Statins and subarachnoid hemorrhage in Medicare patients with unruptured cerebral aneurysms

Abstract

Background

Statins have been shown to decrease aneurysm progression and rupture in two experimental settings: animals with cerebral aneurysm and humans with abdominal aortic aneurysms.

Aims

To investigate statin use and outcomes in humans with unruptured cerebral aneurysms through Medicare administrative data.

Methods

We used a 40% random sample Medicare denominator file and corresponding inpatient, outpatient (2003–2011) and prescription (2006–2011) claims to conduct a retrospective cohort study of patients diagnosed with unruptured cerebral aneurysms, between 2003 and 2011. We used propensity score-adjusted models to investigate the association between statin use and risk of subarachnoid hemorrhage (SAH). Secondary analyses repeated the main models stratified on tobacco use status and separately assessed other composite outcomes.

Results

We identified 28,931 patients with unruptured cerebral aneurysms (average age 72.0 years, 72.6% female); mean follow up was 30.0 months; 41.3% used statins. Overall, 593 patients developed SAH and 703 underwent treatment before SAH. Current recent statin use was not associated with a difference in SAH risk (OR, 1.03; 95% CI 0.86–1.23); models stratified on tobacco use status were nearly identical. No association was observed between statin use and the composite outcome of SAH or aneurysm treatment (OR, 0.94; 95% CI, 0.84–1.06). The risk of SAH or out-of-hospital death was lower among statin users (OR, 0.69; 95% CI, 0.64–0.74).

Conclusions

Statin use by patients with unruptured cerebral aneurysms was not associated with SAH risk. Given the prior animal experimental studies demonstrating a protective effect, further prospective studies are needed to investigate the potential relationship.

INTRODUCTION

The one-month case-fatality for patients hospitalized with subarachnoid hemorrhage (SAH) ranges from 15% to 30%; more than one third of those who survive have major neurological deficits,(1, 2) and approximately 12% of patients die before receiving medical attention.(3–5) Most cerebral aneurysms are discovered incidentally on an imaging study of the brain done for unrelated indications. Options to prevent rupture are surgical clipping and endovascular coiling.(4, 6, 7) These treatments are invasive and carry significant risks.(4, 6–8) Many patients elect to have their aneurysms monitored and undergo treatment only with evidence of progression.(4, 6–8) Currently no medical treatment is known to prevent cerebral aneurysm progression or rupture in patients being monitored.

The etiology of cerebral aneurysm rupture remains poorly understood. Aneurysm size and location,(4, 7–10) as well as smoking history(4, 8, 10–13) have been associated with an increased risk of SAH. Several studies of human aneurysm tissue have demonstrated a role of inflammatory mediators in the development and rupture of cerebral aneurysms.(14–18) In experimental rodent aneurysm models, activation of nuclear factor kB (NF-kB) in the aneurysm wall has been suggested as an early signal of the inflammatory cascade resulting in aneurysm progression.(19, 20)

In addition to lowering cholesterol, statins (HMG-CoA reductase inhibitors) have been shown to confer general vascular protective effects, by inhibiting several inflammatory pathways. In experimentally induced rodent cerebral aneurysms statin administration for periods as short as 3 to 12 weeks has been associated with decreased expression of macrophage chemo-attractants, and vascular remodeling factors,(21–24) preventing aneurysm progression and SAH.(23–25) A similar effect of statins has been seen in human subjects with abdominal aortic aneurysms (AAA).(26–31) In AAA patients statin use is associated with slower progression and lower rupture rate.(26–31) The data for statin use in patients with cerebral aneurysms is less clear.(32, 33) We performed an observational study of Medicare patients with unruptured cerebral aneurysms, to investigate the potential protective effect of statin use in this context.

