Pathogenetical Subtypes of Recurrent Intracerebral Hemorrhage
Designations by SMASH-U Classification System
Background and Purpose—Pathogenetic classification of intracerebral hemorrhage (ICH), using systems such as SMASH-U (structural vascular lesions, medication, cerebral amyloid angiopathy [CAA], systemic disease, hypertension, or undetermined), is important in predicting functional outcomes and mortality in patients with ICH. This study aimed to compare pathogenetic subtypes between the first and recurrent ICH.
Methods—This study obtained data related to 4578 consecutive acute patients with ICH from the National Taiwan University Hospital Stroke Registry during January 1995 to December 2013. Using the SMASH-U method, patients were classified into 6 subtypes. We then analyzed the outcomes of first-ever ICH cases and pathogenetic classification of recurrent ICH.
Results—Among 3785 patients who experienced first-ever ICH (male, 63.3%; mean age, 58.7±17.0 years), the most common cause was hypertensive angiopathy (54.9%), followed by CAA (12.2%), systemic disease (12.1%), undetermined (10.1%), structural vascular lesions (7.8%), and medication related (2.9%). In 185 cases of recurrent ICH, pathogenetic differences between the 2 ICH events were observed in 34 (18.4%) cases, most of which were CAA to hypertensive angiopathy (n=10) or vice versa (n=7). The rates of ICH recurrence were highest for systemic disease-related and CAA-related ICH at 1, 5, 10, and 15 years after the indexed ICH event.
Conclusions—In approximately one fifth of the recurrent patients with ICH, pathogenetic differences were observed between initial and recurrent events, particularly among those with CAA. It is possible that some patients with ICH with concomitant hypertensive angiopathy and CAA may have been categorized as CAA by the SMASH-U method.
Intracerebral hemorrhage (ICH) accounts for 10% to 20% of all incidents of stroke and is associated with high risks of mortality and morbidity, neither of which has been significantly reduced in the past several decades.1,2 Obtaining an pathogenetic diagnosis of ICH is crucial to the accuracy of predictions related to prognosis and the prevention of recurrent ICH. Numerous scoring systems have been proposed for predicting the outcome of ICH3,4; however, only a few of these are based on pathogenetic classification.5,6 The SMASH-U (structural vascular lesions, medication, cerebral amyloid angiopathy, systemic disease, hypertension, or undetermined) classification system is one such method, in which ICH pathogeneses are classified according to the following: structural lesions, medication related, amyloid angiopathy, systemic disease, hypertension, and undetermined pathogenesis. This approach provides a simple and practical means with which to classify the pathogenesis of ICH.5
Hypertensive angiopathy (HA) and cerebral amyloid angiopathy (CAA) are the 2 most common pathogeneses, accounting for 78% to 88% of all cases of primary ICH.1,7 HA-related ICH tends to be localized in deep areas, resulting from the rupture of degenerated arterioles, a condition induced by uncontrolled hypertension.8 CAA-related ICH is predominantly superficial or lobar in location, resulting from the rupture of small or medium sized arteries in the cortical and leptomeningeal regions.9 It has been proposed that the Boston criteria be used for the clinical diagnosis of CAA based on the tendency of CAA-related ICH to result in multiple hemorrhages, to locate in the lobar, cortical, or subcortical regions, and to occur among older patients.10 However, multiple simultaneous hemorrhages or recurrent ICH can also be caused by HA.11,12 The differences between HA- and CAA-related ICH with regard to recurrence rates and secondary preventive strategies make it necessary to differentiate between these 2 types of ICH.13,14 Using the SMASH-U classification system,5 this study investigated the influence of pathogenetic subtypes on the ICH status and outcomes in patients with first-ever ICH and then compared pathogeneses of recurrent ICH to the initial ICH events in patients with recurrent ICH.
