Retrospective study of long-term outcome after brain arteriovenous malformation rupture: the RAP score

View More View Less
  • 1 Departments of Interventional Neuroradiology,
  • 2 Neurosurgical Anesthesiology and Critical Care, and
  • 4 Neurosurgery, Pitié-Salpêtrière Hospital; and
  • 3 Paris VI University, Pierre et Marie Curie, Paris, France
Full access

OBJECTIVE

The authors aimed to design a score for stratifying patients with brain arteriovenous malformation (BAVM) rupture, based on the likelihood of a poor long-term neurological outcome.

METHODS

The records of consecutive patients with BAVM hemorrhagic events who had been admitted over a period of 11 years were retrospectively reviewed. Independent predictors of a poor long-term outcome (modified Rankin Scale score ≥ 3) beyond 1 year after admission were identified. A risk stratification scale was developed and compared with the intracranial hemorrhage (ICH) score to predict poor outcome and inpatient mortality.

RESULTS

One hundred thirty-five patients with 139 independent hemorrhagic events related to BAVM rupture were included in this analysis. Multivariate logistic regression followed by stepwise analysis showed that consciousness level according to the Glasgow Coma Scale (OR 6.5, 95% CI 3.1–13.7, p < 10−3), hematoma volume (OR 1.8, 95% CI 1.2–2.8, p = 0.005), and intraventricular hemorrhage (OR 7.5, 95% CI 2.66–21, p < 10−3) were independently associated with a poor outcome. A 12-point scale for ruptured BAVM prognostication was constructed combining these 3 factors. The score obtained using this new scale, the ruptured AVM prognostic (RAP) score, was a stronger predictor of a poor long-term outcome (area under the receiver operating characteristic curve [AUC] 0.87, 95% CI 0.8–0.92, p = 0.009) and inpatient mortality (AUC 0.91, 95% CI 0.85–0.95, p = 0.006) than the ICH score. For a RAP score ≥ 6, sensitivity and specificity for predicting poor outcome were 76.8% (95% CI 63.6–87) and 90.8% (95% CI 81.9–96.2), respectively.

CONCLUSIONS

The authors propose a new admission score, the RAP score, dedicated to stratifying the risk of poor long-term outcome after BAVM rupture. This easy-to-use scoring system may help to improve communication between health care providers and consistency in clinical research. Only external prospective cohorts and population-based studies will ensure full validation of the RAP scores' capacity to predict outcome after BAVM rupture.

ABBREVIATIONS AUC = area under the ROC curve; BAVM = brain arteriovenous malformation; BP = blood pressure; EVD = external ventricular drain; GCS = Glasgow Coma Scale; ICH = intracranial hemorrhage; IVH = intraventricular hemorrhage; mRS = modified Rankin Scale; RAP = ruptured AVM prognostic; ROC = receiver operating characteristic.

OBJECTIVE

The authors aimed to design a score for stratifying patients with brain arteriovenous malformation (BAVM) rupture, based on the likelihood of a poor long-term neurological outcome.

METHODS

The records of consecutive patients with BAVM hemorrhagic events who had been admitted over a period of 11 years were retrospectively reviewed. Independent predictors of a poor long-term outcome (modified Rankin Scale score ≥ 3) beyond 1 year after admission were identified. A risk stratification scale was developed and compared with the intracranial hemorrhage (ICH) score to predict poor outcome and inpatient mortality.

RESULTS

One hundred thirty-five patients with 139 independent hemorrhagic events related to BAVM rupture were included in this analysis. Multivariate logistic regression followed by stepwise analysis showed that consciousness level according to the Glasgow Coma Scale (OR 6.5, 95% CI 3.1–13.7, p < 10−3), hematoma volume (OR 1.8, 95% CI 1.2–2.8, p = 0.005), and intraventricular hemorrhage (OR 7.5, 95% CI 2.66–21, p < 10−3) were independently associated with a poor outcome. A 12-point scale for ruptured BAVM prognostication was constructed combining these 3 factors. The score obtained using this new scale, the ruptured AVM prognostic (RAP) score, was a stronger predictor of a poor long-term outcome (area under the receiver operating characteristic curve [AUC] 0.87, 95% CI 0.8–0.92, p = 0.009) and inpatient mortality (AUC 0.91, 95% CI 0.85–0.95, p = 0.006) than the ICH score. For a RAP score ≥ 6, sensitivity and specificity for predicting poor outcome were 76.8% (95% CI 63.6–87) and 90.8% (95% CI 81.9–96.2), respectively.

CONCLUSIONS

The authors propose a new admission score, the RAP score, dedicated to stratifying the risk of poor long-term outcome after BAVM rupture. This easy-to-use scoring system may help to improve communication between health care providers and consistency in clinical research. Only external prospective cohorts and population-based studies will ensure full validation of the RAP scores' capacity to predict outcome after BAVM rupture.

ABBREVIATIONS AUC = area under the ROC curve; BAVM = brain arteriovenous malformation; BP = blood pressure; EVD = external ventricular drain; GCS = Glasgow Coma Scale; ICH = intracranial hemorrhage; IVH = intraventricular hemorrhage; mRS = modified Rankin Scale; RAP = ruptured AVM prognostic; ROC = receiver operating characteristic.

Rupture is the most significant source of morbidity and mortality in the natural history of brain arteriovenous malformations (BAVMs) and is a leading cause of social and health care costs.11,18 Much effort has been dedicated to identifying risk factors predictive of BAVM rupture.9 Less is known about predictive factors for clinical outcome after BAVM rupture. The intracranial hemorrhage (ICH) score is a reliable predictor of outcome after ICH.13 However, patients with ICH secondary to BAVM rupture tend to be younger, have lower pre-stroke and admission blood pressure (BP), have a higher admission Glasgow Coma Scale (GCS) score, and are more likely to have an ICH in a lobar location than patients with spontaneous ICH.23 Moreover, ICH is absent in more than 20% of BAVM ruptures, and patients with BAVM-related hemorrhage appear to have a better long-term outcome than those with spontaneous hemorrhage.3,5,23 Therefore, identifying specific predictive factors for long-term neurological outcome is required in BAVM-related hemorrhage.