METHODS

Data source

We used a cohort of Medicare patients as the basis for our analysis. Medicare is a national social insurance program, administered by the U.S. federal government since 1966. Medicare provides health insurance for Americans aged 65 and older, who have worked and paid into the system. It also provides health insurance to younger patients, who are disabled. Medicare has 4 parts. Part A covers inpatient stays. Part B helps pay for some services and products not covered by Part A, generally on an outpatient basis. Part C is additional to other coverage and provides advantage plans. Part D took effect in 2006 and covers prescription drugs. Medicare patients over 65 years old are representative of the entire elderly US population. However, younger patients are only eligible for Medicare if they are disabled. We do not expect that disability would confound the association of our primary outcome (SAH) or exposure (statin use). In sensitivity analysis our results were identical for patients above or below 65 years of age.

Cohort

This study was approved by the Dartmouth Committee for Protection of Human Subjects. The data was anonymized and de-identified prior to use and therefore no informed consent was required. We used a 40% Medicare random sample Denominator file and corresponding Medicare inpatient and outpatient claims, Parts A and B, 2003–2011 (MedPAR, Carrier and Outpatient files) to identify patients with unruptured cerebral aneurysm diagnosis. Aneurysm patients were identified based on one or more inpatient or outpatient diagnoses (International Classification of Diseases, Ninth Revision ((ICD-9) diagnosis code 437.3) between May 1^st 2006 and June 30^th 2011. For cohort inclusion, patients were required to be (1) continuously enrolled in fee-for service Medicare Parts A, and B for 24 months before index diagnosis and at least 3 months after index diagnosis, (2) continuously enrolled in a stand-alone (non-managed care) Medicare Part D Prescription plan for at least 4 months before and 3 months after the index diagnosis, and (2) be age 40 or older at the time of index diagnosis.

Exposure

We used 2006–2011 Medicare Part D Prescription Event (PDE) files to identify prescription use among cohort members. A statin use category was assigned for each patient for each month of observation (i.e., longitudinally) based on fill date and days supply dispensed. Exposure was assumed to begin on the date of dispensing; overlapping supplies were carried over to the following month (e.g. a 90 day supply filled 80 days after a previous 90 day supply resulted in a 10 day supply carried forward). Patients were considered to have stopped statin use if no repeat fill occurred within a time interval 20% longer than the days supply on the last fill (e.g. no repeat fill within 108 of a 90 day fill was deemed discontinuation 108 days after last observed fill). Statin product (generic name) and dose were recorded for each month of observation. Statin exposure for each person month was subsequently categorized dichotomously as current/recent (fill covering the current month or some portion of the past 6 months) or none/remote (no use or no use for the past 6 months).

Our statin exposure measurement depends on prescription fill records in the claims data. Prescription fills have been shown to be a reliable measure of medication use.(34–36) Supplemental Table I provides descriptive statistics on statin exposure of individuals for the duration of the observation period.

Outcome

Because initial diagnosis of aneurysm is commonly followed by near-term treatment, in this assessment of statin-associated aneurysm outcomes, patients with fewer than 90 days follow-up were excluded to achieve a cohort of patients likely under aneurysm monitoring. The primary outcome of interest was SAH due to aneurysm rupture. For main analyses we defined this as a hospital admission with a discharge diagnosis of subarachnoid hemorrhage (ICD-9 of 430, 431, or 432). Coding for subarachnoid hemorrhage in claims data has been validated through medical record review by others.(37, 38) Patients with arteriovenous malformations (ICD-9 747.81) or trauma (ICD-9 800.0–801.9, 803.0–804.9, 850.0–854.1, and 873.0–873.9), coded during the index admission for SAH or during the 24-month look-back, were not included in the cohort. Inpatient deaths (discharge disposition of “death”) with a discharge diagnosis of SAH were treated as SAH events. Multiple secondary outcomes were examined: (1) SAH or out-of-hospital death (defined as death, from Denominator file, without a corresponding hospital admission), and (2) SAH or surgical treatment of aneurysm, defined as coiling [ICD-9-CM procedure code 39.72, 39.79, or 39.52 and 88.41 during the same hospitalization, but no documentation of code 39.51 (aneurysm clipping)] or clipping (ICD-9-CM procedure code 39.51). These were considered together as indicators of aneurysm growth/progression. Secondary analyses also explored these distinct outcomes: (1) SAH or out-of-hospital death or surgical treatment and (2) surgical intervention only, and (3) death only. A sensitivity analysis included a more conservative definition of SAH, using only the single most specific diagnosis code (ICD-9 430).

