Withdrawal of Statins and Risk of Subarachnoid Hemorrhage
Background and Purpose—Vascular endothelium, which can be affected by statins, is believed to play a substantial role in subarachnoid hemorrhage (SAH). Our objective was to estimate the association between use and withdrawal of statins and the risk of SAH.
Methods—We conducted a population-based case–control study within the PHARMO database. A case was defined as a person hospitalized for SAH (ICD-9-CM code 430) in the period January 1, 1998 to December 31, 2006. Ten randomly chosen controls were matched to each case on age, gender, and calendar date.
Results—During the study period 1004 incident cases of SAH were identified. Current use of statins did not significantly decrease the risk of SAH (OR=0.77, 95% CI 0.55 to 1.07). The odds ratio for recent withdrawal compared to nonusers was 1.62 (95% CI 0.96 to 2.73). Compared to current use, recent withdrawal was associated with an increased risk of SAH (OR=2.34, 95% CI 1.35 to 4.05). Interaction analysis showed that the effect of statin withdrawal was highest in patients who had also recently stopped antihypertensive drugs (OR=6.77, 95% CI 2.10 to 21.8).
Conclusions—Current use of statins seems to lower the risk of SAH, although the reduction was not significant in new users. Statin withdrawal increased the risk of SAH by a factor 2, even more in patients who had also recently stopped their antihypertensive treatment.
- subarachnoid hemorrhage
- case–control study
- 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors
Subarachnoid hemorrhage (SAH) is a severe neurological disorder.1 Modifiable risk factors for SAH are hypertension, smoking, and excessive alcohol intake.1 Drugs may constitute another class of risk factors, but their role in SAH etiology has not been systematically evaluated. Considering the potential role of vascular endothelium in SAH,2 drugs affecting the endothelium might be expected to alter SAH risk. Statins affect the endothelium through increased nitric oxide bioavailability3 and have been shown to ameliorate SAH-induced vasospasm.4 Although statins decrease cerebrovascular event rates during use, withdrawal of statins causes a rebound effect leading to increases in vascular event rates beyond that of the absence of prevention alone.5 This observed rebound effect has been attributed to the inhibition of endothelium NOS (eNOS)-mediated pathways by statins.5 Withdrawal of statins is supposed to impair the eNOS-mediated tension regulation, which in turn may lead to short-term increased stress on the vessel wall and risk of SAH.
We hypothesized that statins influence the risk of SAH during use (decrease) and shortly after withdrawal (increase). To examine these hypotheses we conducted a population-based case–control study in the general Dutch population.
Patients and Methods
Data were obtained from PHARMO Record Linkage System (RLS), a population based system with dispensing data and hospitalizations regarding a well-enumerated population and follow-up. For a detailed description of the database we refer to previous work.6
The source population included all persons in the PHARMO RLS database who had at least 3 years of valid database history before the index date (see below). Persons were followed from inception or start of the study period (January 1, 1998) until transferral out of the database, death, diagnosis of SAH, or the end of the study period (December 31, 2006).
Cases and Controls
Study outcome was hospitalization for a first SAH, as recorded in the Dutch National Medical Registration (LMR) with ICD-9-CM code 430. To assess the positive predictive value (PPV) of the LMR information, discharge letters of 42 patients were evaluated.
For each case, 10 controls were obtained from the source population and matched to the case on year of birth, gender, and index date.
Drug Exposure and Covariates
The drugs of primary interest were statins. Use was classified as “no,” “current” (use at index date), “recent withdrawal” (stopping up to 30 days before index date), and “past” (stopping >30 days before index date). To inspect protopathic bias we looked at “recent starters” (current use less than 30 days), to limit prevalent user bias (healthy users) current use was restricted to short term use (< 3 years).
