Association of Selective Serotonin Reuptake Inhibitors With the Risk for Spontaneous Intracranial Hemorrhage

Key Points

Question  What is the risk for intracranial hemorrhage associated with selective serotonin reuptake inhibitors and with antidepressants according to the strength of the inhibition of serotonin reuptake?

Findings  In this population-based cohort study, the use of selective serotonin reuptake inhibitors and more generally of antidepressants that are strong inhibitors of serotonin reuptake were associated with an increased risk for intracranial hemorrhage compared with tricyclic antidepressants, particularly in the first 30 days of use. Concomitant use of oral anticoagulants further increased this risk.

Meaning  Antidepressants with strong serotonin reuptake inhibition properties increase the risk for intracranial hemorrhage, and caution must be exerted with concomitant use of anticoagulants.

Importance  Selective serotonin reuptake inhibitors (SSRIs) may increase the risk for spontaneous intracranial hemorrhage (ICH), an effect that is in theory linked to the strength of inhibition of serotonin reuptake of an antidepressant. However, whether antidepressants that are strong inhibitors of serotonin reuptake actually increase the risk for ICH and the effect of concomitant use of antithrombotics are unknown.

Objectives  To assess the risk for ICH associated with the use of SSRIs compared with tricyclic antidepressants (TCAs) among new users of antidepressants and according to the relative affinity of the antidepressant for the serotonin transporter and to assess whether concomitant use of antithrombotics modifies this risk.

Design, Setting, and Participants  This population-based cohort study included new users of antidepressants 18 years or older from January 1, 1995, to June 30, 2014. More than 650 general practices in the United Kingdom contributing to the Clinical Practice Research Datalink enrolled patients. with use of a nested case-control approach, each case of a first ICH identified during follow-up was matched with as many as 30 control individuals by age, sex, calendar time, and duration of follow-up. Follow-up was completed on October 31, 2014.

Interventions  Current use of SSRIs compared with TCAs and strong compared with weak serotonin reuptake inhibitors.

Main Outcomes and Measures  Incidence rate ratios (RRs) of ICH.

Results  Among a cohort of 1 363 990 incident users of antidepressants (36.8% male; 63.2% female; mean [SD] age, 47.9 [18.5] years), 3036 cases of ICH were identified during follow-up and matched to 89 702 controls. Current SSRI use was associated with an increased risk for ICH (RR, 1.17; 95% CI, 1.02-1.35) relative to TCAs, highest during the first 30 days of use (RR, 1.44; 95% CI, 1.04-1.99), and translating in very few additional events. Similarly, the risk was increased by 25% with strong inhibitors (RR, 1.25; 95% CI, 1.01-1.54) and highest during the first 30 days of use (RR, 1.68; 95% CI, 0.90-3.12). Concomitant use of anticoagulants may increase the risk substantially (RR, 1.73; 95% CI, 0.89-3.39).

Conclusions and Relevance  The use of SSRIs and more generally of antidepressants with strong inhibition of serotonin reuptake are associated with an increased risk for ICH, particularly in the first 30 days of use and when used concomitantly with oral anticoagulants.

Antidepressants are among the most prescribed drugs in industrialized countries,^1,2 and a third of antidepressant prescriptions treat nonpsychiatric indications.³ Among the various classes, selective serotonin reuptake inhibitors (SSRIs) have an interesting profile because of their better tolerability and safety compared with older antidepressants such as tricyclics (TCAs) and because of their comparable efficacy in the treatment of depression. However, SSRIs increase the risk for abnormal bleeding, in particular, gastrointestinal tract bleeding.^4,5 Several recent studies also suggested an increased risk for intracranial hemorrhage (ICH) in patients treated with SSRIs, although this risk was not confirmed by others,^6-14 possibly owing to some methodologic limitations. Moreover, even if this risk exists, the comparison with a nontreated group may exaggerate the strength of a potential association because of indication bias.