Covariates

Covariates included age (at time of index diagnosis), gender, and race/ethnicity (Black race, Hispanic ethnicity, or other from Medicare Denominator file), and Part D low-income subsidy status (used as a dichotomous poverty indicator). The following comorbidities (Supplemental Table II), diagnosed once or more at any time in the 24-month look-back were also obtained: tobacco exposure (defined as tobacco use ICD-9 diagnosis codes and/or chronic obstructive lung disease ICD-9 diagnosis codes), obesity, hypertension, and Charlson comorbidities.(39) Because medication adherence itself can identify patients with a higher tendency to seek health care and preventive care, and because use of potentially cardio-protective medications other than statins may affect risk of death (a secondary outcome), we ascertained from the PDE file use of select cardiovascular medications (beta blockers, and calcium channel blockers) as well as use of medications which could, theoretically affect the inflammation-mediated process of aneurysm progression and rupture (selective and non-selective non-steroidal anti-inflammatory drugs).

Statistical analysis

The primary analysis modeled the dependence of time to SAH on current statin use (a time-dependent exposure) adjusted for the covariates listed above, using logistic regression. We followed a cohort study design. The unit of analysis was a person month, in which the dependent variable is a SAH event, and the exposure is statin use status. Models also included the covariates listed above. We used logistic regression, which in this approach is equivalent to Cox proportional hazards model for time-dependent covariates, as the unit interval (month) gets smaller. Because there are many indications for statin use, including vascular disease, associated with the risk of death (one of our outcomes), we used logistic regression to generate individual-level baseline statin use propensity scores and included decile of this propensity score as a categorical variable in the models (e.g. adjusted for propensity score). This method matched our patients on propensity scores in the sense that it compared risk between individuals within the same decile of propensity score, and furthermore within each decile we controlled for possible confounders using regression. One advantage of our approach is that it does not arbitrarily exclude information, as do direct matching algorithms. Propensity score calculations included baseline exposure to: beta blockers, calcium channel blockers, prescription NSAIDs, clopidogrel, or glucocorticoids, and baseline status of the following risk factors: asthma, cancer, dementia, diabetes, hyperlipidemia, acute myocardial infarction, peripheral vascular disease, renal insufficiency, and stroke. Baseline characteristics were identified in the time immediately before the start of the observational period.

Covariates included directly in all models were limited to key risks and variables of interest (age, gender, race Part D low-income subsidy, baseline hypertension, and tobacco exposure, and current use of NSAID, calcium channel blocker, and beta-blockers) and baseline statin use propensity score decile. Tobacco use is the key risk factor for aneurysm formation and rupture. For this reason we expected the effect of statins might be distinct in patients with ongoing or past tobacco use compared to those without this key risk; all main analyses were thus repeated stratified on our tobacco exposure variable (tobacco use diagnosis or chronic obstructive pulmonary disease (COPD)). We also stratified the models on age < 65 and ≥ 65, assuming competing risks for death would differ in these two age groups and could thus alter the effect of statins. After applying the Bonferroni correction to control for the effect of multiple comparisons, the adjusted level of significance was set at 0.0125. All the values are the result of two sided tests. SAS version 10 (SAS Institute, Cary, NC) and STATA version 13 (StataCorp, College Station, TX) were used for statistical analysis.