The primary data used in this study was obtained from the National Taiwan University Hospital Stroke Registry, which was initiated in 1995. We prospectively captured all cases of stroke in our hospital by daily screening all of the patients receiving head computerized tomography or with a diagnosis of stroke at emergency department or during hospitalization and also screening the diagnosis at discharge as International Classification of Diseases codes, International Classification of Disease, Ninth Revision 430 to 437, excluding 430 (subarachnoid hemorrhage), 432 (subdural hemorrhage), and 435 (transient ischemic attack). We had detailed records regarding pathogenetic factors, clinical course, prognosis, and stroke complications.15,16 The registry enrolled all patients who had a stroke within 2 weeks of admission or during hospitalization. All registered patients were reviewed by neurologists. The Institutional Review Board of National Taiwan University Hospital approved the stroke registry.
ICH was diagnosed according to clinical features,17 and images were obtained from computerized tomography and MRI. Patients with traumatic ICH, ICH colocalized with tumor, primary subdural/epidural/subarachnoid hemorrhage, and postinfarct hemorrhagic transformation were excluded. Patients who had experienced stroke before the study were also excluded.
The following data were obtained for all of the enrolled patients: (1) preadmission data including demographic data, past medical history, and medication history (particularly antiplatelet, anticoagulant, or thrombolytic agents); and (2) admission data including brain imaging studies (head computerized tomography/MRI/cerebral angiography), the National Institute of Health stroke scale scores, coagulation profile, platelet count, as well as liver and renal functions. In the images, we identified the location of hematoma, as well as the size calculated by (length×width×height)÷2 in the largest area,18 and intraventricular hemorrhage. The ICH pathogenesis of each patient was classified using the SMASH-U method. The classifications were as follows: structural lesions, systemic disease related, medication related, CAA, HA, or undetermined.5 Vascular lesions in the locations of ICH were defined as an ICH subtype of structural lesions.5 Our study added renal failure as a kind of systemic disease (in addition to thrombocytopenia, liver cirrhosis, and nondrug-induced coagulopathy proposed by the SMASH-U classification system)5 because it is a risk factor for spontaneous ICH, defined as chronic kidney disease stage 5 or as that requiring dialysis.19 An ICH incident was deemed medication related if it had time-course related warfarin use with international normalized ratio ≥2, new oral anticoagulant use within 3 days, full-dose heparin, or thrombolytic agents.5 CAA-related ICH was defined as lobar, cortical, or subcortical hemorrhage among individuals aged ≥55 years, according to the Boston criteria.5,10 HA was defined as deep or infratentorial hemorrhage with pre-ICH hypertension.5 For the cases with different classifications between the 2 raters (S.J.Y. and J.S.J.), discussion was made to reach a consensus, and the cases without a consensus for classification were rated as undetermined cause.
We obtained the patients with recurrent ICH who admitted to our hospital for recurrent stroke with head image confirmation, and pathogenetic work-up were performed routinely in all cases of ICH. The survival state of all of the patients with ICH was determined by the concomitant Taiwan Stroke Registry through telephone interviews,20 medical records, and cross-linkage with the National Mortality Registry (via the Collaboration Center of Health Information, Department of Health, Executive Yuan in Taiwan) on the basis of each patient’s individual identification number. Recurrence of ICH or infarct at 1, 5, 10, and 15 years after an index ICH event was recorded according to pathogenetic subtype, which was censored by the date of last follow-up, death or December 31, 2013.
Power and Sample Size Estimation
Calculation of the sample size was based on the assumption of survival difference between CAA- and HA-related ICH. According to one previous analysis, the 1-year mortality rates for CAA- and HA-related ICH were 49.4% and 39.8%, respectively.21 A sample size of 422 CAA- and 844 HA-related patients with ICH can provide a 90% power to detect a 2-sided significant threshold of P=0.05.
Categorical variables are presented as percentages, and the continuous or discrete variables are presented as mean±SD or median (25th–75th percentile). Student t test, χ2 test, ANOVA, and Kruskal–Wallis test were used for univariate analysis between groups with relevant variables as indicated. Interaction models of ICH subtypes with covariates were performed using the Cox proportional hazards analysis to ensure overall associations on outcome. All cases with recurrent ICH were retrieved for pathogenetic categorization using the SMASH-U classification method for the analysis of concordant rates of pathogenetic subtype between the 2 ICH events. Kaplan–Meier curves were used to evaluate the cumulative survival of all first-ever ICH cases respectively and those with recurrent ICH, according to pathogenetic subtype. Inter-rater agreement between 2 independent raters was evaluated in the 268 consecutive patients with ICH identified in 2012 using nonweighted Cohen κ analysis. Statistical analysis was performed using the SPSS software package version 17.0 (SPSS Inc, Chicago, IL).
