SSRI and statin use increases the risk for vasospasm after subarachnoid hemorrhage

Cerebral vasospasm occurs in approximately two-thirds of patients with subarachnoid hemorrhage (SAH), and over half of these patients develop delayed ischemic deficits.¹ Options to prevent vasospasm are limited; furthermore, accurate and timely diagnosis of vasospasm remains a challenge. Despite close observation in intensive care units, vasospasm is often diagnosed after the onset of neurologic deterioration, which impedes the efficacy of therapeutic interventions. It is thus important to identify risk factors and improve the ability to predict vasospasm. Patient characteristics such as poor clinical grade, thick blood clots on head CT scan (Fisher Group 3 SAH), type and timing of aneurysm treatment, young age, cigarette smoking, and fever have previously been identified as independent predictors of vasospasm.^1–8

Cocaine use, which induces vasoconstriction via potent sympathomimetic and serotonergic effects, increases the risk for SAH-related vasospasm.^9,10 Theoretically, commonly used prescription medications that have vasoconstrictive or vasodilatory effects (vasoactive medications, e.g., calcium channel blockers, beta-receptor blockers, angiotensin converting enzyme [ACE] inhibitors, angiotensin-1 [AT1] receptor blockers, HMG-CoA reductase inhibitors [statins], certain antidepressants) can also affect cerebral vascular tone. In this study we evaluated the risk for developing vasospasm and poor clinical outcome in patients documented to be using different classes of vasoactive medications at the time of SAH.

Methods.

Under a research protocol approved by our Institutional Review Board, we reviewed medical records of 918 patients with International Classification of Diseases–9 (ICD-9) diagnostic code 430.0 (SAH) admitted to our hospital between November 1995 and February 2003. In addition, we prospectively identified and collected vasospasm data on SAH patients admitted to our neurologic intensive care unit (NICU) after November 2000. We included 514 patients with evidence for SAH on head CT or CSF examination, and intracranial aneurysms on transfemoral, CT, or MR angiography. We excluded 404 patients with nonaneurysmal SAH (e.g., brain trauma, arteriovenous malformation, coagulopathy, vasculitis, dissection; n = 208), incorrect coding for SAH (n = 49), tests for vasospasm not performed (n = 1), SAH from an infected aneurysm (n = 1), untreated aneurysmal SAH (n = 8), admission after day 14, i.e., after the typical period of vasospasm (n = 46), and moribund clinical status on admission who either died by day 4, i.e., before the typical period of vasospasm, or in whom aneurysm treatment was not attempted (n = 91).

We collected data on the following covariables: patient age; sex; Hunt-Hess grade¹¹; thickness of clot on head CT scan (Fisher group³); aneurysm number, size, and location; postictal day and mode (clipping or endovascular) of aneurysm treatment; and history of smoking and hypertension (coded as missing if not documented).

Vasoactive medication exposure was considered present if the documented admission medications belonged to one of the following classes: calcium channel blockers; ACE inhibitors or AT1 receptor blockers; beta-receptor blockers; selective serotonin reuptake inhibitors (SSRIs); antidepressants with predominant sympathomimetic effects (non-SSRI antidepressants, e.g., venlafaxine); and statins. Since statin discontinuation has been implicated in worsening outcomes after myocardial infarction,^12,13 we recorded whether statins were discontinued after hospitalization. In addition, we collected data on exposure to aspirin, which has been shown to improve clinical outcomes after SAH^14,15 and may influence SAH-related vasospasm by inhibiting the arachidonic acid pathway.

Vasospasm was considered present, and its severity noted, based upon the results of direct or indirect angiography or transcranial Doppler (TCD) studies. At our hospital, all SAH patients are initially admitted to the NICU, thoroughly evaluated for SAH etiology,¹⁶ and treated with nimodipine.¹⁷ TCDs are performed at least once daily and when there is a change in the neurologic examination or a new neurologic deficit. When vasospasm is suspected, hypertensive hypervolemic therapy is instituted¹⁸ and cerebral angiography is performed. Angiographic vasospasm is graded as per the degree of arterial narrowing as mild (25 to 49%), moderate (50 to 69%), or severe (>70%). On TCDs, vasospasm in the anterior, middle, and posterior cerebral arteries and the intracranial internal carotid artery is graded as per the peak systolic blood flow velocity as mild (200 to 249 cm/second), moderate (250 to 299 cm/second), or severe (≥300 cm/second); the corresponding values for the basilar artery are 120 to 149 cm/second, 150 to 199 cm/second, and ≥200 cm/second. In most patients, Lindegaard’s hemispheric ratio¹⁹ is used as an additional criterion for vasospasm in the middle cerebral artery. If confirmed, symptomatic vasospasm refractory to medical therapy is treated aggressively with balloon angioplasty and intra-arterial medications.²⁰ Decisions regarding neurosurgical clipping vs endovascular coiling are made by a combined neurovascular team as previously described.²⁰ Vasospasm management is uniform and does not differ according to the aneurysm treatment modality.