With 28,931 patients observed for a mean follow up of 30 months, experiencing 593 total events, and some statin use detected in about 40%, we calculated a minimum detectable risk ratio (with 80% power) of about 0.75 overall, and about 0.60 for tobacco users.

RESULTS

Cohort

Of the 33,694 patients with a diagnosis of cerebral aneurysm, 4,763 (14.1%) developed events within 90 day of the diagnosis, and therefore were not included in analyses. Overall 28,931 patients were followed over time (Figure 1); average age was 72.0 years (standard deviation (SD), 11.5); 72.6% were female; 41.3% filled a statin prescription in the 4 months preceding index diagnosis. Mean follow up time was 30 months (SD). Compared to those without tobacco exposure, patients with evidence of tobacco exposure on average were younger, had more comorbidities, and were more likely to use cardiac medications. The entire cohort had a high burden of atherosclerotic disease (Table 1). Supplemental Table III and IV demonstrate patient characteristics stratified on baseline (at the beginning of the study) statin use.

Figure 1.

Study flow diagram

Table 1.

Patient characteristics overall and stratified on tobacco exposure

	TOTAL		Tobacco exposed***		Non-Tobacco exposed***
	N	%	N	%	N	%
N	28,931	100.0	8,861	30.6	20,070	69.4
	Mean	SD	Mean	SD	Mean	SD
Follow up (months)	30.0	18.1	28.0	17.9	31.0	18.2
Age (years)	72.0	11.5	70.0	11.7	73.0	11.3
Age groups*	N	%	N	%	N	%
40–54	2,863	9.9	1,152	13.0	1,711	8.5
55–64	3,084	10.7	1,272	14.4	1,812	9.0
65–74	9,589	33.1	3,004	33.9	6,585	32.8
75–79	5,368	18.6	1,483	16.7	3,885	19.4
80–84	4,278	14.8	1,132	12.8	3,146	15.7
≥85	3,749	13.0	818	9.2	2,931	14.6
Female	20,991	72.6	6,163	69.6	14,828	73.9
Race/Ethnicity
Black race	3,430	11.9	1,089	12.3	2,341	11.7
Hispanic ethnicity	2,028	7.0	477	5.4	1,551	7.7
Other	23,473	81.1	7,295	82.3	16,178	80.6
Part D low income subsidy	14,035	48.5	5,271	59.5	8,764	43.7
Baseline medication use
Statin	11,956	41.3	3,760	42.4	8,196	40.8
Beta blocker	10,970	37.9	3,596	40.6	7,374	36.7
Calcium channel blocker	8,574	29.6	2,744	31.0	5,830	29.0
NSAID	3,596	12.4	1,209	13.6	2,387	11.9
Charlson Comorbidity Count**
0–1	17,089	59.1	3,782	42.7	13,307	66.3
2–3	9,234	31.9	3,585	40.5	5,649	28.1
≥4	2,608	9.0	1,494	16.9	1,114	5.6
Diabetes Mellitus	7,859	27.2	2,729	30.8	5,130	25.6
Hypertension	22,557	78.0	7,426	83.8	15,131	75.4
Hyperlipidemia	15,787	54.6	5,169	58.3	10,618	52.9
Prior Myocardial infarction	1,074	3.7	591	6.7	483	2.4
Prior stroke	8,562	29.6	3,140	35.4	5,422	27.0

SD: standard deviation, NSAID: Prescription Non Steroidal Anti Inflammatory Drugs, Part D low-income subsidy: Part D low- income subsidy is dichotomized as any/none; this is an indicator of poverty (< 150% of federal poverty level), baseline medication use: use of a medication in the 4 months immediately preceding index diagnosis

^*Age is at index diagnosis

^**Charlson Comorbidity Count: Charlson et al.. J Chronic Dis. 1987;40:373–383

^***Differences between tobacco-exposed (defined as tobacco use ICD-9 diagnosis codes and/or chronic obstructive lung disease ICD-9 diagnosis codes) and non-tobacco exposed were significant (P < 0.05) by Chi-square test or t-test as appropriate, unless otherwise stated

Exposure

Overall, 61.7% of cohort members used some statin after the index diagnosis and the mean exposure was 24.9 months. The most common statins used were simvastatin, at a mean daily dose of 32.8 mg, and atorvastatin, at a mean daily dose of 25.6 mg (Supplemental table I).