A total of 4578 patients with acute ICH with 4837 ICH events were included in this study. After excluding 701 patients with previous stoke before the index ICH event and 92 patients with ICH because of tumor bleeding, this study included 3785 patients presenting first-ever ICH (male, 63.3%; mean age, 58.7±17.0 years). The inter-rater agreement for ICH pathogenetic classification was high (κ=0.89; 95% confidence interval [CI]=0.84–0.94). Table 1 presents the baseline characteristics of first-ever patients with ICH according to SMASH-U pathogenetic classification. The most common pathogenesis was HA (n=2078, 54.9%), followed by CAA (n=463, 12.2%), systemic disease (n=458, 12.1%), undetermined (n=382, 10.1%), structural vascular lesions (n=294, 7.8%), and medication (n=110, 2.9%). The case numbers of CAA- and HA-related patients with ICH assumed that there was ≥90% power to detect a significant difference of 1-year mortality rates. One quarter of these first-ever patients with ICH received angiographic studies.
The mean follow-up period was 5.5±5.3 years (range, 0–19 years). One-year follow-up rate was 86.6%, and the complete follow-up (till death or December 31, 2013) rate was 70.1%. There was a difference in age distribution between those with and without complete follow-up, with higher percentage for age <45 years in those without complete follow-up (25.4% vs 15.6%; P<0.001), which might lead to some degree of overestimation of the mortality rate after ICH. Overall mortality rate after ICH was 19.5% at 1 month, 22.1% at 3 months, and 26.3% at 1 year (Table 1). Systemic disease-related and medication-related ICH had highest 1-year mortality (62.2% and 60%, respectively).
As shown in Table 2, long-term survival which was evaluated by Cox’s model with interaction model revealed that the hazard ratio of mortality was highest in the systemic disease subtype or medication-related subtype compared with the reference subgroup of undetermined cause. Furthermore, the association between CAA subtype with the risk of mortality was modified by sex and age. Compared with the reference, the CAA subtype is inversely associated with the risk of mortality among female (adjusted hazard ratio, 0.53; 95% CI, 0.35–0.81) but not among male patients (adjusted hazard ratio, 0.79; 95% CI, 0.57–1.11). Compared with the reference, the CAA subtype is inversely associated with the risk of mortality among patients >65 years (adjusted hazard ratio, 0.64; 95% CI, 0.45–0.92) but not among those <65 years (adjusted hazard ratio, 1.06; 95% CI, 0.61–1.83).
The long-term survival curves were well differentiated among the 6 ICH pathogeneses, including HA and CAA (P<0.001 by log-rank test, Figure 1). Structural lesion-related ICH presented the best survival rates, whereas systemic disease-related and medication-related ICH presented the worst short-term and long-term survival rates (Figure 1; Table 1).
The recurrence rates of stroke, as well as stroke subtypes of infarct, and ICH in the first 1, 5, 10, and 15 years after ICH are presented in Table 3. A total of 185 patients presented recurrent ICH events during the follow-up period (male, 64.1%; mean age 59.2±15.3 years), and the mean interval between the 2 ICH events was 3.3±3.2 years. Significant differences in ICH recurrence were observed among the 6 ICH pathogeneses (P<0.001 by log-rank test, Figure 2). The recurrence rates of ICH in the first 1, 5, 10, and 15 years after index ICH were the highest in the subgroups of systemic disease and CAA (systemic disease, 5.2%, 16.2%, 23.3%, and 31.8%, respectively; CAA, 5.6%, 13.9%, 25.8%, and 35.6%, respectively). It is worthy of note that the recurrence rates for infarct and ICH differ dramatically in patients of medication-related ICH. None in the patients with medication-related ICH had recurrent ICH, whereas this group had highest rate of infarct after the index ICH during short-term and long-term follow-up, even reaching 40.2% in the 15-year follow-up.