We hypothesized that a composite score combining clinical, radiological, and biological characteristics at admission would be better for predicting long-term neurological outcome in patients with BAVM rupture than individual factors. Therefore, in this study we aimed to identify individual predictive factors of a poor outcome after BAVM rupture and to develop an admission score by combining these factors in a single-center cohort.

Methods

Ethics Statement

The ethics committee of our institution approved this study. The need for patients' informed consent was waived since we performed retrospective analyses of records and imaging data. Patients gave oral consent for the outcome interviews conducted by telephone.

Patients

We retrospectively reviewed the medical records of consecutive patients with BAVM rupture–related hemorrhage who had been admitted to our institution in the period from January 1, 2003, to January 31, 2014. Our institution is a tertiary care center that receives patients with acute and severe neuroradiological, neurosurgical, and neurological conditions. A list of patients was generated by searching the retrospective and prospective interventional neuroradiology department BAVM database. We also searched the discharge databases of the neurosurgery, neurosurgical intensive care, and neurology departments. All BAVM hemorrhagic events were eligible for inclusion, even cases of rupture following endovascular embolization. All of the patients had been admitted within 24 hours after hemorrhage. Each BAVM had been confirmed by digital subtraction angiography (DSA), CT angiography, MRI, or pathological analysis.

Management of Patients

All of the patients who were admitted to the intensive care unit or neurosurgery ward were managed using the same algorithm. At admission, patients with evidence of intraventricular hemorrhage (IVH) with hydrocephalus according to CT or a transcranial Doppler pulsatility index greater than 1.4 received an external ventricular drain (EVD). For patients without these criteria but with a degradation of consciousness or a tetraventricular dilation with associated brain atrophy, a second CT study was performed within 24 hours to diagnose a possible increase in ventricular size, leading to EVD placement. Any EVD infection was controlled with a protocol that included a single dose of prophylactic antibiotics in the operating room, minimal handling of the catheter, and no rinsing. Surgical evacuation of intracerebral hemorrhage was indicated when the hematoma was responsible for severe mass effect leading to a major midline shift, brain or cerebellar herniation, and/or uncontrollable intracranial hypertension. Emergent embolization was performed only in cases with an estimated high risk of early rebleeding, including those with the presence of intranidal/pedicular aneurysms or subependymal topography of BAVMs with IVH. Open surgery and embolization procedures were performed with the patient under general anesthesia. Severe intracranial hypertension (intracranial pressure > 20 mm Hg under sedation) was treated with CSF drainage, deepening of sedation, and, rarely, moderate hypothermia.

Admission Characteristics

The following characteristics were recorded by 1 of 2 authors (E.S. or M.D.): baseline (prehemorrhage) modified Rankin Scale (mRS) score, age, sex, history of high BP, antihypertensive therapy, anticoagulation or antiplatelet therapy, previous BAVM hemorrhage, seizure, headache, incidental discovery of the BAVM, past BAVM treatment (surgery, stereotactic radiosurgery, and/or embolization), and clinical characteristics at presentation (GCS score before sedation or treatment of hydrocephalus, heart rate, BP, oxygen saturation, and body temperature). Glycemia, creatinine, and troponin blood-test results at admission were recorded. Radiological data obtained from an admission CT scan or MR angiogram and DSA study were analyzed by a senior (F.C.) and a junior (E.S.) neuroradiologist blinded to the patient's clinical data. These radiological data were as follows: presence of an ICH, supra- or infratentorial location of the ICH, midline shift (> 5 mm), ICH volume, IVH, hydrocephalus, subdural hemorrhage, subarachnoid hemorrhage, maximal BAVM nidus diameter, superficial and/or deep venous drainage, BAVM in eloquent location, Spetzler-Martin grade,21 and indication of proximal, pedicular, or nidal aneurysm. Disagreements between the 2 reviewers were resolved by consensus. The ICH volume was measured using the ABC/2 method, in which A is the greatest diameter on the slice on which the hematoma appears the largest, B is the diameter perpendicular to A, and C is the approximate number of axial slices, with hemorrhage multiplied by the slice thickness.13,16 The ICH volume was further subdivided into 1 of 4 categories: < 30 ml, ≥ 30 and < 60 ml, ≥ 60 and < 90 ml, and ≥ 90 ml. If the ICH was absent, its volume was noted as 0.

End Points

The primary end point was long-term disability as assessed by the mRS beyond 1 year after admission.19,24 Clinical outcome was evaluated by 1 of 2 authors (E.S. or M.D.) by phone interview or examination of the medical records. Four patients were included twice for 2 independent BAVM hemorrhagic events. For one of these patients, clinical outcome before the second hemorrhagic event was retrieved from the patient's file. For the 3 others, the final clinical outcome was used for the 2 independent hemorrhagic events. The secondary end point was inpatient mortality. Delta mRS (difference between long-term mRS score and baseline prehemorrhage mRS score) was calculated to account for the heterogeneity of the patients at baseline.