Primary outcome

We identified (Table 2) 593 cases of SAH (0.81% risk of rupture per year). In propensity score adjusted logistic regression models (Table 3, Supplemental Table III), compared to remote or no use, current or recent statin use was not associated with risk of SAH (OR, 1.03; 95% CI, 0.86–1.23). In models stratified on tobacco status, risk of SAH risk was not associated with statin use among tobacco exposed (OR, 1.11; 95% CI, 0.80–1.55), or non-tobacco exposed (OR, 1.00; 95% CI, 0.81–1.24). Beta-blocker use (Table 3) was associated with an increased risk of SAH (OR, 1.52; 95% CI, 1.24–1.85); this persisted after stratification for tobacco status. Analyses stratified by age group (< 65 and ≥ 65 years) produced estimates similar to the main model (Supplemental Table IV). Sensitivity analysis (Supplemental Table III), defining SAH most precisely as ICD-9 code 430, resulted in nearly identical estimates overall (OR, 1.08; 95% CI, 0.78–1.48), in tobacco-exposed stratum (OR, 1.96; 95% CI, 1.05–3.67, P = 0.04, not significant after applying multiple comparisons correction), and in the non-tobacco exposed stratum (OR, 0.86; 95% CI, 0.59–1.27).

Table 2.

Crude primary and secondary outcomes overall and stratified on tobacco exposure

		TOTAL		Tobacco exposed		Non-Tobacco exposed
		N	Rate	N	Rate	N	Rate
Subarachnoid hemorrhage		593	8.1	189	9.0	404	7.7
Neurosurgical intervention
Aneurysm coiling		491	6.7	183	8.7	308	5.9
Aneurysm clipping		212	2.9	84	4.0	128	2.5
Out-of-hospital death		3,544	48.4	1,422	67.6	2,122	40.6
In-hospital death*		1,963	26.8	887	42.2	1,076	20.6

SAH: subarachnoid hemorrhage

^*In-hospital death: not related to SAH

Rate = events per 1,000 patient-years

Table 3.

Propensity score adjusted multivariable logistic regression models examining the association between statin use and primary and secondary outcomes, overall and stratified on tobacco exposure