This bleeding risk would result from the mechanism of action of SSRIs on the serotonin transporter expressed by platelets, which inhibits serotonin reuptake into platelets and is responsible for the antiplatelet action of this class of antidepressants.¹⁵ Therefore, the potential bleeding effect of antidepressants is linked to the strength of serotonin inhibition reuptake, and antidepressants can be distinguished according to their degree of inhibition. However, few studies assessed the risk for ICH while classifying antidepressants accordingly, with inconsistent results.^13,16,17 Conversely, strong inhibitors of serotonin reuptake have been associated with the risk for gastrointestinal tract or abnormal bleeding compared with weak inhibitors.^18-20 The objective of our study was therefore to assess the risk for spontaneous ICH associated with the use of SSRIs compared with TCAs in a population-based cohort of new users of antidepressants and according to the relative affinity of the antidepressant for the serotonin transporter.

We conducted a retrospective cohort study with a nested case-control approach to analysis. The source population was identified using the United Kingdom’s Clinical Practice Research Datalink (CPRD), one of the world’s largest computerized databases of anonymized primary care medical records from a population of more than 12 million patients enrolled with more than 650 general practices.^21,22 Age and sex distributions of patients broadly reflect those of the UK population.^21,23 Information collected includes demographic data, lifestyle factors, medical symptoms and diagnoses, laboratory test results, prescriptions, and referrals to specialists and hospitals. Prescriptions dispensed by the general practice physician are automatically recorded in the computerized file. Read codes are used to enter medical diagnoses,²⁴ and a coded drug dictionary based on the UK Prescription Pricing Authority Dictionary²⁵ is used for recording prescriptions. Quality control of data are performed regularly, and practices fulfilling predefined quality criteria are labeled up to standard. Numerous studies have shown the validity and high quality of the recorded data.^26,27 The study protocol was approved by the independent scientific advisory committee of the CPRD and the research ethics committee of the Jewish General Hospital, Montréal, Québec, Canada, which also waived the need for patient informed consent because data were deidentified.

The cohort was formed of all patients in the CPRD 18 years or older and members of an up-to-standard practice with a first prescription for an antidepressant (eMethods 1 in the Supplement) between January 1, 1995, and June 30, 2014. Cohort entry was taken as the date of the first antidepressant prescription within the study period. Patients with less than 1 year of information in the CPRD before cohort entry were excluded, as were patients who had been dispensed any antidepressant before cohort entry, to form a cohort of new users of antidepressants. We also excluded patients with a history of stroke (ischemic or hemorrhagic) or transient ischemic attack before cohort entry. Patients were followed up until the date of their first outcome event, the date they transferred out of the up-to-standard practice, death, or end of the study period (October 31, 2014), whichever occurred first. Patients with a first code for ischemic stroke or traumatic cerebral hemorrhage during follow-up were censored at the date of diagnosis.

Within the study cohort, we identified all patients with a first-time Read code related to ICH, including ICH not otherwise specified, nontraumatic subdural or extradural hematoma, hemorrhagic stroke, intracerebral hemorrhage, or subarachnoid hemorrhage, during follow-up. The index date was defined as the calendar date of the first recorded ICH.

For each case, we randomly selected as many as 30 controls among the cohort members in the risk sets defined by the case, after matching for sex, age, calendar year of cohort entry, and duration of follow-up. Because they were selected from the risk set defined by each case, all controls were necessarily alive, active in the practice, and event free when matched to their corresponding case. The date resulting in the same duration of follow-up for the case and controls was the index date for the controls.

For all cases and their matched controls, we identified all prescriptions for an SSRI, a TCA, or another antidepressant from the computerized medical records from cohort entry to the index date. Patients were considered current users if the duration of their last prescription included the index date or ended within 30 days before the index date. Past users included patients whose last prescription ended 30 to 90 days before the index date. Nonusers included patients with no antidepressant prescription in the 90 days preceding the index date. Patients currently exposed to more than 1 class of antidepressant were classified into a separate category (multiple users). Consequently, current, past, and multiple users and nonusers represented mutually exclusive exposure categories. The reference category was current use of TCAs. The choice of a treated comparator group allowed the estimation of the potential extra risk associated with SSRIs while minimizing confounding by indication (severity). The choice of TCAs ensured a large number of patients in the reference category, and TCAs as a class are not known to potentially increase the risk for bleeding based on their biological properties. In a subsequent analysis, patients currently exposed were categorized according to duration of use, stratified in 3 groups according to the tertiles of the distribution of use among the controls.