Statistical Analysis

To identify factors independently associated with poor long-term outcome (beyond 1 year after admission), univariate then multivariate models were evaluated using stepwise logistic regression. Only available data were analyzed. A variable was included in the multivariate logistic regression if it was clinically relevant and associated with a poor outcome in the univariate analysis (p < 0.05). Systolic BP was chosen to represent the hemodynamic state at admission. Intraventricular hemorrhage rather than hydrocephalus was included because it is less subject to interobserver discrepancy. The following variables were included in the multivariate logistic regression and stepwise analyses: initial glycemia > 8 mmol/L, systolic BP ≥ 160 mm Hg, midline shift (> 5 mm), deep venous drainage of the BAVM, 4-level ICH volume (< 30 ml, ≥ 30 and < 60 ml, ≥ 60 and < 90 ml, and ≥ 90 ml), 3-level consciousness score according to the GCS (GCS Scores 3 and 4, Scores 5–12, and Scores 13–15), and IVH. The GCS score categories were chosen to match those in the ICH score.13 The different ICH volume cutoff points are multiples of the 30-ml cutoff point in the ICH score. The variable with the highest p value was removed at each step until all variables were significantly associated with a poor outcome (p < 0.05). The weights of the ruptured AVM prognostic (RAP) score components were derived from the odd radios of the most parsimonious model. We constructed receiver operating characteristic (ROC) curves and compared the area under the ROC curve (AUC) of the RAP score and the ICH score to predict poor long-term outcome and inpatient mortality. The AUCs of the 2 scores were also compared for the prediction of the change in the mRS score (delta mRS). Sensitivity analysis was performed by comparing AUCs after excluding recurrent hemorrhagic events, BAVM ruptures without ICH, patients with past BAVM rupture, and BAVM ruptures following embolization. The sensitivity, specificity, positive predictive value, and negative predictive value of the RAP score were extracted from the different ROC curves constructed.

Data are expressed as a percentage (with 95% confidence intervals [CIs]) for binary variables and as the mean ± standard deviation (SD) for continuous variables. All tests were 2-sided, and all provided probability values are uncorrected. Probability values < 0.05 were considered significant. Statistical analyses were performed using STATA version 11 (StataCorp).

Results

Patient Characteristics

One hundred thirty-five consecutive patients with 139 independent hemorrhagic events related to BAVM rupture were included in our analysis (Fig. 1). Outcome was good after 79 AVM ruptures (56.8%) and poor in 60 cases (43.2%). The mean age at admission was 42 ± 14.9 years (range 15–73 years; Table 1). In 9 patients, BAVM rupture occurred after partial endovascular embolization; the mean delay after embolization was 3.3 ± 3.3 days. Twenty-five deaths occurred during the hospital stay (18% among 139 hemorrhagic events). Two additional deaths occurred during the follow-up period, after recurrent hemorrhages that were managed at another institution (at 6 and 33 months after the previous hemorrhagic event). One patient died of peritonitis 6 months after the hemorrhagic event. Overall mortality of the patients was 20.7% (28 of 135 patients).

FIG. 1.
FIG. 1.

Recruitment flowchart.

TABLE 1.

Summary of characteristics of the study population

ParameterValue
Demographics
  Mean age at admission in yrs (n = 139)42 ± 14.9
  Male sex (n = 139)72 (51.8%)
  Past BAVM rupture (n = 138)20 (14.5%)
Baseline prehemorrhage mRS score (n = 139)
  0113 (81.3%)
  1 & 223 (16.5%)
  3–53 (2.2%)
Clinical characteristics at presentation (n = 136)
  Initial GCS score <514 (10.3%)
  Initial GCS score <1356 (41.2%)
BAVM Spetzler-Martin grade (n = 126)
  I & II84 (66.7%)
  III25 (19.8%)
  IV & V17 (13.5%)
CT scanning data
  ICH (n = 138)117 (84.8%)
  Infratentorial location of ICH (n = 138)19 (13.8%)
  Midline shift (>5 mm) (n = 137)48 (35%)
  Mean ICH vol in ml (n = 135)41.2 ± 41
  IVH (n = 137)85 (62%)
  Hydrocephalus (n = 137)69 (50.4%)
  SDH (n = 138)9 (6.5%)
  SAH (n = 138)31 (22.5%)
Management in acute phase (n = 139)
  Surgery61 (43.9%)
  Ventricular drain77 (55.4%)
  Embolization26 (18.7%)
Outcome (n = 139)
  Inpatient death25 (18%)
  Mean delay to mRS scoring in mos (excluding inpatient mortality)33 ± 24
  Outcome mRS score <379 (56.8%)
  Outcome mRS score ≥360 (43.2%)

SAH = subarachnoid hemorrhage; SDH subdural hematoma; n = number of cases with available data for each variable.

Values expressed as mean ± standard deviation for quantitative variables or as percentage for qualitative variables.

Factors Associated With Poor Outcome After BAVM Rupture

Univariate analysis for long-term outcome showed that deep venous drainage, low initial GCS score, high BP, midline shift (> 5 mm), high ICH volume, IVH, hydrocephalus, and elevated glycemia were associated with a poor outcome (Table 2). Conversely, age, presence of an ICH, infratentorial location of ICH, subdural hemorrhage, and subarachnoid hemorrhage were not associated with a poor outcome. Multivariate logistic regression followed by stepwise analysis showed that only 3-level GCS (GCS Scores 3 and 4, 5–12, and 13–15), 4-level ICH volume (< 30 ml, ≥ 30 and < 60 ml, ≥ 60 and < 90 ml, and ≥ 90 ml), and IVH were independently associated with poor long-term outcome (Table 3).

TABLE 2.