We considered diabetes mellitus, a history of vascular events, and hypertension as covariates. Use of insulin without a history of use of oral antidiabetic agents was considered a proxy for having diabetes mellitus type I. Use of oral antidiabetic agents either or not followed by use of insulin was considered a proxy for having diabetes mellitus type II. A history of vascular events was defined by use of 2 or more prescriptions of nitrates, or of either aspirin or clopidogrel or hospitalization for ischemic cardio- or cerebrovascular events (CVD). Past use of antihypertensives was considered a proxy for uncontrolled hypertension.
All drugs with potential effect on the eNOS pathway were considered as covariates (Table 1⇓). We also studied agents known for increased bleeding risk, as well as NSAIDs, coxibs, tricyclic antidepressants, benzodiazepines, antipsychotics, antiepileptic drugs, female sex hormones, and antibiotics.
Conditional logistic regression analysis was used for matched case–control data to estimate odds ratios (OR) and 95% confidence intervals (CI) for SAH associated with exposure to drugs. Analyses were performed with reference to “no use,” and subsequently to “current use” to better adjust for confounding by indication. In the adjusted model we included all factors that changed the effect estimate with more than 5%.
Sensitivity analyses were conducted on time intervals for recent withdrawal, information bias attributable to hospitalization (patients hospitalized in days before), and new users to avoid a healthy user bias. Effect modification was investigated for gender, a history of ischemic CVD, and use of antihypertensives.
Association measures for products with exposure in less than 3 cases or 3 controls are not reported.
Our study population comprised 1004 cases of SAH and 10 033 matched controls. The PPV of the discharge diagnosis was 88% (95% CI 75% to 95%). The PPV did not differ significantly (P=0.26) between statin-exposed and unexposed cases.
Past use of antihypertensives was associated with an increased risk of SAH (Table 1⇑). NSAIDs, coxibs, and benzodiazepines increased the risk; however, exclusion of recent starters weakened the effect, which points to protopathic bias. Current use of statins was associated with a reduction in SAH risk, but this effect was not statistically significant when restricting to new users (Table 2). Exclusion of patients hospitalized in the 30 days prior further attenuated the effect (OR=0.79, 95% CI 0.56 to 1.12).
Compared to no use, recent withdrawal of statins was associated with a nonsignificant increased risk of SAH (Table 2). Recent withdrawal of statins was associated with a 2-fold increase in risk when compared to current users of statins: ORadjusted=2.34 (95% CI 1.35 to 4.05) (Table 2), this remained when current users were restricted to new users (OR=2.23, 95% CI 1.24 to 4.02) (not shown). The effect of withdrawal was most outspoken after withdrawal of atorvastatin. Sensitivity analyses on different risk windows of withdrawal showed that the association measure is highest when withdrawal is closest to the index date, ie, persons who stopped more recently.
Stratification by gender did not reveal significant differences, although women seem to be more susceptible to statin withdrawal (Table 3). A history of CVD did not modify the association, although the effects of current use and withdrawal were stronger in patients with CVD (Table 3). Use of antihypertensives modified the effect of statins; in persons who stopped statins and antihypertensives the risk of SAH was more than 6-fold increased compared to current users of both statins and antihypertensives (Table 3).
In this study we showed that statins are associated with the risk of SAH. Whereas current use of statins may protect against SAH, withdrawal is associated with a clear increase in risk, especially if patients have also stopped antihypertensive drugs. The observed results support our hypothesis that statins have an effect on vascular endothelium. An important finding is the interaction between stopping of statins and having stopped use of antihypertensives, especially because statins and antihypertensives are often prescribed concomitantly and nonadherence frequently affects both types of drugs.7 The apparent protective effect during current use of statins requires confirmation in additional studies with richer clinical details as the results were susceptible to analyses aiming to evaluate healthy user bias.
Results of this study should be seen in the light of limitations inherent to observational studies with nonrandom treatment-assignment. Information bias may occur on the level of outcome or exposure. Despite errors in exposure assessment attributable to noncompliance, information bias is likely to be nondifferential because data were gathered prospectively without knowledge of the hypothesis studied and withdrawal effects were not observed for other non-CNS and nonendothelium affecting drugs.8,9 Misclassification of exposure because of hospitalization before the index date was ignorable.
