was determined as lobar (frontal, parietal, temporal, occipital), basal ganglionic (putamen, thalamus, caudate nucleus), brainstem, and cerebellar.²² Diffuse white matter hyperintensities on T₂-weighted MRI were evaluated in five grades, using the criteria purposed by Van Swieten et al.³⁴ Grade 0 represented none or a single lesion, grade 1 and 2 multiple focal lesions, and grade 3 and 4 multiple confluent lesions in one or both of the anterior and posterior regions of the brain.

The intervals between the bleeding, and the functional and cognitive prognosis, were analyzed.

To compare the data of these 24 patients with recurrent bleeding, 160 patients admitted to our department during the same period and picked from the 351 patients with a single episode of spontaneous ICH, were chosen at random. Controls who died within 2 years (69 patients) or after 2 years of follow-up but of unknown cause (10 patients) and controls with known causes of recurrent intracranial hemorrhage, as previously defined, were excluded. The 81 controls without second bleeding episodes were thus followed at our department for at least 2 years.

The age and gender, the location (lobar or ganglionic) of hematoma, and the incidence of the possible known risk factors or confounders were compared in the two groups by using Fisher's exact test, Mann-Whitney's test, or stratified Mantel-Haenszel's test. To obtain individual odds ratios, the continuous variables were dichotomized. Arterial hypertension, diabetes mellitus, and heavy drinking were dichotomized as yes versus no and age from selected cutoff point (≤65 years and >65 years).

Results. Clinical features. Twenty-four patients presented a recurrent ICH (table 1). There were 15 women and nine men with a mean age of 64.7 ± 9.4 years (range 49-81) for the first ICH and 68.7 ± 7.5 years (range 57-83) for the second ICH. The mean interval between the two bleeding episodes was 47.5 ± 30.5 months and the median time 46 months (range 3 months to 14.8 years).

The men (59.2 ± 7.9 years) were significantly (p = 0.02) younger than the women (67.9 ± 8.8 years).

Fifteen patients presented with two episodes of ICH, six with three episodes of ICH, and three patients with four episodes of ICH.

The major neurologic deficit associated with the first ICH was hemiplegia in 14 patients, associated with aphasia in two patients, isolated aphasia or associated with headache or seizures or homonymous hemianopia in seven patients, headaches in two patients, and isolated homonymous hemianopia in one patient.

The second bleeding was associated with hemiparesis in 13 patients associated with aphasia in four patients, isolated aphasia in five patients, seizures in three patients associated with homonymous hemianopia in one, cortical blindness in one patient, akinetic mutism in one patient, and no clinical sign for patient 18 (systematic control of CT after the first ICH).

After the first ICH, treatment was conservative in all patients, except two for whom surgical evacuation was performed (cases 8 and 24). After the recurrent bleeding, treatment was always nonsurgical.

After the first bleeding, recovery was good, and all patients could walk without aids. Nine patients suffered from an aphasia of variable severity. After the second bleeding, 18 patients could walk (75%) without aids, while only five (21%) presented no neuropsychological disturbances. One patient died directly as a result of the recurrent bleeding. Fifteen patients (62.5%) developed seizures with a mean time of 37.4 ± 38.6 months (range 0-126 months).

Biological features and risk factors. Seventeen patients had hypertension (nine women, eight men). Hypertension had been diagnosed previously in 14 patients and was disclosed during the first hospitalization in three patients. Seven patients (29%) did not present with hypertension and did not receive any regular antihypertensive therapy during the follow-up even after the first stroke. There was no statistical difference of the mean age between the hypertensive patients and the nonhypertensive patients. Furthermore, the mean age was not statistically different between nonhypertensive and hypertensive women, and nonhypertensive and hypertensive men. The median interval between the two bleeding episodes was not statistically different between hypertensive patients and nonhypertensive patients. Fifty-seven percent of patients with first bleeding in lobar areas were hypertensive, whereas 100% of the patients with deep hemorrhages were hypertensive.