Univariate analysis of admission characteristics by outcome (mRS score ≥ 3)

ParameterOR (95% CI)p Value
Demographics, medical history, and BAVM characteristics
  Age at admission1.02 (0.996–1.04)0.097
  Male sex1.41 (0.718–2.77)0.317
  History of high BP2.37 (0.947–5.92)0.065
  Antihypertensive therapy2.15 (0.852–5.44)0.105
  Anticoagulation therapy1.32 (0.081–21.6)0.845
  Antiplatelet therapy1.32 (0.081–21.6)0.845
  Knowledge of previous BAVM1.83 (0.887–3.79)0.102
  Previous BAVM rupture1.72 (0.663–4.47)0.265
  Previous BAVM treatment
    Surgery1.32 (0.081–21.6)0.845
    Radiosurgery0.874 (0.141–5.4)0.884
    Embolization1.85 (0.809–4.21)0.145
    Multiple embolization sessions1.57 (0.566–4.33)0.388
BAVM rupture complicating or following embolization2.35 (0.538–10.2)0.257
Superficial venous drainage0.648 (0.286–1.47)0.298
Deep venous drainage2.59 (1.22–5.5)0.014
Mixed venous drainage2.18 (0.94–5.03)0.069
BAVM in eloquent region (n = 135)0.754 (0.375–1.52)0.43
Max nidus diameter1.01 (0.994–1.03)0.18
Proximal or pedicular arterial aneurysms2.08 (0.916–4.73)0.08
Proximal, pedicular, or nidal aneurysms1.48 (0.704–3.13)0.3
Admission characteristics
  Initial GCS <520 (4.44–90.1)<10−3
  Initial GCS <1312 (5.31–27.1)<10−3
  Heart rate in bpm0.966 (0.941–0.991)0.009
  Systolic BP, mmHg1.02 (1–1.03)0.016
  Diastolic BP, mmHg1.03 (1–1.05)0.025
  Mean BP, mmHg1.03 (1–1.05)0.014
  SpO2 (%)0.87 (0.747–1.01)0.074
  Body temperature in °C0.99 (0.927–1.06)0.765
  ICH0.504 (0.197–1.29)0.153
  Infratentorial location of ICH0.846 (0.306–2.34)0.747
  Midline shift (>5 mm)3.63 (1.74–7.58)0.001
  ICH volume, ml1.02 (1.01–1.03)0.001
  ICH vol ≥301.76 (0.837–3.76)0.11
  ICH vol ≥606.85 (2.74–18.4)<10−3
  ICH vol ≥90 ml7.45 (2.23–32.5)<10−3
  IVH6.98 (3.02–16.1)<10−3
  Hydrocephalus5.84 (2.75–12.4)<10−3
  SDH0.652 (0.156–2.72)0.557
  SAH1.34 (0.601–3)0.472
  Glycemia, mmol/l1.35 (1.13–1.62)0.001
  Troponin I >0.04 µg/L1.32 (0.718–2.42)0.374
  Creatinine blood level, mg/dl0.988 (0.971–1.01)0.203

SpO2 = oxygen saturation.

TABLE 3.

Multivariate logistic regression of admission characteristics by outcome (mRS score ≥ 3)

ParameterMultivariate AnalysisStepwise Multivariate Analysis
p ValueOR (95% CI)p Valuep Valuep Valuep ValueOR (95% CI)
Initial glycemia >8 mmol/L0.8731.1 (0.354–3.4)
Systolic BP ≥160 mm Hg0.5241.49 (0.44–5.02)0.46
Midline shift (>5 mm)0.4431.59 (0.484–5.24)0.450.511
Deep venous drainage0.182.16 (0.7–6.69)0.180.2220.275
Four-level ICH volume0.0641.62 (0.971–2.72)0.0640.0460.0210.0051.84 (1.21–2.8)
Three-level GCS<10−36.42 (2.645–15.6)<10−3<10−3<10−3<10−36.5 (3.09–13.7)
IVH0.0026.58 (1.99–21.7)0.0020.0010.001<10−37.46 (2.66–21)

The last 5 columns represent backward stepwise logistic regression and show p values at each step after the highest value was removed.

Development of RAP Score and Comparison of RAP Score With ICH Score

The RAP score was designed by taking into account the odds ratios of the most parsimonious model. This score combines IVH, GCS score, and ICH volume into a single value between 0 and 11 (Table 4). Complete data were available for calculating the RAP score in 132 of 139 cases (Fig. 2). In our study population, a RAP score < 6 was associated with a poor outcome in 15.8% of cases (13 of 82 hemorrhagic events), whereas a RAP score ≥ 6 was associated with a poor outcome in 86% of cases (43 of 50 rupture events). For a RAP score ≥ 6, the sensitivity, specificity, and positive and negative predictive values for predicting a poor outcome were 76.8% (95% CI 63.6–87), 90.8% (95% CI 81.9–96.2), 86% (95% CI 73.3–94.2), and 84.1% (95% CI 74.3–91.3), respectively. Comparison of the AUCs showed that the RAP score (0.87, 95% CI 0.8–0.92) was superior to the ICH score (0.81, 95% CI 0.74–0.88) in predicting a poor long-term outcome (p = 0.009; Fig. 2C). The inpatient mortality rate was available for the entire BAVM hemorrhagic event source cohort (151 cases). Among 147 patients with 151 BAVM hemorrhagic events, 25 deaths occurred during the hospital stay (16.6% inpatient mortality per hemorrhagic event; 17% global patient mortality). Comparison of the AUCs showed that the RAP score (0.91, 95% CI 0.85–0.95) was superior to the ICH score (0.85, 95% CI 0.78–0.9) in predicting inpatient mortality (p = 0.006; Fig. 2D).

FIG. 2.
FIG. 2.

Distribution of RAP score and comparison with the ICH score according to the AUC. A: Distribution of RAP score values in the study population (n = number of hemorrhagic events). B: Observed and best-fit curve of the percentage of poor outcome for each value of the RAP score. C: Comparison of AUCs for RAP and ICH scores for predicting poor long-term outcome. D: Comparison of AUCs for RAP and ICH scores for predicting inpatient mortality. Figure is available in color online only.

