Selective Serotonin Reuptake Inhibitors and the Risk of Stroke
A Population-Based Case-Control Study
Background and Purpose— Selective serotonin reuptake inhibitors (SSRIs) have been associated with increased risk of bleeding complications, possibly as a result of inhibition of platelet aggregation. Little is known about the risk of intracerebral hemorrhage in users of SSRIs and whether the effect on platelet aggregation reduces the risk of ischemic stroke. We used population-based data to estimate the risk of hemorrhagic and ischemic stroke in users of SSRIs.
Methods— We performed a nested case-control study in Funen County (465 000 inhabitants), Denmark. All patients with a first-ever stroke discharge diagnosis in the period of 1994 to 1999 were identified, and a validated diagnosis of stroke was reached in 4765 cases. In all, 40 000 controls were randomly selected from the background population. Information on drug use for cases and controls was retrieved from a prescription registry with full coverage of the county. Odds ratios were adjusted for age, sex, calendar year, and use of other medication. To evaluate the effect of various potential confounders not recorded in the register data, we performed separate analyses on data from 2 large population-based surveys with more detailed information on risk factors.
Results— Of 659 patients with hemorrhagic stroke, 21 were current users of SSRIs. The adjusted odds ratio of hemorrhagic stroke in current SSRI users compared with never users was 1.0 [95% confidence interval (CI), 0.6 to 1.6]. Of 2717 patients with ischemic stroke, 100 were current users of SSRIs, and the adjusted odds ratio of ischemic stroke in cases compared with controls was 1.1 (95% CI, 0.9 to 1.4). The survey data indicated that additional confounder control would not have led to an increase in the relative risk estimates.
Conclusions— Current exposure to SSRIs is not associated with increased risk of intracerebral hemorrhage and is probably not associated with a decreased risk of ischemic stroke.
Selective serotonin reuptake inhibitors (SSRIs) are widely used in the treatment of depression because of their tolerability, ease of dosing, and safety. However, treatment with SSRIs has been associated with bleeding complications,1 and it was recently shown that treatment with SSRIs increases the risk of upper gastrointestinal bleeding.2,3⇓ It is thus a major concern whether SSRIs also increase the risk of hemorrhagic stroke. A recent case-control study found no association between the use of SSRIs and the risk of intracranial hemorrhage (intracerebral, subarachnoidal, subdural, and unspecified intracranial hemorrhages). However, intracerebral hemorrhage, which is the most frequent type of intracranial hemorrhage, was identified in only 3 cases with current exposure to SSRIs.4
Serotonin is released from platelets to promote vasoconstriction and platelet aggregation in response to vascular tears.1 During treatment with SSRIs, the intraplatelet serotonin content is reduced as a result of inhibition of the serotonin reuptake mechanism.5 Bleeding complications during SSRI treatment could be due to an impaired hemostatic response. Conversely, it is conceivable that SSRIs might reduce the risk of thrombus formation, although the effect of SSRI exposure on the risk of ischemic stroke has not been previously studied. We conducted a population-based register study to estimate the risk of hemorrhagic and ischemic stroke in users of SSRIs and other antidepressants.
Subjects and Methods
We performed a nested case-control study based on information from a patient registry, a residence history registry, and a prescription registry in Funen County, Denmark (population, 465 000, 9% of the Danish population). The civil registration number, a unique and permanent identifier of all Danish residents, enabled simple and correct linkage across the registries.
We also analyzed data from 2 large population surveys, the Middle Aged Twins Survey and the Longitudinal Study of Ageing Danish Twins, which allowed us to study the association of use of SSRIs with various potential confounders not included in the nested case-control study.
Case Ascertainment and Validation
The Patient Registry contains information on all discharges from nonpsychiatric hospitals in Funen County since 1973. Recorded data include dates of admission and discharge, county of residency at the time of admission, and up to 20 discharge diagnoses coded according to the International Classification of Diseases (ICD), 8th revision for 1973 to 1993 and 10th revision for 1994 to 1999.