Outcome		TOTAL				Tobacco exposed				Non-Tobacco exposed
Main Models		OR	LL95	UL95	P-value	OR	LL95	UL95	P-value	OR	LL95	UL95	P-value
SAH	Statin	1.03	0.86	1.23	0.730	1.11	0.80	1.55	0.516	1.00	0.81	1.24	0.982
	NSAID	0.84	0.62	1.12	0.234	0.57	0.32	1.04	0.067	0.98	0.69	1.37	0.887
	Calcium channel blocker	1.11	0.91	1.34	0.302	1.17	0.83	1.66	0.375	1.07	0.85	1.36	0.545
	Beta blocker	1.52	1.24	1.85	<0.001	1.49	1.03	2.15	0.036	1.52	1.19	1.94	0.001
	Female	0.82	0.68	0.98	0.029	0.78	0.58	1.06	0.117	0.83	0.66	1.03	0.093
	Black race*	1.29	1.00	1.67	0.050	1.25	0.79	1.98	0.338	1.31	0.96	1.79	0.092
	Hispanic ethnicity*	1.34	0.98	1.82	0.068	1.07	0.55	2.08	0.835	1.38	0.97	1.98	0.076
	Hypertension	1.01	0.81	1.26	0.911	0.94	0.61	1.43	0.763	1.04	0.81	1.34	0.756
	Tobacco exposure	1.20	1.00	1.44	0.045
Secondary Models
SAH or Treatment	Statin	0.94	0.84	1.06	0.345	0.96	0.78	1.18	0.704	0.94	0.81	1.09	0.396
	NSAID	1.00	0.84	1.19	0.977	1.07	0.81	1.41	0.644	0.94	0.75	1.19	0.624
	Calcium channel blocker	1.06	0.93	1.21	0.383	1.18	0.95	1.48	0.132	1.00	0.85	1.18	0.998
	Beta blocker	1.08	0.94	1.24	0.292	1.16	0.91	1.48	0.241	1.03	0.86	1.23	0.770
	Female	1.07	0.95	1.22	0.276	1.04	0.85	1.28	0.692	1.09	0.93	1.28	0.307
	Black race*	1.06	0.88	1.26	0.552	0.92	0.68	1.26	0.621	1.15	0.92	1.43	0.225
	Hispanic ethnicity*	1.12	0.90	1.40	0.302	1.24	0.84	1.83	0.278	1.08	0.83	1.41	0.571
	Hypertension	0.94	0.81	1.08	0.384	0.85	0.65	1.10	0.214	0.97	0.82	1.16	0.762
	Tobacco exposure	1.31	1.16	1.47	<0.001
SAH or Out-of-Hospital Death	Statin	0.69	0.64	0.74	<0.001	0.71	0.63	0.80	<0.001	0.67	0.61	0.74	<0.001
	NSAID	0.79	0.70	0.89	<0.001	0.74	0.61	0.90	0.002	0.82	0.70	0.95	0.010
	Calcium channel blocker	0.98	0.91	1.06	0.679	1.04	0.92	1.17	0.527	0.94	0.86	1.04	0.219
	Beta blocker	1.40	1.30	1.51	<0.001	1.43	1.26	1.62	<0.001	1.38	1.25	1.52	<0.001
	Female	0.73	0.68	0.78	<0.001	0.83	0.75	0.93	0.001	0.67	0.61	0.73	<0.001
	Black race*	0.92	0.83	1.03	0.135	0.84	0.70	0.99	0.043	0.98	0.86	1.12	0.777
	Hispanic ethnicity*	0.66	0.57	0.76	<0.001	0.63	0.49	0.82	0.001	0.69	0.58	0.82	<0.001
	Hypertension	1.10	1.01	1.19	0.032	1.16	1.00	1.34	0.051	1.05	0.95	1.17	0.334
	Tobacco exposure	1.68	1.57	1.79	<0.001

LL95: lower limit of 95% Confidence Interval, UL95: upper limit of 95% Confidence Interval, SAH: subarachnoid hemorrhage, NSAID: Prescription Non Steroidal Anti Inflammatory Drugs, treatment: aneurysm coiling or clipping, medication: current/recent use (fill covering the current month or some portion of the past 6 months) with none/remote (no use or no use for the past 6 months)

^*White is reference category for race/ethnicity. Race/Ethnicity is based on the Medicare denominator file Research Triangle Institute variable Multivariable logistic regression models also included the following variables: Part D low-income subsidy, age, deciles of propensity score for statin use Factors used for propensity score calculation: baseline exposure to beta blockers, calcium channel blockers, prescription NSAIDs, clopidogrel, or glucocorticoids, and baseline status of the following risk factors: asthma, cancer, dementia, diabetes, hyperlipidemia, acute myocardial infarction, peripheral vascular disease, renal insufficiency, and stroke

The level of significance was set at 0.0125 after using Bonferroni correction for multiple comparisons