TABLE 4.

The RAP score

ComponentAssigned Points
IVH4
GCS score
  3–44
  5–122
  13–150
ICH volume (ml)
  ≥903
  ≥60 & <902
  ≥30 & <601
  <300

The total RAP score is the sum of the points corresponding to each component (the score ranges from 0 to 11).

Prediction of Delta mRS and Sensitivity Analysis

The RAP score was superior to the ICH score in predicting delta mRS ≥ 3 (Supplementary Fig. 1). Sensitivity analysis was performed to test the robustness of RAP superiority over the ICH score. The study population comprised 4 patients who were included twice for 2 independent hemorrhagic events each. Moreover, the ICH score was created to predict outcome after ICH. Comparison of the AUCs showed that the RAP score was still superior to the ICH score in predicting poor long-term neurological outcome (Table 5) and inpatient mortality (Table 6) after excluding recurrent hemorrhage, BAVM ruptures without ICH, patients with prior BAVM rupture, and BAVM ruptures following embolization. Sensitivity, specificity, positive predictive value, and negative predictive value of the RAP score were calculated for these different subgroups (Table 7).

TABLE 5.

Sensitivity analysis for the prediction of poor long-term outcome defined as mRS score ≥ 3 more than 1 year after discharge

ParameterPoor Outcome (%)RAP Score AUC95% CIICH Score AUC95% CIp Value
Exclusion of recurrent hemorrhagic events (n = 4)57/135 (42.2%)0.870.8–0.920.810.73–0.870.012
Exclusion of patients w/past BAVM rupture (n = 20)47/119 (39.5%)0.840.73–0.910.790.70–0.860.022
Exclusion of BAVM ruptures w/o ICH (n = 21)48/118 (40.7%)0.90.82–0.950.830.75–0.90.006
Exclusion of patients w/BAVM rupture after embolization (n = 9)49/130 (37.7%)0.820.75–0.890.770.69–0.840.026

n = number of patients excluded from each subgroup.

TABLE 6.

Sensitivity analysis for the prediction of inpatient mortality

ParameterInpatient Mortality (%)RAP Score AUC95% CIICH Score AUC95% CIp Value
Exclusion of recurrent hemorrhagic events (n = 4)24/147 (16.3%)0.910.85–0.950.840.77–0.900.005
Exclusion of patients w/past BAVM rupture (n = 21)19/130 (14.6%)0.90.84–0.950.840.77–0.900.025
Exclusion of BAVM ruptures w/o ICH (n = 23)22/128 (17.2%)0.910.84–0.950.860.79–0.910.02
Exclusion of patients w/BAVM rupture after embolization (n = 10)21/141 (14.9%)0.920.86–0.960.860.79–0.920.021

n = number of patients excluded from each subgroup.

TABLE 7.

Sensitivity, specificity, positive predictive value, and negative predictive value of RAP score ≥ 6 for predicting a poor outcome

ParameterSensitivity (95% CI)Specificity (95% CI)Positive Predictive Value (95% CI)Negative Predictive Value (95% CI)
Exclusion of recurrent hemorrhagic events (n = 4)77.4% (63.8–87.7)90.8% (81.9–96.2)85.4% (72.1–94)85.2% (75.6–92.1)
Exclusion of patients w/past BAVM rupture (n = 20)77.8% (62.9–88.8)89.6% (79.7–95.7)83.3% (68.4–93.1)85.7% (75.3–92.9)
Exclusion of BAVM ruptures w/o ICH (n = 21)80.4% (66.1–90.6)91% (81.5–96.6)86% (72.1–94.7)87.1% (76.9–94)
Exclusion of patients w/BAVM rupture after embolization (n = 9)75% (61.1–86)90.5% (81.5–96.1)84.8% (71.1–93.7)83.7% (73.8–91.1)

n = number of patients excluded from each subgroup.

Discussion

In this study we aimed to identify individual factors predictive of a poor long-term neurological outcome after BAVM rupture and to develop an admission prognostic score. The resulting RAP score was a better predictor of death and dependence (mRS score ≥ 3 beyond 1 year after admission) than the ICH score in our study population.

Brain AVM rupture is associated with high morbidity and mortality. After 139 hemorrhagic events, we observed a 43.2% rate of poor outcome, defined as an mRS score ≥ 3 beyond 1 year after admission. Comparable rates of death and dependence (40%) after BAVM rupture have been reported by the Scottish Intracranial Vascular Malformation Study.23

Clinical grading scales enhance communication between health care providers and the quality of clinical research. They also provide a basis for decision making as regards aggressive neurocritical care. However, their use in clinical practice must be balanced against the risk of self-fulfilling prophecies of a poor outcome.15,25 Grading scales have been developed for disorders such as acute ischemic stroke (NIH Stroke Scale) and brain trauma (GCS). Development of the original ICH score included patients with secondary nontraumatic ICH; therefore, the score was intended to predict outcome after primary and secondary ICH.13,14 However, both initial studies and prospective external validation studies included low numbers of patients with ICH from secondary causes or excluded patients with secondary ICH.4,6–8,13,14 Attempts have been made to adapt the ICH score to BAVM rupture, but without improving performance in predicting outcome as compared with the original ICH score.2