We identified all patients ≥20 years of age with a discharge diagnosis of hemorrhagic stroke (ICD 431 and DI61), ischemic stroke (432 to 434, DI63), subarachnoid hemorrhage (430, DI60) or unspecified stroke (436.01, 436.90, DI64, DG46) from January 1, 1994, to December 31, 1999, and who were free of such diagnoses during the period of January 1, 1973, to December 31, 1993. Only patients who were residents of Funen County at the time of discharge were included (n=5964).
The date of first admission for stroke was defined as the index date. Discharge records for all first stroke admissions were reviewed. For patients transferred to the neurology or neurosurgery department within 21 days after the index date, we also reviewed discharge records from these subsequent admissions. Information on use of medication was suppressed, and 3 specialists in neurology then evaluated the discharge records according to predefined criteria.
A diagnosis of intracerebral hemorrhage was reached if the patient had clinical signs of stroke according to the World Health Organization (WHO) definition6 and the neurovisualization description (CT scan or MRI scan) or autopsy findings were compatible with intracerebral hemorrhage. Cases of ischemic stroke were divided into definite ischemic stroke (clinical signs according to the WHO definition and neurovisualization description or autopsy showing a relevant cerebral infarction) and probable ischemic stroke (cases with appropriate clinical signs but no relevant lesion on neurovisualization or autopsy). A diagnosis of probable ischemic stroke was frequently reached because of early neurovisualization at which time many infarcts would not be visible.
We included only definite cases of subarachnoid hemorrhage defined as an appropriate clinical history and signs compatible with a diagnosis of subarachnoid hemorrhage from at least 1 of the following: CT or MRI scan, lumbar puncture (xanthochromia or frank bleeding), cerebral angiography, or autopsy.
In some cases, because of insufficient diagnostic workup, patients could not be classified according to one of the above diagnoses. Patients with clinical signs of stroke for whom neuroradiological examination or autopsy had not been performed were thus labeled clinical stroke cases. Patients with some clinical signs of stroke but inadequate information for fulfillment of the WHO definition were defined as possible cases of stroke.
Patients with a register diagnosis of stroke that did not fulfill any of the above criteria were excluded (n=307). Cases of stroke occurring during hospitalization (n=439) and subdural or epidural hematoma and head injuries (n=83) were also excluded. A further 370 cases were excluded for various reasons, mainly previous stroke. A total of 4765 cases were included in the case-control study.
The Funen Residence History Registry contains information on the civil registration number and residence history of all inhabitants of Funen County. Information on immigration and emigration from Funen County has been continuously updated by retrieval of data from a national register (Central Person Register). We identified all residents of Funen County during the period of 1994 to 1999 who were ≥20 years with no hospitalization for stroke between 1973 and 1993. These residents were allocated a random index date for the period of January 1, 1994, to December 31, 1999, and 40 000 controls with no stroke discharge diagnosis before their index date were chosen at random.
Information on reimbursed medicine in the County of Funen has been recorded in the Odense University Pharmacoepidemiological Database since October 1990.7 The coverage of the county by this database increased gradually and was complete by November 1992. For each prescription, the registry includes information on the civil registration number, the date the prescription was presented, and the package identification number, which enables identification of the brand, quantity, and form of the drug. The total package content of the drug is recorded as the number of defined daily doses (DDD).8 The DDD is established by an expert panel as the typical maintenance dose required when the drug is used for its main indication in an adult. Drugs used for the same indication are in principle equipotent when measured in DDD. All drugs are classified according to the anatomical therapeutical chemical (ATC) system.8 The indication for treatment and the prescribed dosing are not recorded.
We retrieved all available information from the prescription registry on the use of antidepressants and other drugs in cases and controls before the index date. Antidepressants were classified according to their action on the serotonin and norepinephrine reuptake mechanisms into 3 groups. The first group comprised SSRIs, including citalopram, fluoxetine, sertraline, paroxetine, clomipramine, and fluvoxamine. In the studies of de Abajo et al,2,4⇓ clomipramine was included in the SSRI group because of a rather selective effect on serotonin transport mechanisms. We followed the same strategy to enhance comparability with these previous studies. The second group comprised antidepressants with an inhibitory action on both serotonin and norepinephrine reuptake mechanisms. This group was made up of amitriptyline, imipramine, venlafaxine, lofepramide, and imipramine oxide. Antidepressants in the third group were characterized by either a selective inhibitory action on norepinephrine reuptake mechanisms or no effect on any reuptake mechanism. This third group comprised nortriptyline, mianserine, amoxapine, mirtazepine, opipramol, doxepin, maprotiline, dosulepin, trimipramine, reboxetine, desipramine, and protriptyline. Monoamine oxidase inhibitors were not included.