In a separate model, we assessed whether the risk for ICH varied according to the degree of serotonin reuptake inhibition of an antidepressant (strong, intermediate, or weak), and all antidepressants (SSRIs, TCAs, and others) were therefore reclassified accordingly (eMethods 2 in the Supplement). This classification is based on the affinity of the antidepressant for the serotonin transporter.²⁸ A strong affinity for the serotonin transporter indicates a higher inhibition of serotonin reuptake and thus higher antiplatelet action. The reference category included antidepressants with a weak degree of inhibition.

We used a nested case-control approach because of the time-varying nature of exposure and the covariates, the size of the cohort, and the long duration of follow-up.^29,30 Thus, we used conditional logistic regression to compute odds ratios that are unbiased estimators of incidence rate ratios (RRs), with little or no loss in precision.³¹

In the primary analysis, we estimated the RR of ICH associated with current use of SSRIs relative to current use of TCAs and their corresponding 95% CIs. We also estimated the risk associated with SSRIs compared with no current use to assess whether point estimates were indeed higher when using a nontreated comparison group. According to the matching process, all RRs were inherently adjusted for sex, age, duration of follow-up, and calendar time and adjusted for potential confounders (eMethods 3 in the Supplement). In a separate model, we assessed the risk for ICH according to the duration of current SSRI use compared with the same duration of TCA use. This analysis was restricted to current users ever exposed to only 1 class of antidepressants to exclude potential remnant effects of other classes of antidepressants. We also calculated an approximation of the absolute adjusted rate differences and corresponding 95% CIs as measures of the excess risk associated with current use of SSRIs. Rate differences were calculated as R0 × (RR − 1), where R0 indicates the incidence rate of ICH among the TCA-exposed person-time in the cohort and RR, the adjusted RR for the SSRIs vs TCAs. In separate models, we assessed the risk for different subtypes of ICH. Stratified analyses were conducted to assess effect modification by concomitant use of anticoagulants and antiplatelets in patients currently exposed to SSRIs. In this analysis, the risk was first estimated with as many as 30 controls per case and then by taking all the available controls in the risk sets. All analyses were repeated while reclassifying exposure to antidepressants according to the degree of serotonin reuptake inhibition and assessing the risk for ICH with current use of strong inhibitors compared with weak inhibitors. We performed several sensitivity analyses to assess the robustness of our results (eMethods 4 in the Supplement). All computations were performed using the SAS software (version 9.4; SAS Institute Inc).

The cohort consisted of 1 363 990 new users of antidepressants, including 773 364 (56.7%) new users of SSRIs, 534 587 (39.2%) new users of TCAs, and 56 039 (4.1%) new users of other antidepressants (eFigure in the Supplement). The mean (SD) age at cohort entry was 47.9 (18.5) years; 63.2% were female; and 36.8% were male. During a mean (SD) follow-up of 5.8 (4.6) years, 3036 patients were diagnosed with ICH, yielding an overall incidence rate of 3.8 (95% CI, 3.7-3.9) per 10 000 persons per year. Table 1 shows the characteristics of the cases and their matched controls (n = 89 702). Risk factors for bleeding and most comorbidities were more prevalent in cases than controls, as expected. At the index date, 1120 cases (36.9%) and 28 514 controls (31.8%) were currently exposed to antidepressants; among them, 678 (60.5%) of the cases and 19 171 (67.2%) of the controls remained exposed to the same class of antidepressant during follow-up.

Current SSRI use was associated with an increased risk for ICH (RR, 1.17; 95% CI, 1.02-1.35) relative to TCAs, corresponding to an absolute adjusted rate difference of 6.7 (95% CI, 0.2-13.2) per 100 000 persons per year. The risk was highest during the first 30 days of use (RR, 1.44; 95% CI, 1.04-1.99) (Table 2). Similarly, when classified by degree of inhibition of serotonin reuptake, use of strong inhibitors was associated with a 25% increased risk when compared with weak inhibitors (RR, 1.25; 95% CI, 1.01-1.54), corresponding to an absolute adjusted rate difference of 9.5 (95% CI, −0.5 to 19.6) per 100 000 persons per year. The risk was also highest during the first 30 days of use (RR, 1.68; 95% CI, 0.90-3.12) (Table 3). The same pattern of increased risk was generally observed for all types of ICH with current use of SSRI and current use of strong serotonin reuptake inhibitors (eTables 1-6 in the Supplement). When compared with nonuse rather than use of TCAs or weak inhibitors, the risk for ICH was further increased with current use of SSRIs (RR, 1.35; 95% CI, 1.22-1.50) and current use of strong inhibitors (RR, 1.39; 95% CI, 1.22-1.59).