Only four patients (44%) of the nine patients who presented with more than two ICH were hypertensive. When the pattern of hemorrhage was "lobar-lobar," the prevalence of hypertension was 64.3%, whereas it was 88.9% when the pattern included a ganglionic hemorrhage.

One patient presented with impairment of liver functions and two with an abnormal thrombocyte count (100.000/mm³ < thrombocyte count<150.000/mm³), but all had normal coagulation studies. Five patients (20.8%) had diabetes mellitus, five had a hypercholesterolemia, and five were heavy drinkers and five cigarette smokers.

Radiologic data. All patients underwent CT. Initial angiograms, performed in 15 patients, demonstrated only an avascular mass without any vascular malformations. Eleven patients underwent MRI in order to appreciate lacunar infarcts, periventricular and deep white matter hyperintensities,³⁴ and to exclude secondary ICH. The 24 CT studies confirmed the presence of the hematoma; CT findings are summarized in table 1.

Fifteen first hemorrhages were in the left hemisphere and nine in the right. The site of the first hemorrhage was the putamen and/or thalamus in six patients and lobar areas in 17 patients. No first bleeding was found in the cerebellum or in the brainstem. A subacute subdural hemorrhage was the first cerebral bleeding in a patient who presented with two further frontal ICH. Lobar hemorrhages were located in the parietal lobe in four patients, in the frontal lobe in three patients, in the temporal lobe in two patients, in the occipital lobe in one patient, and in the temporo-occipital or parieto-occipital area in seven patients.

The location of first ICH was lobar in 92.8% of women and deep in 55.6% of men.

In all cases, the recurrent hemorrhage was located at a different site from the previous ICH. The site of the second attack was the putamen and/or thalamus in eight patients and the lobar areas in 16 patients (six parietal lobe, three frontal lobe, two temporal lobe, one occipital lobe, and four parieto-occipital area).

The patterns of the recurrent bleeding were "ganglionic-ganglionic" (five patients) (figures 1 and 2), "ganglioniclobar"(one patient), "lobar-ganglionic" (three patients), and "lobar-lobar" (14 patients) (figures 3 and 4). The recurrent ICH was contralateral to the initial side of the hemorrhage in 14 of the 24 patients, but was always observed in "ganglionic-ganglionic" pattern. The site of the third or the fourth hemorrhage was almost always lobar or cerebellar.

MRI, performed in 11 patients after the second bleeding episode, disclosed no vascular malformations (cavernous angioma) or cerebral tumors. One or several lacunar infarcts were observed in six patients and a small cortical middle cerebral artery territory infarct in one of them, all but one with known hypertension. Five patients had no history of stroke except for the first bleeding, but one patient presented 8 years before the first bleeding episode with a lacunar infarct with normal CT. One patient exhibited a corticosubcortical petechial hemorrhage associated with lobar hemorrhage. White matter hyperintensities were severe in five patients (grade 3-4) and mild to moderate (grade 1-2) in six patients without any correlation with hypertension.

Comparison with control group. The characteristics of these 24 patients with recurrent ICH were compared with 81 patients admitted with ICH during the same period and without second hemorrhage (table 2). These patients (44 men, 37 women) were followed for at least 2 years(47.9 ± 22.2 months) at our hospital. Thirty-eight hemorrhages were in the left hemisphere and 43 in the right. The site of the hemorrhage was the putamen and/or thalamus in 41 patients, lobar areas in 37 patients, and brainstem in three patients. Lobar hemorrhages were located in the frontal lobe in four patients, in the temporal lobe in nine patients, in the occipital lobe in eight patients, and in the parietal lobe or parieto-occipital area in 16 patients. As in the group with recurrent ICH, the mean age of women (72.5± 9.6 years) was significantly higher (p < 0.01) than the mean age of men (66.7 ± 11.7 years).

