Predictors and Impact of Aneurysm Rebleeding After Subarachnoid Hemorrhage FREE

Aneurysm rebleeding has long been recognized as a catastrophic complication of subarachnoid hemorrhage (SAH) and a major cause of in-hospital mortality.¹ Even in centers that aim for early definitive treatment, rebleeding has been reported to be the predominant preventable cause of poor outcome,² which may in part be the result of advances in intensive care management that reduce damage from cerebral ischemia and hydrocephalus.³

Disagreement exists about the period of greatest risk for rebleeding. Most studies show the highest risk of rerupture to be within the first 24 hours^{4 - 5} (often within the first 6⁶ or 12⁷ hours), whereas others have shown the highest risk to be between days 4 and 9⁸ or after day 10⁹ or found no period of highest risk.¹⁰ Some authors have found an association between rebleeding and poor clinical grade,^{8 ,11 - 15} whereas others have not.^{16 - 17} Loss of consciousness at ictus has been reported to be a predictor,⁴ and a relationship between ventricular drainage and rebleeding has been reported.^{18 - 19} There is also evidence for^{14 - 15} and against²⁰ large aneurysm size as a risk factor for rebleeding.

The widespread adoption of treatment strategies that emphasize early aneurysm repair has been cited as a critical factor in explaining the progressive reduction in mortality associated with SAH during the past 10 to 20 years.²¹ Accordingly, it might be assumed that with current aggressive endovascular therapy, surgical therapy, and neurocritical care, rebleeding is no longer a major clinical concern. We sought to identify the frequency, predictors, and impact of rebleeding on outcome in the modern era of aneurysm care.

We enrolled 574 patients with SAH admitted to the Neurological Intensive Care Unit of Columbia-Presbyterian Medical Center, New York, NY, from August 1996 to June 2002 in the Columbia University SAH Outcomes Project. Subarachnoid hemorrhage was diagnosed by the admission computed tomogram (CT) or by xanthochromia of the cerebrospinal fluid if the CT was nondiagnostic. Patients admitted within 14 days of spontaneous aneurysmal and nonaneurysmal SAH were included. We prospectively recorded demographic, historical, medical, and 3-month outcomes data as previously described.²² The study was approved by the hospital’s institutional review board. Written informed consent was obtained from the patient or a surrogate in all cases except when the patient died in the hospital, in which case the hospital’s institutional review board approved collection of data in an anonymous registry if the patient’s family declined consent.

Our management protocol calls for diagnostic angiography within 12 hours and aneurysm repair within 24 hours of admission whenever possible. Decisions regarding the modality and timing of treatment are made by an interdisciplinary team, including a neurosurgeon, interventional neuroradiologist or endovascular neurosurgeon, and neurointensivist. Reasons for deferring treatment within 2 days included deep coma (Glasgow Coma Scale [GCS] score <5), hemodynamic instability, severe concurrent medical complications, anatomic features of the aneurysm that precluded safe treatment during the acute phase, severe coexisting vasospasm, and patient or family refusal. Oral nimodipine was given unless systolic blood pressure was less than 120 mm Hg. To maintain a central venous pressure of 5 mm Hg or higher (or pulmonary artery diastolic pressure ≥7 mm Hg), patients were treated with isotonic sodium chloride solution, 1 mL/kg hourly, and 250 mL of 5% albumin solution every 2 hours if needed. An external ventricular drain (EVD) was placed in all patients with symptomatic hydrocephalus or intraventricular hemorrhage with depressed level of consciousness. Vasopressors were given to patients after securing the aneurysm to maintain a systolic blood pressure of more than 140 mm Hg if needed, and patients with clinical signs of delayed cerebral ischemia were treated with hypertensive hypervolemic therapy to a target systolic blood pressure of 180 to 220 mm Hg. When significant clinical symptoms persisted despite this therapy for more than 2 hours, balloon angioplasty of vasospastic vessels was performed or intra-arterial verapamil hydrochloride was administered if feasible. We did not use antifibrinolytic treatment.

Day 0 refers to the calendar day of the index SAH, defined as the clinical bleeding event that immediately preceded hospital admission. Rebleeding was defined as an acute deterioration in neurological status in conjunction with new hemorrhage apparent on CT while in the hospital. We recorded rebleeding episodes at referring hospitals before transfer, as well as those at our institution. We did not classify intraprocedural aneurysm rupture or sentinel events followed by an index SAH as rebleeding. Postprocedure rebleeding episodes are not considered in this report. In each case of rebleeding, the clinical data were prospectively recorded by the study investigators and adjudicated by consensus. Each episode of rebleeding was reconfirmed independently by a second investigator (A.M.N.) by review of the CTs and medical record.