Table 4 displays the results of stratified analyses assessing whether concomitant use of antithrombotic medications alters the risk for ICH in current SSRI users. Among 149 cases and 2232 controls who were current users of anticoagulants in the month before the index date, 136 cases (91.3%) and 2133 controls (95.6%) used vitamin K antagonists, and the mean (SD) duration of use at the index date was 151 (285) days for cases and 180 (364) days for controls. Among 426 cases and 11 928 controls who were current users of antiplatelets, 375 cases (88.0%) and 10 855 controls (91.0%) used aspirin, with a mean (SD) duration of use of 291 (483) and 365 (584) days, respectively. Among users of anticoagulants, concomitant use of SSRIs increased the risk 3-fold compared with concomitant use of TCAs, although the result did not reach statistical significance. Conversely, among antiplatelet users, SSRI use did not increase the risk. When we included all controls in the risk sets, the findings were similar, although the point estimate was substantially lower for anticoagulants (RR, 1.73; 95% CI, 0.89-3.39). When classified by degree of inhibition, both analyses (using a sample of controls or all controls in the risk sets) showed a substantial increased risk in patients exposed concomitantly to strong serotonin reuptake inhibitors and anticoagulants, although the number of exposed cases was small (Table 5). Use of antiplatelets did not modify the risk. In sensitivity analyses, changing the exposure time window to 15 or 60 days, measuring covariates within 6 to 18 months before the index date, or controlling for confounding using high-dimensional propensity scores did not materially change the point estimates in the primary analyses.

In a large population-based cohort of new users of antidepressants, current use of SSRIs compared with TCAs was associated with an increased risk for ICH, particularly during the first 30 days of use. These findings were corroborated by a slightly higher risk when antidepressants were classified by the degree of serotonin reuptake inhibition. However, this relative increase translated to very few additional events. Concomitant use of oral anticoagulants further increased the risk for ICH substantially.

The association between ICH and SSRIs has been assessed in several observational studies with inconsistent results.^6-14 Possible explanations include limited statistical power, a heterogeneous comparator group, the inclusion of prevalent users, various definitions of exposure, or substantial residual confounding. All studies included a comparator group of nonusers, and some studies focused on hemorrhagic stroke only or intracerebral hemorrhage only. In the Women’s Health Initiative cohort study,¹² exposure to SSRIs in the year before cohort entry was associated with a 2-fold increased risk for ICH compared with nonuse. Finally, a recent meta-analysis of the main observational studies comparing patients exposed to SSRIs with patients not exposed to antidepressants³² showed an increased risk for ICH associated with SSRI use (RR, 1.51; 95% CI, 1.26-1.81). The comparison of SSRIs users with nonusers may introduce indication bias because the condition being treated may be a risk factor for bleeding. For instance, an association of depression with an increased risk for hemorrhagic stroke has been suggested.³³ This possibility precludes separating the risk due to antidepressants from the risk related to the condition being treated and may artificially increase the risk estimate. Indeed, we found a higher point estimate when SSRI use was compared with no use rather than an active treatment group (TCAs). However, we cannot exclude that at least part of this increased risk is attributable to residual confounding. On the other hand, our results are still compatible with a true increased risk based on the biological plausibility of and the well-established increased risk for bleeding in other body systems.^4,5