The gender ratio was not statistically different between patients with recurrent ICH and those with a unique bleeding episode (p = 0.17). The mean age of this control group (69.3 ± 11.1 years; range 32-87) was statistically different from the mean age (64.7 ± 9.4 years, range 49-81) at the first hemorrhage of the group with recurrent ICH(p < 0.02).

The comparison of the prevalence of possible risk factors and location of bleeding between patients with recurrent ICH and those with a unique bleeding episode is shown in table 2. There was no statistical difference in the prevalence of hypertension, diabetes mellitus, hypercholesterolemia, and smoking and drinking habits between the two groups. However, the prevalence of lobar hematomas was statistically higher(p < 0.02) in the group of patients with recurrent bleeding, especially in the absence of hypertension. Lobar hematoma was a significant risk factor for recurrence of bleeding (OR: 3.57, 95% CI 1.2-12.0, p < 0.02), even after adjustment for age (p < 0.02) or sex (p < 0.05). A logistic regression analysis by backward elimination procedure with recurrence of bleeding as dependent variable and lobar hemorrhage, sex, and age (continuous variable) as independent variables confirmed that only lobar hemorrhage remained significant (OR: 3.6, 95% CI 2.1-6.2, p < 0.02).

Discussion. We found two main patterns of recurrence, which differ by frequency and stroke mechanisms: "ganglionic-ganglionic" and"lobar-lobar" patterns. The "ganglionic-ganglionic" pattern, encountered in five patients, is less frequent in European series²⁴ than in other series.^21,22 In Asian studies, this pattern represents more than half the patients^21,22 and occurs frequently on the contralateral side. It always occurs in hypertensive patients. Rupture of microaneurysms due to chronic hypertension in the deep small penetrating vessels explains the location of hemorrhage in the thalamus, putamen,³⁵ or cerebellum.³⁶ The destruction of the microaneurysms of lenticulostriate vessels after the first ICH could explain why the second ICH occurred in a different location.²² In our series, the most common pattern of recurrence was "lobar-lobar," as in other European reports.^23,24,37 However, in two of these reports, clinical and radiologic data were lacking.^23,37 In the Oxfordshire Community Stroke Project, four patients(two men and two women of mean age 70 years) had recurrent ICH (6%), all with a "lobar-lobar" pattern.²³ One patient exhibited a fatal recurrence, two patients had two recurrences (one fatal), and one three recurrences. Hypertension was the suspected etiology in one patient and congophilic angiopathy in two.²³ In the second European study, only 16% of the patients with recurrent bleeding were hypertensive.³⁷ A third European study reported a"lobar-lobar" pattern in 44% and a "ganglionic-ganglionic" pattern in 26% of patients with recurrences.²⁴

The lack of hypertension and the lobar localization of the hemorrhage are strongly suggestive of cerebral amyloid angiopathy (CAA).^38,39 CAA is a frequent cause of spontaneous, recurrent, and sometimes multiple ICH in elderly nonhypertensive patients and represented the cause in 2%⁴⁰ to 9.3%⁴¹ of cases of ICH. CAA increases with advancing age: 5% in the seventh decade, 43% in the eighth decade, and 57% in persons over age 90.^42-44 In our population, 17 patients were older than 60, seven patients were older than 70, and three were older than 80. Although CAA is a disease of the elderly, a large number of patients are in the sixth and in the seventh decade.^25,38,45,46 CAA most commonly affects the walls of cortical, subcortical white matter, and leptomeningeal arteries, arterioles, and capillaries with posterior predominance (occipital and parietal gray matter). CAA characteristically spares the basal ganglia, corpus callosum, thalamus, cerebellum, and brainstem,⁴³ and is associated with clinical dementia and Alzheimer's disease changes at autopsy in more than 40% of cases. Despite these anatomic changes, lobar hematomas are most common in the frontal (35%) and parietal (26%) regions, less common in the temporal (14%) and occipital(19%) regions, and unusual in the deep central gray matter (4%) and cerebellum (2%).³⁸ The location of ICH in our nonhypertensive patients was most commonly the frontal and parietal lobes. One nonhypertensive patient had a subdural hematoma^38,47,48 as first the bleeding episode, followed by two frontal ICH.