Several individual predictors of a poor outcome after BAVM rupture have been reported.2,3,12,17 The GCS score is an independent predictor of death or dependence after BAVM rupture.23 It was found to be a strong predictor of outcome in the original ICH score and was thus subdivided into 3 groups according to the GCS score.13 Our results confirm this finding. Surprisingly, IVH was the strongest independent predictor of a poor outcome in our study population. Parenchymal BAVM hemorrhage has been associated with the highest stroke morbidity compared with subarachnoid hemorrhage and IVH.3,12 However, only isolated IVH (that is, not associated with ICH) was categorized as IVH in previous studies, resulting in 14%–16% rates of IVH after BAVM rupture.3,12 In the present study we considered all forms of IVH after BAVM rupture, whether or not it was associated with ICH. Our rate of IVH after BAVM rupture was 62%, which is similar to the rate noted by Gross et al., who specifically studied IVH and hydrocephalus after BAVM rupture.10 Intraventricular hemorrhage after AVM rupture may be a surrogate for deep as opposed to more superficial AVMs. Nevertheless, IVH morbidity was independent from ICH, as demonstrated by multivariate analysis in our study. In contrast to previous reports, our study revealed that an infratentorial location of ICH from BAVM rupture was not associated with a poor outcome.1 This finding may reflect the low incidence of infratentorial ICH in our series (13.8%) or a selection bias due to increased prehospital mortality in infratentorial ICH. An alternative explanation may be that infratentorial ICH could be more frequently associated with IVH, which was found to be the strongest predictor of a poor clinical outcome in our study. Note that we also found no effect of age on clinical outcome. The effect of age on outcome after BAVM rupture remains controversial.3,17,22

It is noteworthy that none of the individual components of the Spetzler-Martin grade affected outcome. A possible explanation is that prognosis is determined much more by the initial gravity of a BAVM rupture than by therapeutic management. However, it is possible that in a larger cohort, which has more statistical power, factors with a more subtle effect on long-term outcome may emerge, including treatment-related characteristics. Using the identified predictors, we developed a RAP score that is accurately associated with a poor outcome in our study population. Furthermore, the RAP score predicted a poor long-term outcome and inpatient mortality more accurately than the ICH score.

The major limitations of our study are its retrospective design with the inherent risk of bias and lack of a validation cohort. Referral bias also limited our study, which was conducted in a specialized tertiary care center. Indeed, the study population may have been biased toward more severe forms of hemorrhage. Future external prospective cohorts and population-based studies are needed to ensure full validation of RAP scores' capacity to predict outcome after BAVM rupture. Only patient characteristics at admission were evaluated in our study. Consequently, several factors with potential impact on outcome are not analyzed, including rebleeding, vasospasm, and therapeutic interventions. Our study was also limited by the choice of long-term mRS score as the outcome. Follow-up data were unavailable in 12 of 147 patients with 151 BAVM hemorrhagic events. Moreover, a heterogeneous time point for outcome evaluation was chosen because of the follow-up heterogeneity and retrospective study design. However, we believe that long-term outcome is the most clinically relevant time point for outcome evaluation because a substantial proportion of patients improve throughout the 1st year post-ICH.14 Inclusion criteria were purposely wide with the inclusion of patients with previous BAVM rupture, baseline disability, partial BAVM treatment, and hemorrhage following BAVM embolization, which may differ from spontaneous BAVM rupture in terms of mechanism. This choice may have increased the population's heterogeneity. However, limiting exclusion criteria also favors generalizability (external validity) of the study. Moreover, sensitivity analysis showed that the RAP score was still superior to the ICH score after excluding recurrent hemorrhages, BAVM ruptures without ICH, patients with prior BAVM rupture, and BAVM ruptures following embolization. The ICH score is both easy to use and accurate in predicting outcome after BAVM rupture, and it remains to be proved that the RAP score's superiority in the current study translates into a clinical benefit. Finally, the RAP score must be validated for other end points, such as functional outcome at hospital discharge after BAVM rupture or long-term neurological deficit (for example, the NIH Stroke Scale), activities of daily living (for example, the Barthel Index), global disability (for example, the Glasgow Outcome Scale),20 and neuropsychological outcome.

Conclusions

In summary, BAVM rupture was associated with high morbidity and mortality in this study. Factors independently associated with a poor outcome after BAVM-related hemorrhage were IVH, GCS score, and ICH volume. These factors were combined to form a new admission score, the RAP score, which is better than the ICH score in stratifying patients based on the risk of a poor outcome after BAVM rupture.

Acknowledgments

Financial support was provided by the Fondation des Gueules Cassées.

Disclosures

Dr. Clarençon is a consultant for Medtronic and Codman Neurovascular. Dr. Sourour is a consultant for Medtronic and Microvention.

Author Contributions

Conception and design: Clarençon, Shotar, Debarre, Degos. Acquisition of data: Clarençon, Shotar, Debarre. Analysis and interpretation of data: Clarençon, Shotar, Degos. Drafting the article: Shotar. Critically revising the article: Clarençon, Shotar, Sourour, Di Maria, Gabrieli, Nouet, Chiras, Degos. Reviewed submitted version of manuscript: Clarençon, Shotar, Sourour, Di Maria, Gabrieli, Nouet, Chiras, Degos. Approved the final version of the manuscript on behalf of all authors: Clarençon. Statistical analysis: Shotar, Degos. Administrative/technical/material support: Clarençon, Chiras, Degos. Study supervision: Clarençon, Degos.

Supplemental Information

Online-Only Content

Supplemental material is available with the online version of the article.