We recorded baseline demographic data (age and sex), social history (tobacco and alcohol use), relevant medical history, clinical features at onset (sentinel bleeding, initial misdiagnosis, loss of consciousness, or seizure at ictus), and interval from symptoms of SAH to diagnosis. A study neurologist performed a neurologic and general medical evaluation on admission. Neurologic status on admission was assessed with the GCS²³ and the Hunt-Hess scale.²⁴ We also calculated admission Acute Physiology and Chronic Health Evaluation II (APACHE II)²⁵ physiologic subscores by subtracting the GCS contribution, age, and chronic health points from the total score. At the end of each patient’s hospital stay, we recorded all clinical adverse events, diagnostic studies, procedures, and outcomes.

The CTs were obtained on admission, at every major clinical event, and at 14 days or discharge, whichever was sooner. We also reviewed CTs performed at other institutions for patients transferred to our institution. Admission scans were prospectively rated for acute hemorrhage using the Columbia SAH scale²⁶ and Hijdra SAH sum score.²⁷ In patients with negative angiogram results, CTs were rated as showing a classic perimesencephalic or nonperimesencephalic pattern according to standard criteria.²⁸ Angiograms were rated for aneurysm location and maximal diameter on digital subtraction angiography.

Recorded events during hospitalization included mode of aneurysm treatment (clip, coil, or none), timing of aneurysm repair (within 2 days of admission, later repair, or never), EVD placement, use of intravenous vasopressors and antihypertensive agents, cerebral infarction on follow-up CT, and in-hospital mortality.

Functional and cognitive outcomes at 3 months were assessed with the modified Rankin Scale (mRS), a global disability scale with scores that range from 0 (no symptoms) to 6 (death); and the Telephone Interview for Cognitive Status,²⁹ a 10- to 15-minute telephone-administered test of global cognitive function with scores that range from 51 (best) to 0 (worst).

Variables on admission were tested for an association with rebleeding with the χ² test (for categorical variables), analysis of variance (for continuous normally distributed variables), or the Mann-Whitney U test (for continuous, nonnormally distributed variables). To derive a multiple logistic regression model for the prediction of rebleeding, forward selection algorithms were used to define the most important baseline predictors. Two-way interactions were tested when appropriate. Treatment factors (coiling vs clipping, aneurysm repair within 2 days of arrival to our hospital, vasopressor use, and EVD placement) were then tested for significant effects when added to the model of baseline predictors. To determine the validity of our findings in patients who were stable enough to undergo early aneurysm repair, we performed this analysis again in patients who underwent a repair procedure within 2 days of admission.

Kaplan-Meier analysis was performed for analyses concerning time to rebleeding episode. Predictors of functional outcome (mRS) were tested in a multiple logistic regression model, with survival with no disability defined as an mRS score of 0 or 1 or survival with functional independence defined as an mRS score of 0 to 3. Results with P<.05 were considered statistically significant. Statistical calculations were made with standard commercial software (SPSS Institute, Chicago, Ill).

Baseline characteristics of the 574 patients are given in Table 1. The use of diagnostic angiography, treatment, and clinical course are shown in Figure 1. Of the 574 study patients, 93% underwent angiography; aneurysms were found in 88% of these patients. Among patients with a defined aneurysm, 76% underwent clipping (median SAH day 2, median admission day 1), 18% underwent coiling (median SAH day 2, median admission day 1), and 6% were not treated. Of those 455 in whom angiography revealed an aneurysm and who underwent a repair procedure, 30 had their aneurysms repaired on the day of ictus, 168 on day 1, and 94 on day 2. A total of 454 (79%) of the 574 patients were transferred from an outside hospital. When analyzed with respect to admission to our hospital, 153 had their aneurysms repaired on day 0, 203 on day 1, and 39 on day 2.

Forty (6.9%) of the 574 patients experienced a rebleeding episode. Univariate baseline predictors of rebleeding included sentinel bleeding, seizure or loss of consciousness at ictus, higher National Institutes of Health Stroke Scale score, higher Hunt-Hess grade, lower GCS score, higher APACHE II physiologic subscores, larger aneurysm size, and intraventricular or intracerebral hemorrhage (Table 2). Sex, age, blood pressure, aneurysm size and location, and Hijdra SAH sum score were not associated with rebleeding. Predictors of rebleeding in a multiple logistic regression model of factors present on admission included only Hunt-Hess grade (odds ratio [OR], 1.92 per grade; 95% confidence interval [CI], 1.33-2.75; P<.001) and larger aneurysm size (OR, 1.07/mm; 95% CI, 1.02-1.13; P = .005) without interaction. As expected, adjustment for aneurysm repair (coded yes or no) showed that patients who had their aneurysms repaired were less likely to experience a rebleeding episode (OR, 0.21; 95% CI, 0.07-0.613; P = .004). Figure 2 illustrates the predicted risk of rebleeding, depending on Hunt-Hess grade and aneurysm size from this model.