Each recorded prescription was assumed to last a number of days equivalent to the number of issued DDDs. We defined a person as a current user of antidepressants if the supply of the prescription ended after 30 days before the index date. Persons were defined as recent users if the supply of the prescription ended between 31 and 60 days before the index date and as past users if the supply ended before 61 days before the index date. Persons with no prescriptions of antidepressants before the index date were defined as never users.
Duration of use was defined by the interval between presentation of the first prescription and end of supply of the last prescription in a series of consecutive prescriptions. Prescriptions were regarded as consecutive when the supply of an antidepressant ended <7 days before presentation of a new prescription of an antidepressant from the same group of antidepressants. An estimate of the daily antidepressant dose in a series of consecutive prescriptions was calculated as the total number of issued DDDs divided by duration of use.
Age, sex, hypertension, diabetes, smoking, and other stroke risk factors are all possible confounders of the association between antidepressant use and stroke. The potential confounding effects of age, sex, and calendar time were accounted for through the statistical analysis. Information on diseases influencing the risk of stroke was not available from the registries. Instead, we retrieved information on the use of drugs, which can be regarded as proxy measures for certain risk factors. Furthermore, we used information on drugs that might increase the risk of stroke through their pharmacological action, eg, anticoagulants. The effects of these variables were also controlled for in the analysis.
We used prescriptions of diuretics (ATC code C03), β-blockers (C07), calcium channel blockers (C08), angiotensin-converting enzyme (ACE) inhibitors (C09), antiarrhythmics (C01A and C01B), and antianginal drugs (C01D) as proxy measures for hypertension, cardial arrhythmia, and ischemic heart disease. Antidiabetics (A10) and lipid-lowering drugs (C10A) were used as proxy measures for diabetes and hyperlipidemia, respectively. A person with at least 1 prescription of the drug before the index date was defined an ever user. A never user was a person with no recorded prescriptions.
Anticoagulants (B01AA03, B01AA04), low-dose acetyl salicylic acid (B01AC06, N02BA01), and other nonsteroidal antiinflammatory drugs (M01A) might increase the risk of stroke through their action on hemostasis. Current use and never use of each of these drugs were defined as for antidepressants. Past users were persons whose supply of the prescription ended >30 days before the index date.
Data From Twin Surveys
The data from the nested case-control study did not allow for sufficient control of a number of potential confounders. To estimate the magnitude and direction of the association between these potential confounders and use of antidepressants, we used data from 2 large population-based surveys conducted in 1998 to 1999, the Middle Aged Twins Survey9 and the Longitudinal Study of Ageing Danish Twins.10 Participants completed a comprehensive structured interview, and the response rates were high (83% for middle-aged and 70% for elderly twins). Using responses from both twins in a pair could, however, affect our results in an unpredictable way. Therefore, in twin pairs with participation of both twins, we randomly selected only 1 of the twins for the present study.
A total of 4556 participants were identified. One hundred forty-five were excluded because of a diagnosis of stroke in a national patient register recorded before the interview date. We used the same diagnostic codes as previously described to identify incidents of stroke in the national patient register. Thus, the exclusion procedure in this study was the same as that used in the nested case-control study. In all, 4411 participants were included in the present study.
Self-reported current medication was recorded at the interview and coded afterward by a pharmacist according to the ATC system. Antidepressants and other drugs were identified by use of the ATC codes used in the nested case-control study. Participants were specifically asked whether a physician had ever told them that they suffered from a number of diseases, including hypertension, diabetes, and myocardial infarct. Lifestyle questions included information on smoking (never, former, or current smoker) and alcohol intake per week [no regular intake (<1 drink), moderate intake (men, 1 to 21 drinks; women, 1 to 14 drinks), and high intake (men, >21 drinks; women, >14 drinks)]. Self-reported height and weight were used to calculate body mass index.