Misclassification of the outcome was a concern in our study as there is no information on the overall PPV of the discharge diagnosis. We demonstrated that the PPV was quite good and unrelated to exposure. Because we could not validate all cases, outcome misclassification attributable to inclusion of false-positives is a fact but limited and nondifferential, resulting in an attenuation of the association. SAHs that were fatal before reaching the hospital were not included in our study because of the nature of the hospitalization database. Therefore the results cannot be generalized to these types of fatal SAH. Confounding by indication might play a role in studying relations between statins and SAH, especially when evaluating potential beneficial effects. For the withdrawal analysis we addressed confounding by indication by comparing withdrawers to current users instead of no users. For current use of statins we conducted several analyses to inspect healthy user effects: exclusion of prevalent users and stratification for primary/secondary prevention. Healthy user bias may explain at least part of the association between current use of statins and SAH. Hypertension is a known risk factor for SAH. In this study we did not have information on actual blood pressure; we took stopping antihypertensives as proxy. Recent studies show that nonadherence with antihypertensives is indeed associated with increased systolic blood pressure,10 which supports the assumption that patients who stopped have uncontrolled blood pressure.
The observation that the withdrawal effect is specific for statins, and does not occur in other drugs, encourages us to think of it as a true association; were the withdrawal effect attributable to reversed protopathic bias or immeasurable time bias, the ORs would have increased for other drugs.
In conclusion, we found that statins are associated with SAH, the risk increases substantially in patients who stop their statins, and this effect was particularly pronounced in patients who had also stopped antihypertensive drug use.
Sources of Funding
This study is part of the @neurIST-project (@neurIST, Integrated biomedical informatics for the management of cerebral aneurysms, www.aneurist.org), funded by the European Commission 6th Framework Programme. Financial support of the foundation Vereniging Trustfonds Erasmus Universiteit Rotterdam is gratefully acknowledged.
H.S.: Statistician at PHARMO Institute, Utrecht, the Netherlands (www.pharmo.nl). Clients are amongst others: universities, governmental agencies and pharmaceutical companies. A.v.d.L.: Member of Bayer Schering Pharma expert panel on radiological contrast media and investigator of an Investigator Sponsored Study on radiological contrast media of Bayer Schering Pharma. W.J.N.: Member of Scientific Advisory Board Cardialysis. R.M.C.H.: Director of PHARMO Institute, Utrecht, the Netherlands (www.pharmo.nl). Clients are amongst others: universities, governmental agencies and several pharmaceutical companies. M.C.J.M.S.: Scientific coordinator of the Integrated Primary Care Information (IPCI) group which is partially funded through unconditional research grants from the pharmaceutical industry, namely Pfizer, Merck, Astra Zeneca, Eli Lilly, GSK and Altana. She has been consultant to Pfizer, Novartis Consumer Health, Servier, Celgene and Lundbeck. Funding of PHARMO Institute takes place on project basis. Clients are amongst others: universities, governmental agencies, and pharmaceutical companies.
- Received March 16, 2009.
- Accepted May 5, 2009.
Wolfrum S, Jensen KS, Liao JK. Endothelium-dependent effects of statins. Arterioscler Thromb Vasc Biol. 2003; 23: 729–736.
Rosengarten B, Auch D, Kaps M. Effects of initiation and acute withdrawal of statins on the neurovascular coupling mechanism in healthy, normocholesterolemic humans. Stroke. 2007; 38: 3193–3197.
Herings RM, Bakker A, Stricker BH, Nap G. Pharmaco-morbidity linkage: a feasibility study comparing morbidity in two pharmacy based exposure cohorts. J Epidemiol Community Health. 1992; 46: 136–140.