The rebleeding rate among patients who underwent aneurysm repair was 5% (22/444); this rate was higher among patients with clipped rather than coiled aneurysms (5.7% vs 2.4%, Table 3, P = .28). Aneurysm repair was performed within 2 calendar days of hospital admission in 18 (82%) of 22 of these patients. The rebleeding rate was 38% (11/29) in patients with an angiographically defined aneurysm that could not be repaired, 13% (5/38) in patients who were too unstable to undergo angiography, and 3% (2/63) in patients without an aneurysm on angiography. Forty-two of the 63 patients with a negative angiogram result had a nonperimesencephalic pattern of SAH, including both of the patients with a negative angiogram result who experienced a rebleeding episode. No patient experienced a rebleeding episode after 2 negative angiogram results a week apart.

The daily risk of rebleeding among patients who had not already experienced rebleeding or died was 2.6% on day 0 (N = 15/574), 2.1% on day 1 (N = 12/588), 0.3% on day 2 (N = 2/527), 0.7% on day 3 (N = 4/518), and 0.1% on day 4 (N = 1/510), with a similar low rate thereafter. Among those who experienced rebleeding, Kaplan-Meier analysis revealed that patients with Hunt-Hess grade IV and V neurologic status experienced rebleeding earlier than patients with other grades (P = .001 by log-rank test); all patients with grade IV or V neurologic status experienced rebleeding by day 8 (Figure 3). There was no relation between time to rebleeding and aneurysm size.

After controlling for Hunt-Hess grade and aneurysm diameter on admission, patients who experienced rebleeding were more likely to have an EVD placed (OR, 5.10; 95% CI, 1.62-15.99; P = .005) than those who did not (Table 3). The multiple logistic regression model that predicted rebleeding from Hunt-Hess grade, aneurysm size, and EVD placement was improved by adding an EVD times Hunt-Hess interaction (P = .002, Nagelkerke R² increased from 0.21 to 0.28), indicating that the relationship between placement of an EVD and rebleeding depended on Hunt-Hess grade, in that the association between rebleeding and EVD placement was limited to patients with good grades (I-III). Rebleeding was also associated with an increased frequency of vasopressor use and cerebral infarction. However, in no instances were vasopressors used to induce hypertension before aneurysm repair.

When we repeated the univariate analyses for associations with baseline variables and rebleeding (Table 2), as well as the associations between in-hospital treatments and rebleeding (Table 3), the results were broadly similar. In these patients only Hunt-Hess grade was associated with rebleeding (OR, 1.9 per grade; 95% CI, 1.24-2.9; P = .003).

Patients who experienced rebleeding were more likely to die in the hospital than those who did not (OR, 6.5; 95% CI, 3.3-12.6; P<.001). Among 417 patients for whom 3-month mRS scores were available (Table 4), rebleeding significantly increased the odds of death (OR, 5.1; 95% CI, 2.5-10.5; P<.001), reduced the odds of survival without disability (mRS score <2; OR, 0.05; 95% CI, 0.007-0.35; P = .003), and reduced the odds of survival with functional independence (mRS score <4; OR, 0.05; 95% CI, 0.02-0.15; P<.001). The reduced probably of survival with functional independence persisted after correction for admission Hunt-Hess grade and aneurysm size (OR, 0.08; 95% CI, 0.02-0.34; P<.001). Telephone Interview for Cognitive Status scores among survivors at 3 months were lower in patients who experienced rebleeding (21.8 vs 33.6, P = .03).

We sought to examine the frequency and impact of rebleeding in the modern era of aneurysm care. We found that rebleeding still occurs in nearly 1 of every 15 patients, is predicted by poor admission Hunt-Hess grade and large aneurysm size, occurs most often within 72 hours of ictus (73%), and is associated with a dismal prognosis. However, the high rebleeding rate (6.9%) found at our urban tertiary care referral center may not be generalizable to all hospitals. Larger and more anatomically complex aneurysms were likely overrepresented in our study population, and the frequency of posterior circulation aneurysms (20%) was 4 times higher than that reported in population-based studies of SAH.³⁰

Rebleeding was strictly defined as an in-hospital event. Antecedent thunderclap headaches before an index SAH are generally termed sentinel headaches. Not all authors have found an association between sentinel headache and subsequent in-hospital rebleeding, and in the past such events have not been associated with poor outcome.³¹

In our study 2 patients with a negative initial angiogram result experienced rebleeding. Both of these patients had good neurologic grades (Hunt-Hess grades I and III) on admission, and both died in the hospital after rerupture. However, their CTs were not typical of perimesencephalic SAH. The most likely explanation is that the aneurysm thrombosed after the initial SAH and was angiographically invisible. These cases lend support to the practice of performing additional angiograms after several days in suspicious cases.