We used unconditional logistic regression to calculate crude and adjusted odds ratios (ORs) and 95% confidence intervals (CIs) of the risk of hemorrhagic, ischemic, clinical, and possible stroke. We adjusted for the effect of age (15-year bands), sex, calendar time (1-year bands), and use of other drugs. Analyses according to stroke subtype and antidepressant group were performed separately.
To minimize information bias resulting from misclassification of prevalent cases of stroke as incident cases, we repeated all analyses including only cases and controls that had resided in Funen County for ≥10 years. Furthermore, patients with current use of >1 type of antidepressant could affect our results in an unpredictable fashion, so we performed separate analyses after exclusion of such patients. To study the effect of treatment duration, we calculated the risk of stroke in never users of SSRIs compared with current users treated for <91 days and current users treated for ≥91 days. The effect of dose was estimated by comparison of the risk of stroke in never users of SSRIs, current users of lower daily doses of SSRIs (≤1 DDD), and current users of higher daily doses of SSRIs (>1 DDD).
To study the interaction between SSRIs and nonsteroidal antiinflammatory drugs, we used all persons with never use of either drug as the reference group. The independent effect of SSRI exposure was calculated as the risk of stroke in persons who were current users of SSRIs and never users of nonsteroidal antiinflammatory drugs compared with the reference group. The independent effect of exposure to nonsteroidal antiinflammatory drugs was calculated in the same way. Finally, we compared the joint effect of SSRIs and nonsteroidal antiinflammatory drugs with the risk of stroke in current users of both drugs. The calculations were performed separately for hemorrhagic and ischemic stroke.
In the twin surveys, unconditional logistic regression was used to study the association between potential stroke risk factors and current use of SSRIs. We adjusted for age (15-year bands), sex, hypertension, diabetes, myocardial infarction, smoking, alcohol intake, and body mass index.
The study was approved by the local ethics committee and the Danish registry board.
We identified 4765 patients with a valid diagnosis of first-ever stroke during the period of January 1, 1994 to December 31, 1999. In all, 659 (13.8%) had intracerebral hemorrhage, and 2717 (57%) had ischemic stroke (Table 1). Ever use of antidepressants was registered in 610 cases and 2910 controls. The most commonly prescribed antidepressant was citalopram, which was used by 266 cases and 1040 controls (Table 2).
Among patients with intracerebral hemorrhage, 21 (3.2%) were current users of SSRIs compared with 742 controls (1.9%) (Table 3). The relative risk of intracerebral hemorrhage in current users of SSRIs was 1.0 (95% CI, 0.6 to 1.6) compared with never users after adjustment for the effect of age, sex, calendar time, and use of other medication (diuretics, β-blockers, calcium channel blockers, ACE inhibitors, antiarrhythmics, antianginal medication, warfarin, phenprocoumon, antidiabetics, lipid-lowering drugs, low-dose acetyl salicylic acid, and other nonsteroidal antiinflammatory drugs). Current use of anticoagulants and low-dose acetyl salicylic acid was independently associated with an increased relative risk of intracerebral hemorrhage. Use of drugs that functioned as proxy measures of risk factors, ie, β-blockers, calcium channel blockers, and ACE inhibitors, also increased the relative risk of intracerebral hemorrhage.
Of 2717 cases with ischemic stroke, 100 (3.7%) were current users of SSRIs compared with 742 controls (1.9%) (Table 4). The adjusted relative risk of ischemic stroke in current users of SSRIs was 1.1 (95% CI, 0.9 to 1.4) compared with never users (Table 5). Separate analysis of definite and probable ischemic stroke gave essentially the same results. As for hemorrhagic stroke, male sex and advancing age increased the risk of ischemic stroke. Past use of anticoagulants and ever use of diuretics, β-blockers, calcium channel blockers, ACE inhibitors, antidiabetics, low-dose acetyl salicylic acid, and other nonsteroidal antiinflammatory drugs all increased the risk of ischemic stroke.
We also performed a combined analysis including all cases of stroke (n=4557) except SAH that yielded an adjusted OR of 1.0 (95% CI, 0.8 to 1.2). We did not perform analyses of cases with SAH because of small numbers (only 2 cases were current users of SSRIs).