Symptomatic vasospasm was defined as a focal neurologic deficit in the territory of the basal or cortical arteries from 4 to 14 days after ictus or a decline in level of consciousness; the neurologic worsening had to be associated with cerebral vasospasm and unrelated to other causes such as hydrocephalus or rebleeding. Other outcome measures included all vasospasm (i.e., the total of symptomatic and asymptomatic vasospasm, the latter identified by angiography, TCDs, or parenchymal hypodensity on head CT scan); length of stay (LOS) in the NICU and in the hospital; discharge modified Rankin scale²¹ (obtained from the last physical therapist’s note and the discharge summary); and the incidence of death on or after postictal day 4.

Results.

Table 1 shows characteristics of the final study population (n = 514). Vasospasm was documented in 62%, and 29% (47% of those with vasospasm) developed symptomatic vasospasm. Ninety percent of patients with symptomatic vasospasm were diagnosed by transfemoral angiography. At the time of admission, 45 patients were on calcium channel blockers (diltiazem 12, amlodipine 12, nifedipine 11, verapamil 10); 51 on beta-receptor blockers (atenolol 22, metoprolol 21, propranolol 4, nadolol 2, sotalol 1, bisoprolol 1); 38 on ACE inhibitors or AT1 receptor blockers (lisinopril 18, enalapril 6, captopril 2, fosinopril 2, valsartan 2, trandolapril 2, ramipril 1, moexipril 1, benazepril 1, enalapril and lisinopril 2, lisinopril and captopril 1); 39 on aspirin; 40 on SSRIs (paroxetine 13, fluoxetine 12, sertraline 8, trazodone 2, citalopram 2, fluoxetine and trazodone 1, paroxetine and trazodone 1, citalopram and trazodone 1); 15 on non-SSRI antidepressants (bupropion 7, mirtazepine 3, venlafaxine 3, nefazodone 1, milnacipran 1); and 36 on statins (atorvastatin 23, simvastatin 5, pravastatin 4, lovastatin 2, fluvastatin 2). Statins were discontinued in 19 patients after admission.

Table 2 shows results of the univariate analysis for vasospasm. Use of calcium channel blockers, beta-receptor blockers, ACE inhibitors/AT1 receptor blockers, aspirin, and non-SSRI vasoactive antidepressants did not affect the risk for all vasospasm or symptomatic vasospasm. However, SSRI users tended to have a higher risk for all vasospasm, and a significantly higher risk for symptomatic vasospasm. Statin users tended to have a higher risk for total vasospasm, but not symptomatic vasospasm. As compared to patients who were either not using statins or who continued statins, patients who discontinued statins had a marginally higher risk for total vasospasm and symptomatic vasospasm.

The Cochran-Armitage trend test (table 3) showed that the proportion of patients with age ≤ 50 years, Hunt-Hess grade 4 or 5, Fisher Group 3 SAH, poor clinical status at discharge, and death increased significantly across the three categories of worsening vasospasm. The proportion of patients using calcium channel blockers, ACE inhibitors/AT1 receptor blockers, beta-receptor blockers, non-SSRI vasoactive antidepressants, and aspirin did not show any significant trend across the vasospasm categories. However, the trend toward worsening vasospasm increased for SSRI users (p = 0.019), and tended to increase for statin users (p = 0.05). Among statin users, the risk for worsening vasospasm tended to be higher (p = 0.054) if statins were discontinued after admission.

Table 4 shows results of the multivariate logistic regression analyses. The model included recognized risk factors for vasospasm, and medication classes identified as potential risk factors from the univariate analysis. Use of SSRIs prior to SAH increased the incidence of symptomatic vasospasm from 27.6% to 45%, and of all vasospasm from 61% to 75%. SSRI use proved to be an independent predictor of symptomatic vasospasm, and tended to increase the risk for all vasospasm. Use of statins prior to SAH increased the incidence of symptomatic vasospasm from 28.2% to 38.9%, and of all vasospasm from 61% to 78%. Statin use increased the risk for all vasospasm, but not symptomatic vasospasm. In a second multivariate analysis where statin exposure was replaced with statin discontinuation, statin discontinuation did not increase the risk for all vasospasm (OR 2.54 [0.78 to 8.28], p = 0.12) or symptomatic vasospasm (OR 1.87 [0.68 to 5.11], p = 0.22). When the multivariate analysis was restricted to patients admitted within 3 days postictus, the effect of SSRIs on all vasospasm increased (OR 3.7, p = 0.01); however, there was no substantial change in the other results. Neither SSRI nor statin use increased the risk for poor clinical outcome or death.

Linear regression analysis showed that patients with Hunt-Hess scores 4 or 5 had longer NICU and hospital LOS (p < 0.0001), and patients with Fisher 3 SAH had longer hospital LOS (p = 0.03), however medication exposures had no effect on LOS.

Discussion.