Secondary outcomes

During the study period, 703 patients underwent neurosurgical treatment of an unruptured aneurysm (491 coiling, and 212 clipping) 90 or more days after index diagnosis (Table 2). In addition, 3,544 out-of-hospital deaths occurred, with unknown underlying cause. Overall 1,963 in-hospital deaths not attributable to SAH (based on discharge diagnoses) occurred as well. The secondary outcomes were modeled like the primary models, propensity score adjusted logistic regression (Table 3, Supplemental Table III). Statin use was not associated with risk of the composite outcome SAH or aneurysm treatment (OR, 0.94; 95% CI, 0.84–1.06). Stratification on tobacco use status produced similar results. In contrast, risk of the composite outcome of SAH or out-of-hospital death was significantly lower among statin exposed (OR, 0.69; 95% CI, 0.64–0.74). This result persisted after stratification on tobacco status; a protective effect was observed among tobacco exposed (OR, 0.71; 95% CI, 0.63–0.80) and non-tobacco exposed (OR, 0.67; 95% CI, 0.61–0.74).

Further sensitivity analyses included models with the following outcomes: SAH or out-of-hospital death or surgical treatment, surgical intervention only, and out-of-hospital death only. These demonstrated that out-of-hospital death events drove the protective effect of statins observed in models on the composite outcome SAH or out-of-hospital death. Statin use was not associated with a difference in odds of receiving surgical treatment. Further stratification of all the models by age group (< 65 and ≥ 65 years) yielded similar results (Supplemental Table IV).

DISCUSSION

In this observational study, we find no association between current or recent statin use and the risk of SAH among Medicare patients with unruptured cerebral aneurysms. The lack of association persisted after stratification on age and tobacco exposure status. The relationship between statin use and risk of SAH has been a matter of debate in recent years. Our observational study does not reinforce evidence from animal models and case control studies suggesting a SAH protective effect of statins.

Prior clinical investigations have demonstrated conflicting conclusions regarding the effect of statins on SAH risk. Risselada et al(32) in a case-control study of Dutch cerebral aneurysm patients demonstrated that statin withdrawal was associated with SAH, although there was no significant protective effect for new statin users. On the contrary, Yoshimura et al,(33) in a small case-control study of Japanese patients demonstrated a significantly lower rate of statin use in patients with SAH. An important limitation in the latter investigation, recognized by the authors, is the possibility that SAH cases might not have interacted with the medical system regularly, resulting in a lower baseline statin use. Overall, both studies differ in design from our analysis. They compare the prevalence of statin use (without matching or controlling for the propensity of statin use), among cases and controls. Our cohort study follows patients with unruptured aneurysms and assesses the association between statin use and SAH over time and therefore can discern temporal relationships.

Our results diverge from experimental findings in animals. Several investigators(21–25) have demonstrated that continuous statin use in rodent cerebral aneurysm models was associated with a decreased risk of aneurysm formation, progression, and rupture. Although the presence of publication bias in the animal studies cannot be excluded, the result discordance warrants further investigation in future prospective studies.

Differences in pathophysiology may explain the divergence of our results from what has been observed in most studies of abdominal aortic aneurysms (AAA) in human subjects. Although many studies have shown a protective effect of statin use on the growth rate of small AAAs,(26–29) some authors have seen a less clear impact.(40) Meta-analyses of AAA study data have suggested a protective effect of statins.(30, 31) AAAs are atherosclerotic in nature, and their development and progression have been linked to common cardiovascular risk factors.(27) Cerebral aneurysms are different; hyperlipidemia and diabetes have not been associated with cerebral aneurysm formation or rupture.(4, 7, 8) In addition, unlike AAAs, the histology of cerebral aneurysms rarely involves the formation of atheromatous plaques. Instead, it is likely a structural deficiency of the vascular bed promotes the formation of cerebral aneurysms, in areas of intracranial arteries with turbulent blood flow.(27)

We investigated the association of other factors with SAH. Our variable for tobacco exposure was associated with an increased risk of SAH, as previously observed by others.(7, 8) However, no significant differential effect of statin use was observed after stratification on this variable. Higher age demonstrated a negative association with SAH, possibly secondary to a survivorship bias. This can lead to optimistic estimates, because the patients who have died cannot be examined. We did not observe an association between hypertension and risk of SAH. This issue has been a subject of controversy. Although hypertension has been associated with higher SAH risk in population studies,(11–13) the large natural history studies failed to demonstrate a relationship.(7, 8, 41) A meta-analysis by Greving et al,(9) has shown a marginal association between hypertension and SAH risk, but a more recent high-quality study(10) of lifelong risk did not support that. The effect of hypertension on aneurysm rupture deserves further investigation.