Restricting the analyses to 3950 cases (82.9%) and 25 613 controls (64.0%) who had resided in Funen County for >10 years yielded highly comparable results (data not shown). Exclusion of 12 cases and 30 controls with current use of >1 type of antidepressant had no effect on the results. Treatment duration (<91 versus ≥91 days) and daily dose of SSRIs (≤1 versus >1 DDD) had no influence on the risk of ischemic or hemorrhagic stroke (results not shown).
A trend toward increased risk of hemorrhagic stroke and decreased risk of ischemic stroke was observed for persons with current exposure to both SSRIs and nonsteroidal antiinflammatory drugs (Table 6). However, because of the small number of cases with concomitant exposure to both drugs, the results should be interpreted with caution.
In the survey study of twins, we identified 102 current users and 4309 nonusers of SSRIs (Table 7). Mean age was 66.8 years (SD, 12.6 years). Current users of SSRIs were predominantly women, current smokers, and persons with either no regular alcohol intake or high alcohol intake.
Our results suggest that exposure to SSRIs is not a risk factor for intracerebral hemorrhage. Furthermore, we found no evidence that current exposure to SSRIs is associated with a decreased risk of ischemic stroke. The main strengths of the present study were the application of a longstanding patient registry in a setting with equal access to medical care, the review of discharge records by neurologists who were blinded to drug exposure information, and the use of a prescription database that prospectively recorded drug information during the study period. The study design minimized selection and information bias and eliminated recall bias.
Our results are in line with those of the only previous study on the risk of hemorrhagic stroke in users of SSRIs. The previous study, however, identified only 3 cases with current SSRI exposure; consequently, the 95% CI for the relative risk estimate was wide (OR, 0.8; 95% CI, 0.1 to 5.6).4 As expected, we found that advancing age and male sex increased the risk of intracerebral hemorrhage. Current anticoagulation therapy (warfarin and phenprocoumon) and medication acting as proxy measures for hypertension (β-blockers, calcium channel blockers, and ACE inhibitors) were also associated with an increased risk of hemorrhagic stroke. We found a ∼2-fold–increased risk of intracerebral hemorrhage among current users of low-dose acetyl salicylic acid compared with never users. This is in line with results from randomized, controlled trials11,12⇓ but contrary to a recent case-control study.13
Our study, which is the first to examine the risk of ischemic stroke in SSRI users, showed no association between current use of SSRIs and ischemic stroke. These findings are in keeping with those recently reported in a case-control study on the risk of first-time acute myocardial infarction.14 The adjusted OR of acute myocardial infarction in current SSRI users compared with nonusers was 0.9 (95% CI, 0.5 to 1.8).
We found, as expected, that advancing age and male sex increased the risk of first-ever ischemic stroke. Furthermore, past use of SSRIs was associated with a small increased risk of ischemic stroke. We speculate whether this indicates that prior depression might be a risk factor for stroke, a finding reported in a number of studies.15–17⇓⇓ It is, however, still unknown whether depression before a clinical stroke is an independent stroke risk factor or a consequence of silent strokes.18
The antiplatelet effect of SSRIs may occur only during treatment with high doses or after long-term exposure to SSRIs. We found, however, that the average daily dose and treatment duration had no effect on the risk of hemorrhagic or ischemic stroke.
De Abajo et al2 studied the effect of SSRIs on the risk of upper gastrointestinal bleeding and found a multiplicative interaction between SSRIs and nonsteroidal antiinflammatory drugs. The small additive effect we found in intracerebral hemorrhage, which was not statistically significant, may well be a chance finding and should therefore be interpreted cautiously.
Our study has several potential limitations. First, misclassification of prevalent cases of stroke as incident cases could seriously confound our results, because poststroke depression occurs frequently and is primarily treated with SSRIs. We addressed this problem through the design of the study by exclusion of all potential cases and controls with a stroke discharge diagnosis from 1973 until 1993. Furthermore, potential cases with previous strokes were excluded during the validation procedure. Patients with stroke during previous residence outside Funen County might have been hospitalized elsewhere and thus not registered in the Funen patient registry. However, repeating the analyses including only cases and controls who had resided in Funen County for ≥10 years had no influence on the results.