Serotonin stored in platelets represents more than 99% of the total serotonin concentration in the human body. After vascular injury and platelet activation, serotonin is released into the bloodstream and binds to specific receptors to promote vasoconstriction and platelet aggregation and thereby promote hemostasis.³⁴ Reuptake of serotonin into platelets involves a serotonin transporter that is blocked by SSRIs, thus inhibiting serotonin reuptake into platelets.¹⁵ This inhibition would in turn reduce the potential for platelet aggregation and therefore platelet thrombus formation with a subsequent increased risk for bleeding. The antiplatelet mechanisms of the action of SSRIs have been shown in animals and humans studies as early as 1 or 2 weeks after starting treatment.^15,35-37 Antidepressants with the highest degree of serotonin reuptake inhibition have been more frequently associated with abnormal bleeding and modifications of hemostasis markers.¹⁵

A few studies assessed the risk for ICH according to the degree of inhibition of serotonin reuptake of antidepressants. In one study, the definition of exposure (antidepressants used at least once during follow-up vs no use) might have diluted any potential effect.¹⁷ The second study reported a 13% increased risk that did not reach statistical significance.¹⁶ Another study showed an increased risk with all antidepressant classes and with any degree of serotonin inhibition, which might indicate residual confounding.¹³ Although the results concerning the potential risk for ICH have been difficult to show and somewhat contradictory (probably because these events are rare), an association was found between strong inhibitors of the serotonin transporter and gastrointestinal tract hemorrhage, abnormal bleeding, and the need for transfusion after surgery.^18-20

An increased risk for ICH in patients exposed concomitantly to anticoagulants and SSRIs has been suggested previously. In a cohort of new users of coumarins, the risk for non–gastrointestinal tract bleeding increased in users of SSRIs compared with nonusers, including a nonsignificant increased risk for ICH.³⁸ In a case-control study,¹⁰ the use of SSRIs with warfarin was associated with a numerically greater risk for hemorrhagic stroke (odds ratio, 4.7; 95% CI, 1.2-18.4) than use of warfarin alone (odds ratio, 3.0; 95% CI, 1.8-5.0), but neither use of aspirin alone nor the use of SSRIs with aspirin increased the risk. Two other studies^12,16 found no statistical interaction between the use of SSRIs and aspirin and the risk for ICH, in accordance with our results. However, we cannot exclude that an increased risk with concomitant use of SSRIs and antiplatelets exists and that we could not capture it precisely. The number of exposed cases was small, and the upper bound of the 95% CI is 1.37. Also, current antiplatelet use was broadly defined as exposure in the month before the index date.

This observational study was conducted in a population-based setting using the CPRD, which is one of the world’s largest databases of computerized medical files. The CPRD is an asset, considering the rarity of ICH. The nested case-control approach within a well-defined cohort virtually rules out the potential for selection bias. One of the potential threats to validity may be the presence of residual confounding. By design, this bias was minimized by the choice of a treated comparator group, and all models were adjusted for numerous potential confounders, including lifestyle variables, such as smoking and alcohol abuse. Finally, the definition of exposure was based solely on prescriptions issued and not on prescriptions actually filled or taken by the patient. This definition could result in misclassification of exposure that is likely to be nondifferential between cases and controls, thus potentially biasing the results toward the null. Therefore, the finding of an increased risk for ICH likely indicates an increased risk that is potentially underestimated.

Selective serotonin reuptake inhibitors and more generally antidepressants acting as strong inhibitors of serotonin reuptake increase the risk for ICH, particularly during the first months of use. Concomitant use of oral anticoagulants further increases the risk.

Corresponding Author: Christel Renoux, MD, PhD, Lady Davis Research Institute, Jewish General Hospital, 3755 Côte Ste Catherine, H Pavilion, Room 471, Montréal, QC H3T 1E2, Canada (christel.renoux@mcgill.ca).

Accepted for Publication: September 14, 2016.

Published Online: December 5, 2016. doi:10.1001/jamaneurol.2016.4529

Author Contributions: Dr Renoux had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: All authors.

Acquisition, analysis, or interpretation of data: Renoux, Vahey, Dell’Aniello.

Drafting of the manuscript: Renoux.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Vahey, Dell’Aniello, Boivin.

Obtained funding: Renoux.

Study supervision: Renoux.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by operating grant MOP-133553 from the Canadian Institutes of Health Research and a Chercheur-Boursier Award from the Fonds de la Recherche en Santé du Québec (Dr Renoux).

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.