The use of SSRIs and statins has increased steadily over the past decade. While these medications have a remarkable safety profile, the results of this retrospective cohort study indicate that SSRI and statin users have a higher risk for vasospasm after aneurysmal SAH, and that SSRI use increases the risk for developing complications from vasospasm (i.e., symptomatic vasospasm). The Cochran-Armitage test results (see table 3) suggest that both statins and SSRIs increase vasospasm severity. However, calcium channel blockers, beta-receptor blockers, ACE inhibitors/AT1 receptor blockers, and aspirin do not appear to affect the risk or the severity of vasospasm. All patients received nimodipine, a calcium channel blocker routinely administered for vasospasm prophylaxis,¹⁷ which might have masked any potential protective effects of the preadmission calcium channel blocker. Consistent with the results of previous studies, Hunt-Hess grade 4 or 5 and Fisher Group 3 SAH proved to be powerful predictors of vasospasm, poor clinical outcome, and death, and older patients had a lower risk for vasospasm but a higher risk for poor clinical outcome and death.^2,3,5,23

This is the first study showing that prescription medications can alter the risk for SAH-related vasospasm. While there appears to be a higher risk for vasospasm after SAH among SSRI and statin users, there is no evidence that these medications increase the risk for SAH. Previous studies found no evidence of an increased risk for hemorrhagic or ischemic stroke with SSRIs.^24,25 Statins, in fact, have been shown to decrease stroke risk and severity.^26–29 We emphasize that our results should not discourage the use of these relatively safe medications. Rather, awareness of these new vasospasm risk factors should increase vigilance for vasospasm-related complications in patients using these medications.

Because abrupt withdrawal may lead to deleterious rebound effects, it is debatable whether these medications should be discontinued after SAH. Accumulating experimental evidence suggests that abrupt withdrawal of statins suppresses endothelial nitric oxide release, increases oxidative stress, and eliminates the protective cerebrovascular effects of statins.^30–32 Statin withdrawal leads to a rise in C-reactive protein within 2 days,³³ and in patients with acute coronary syndrome, increases the risk for myocardial infarction, congestive heart failure, cardiac arrhythmia, shock, and death.^12,13 In our study, the risk for vasospasm after statin cessation did not reach significance, perhaps due to the small number of patients (3.7%) in whom statins were discontinued. Further studies are needed to determine whether the elevated risk for SAH-related vasospasm with statin exposure is accounted for by statin withdrawal, or related to the underlying disease indication (hypercholesterolemia, atherosclerosis) that prompted therapy. Similarly, with SSRIs and other vasoactive medications, future studies should aim to determine whether factors such as the underlying disease indication, pharmacokinetic properties, drug withdrawal, and the duration of effects on neurotransmitters, vascular receptors, or endothelial function influence the risk for vasospasm.

SAH-related vasospasm is a poorly understood phenomenon. Multiple factors have been implicated, including endothelin, serotonin, calcium, oxyhemoglobin, and phosphokinase-C.³⁴ An analysis of the mechanism of action of vasospasm-promoting medications can provide important insights into the pathophysiology of vasospasm. The present study showed that SSRIs—which enhance neuronal release of serotonin—are independent predictors of vasospasm. We have previously reported that SSRIs, when combined with other serotonergic medications, can precipitate cerebral vasoconstriction and stroke in patients without SAH.³⁵ Together, our observations suggest that serotonergic mechanisms are central to the pathophysiology of vasospasm. Non-SSRI vasoactive antidepressants (that predominantly act via sympathomimetic pathways) did not increase the risk for vasospasm; however, data interpretation is limited due to the small number of exposures to these medications. The increased risk with statins, and perhaps with statin discontinuation, suggests an important role for nitric oxide pathways in vasospasm.

Vasospasm, by inducing cerebral ischemia, worsens clinical outcome.¹ However, in this study, although SSRI and statin exposure increased the risk for vasospasm, use of these medications did not predict poor outcome. These results are similar to those of a previous study where use of cocaine was found to be an independent predictor of vasospasm, but not poor clinical outcome.⁹ The small numbers of exposed patients, the overwhelming effects of Hunt-Hess grade 4/5 and Fisher Group 3 on SAH outcome, and use of relatively insensitive outcome measures such as mRS scores are the most likely explanations for these discordant results.

Our study has several strengths: there was little missing data (see table 1) and 90% of patients with symptomatic vasospasm were diagnosed by angiography. Because our center has high SAH patient volumes with established management protocols, there was consistency in patient care, testing procedures, and diagnosis of vasospasm. However, because of the retrospective study design, we had limited knowledge about exposure to nonprescription drugs such as cocaine (which increase the risk for vasospasm^9,10), and no knowledge of the duration of medication exposure, which can theoretically increase or decrease the risk for vasospasm by up- or downregulating receptors involved with vasospasm. It is also possible that ICD-9 miscoding resulted in our missing a few SAH patients. By excluding patients who died before day 4, we were unable to assess the early effects of medications. In this study we restricted our investigation of medication withdrawal to patients taking statins, because the vascular effects of statin withdrawal are known to occur within 2 to 4 days,^30–33 and because of the emerging data concerning poor outcomes after withdrawal of this class of medications.^13,31,32 However, withdrawal of other medications may pose similar risks. Prospective multicenter studies, with reliably elicited medication history, are warranted not only to validate our results, but also to investigate for differential effects within a given class of medications and to ascertain effects related to dose and treatment duration.