The classification of antidepressants used in previous studies has been debated, and a classification based on drug affinity for the serotonin transporter has recently been proposed.3 The difference between the 2 classifications is minor. We reanalyzed the data using the recently suggested classification and achieved highly comparable results for hemorrhagic and ischemic stroke (results not shown).
Information on potential confounders was not directly available in the nested case-control study. Using prescription data on current and previous use of other medication gave us the opportunity to adjust for other stroke risk factors, including hypertension, diabetes, atrial fibrillation, and ischemic heart disease. Smoking and alcohol intake could not be adjusted for in the case-control study. Results from the population-based surveys indicated that SSRI users were predominantly women, current smokers, and individuals with either no regular alcohol intake or high alcohol intake.
Provided that smoking and high alcohol intake are generally more common in SSRI users, our odds ratio of hemorrhagic stroke in SSRI users compared with nonusers may be overestimated. By the same token, our risk estimate for ischemic stroke in current SSRI users compared with nonusers may represent an overestimate; ie, a beneficial antithrombotic effect of SSRI exposure could have been overlooked. However, if SSRIs were to confer a protective effect, we would expect past users of SSRIs to be at increased risk compared with current users of the drug, because it is unlikely that past and current users of SSRI differ with respect to a large number of other risk factors. We found, however, that the relative risk of ischemic stroke in current users of SSRIs is of the same magnitude as the relative risk of ischemic stroke in recent and past users of SSRIs.
The risk of ischemic stroke in current SSRI users might be overestimated if treatment with SSRIs is generally preferred to other antidepressants in subjects with concomitant diseases. The results from the survey study indicated that hypertension, diabetes, and myocardial infarction were not more common in SSRI users compared with nonusers. Furthermore, we adjusted for several different drugs that are indicators of other diseases acting as risk factors for ischemic stroke.
In conclusion, current exposure to SSRIs is not associated with increased risk of intracerebral hemorrhage and probably is not associated with decreased risk of ischemic stroke.
This study was supported by the Danish Medical Research Council (42865-7042 to Dr Mellerup) and by the National Institute of Aging (grant NIA-PO1-AG08761).
- Received December 20, 2002.
- Revision received March 18, 2002.
- Accepted March 28, 2002.
- ↵De Abajo FJ, García Rodríguez LA, Montero D. Association between selective serotonin reuptake inhibitors and upper gastrointestinal bleeding: population based case-control study. BMJ. 1999; 319: 1106–1109.
- ↵van Walraven C, Mamdani MM, Wells PS, Williams JI. Inhibition of serotonin reuptake by antidepressants and upper gastrointestinal bleeding in elderly patients: retrospective cohort study. BMJ. 2001; 323: 1–6.
- ↵World Health Organization. Recommendations on stroke prevention, diagnosis, and therapy. Stroke. 1989; 20: 1407–1431.
- ↵WHO Collaborating Centre for Drug Statistics Methodology. Anatomical Therapeutic Chemical (ATC) Classification Index With Defined Daily Doses (DDDs). Oslo, Norway: WHO; 2000.
- ↵Christensen K, McGue M, Yashin AI, Iachine IA, Holm NV, Vaupel JW. Genetic and environmental influences on functional abilities among Danish twins aged 75 years and older. J Gerontol A Biol Sci Med Sci. 2000; 55A: M446–M452.
- ↵Thrift AG, McNeil JJ, Forbes A, Donnan GA. Risk of primary intracerebral haemorrhage associated with aspirin and non-steroidal anti-inflammatory drugs: case-control study. BMJ. 1999; 318: 759–764.
- ↵Ohira T, Iso H, Satoh S, Sankai T, Tanigawa T, Ogawa Yimano H, Sato S, Kitamura A, Shimamoto T. Prospective study of depressive symptoms and risk of stroke among Japanese. Stroke. 2001; 32: 903–908.
- ↵Larson SL, Owens PL, Ford D, Eaton W. Depressive disorder, dysthymia, and risk of stroke: thirteen-year follow-up from the Baltimore Epidemiologic Catchment Area Study. Stroke. 2001; 32: 1979–1983.
- ↵Krishnan KRR. Depression as a contributing factor in cerebrovascular disease. Am Heart J. 2000; 140: S70–S76.