The distribution of pathogeneses among patients with recurrent ICH is presented in Table I in the online-only Data Supplement. Among the 185 patients with recurrent ICH, pathogenetic differences between the 2 events were observed in 34 (18.4%) of the cases. Among the first ICH events, 44 were classified as CAA and 93 as HA. Among the 44 patients with CAA, 31 had recurrences with CAA subtype, whereas 10 (23%) were classified as HA in the recurrence event. Among the 93 patients with HA, 78 had recurrent HA subtype, but 7 (8%) were classified as CAA in the recurrent event. The subtype classification of the 34 cases with different ICH pathogeneses between the 2 ICH events are shown in Table II in the online-only Data Supplement. Among the 10 cases diagnosed as CAA in the first and HA in the recurrent event (ages 66.8±9.3 years [range, 55–84 years] at the first ICH event), the locations of the hematomas were all lobar in the first events and in the basal ganglia, pons, or cerebellum in the second events, and all patients had concomitant hypertension. In the 7 cases that changed from HA in the first ICH event to CAA in the recurrent event (ages 70.6±10.2 years [range, 57–84 years] at the first ICH event), the hematomas were all in the basal ganglia in the first events and lobar in the second events, and all patients concomitantly had hypertension.
Our results using the SMASH-U classification method revealed pathogenetic differences between the 2 events in approximately one fifth of the cases, most of which were CAA in the first event and HA in the second. In addition to the possibility that patients concomitantly had CAA and HA,22 our findings indicate the possibility that clinical differentiation between CAA and HA is not straightforward in some cases of ICH. The classification accuracy of HA and CAA based on hemorrhage patterns in brain images has been reported to be only 66.8%.23 The definition of CAA-related ICH in the SMASH-U classification system is based on lobar, cortical, or subcortical hemorrhage in patients >55 years.10 Nevertheless, pathology results prove that CAA is not rare among patients with ICH 50 to 60 years old (14%) and could even strike patients <45 years.24,25 Although a strong association exists between CAA and lobar ICH,26 CAA-related ICH may also involve basal ganglia in 44% of surgically treated ICH cases,27 and pathology reports have indicated that HA-related ICH can also occur in the subcortical region (3%–9%).28 Other clinical or radiological features of CAA could help in the differential diagnosis. First, CAA-related ICH is characterized by a high recurrence rate.7 Second, microbleeds or cortical superficial siderosis observed in brain MRI has recently been regarded as a marker for CAA-associated ICH.29 Considering the ambiguity in differentiating between HA- and CAA-related ICH based on clinical information, the inclusion of microbleeds, cortical superficial siderosis, and recurrence course into the criteria possibly could improve the diagnostic accuracy of CAA-related ICH.
Our analysis revealed that the highest recurrence rates of ICH were associated with CAA subtype (25.2%) and systemic disease subtypes (25.0%), whereas the lowest recurrence rates were associated with medication-related (0%) and structural lesion-related ICH (3.7%). CAA-related ICH is characterized by a high annual recurrence rate reaching 10.5%.7 In addition, patients with CAA have demonstrated increased susceptibility to ICH caused by drugs including anticoagulants and antithrombotic or thrombolytic agents.9 This condition would be categorized as medication-related ICH by the SMASH-U classification system because of the sequence of the diagnostic flow chart,5 which could lead to an underestimation in the percentage of CAA. No proven strategy exists to prevent the recurrence of CAA-related ICH; nonetheless, the avoidance of anticoagulants or thrombolytic agents may be beneficial.30 As for HA-related ICH, the annual recurrence rate is 2% and that can be significantly reduced by controlling blood pressure.31 However, the high recurrence rate of systemic disease-related ICH has never been mentioned before, and prevention of the recurrence is difficult because these diseases are usually untreatable. About medication-related ICH, no recurrence of ICH but instead a high rate of infarct (39.6%) after the ICH was probably because of avoidance of anticoagulants after the ICH event. Furthermore, most structural lesions resulting in ICH are treatable, such that the recurrence of ICH can be effectively prevented.