References

  • 1

    Abla AA, Nelson J, Rutledge WC, Young WL, Kim H, Lawton MT: The natural history of AVM hemorrhage in the posterior fossa: comparison of hematoma volumes and neurological outcomes in patients with ruptured infra- and supratentorial AVMs. Neurosurg Focus 37:3 E6, 2014

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 2

    Appelboom G, Hwang BY, Bruce SS, Piazza MA, Kellner CP, Meyers PM, : Predicting outcome after arteriovenous malformation-associated intracerebral hemorrhage with the original ICH score. World Neurosurg 78:646650, 2012

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 3

    Choi JH, Mast H, Sciacca RR, Hartmann A, Khaw AV, Mohr JP, : Clinical outcome after first and recurrent hemorrhage in patients with untreated brain arteriovenous malformation. Stroke 37:12431247, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Clarke JL, Johnston SC, Farrant M, Bernstein R, Tong D, Hemphill JC III: External validation of the ICH score. Neurocrit Care 1:5360, 2004

  • 5

    Cordonnier C, Al-Shahi Salman R, Bhattacharya JJ, Counsell CE, Papanastassiou V, Ritchie V, : Differences between intracranial vascular malformation types in the characteristics of their presenting haemorrhages: prospective, population-based study. J Neurol Neurosurg Psychiatry 79:4751, 2008

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6

    Fernandes H, Gregson BA, Siddique MS, Mendelow AD: Testing the ICH score. Stroke J Cereb Circ 33:14551456, 2002

  • 7

    Godoy DA, Boccio A: ICH score in a rural village in the Republic of Argentina. Stroke J Cereb Circ 34:e150e151, 2003. (Letter)

  • 8

    Godoy DA, Piñero G, Di Napoli M: Predicting mortality in spontaneous intracerebral hemorrhage: can modification to original score improve the prediction?. Stroke 37:10381044, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Gross BA, Du R: Natural history of cerebral arteriovenous malformations: a meta-analysis. J Neurosurg 118:437443, 2013

  • 10

    Gross BA, Lai PMR, Du R: Hydrocephalus after arteriovenous malformation rupture. Neurosurg Focus 34:5 E11, 2013

  • 11

    Halim AX, Johnston SC, Singh V, McCulloch CE, Bennett JP, Achrol AS, : Longitudinal risk of intracranial hemorrhage in patients with arteriovenous malformation of the brain within a defined population. Stroke 35:16971702, 2004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12

    Hartmann A, Mast H, Mohr JP, Koennecke HC, Osipov A, Pile-Spellman J, : Morbidity of intracranial hemorrhage in patients with cerebral arteriovenous malformation. Stroke 29:931934, 1998

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13

    Hemphill JC III, Bonovich DC, Besmertis L, Manley GT, Johnston SC, Tuhrim S: The ICH score: a simple, reliable grading scale for intracerebral hemorrhage. Stroke 32:891897, 2001

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 14

    Hemphill JC III, Farrant M, Neill TA Jr: Prospective validation of the ICH Score for 12-month functional outcome. Neurology 73:10881094, 2009

  • 15

    Hemphill JC III, Newman J, Zhao S, Johnston SC: Hospital usage of early do-not-resuscitate orders and outcome after intracerebral hemorrhage. Stroke 35:11301134, 2004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    Kothari RU, Brott T, Broderick JP, Barsan WG, Sauerbeck LR, Zuccarello M, : The ABCs of measuring intracerebral hemorrhage volumes. Stroke 27:13041305, 1996

  • 17

    Lv X, Liu J, Hu X, Li Y: Patient age, hemorrhage patterns, and outcomes of arteriovenous malformation. World Neurosurg 84:10391044, 2015

  • 18

    Miller CE, Quayyum Z, McNamee P, Al-Shahi Salman R: Economic burden of intracranial vascular malformations in adults: prospective population-based study. Stroke 40:19731979, 2009

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 19

    Rankin J: Cerebral vascular accidents in patients over the age of 60. II. Prognosis. Scott Med J 2:200215, 1957

  • 20

    Saver JL: Optimal end points for acute stroke therapy trials: best ways to measure treatment effects of drugs and devices. Stroke 42:23562362, 2011

  • 21

    Spetzler RF, Martin NA: A proposed grading system for arteriovenous malformations. J Neurosurg 65:476483, 1986

  • 22

    Taylor B, Appelboom G, Yang A, Bruce E, LoPresti M, Bruce S, : Underlying effect of age on outcome differences in arteriovenous malformation-associated intracerebral hemorrhage. J Clin Neurosci 22:526529, 2015

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 23

    van Beijnum J, Lovelock CE, Cordonnier C, Rothwell PM, Klijn CJM, Al-Shahi Salman R: Outcome after spontaneous and arteriovenous malformation-related intracerebral haemorrhage: population-based studies. Brain 132:537543, 2009

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 24

    van Swieten JC, Koudstaal PJ, Visser MC, Schouten HJ, van Gijn J: Interobserver agreement for the assessment of handicap in stroke patients. Stroke 19:604607, 1988

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 25

    Zahuranec DB, Brown DL, Lisabeth LD, Gonzales NR, Longwell PJ, Smith MA, : Early care limitations independently predict mortality after intracerebral hemorrhage. Neurology 68:16511657, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation

If the inline PDF is not rendering correctly, you can download the PDF file here.

Contributor Notes

Correspondence Frédéric Clarençon, Department of Interventional Neuroradiology, Pitié-Salpêtrière Hospital, 47 Boulevard de l'Hôpital, Paris 75013, France. email: fredclare5@gmail.com.

INCLUDE WHEN CITING Published online January 20, 2017; DOI: 10.3171/2016.9.JNS161431.

Drs. Degos and Clarençon contributed equally to this work.

Disclosures Dr. Clarençon is a consultant for Medtronic and Codman Neurovascular. Dr. Sourour is a consultant for Medtronic and Microvention.

  • View in gallery

    Recruitment flowchart.

  • View in gallery

    Distribution of RAP score and comparison with the ICH score according to the AUC. A: Distribution of RAP score values in the study population (n = number of hemorrhagic events). B: Observed and best-fit curve of the percentage of poor outcome for each value of the RAP score. C: Comparison of AUCs for RAP and ICH scores for predicting poor long-term outcome. D: Comparison of AUCs for RAP and ICH scores for predicting inpatient mortality. Figure is available in color online only.