Rebleeding has previously been associated with ventricular drainage, suggesting that zealous control of ICP with drainage might reduce the tamponade effect on a recently ruptured aneurysm and lead to rebleeding.¹⁸ We too found a relationship between EVD and rebleeding but also found that this association applied only to patients who had good neurologic grades on admission (Hunt-Hess grades I to III). The most logical explanation for this relationship is that EVDs were placed after rebleeding occurred in patients with good neurologic grade when their condition deteriorated. Thus, we believe that rebleeding led to EVD placement and not the other way around.

When patients experienced rebleeding they were unlikely to undergo subsequent coiling but rather underwent clipping if feasible. Since we did not consider postprocedure rebleeding episodes in this analysis, we cannot comment on the relative durability of either procedure. We found a higher risk of pretreatment rebleeding in patients whose aneurysms were clipped rather than coiled (5.7% and 2.4%), although this difference was nonsignificant. The International Subarachnoid Aneurysm Trial³² found a similar trend (2.1% and 1.3%). The higher overall pretreatment rebleeding risk in our study may reflect the fact that we recorded rebleeding episodes at other hospitals before transfer, the fact that we studied all SAH patients as opposed to a subset, and the high acuity level of our referral population.

When we performed univariate analyses for associations with rebleeding in patients whose aneurysms were repaired within 2 days of admission, aneurysm diameter was no longer predictive, probably because patients with aneurysms that were larger and more difficult to repair were excluded from the early treatment group. The APACHE II physiologic subscores were no longer different between patients who experienced rebleeding and those who did not in the early treatment cohort, suggesting that these patients were more medically stable and therefore better candidates for emergency surgery. Interestingly, the daily risk of rebleeding in our cohort of up to 2 days after rupture is close to that described in natural history studies, whereas the daily risk thereafter is lower.

Rebleeding was associated with increased use of vasopressor medications and cerebral infarction, although these factors did not add to the model predicting rebleeding from Hunt-Hess grade and aneurysm size. By adding to the amount of blood in the subarachnoid space, rebleeding increases the risk of vasospasm-associated ischemic stroke, which is generally treated with pressors.

Since rebleeding occurs most frequently on the day of SAH, more rapid securing of unruptured aneurysms might lead to improved outcomes. One institution³³ found that a policy of clipping aneurysms sooner (as opposed to delayed surgery) led to a reduction in the risk of rebleeding, although timing of surgery in other studies^{9 ,34} has not been found to affect overall prognosis independent of patient age and clinical grade. The more liberal application of endovascular therapy at the time of initial angiography may also reduce the risk of early rebleeding, since surgery after angiography is often delayed by 6 to 12 hours.

In addition to performing aneurysm repair as early as possible, our findings suggest that the routine use of antifibrinolytics or other hemostatic pharmacologic interventions should be reconsidered,³⁵ particularly in patients with poor grade and large aneurysms or patients who must be transferred to a tertiary care center. One large, prospective, randomized trial showed that a policy of administering tranexamic acid for up to 72 hours from time of rupture until repair decreased rebleeding by 80% without an increase in thrombotic complications.³⁶ If we had used such a strategy and it were 100% successful in preventing rebleeding up to 72 hours after ictus, we would have prevented or delayed this catastrophic complication in 29 (73%) of 40 patients. This would have resulted in an overall rebleeding rate of 1.9%, with a number needed to treat of 20 to prevent 1 rebleeding episode.

Correspondence: Stephan A. Mayer, MD, Neurological Institute, 710 W 168th St, Box 39, New York, NY 10032 (sam14@columbia.edu).

Accepted for Publication: July 8, 2004.

Author Contributions:Study concept and design: Naidech, Kreiter, and Mayer. Acquisition of data: Mayer. Analysis and interpretation of data: Naidech, Janjua, Kreiter, Ostapkovich, Fitzsimmons, Parra, Commichau, Connolly, and Mayer. Drafting of the manuscript: Naidech, Kreiter, Commichau, and Mayer. Critical revision of the manuscript for important intellectual content: Janjua, Kreiter, Ostapkovich, Fitzsimmons, Parra, Connolly, and Mayer. Statistical analysis: Naidech, Kreiter, and Ostapkovich. Administrative, technical, and material support: Ostapkovich and Mayer. Study supervision: Fitzsimmons, Parra, Commichau, Connolly, and Mayer.

Funding/Support: This research was supported in part by grant-in-aid 9750432N from the American Heart Association, Dallas, Tex (Dr Mayer).