More than 50% of our patients with recurrent lobar ICH were hypertensive. The contribution of hypertension to lobar hemorrhage is uncertain. Several studies demonstrate that the cause of first lobar ICH is hypertension in 20 to 48% of the patients,^49-52 whereas hypertension is the putative mechanism in 70% of putaminal, 57% of thalamic, 90% of pontine, and 50 to 75% of cerebellar hemorrhages.³⁹ Broderick et al.⁵³ however, showed that the frequency and the contribution of hypertension to ICH was similar in patients with lobar hemorrhages compared with deep hemispheric, cerebellar, and pontine hemorrhages, and this remained unchanged with advancing age. In the absence of brain biopsy or anatomic study, it remains difficult to estimate the respective contributions of hypertension and CAA to recurrent lobar ICH. However, in documented CAA-related ICH, more than 30% of patients had hypertension⁵⁴ with mixed microangiopathy (with both hypertensive and CAA changes). Hypertension may increase the tendency to CAA-related hemorrhage, or vice versa.³⁸ Hypertension does not appear to be an additional risk factor in the causation of ICH associated with CAA in the setting of Alzheimer's disease.⁵⁵

Although the diagnosis of CAA, according to previously defined criteria,⁵⁶ was possible in several cases of our series, there were no brain biopsies or autopsies performed to confirm this diagnosis. However, CAA might be diagnosed during life with gradient-echo MRI sequences,⁵⁷ and determination of apoE genotype,⁵⁸ in the absence of brain biopsy. Lobar hemorrhages could radiologically be accompanied by multiple cortical or corticosubcortical petechial hemorrhages,⁵⁷ as in one of our patients. Patients with multiple lobar hemorrhages demonstrated a greater than twofold overrepresentation in frequency of the apoE ϵ4 allele compared with a population-based sample and patients with hemorrhages in deep territories. Patients with this genotype had earlier onset of hemorrhage by approximately 5 years compared with the noncarriers of the genotype; this was similar to the age difference in our study. These data support a specific role for apoE ϵ4 in accelerating the process that leads from deposition of amyloid β-peptide to intracerebral bleeding.⁵⁸

Although the prognosis of recurrent bleeding is much worse than that of the first ICH, since the recurrence occurs in a different location, the functional prognosis remained good, with more than two-thirds of the patients being able to walk. However, the neuropsychological prognosis was usually poor, with progressive dementia or a frontal syndrome.

Our study clearly indicated that only lobar cerebral hemorrhage was a strong risk factor for recurrence,²⁴ hypertension, smoking, diabetes mellitus, and gender did not convey increased risk for recurrence. Patients with recurrence were significantly younger than patients without recurrence; a correlation at variance with a previous study.²⁴

Despite lobar cerebral hemorrhage in elderly non-hypertensive patients predisposing to recurrence, the prediction of a recurrence after a first hemorrhage remains difficult. Lobar cerebral hemorrhages are frequent, accounting for 16%⁸ to 46%⁵³ of non-traumatic ICH in large series,^18,19,59 and even surpassed the putaminal and the thalamic locations in frequency.^19,60-62

In conclusion, recurrent ICH are not rare and their mechanisms seem different in our study from those reported in Asian studies. Although the exact mechanisms of recurrent bleeding are often undetermined, optimal antihypertensive therapy remains important to prevent recurrences, since poor control of hypertension increases the risk of recurrence.^24,36

Acknowledgments

The authors are grateful to Dr. Julien Bogousslavsky (Lausanne, Switzerland) for his critical comments and to Ronald Prendergast for English revision.