  • 1

    Abla AA, Nelson J, Rutledge WC, Young WL, Kim H, Lawton MT: The natural history of AVM hemorrhage in the posterior fossa: comparison of hematoma volumes and neurological outcomes in patients with ruptured infra- and supratentorial AVMs. Neurosurg Focus 37:3 E6, 2014

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 2

    Appelboom G, Hwang BY, Bruce SS, Piazza MA, Kellner CP, Meyers PM, : Predicting outcome after arteriovenous malformation-associated intracerebral hemorrhage with the original ICH score. World Neurosurg 78:646650, 2012

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 3

    Choi JH, Mast H, Sciacca RR, Hartmann A, Khaw AV, Mohr JP, : Clinical outcome after first and recurrent hemorrhage in patients with untreated brain arteriovenous malformation. Stroke 37:12431247, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Clarke JL, Johnston SC, Farrant M, Bernstein R, Tong D, Hemphill JC III: External validation of the ICH score. Neurocrit Care 1:5360, 2004

  • 5

    Cordonnier C, Al-Shahi Salman R, Bhattacharya JJ, Counsell CE, Papanastassiou V, Ritchie V, : Differences between intracranial vascular malformation types in the characteristics of their presenting haemorrhages: prospective, population-based study. J Neurol Neurosurg Psychiatry 79:4751, 2008

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6

    Fernandes H, Gregson BA, Siddique MS, Mendelow AD: Testing the ICH score. Stroke J Cereb Circ 33:14551456, 2002

  • 7

    Godoy DA, Boccio A: ICH score in a rural village in the Republic of Argentina. Stroke J Cereb Circ 34:e150e151, 2003. (Letter)

  • 8

    Godoy DA, Piñero G, Di Napoli M: Predicting mortality in spontaneous intracerebral hemorrhage: can modification to original score improve the prediction?. Stroke 37:10381044, 2006

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Gross BA, Du R: Natural history of cerebral arteriovenous malformations: a meta-analysis. J Neurosurg 118:437443, 2013

  • 10

    Gross BA, Lai PMR, Du R: Hydrocephalus after arteriovenous malformation rupture. Neurosurg Focus 34:5 E11, 2013

  • 11

    Halim AX, Johnston SC, Singh V, McCulloch CE, Bennett JP, Achrol AS, : Longitudinal risk of intracranial hemorrhage in patients with arteriovenous malformation of the brain within a defined population. Stroke 35:16971702, 2004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12

    Hartmann A, Mast H, Mohr JP, Koennecke HC, Osipov A, Pile-Spellman J, : Morbidity of intracranial hemorrhage in patients with cerebral arteriovenous malformation. Stroke 29:931934, 1998

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13

    Hemphill JC III, Bonovich DC, Besmertis L, Manley GT, Johnston SC, Tuhrim S: The ICH score: a simple, reliable grading scale for intracerebral hemorrhage. Stroke 32:891897, 2001

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 14

    Hemphill JC III, Farrant M, Neill TA Jr: Prospective validation of the ICH Score for 12-month functional outcome. Neurology 73:10881094, 2009

  • 15

    Hemphill JC III, Newman J, Zhao S, Johnston SC: Hospital usage of early do-not-resuscitate orders and outcome after intracerebral hemorrhage. Stroke 35:11301134, 2004

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16

    Kothari RU, Brott T, Broderick JP, Barsan WG, Sauerbeck LR, Zuccarello M, : The ABCs of measuring intracerebral hemorrhage volumes. Stroke 27:13041305, 1996

  • 17

    Lv X, Liu J, Hu X, Li Y: Patient age, hemorrhage patterns, and outcomes of arteriovenous malformation. World Neurosurg 84:10391044, 2015

  • 18

    Miller CE, Quayyum Z, McNamee P, Al-Shahi Salman R: Economic burden of intracranial vascular malformations in adults: prospective population-based study. Stroke 40:19731979, 2009

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 19

    Rankin J: Cerebral vascular accidents in patients over the age of 60. II. Prognosis. Scott Med J 2:200215, 1957

  • 20

    Saver JL: Optimal end points for acute stroke therapy trials: best ways to measure treatment effects of drugs and devices. Stroke 42:23562362, 2011

  • 21

    Spetzler RF, Martin NA: A proposed grading system for arteriovenous malformations. J Neurosurg 65:476483, 1986

  • 22

    Taylor B, Appelboom G, Yang A, Bruce E, LoPresti M, Bruce S, : Underlying effect of age on outcome differences in arteriovenous malformation-associated intracerebral hemorrhage. J Clin Neurosci 22:526529, 2015

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 23

    van Beijnum J, Lovelock CE, Cordonnier C, Rothwell PM, Klijn CJM, Al-Shahi Salman R: Outcome after spontaneous and arteriovenous malformation-related intracerebral haemorrhage: population-based studies. Brain 132:537543, 2009

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 24

    van Swieten JC, Koudstaal PJ, Visser MC, Schouten HJ, van Gijn J: Interobserver agreement for the assessment of handicap in stroke patients. Stroke 19:604607, 1988

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 25

    Zahuranec DB, Brown DL, Lisabeth LD, Gonzales NR, Longwell PJ, Smith MA, : Early care limitations independently predict mortality after intracerebral hemorrhage. Neurology 68:16511657, 2007

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation

Metrics

All Time Past Year Past 30 Days
Abstract Views 485 250 0
Full Text Views 397 207 12
PDF Downloads 286 113 7
EPUB Downloads 0 0 0