Our results revealed that systemic disease-related and medication-related ICH presented the worst survival. In this study, patients with the 2 pathogeneses were more likely to experience ICH ≥30 mL (59.6% and 66.4%, respectively), intraventricular hemorrhage (50.9% and 65.5%, respectively), multiple ICH (15.3% and 14.5%, respectively), higher median National Institute of Health stroke scale scores (20 and 24, respectively), and lower median Glasgow Coma Scale (9 and 8, respectively), which have been proven as poor prognosticators for ICH.3,7,32 In addition to the contribution of ICH status to long-term outcomes, comorbidities in the systemic disease subtype, as well as heart conditions or high rate of infarction after medication-related subtype, all contribute to the poor prognosis. As reported by Meretoja et al,5 the survival curves obtained from the SMASH-U classification system revealed similar survival rates among patients with HA- and CAA-related ICH. In contrast, the survival curves in this study were well differentiated between the 2 pathogeneses, with HA-related ICH presenting survival rates better than that of CAA-related ICH. The mean age of onset among patients of HA-related ICH in this study was younger than that reported by Meretoja et al5 (60 years versus 66 years), which could explain the relatively better survival rate.3 However, patients with CAA-related ICH in this study had far larger hematoma than those reported by Meretoja et al5 (28 mL versus 14 mL), which could contribute to the worse survival.3,7
Some noteworthy limitations to this study should be mentioned. First, this study was conducted in a single tertiary medical center with Chinese population that has been shown to have a higher risk of ICH than white one.33 Thus, the generalizability of the findings may be limited. Second, our methods yielded a conservative estimation for the recurrence rate of ICH, and it should be emphasized that the possibility of underestimation of the ICH recurrence rate cannot be excluded. Third, pathological proof of pathogeneses was not available for all patients with ICH, particularly those with CAA-related ICH. This made it difficult to avoid misclassification. Finally, this study made a slight modification to the SMASH-U classification system in the addition of renal failure to the category of systemic disease, which may have influenced the proportion classified in this subtype.
This study has several strengths, including a large number of patients with stroke, an extended follow-up duration, consistency in the study methods across the 19-year period, and highly detailed, systematic registered information.
In conclusion, approximately one fifth of the patients with recurrent ICH were subject to ICH of a different subtype in their second episode, particularly those with CAA. It is possible that some patients with ICH with concomitant HA and CAA could be categorized as CAA in the SMASH-U classification system.
Sources of Funding
This work was supported, in part, by grants from the National Science Council (NSC96-2314-B-002-191-MY3) and National Taiwan University Hospital (A116), Taiwan.
Presented in part at the International Stroke Conference of the American Heart Association, San Diego, CA, February 12–14, 2014.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.005598/-/DC1.
- Received March 25, 2014.
- Revision received July 1, 2014.
- Accepted July 2, 2014.
- © 2014 American Heart Association, Inc.
- Wang W,
- Lu J,
- Wang C,
- Wang Y,
- Li H,
- Zhao X,
- et al
- Meretoja A,
- Strbian D,
- Putaala J,
- Curtze S,
- Haapaniemi E,
- Mustanoja S,
- et al
- Hatano S
- Kothari RU,
- Brott T,
- Broderick JP,
- Barsan WG,
- Sauerbeck LR,
- Zuccarello M,
- et al
- Hsieh FI,
- Lien LM,
- Chen ST,
- Bai CH,
- Sun MC,
- Tseng HP,
- et al
- Poon MT,
- Fonville AF,
- Al-Shahi Salman R
- Samarasekera N,
- Smith C,
- Al-Shahi Salman R
- Mizukami M,
- Kin H,
- Araki G,
- Mihara H,
- Yoshida Y
- Arakawa S,
- Saku Y,
- Ibayashi S,
- Nagao T,
- Fujishima M