We additionally controlled for the effect of other prescription medications. Use of beta-blockers was associated with a higher incidence of SAH. This category of drugs is a marker of cardiovascular pathology (which we also observed in our data), and risk factors such as smoking. It is likely that this relationship, rather than a primary effect of beta-blockers on the aneurysms, contributed to the higher incidence of SAH in this subgroup. Prescription NSAID use was not associated with decreased risk of SAH. The inability to control for over-the-counter NSAID use (such as aspirin) limits the interpretation of this observation. However, the relationship between NSAID use and SAH warrants further prospective investigation, in light of a recent study(42) demonstrating a protective effect of aspirin against SAH for similar patients.

We observed a protective effect of statin use on overall mortality in our cohort. This was expected. The study population had high prevalence of atherosclerotic disease. Our findings likely reflect the protective effect of statins on general cardiovascular mortality. Statin use and adherence may also be a marker of other unmeasured health protecting behaviors or traits.

Our study has several limitations common to administrative databases. First, this is a retrospectively created and analyzed cohort, and the study is subject to limitations common to this research approach including indication bias and residual confounding. Second, our measure of tobacco use and past tobacco exposure depends on diagnosis codes. While clinicians are improving their tendency to explicitly diagnose tobacco use, our ascertainment of this important risk factor is likely incomplete. Similarly, the diagnosis of COPD may be incomplete and when made, likely reflects a broad range of disease and past/current tobacco use. While this variable was associated with substantially higher risk of SAH in our data, as expected, the inability to precisely measure tobacco use status remains an important limitation of this study.

Third, patients who initiate and adhere to statin therapy may be distinct from patients who do not. If medication use were a general indicator of healthier status, the bias introduced by statin use would be expected to result in an exaggerated protective effect; we found no effect. Additionally beta blocker use and calcium channel blocker use, other indicators of medication use and adherence tendency were not associated with lower risk of SAH. Fourth, coding inaccuracies can affect our ascertainment of events. However, coding for subarachnoid hemorrhage has been validated through medical record review, in prior studies.(37, 38) These analyses of randomly selected patients from the medical records of several hospitals and hospital types (academic and community centers in one study,(37) and all hospitals of the Seattle area(38) in the other) demonstrated sensitivity, specificity, and positive predictive value over 90% for the diagnosis of SAH through ICD-9 codes. Fifth, findings among this older, American population may not be generalizable to younger or otherwise dissimilar populations.

Sixth, we have no information on aneurysm size and location, which have been established as major predictors of aneurysm rupture. Although, we do not expect differential distribution of aneurysm characteristics between the statin users and non-statin users, this unmeasured risk of rupture is a limitation of our analysis. Lastly, our measure of statin use was derived from prescription fill records. Prescription fill records have been shown a good proxy measure of prescription use but poor adherence could result in misclassification of exposure and bias our findings toward the null.(34, 35) Our cohort used diverse and varying doses of statin products. We had insufficient statistical power to permit dose-specific or dose stratified sub-analyses. Such analyses should be pursued in future studies examining this question.

Conclusions

Promising experimental evidence from rodent models suggest the use of statins may decrease the risk of cerebral aneurysm progression and rupture. In our study, using propensity score adjusted regression models, current or recent use of statins by patients with unruptured cerebral aneurysms was not associated with decreased risk of SAH. These findings highlight the uncertain relationship between statin use and SAH. Further prospective trials are needed to clarify the association.