Clinical impact of selective serotonin reuptake inhibitors therapy with bleeding risks
Abstract.
The selective serotonin reuptake inhibitors (SSRIs) are extensively used for the treatment of multiple psychiatric conditions. In vitro and ex vivo data with these agents indicate they may have varying degrees of antiplatelet activity via multiple receptors. Reports of bleeding in patients receiving SSRIs appeared soon after their introduction. A review of the literature suggests SSRI therapy may increase the risk of bleeding especially with concomitant aspirin or nonsteroidal anti-inflammatory agents. Clinicians should exercise caution when prescribing these agents in high risk patients and maintain awareness of the potential contribution of SSRIs to unexplained bleeding episodes.
Introduction
The selective serotonin reuptake inhibitors (SSRIs) are a cornerstone of therapy for many psychiatric conditions, including depression, generalized anxiety disorder, panic disorder, and obsessive–compulsive disorder. Depression is the sixth most common cause of healthcare visits, accounting for 20.3 million visits in 2005 [1]. Also in 2005, there were 122 500 000 prescriptions for SSRIs and selective serotonin/noradrenaline reuptake inhibitors (SNRIs) dispensed in the US which ranks fifth amongst all therapeutic classes [2]. These prescriptions are the sixth most costly therapeutic class at an estimated $12 501 029 in yearly sales [3]. With the prevalence of depression and corresponding SSRI use it is important to closely examine the adverse effect profile of these agents, as even rare adverse effects can affect a significant number of patients. We briefly review the SSRI agents, evaluate the current evidence on SSRIs and bleeding complications and discuss the clinical impact of the data on patient management.
Design
We performed Ovid and PubMed searches of English- and foreign-language literature (from 1966 to 1 May 2006) using the following Medical Subject Headings (MeSH) and terms: ‘serotonin uptake inhibitors’ and ‘hemorrhage’. Other references were obtained from citations from retrieved articles. We specifically reviewed all case reports and case series of patients with bleeding associated with SSRI use. No funding was received for this review.
Review of SSRIs
Six SSRIs are currently available in the US. SSRIs were first introduced for the treatment of depression in 1987, and since then their use has steadily increased compared with traditional agents such as tricyclic antidepressants and monoamine oxidase inhibitors (MAOIs) [4]. Since their introduction, the number of conditions for which the SSRIs have indications has increased beyond depression to include obsessive–compulsive disorder, panic disorder, generalized anxiety disorder and bulimia [5–10]. Table 1 lists the currently available SSRIs, FDA-approved indications, adult dosing range and their relative effects on serotonin.
| Generic name | Brand name | FDA-approved indications | Dose range (mg day−1) | Degree of inhibition of serotonin reuptake |
|---|---|---|---|---|
| Fluoxetine | Prozac, Sarafem | Major depressive disorderObsessive–compulsive disorderBulimia nervosaPanic disorderPremenstrual dysphoric disorder | 10–80 | High |
| Fluvoxamine | Luvox | Obsessive–compulsive disorder | 50–300 | Intermediate |
| Paroxetine | Paxil | Major depressive disorderObsessive–compulsive disorderPanic disorderSocial anxiety disorderGeneralized anxiety disorderPost-traumatic stress disorder | 10–60 | High |
| Sertraline | Zoloft | Major depressive disorderObsessive–compulsive disorderPanic disorderPost-traumatic stress disorderSocial anxiety disorderPremenstrual dysphoric disorder | 25–200 | High |
| Citalopram | Celexa | Major depressive disorder | 20–40 | Intermediate |
| Escitalopram | Lexapro | Major depressive disorderGeneralized anxiety disorder | 10–20 | Intermediate |
Common class adverse effects include nausea, vomiting, diarrhoea, headache, insomnia, somnolence, agitation, anxiety and sexual dysfunction [11, 12]. SSRIs have also been associated with serotonin syndrome, although this is usually in combination with other serotonergic agents [4, 11–13]. Less frequently, these medications have also been associated with the development of extrapyramidal symptoms (such as akathesia, parkinsonism, acute dystonia and tardive dyskinesia) [12, 13], syndrome of inappropriate antidiuretic hormone [4, 13], and bleeding complications [13]. We specifically review the current evidence on SSRI and bleeding complications in this article.
Reports of the antiplatelet effects of SSRIs emerged soon after the introduction of this class of agents. Butler and Leonard [14] described decreased serotonin (5-HT) uptake by platelets with subsequent reduction in 5-HT-mediated platelet aggregation in patients treated with sertraline. This led to numerous in vitro and in vivo studies that yielded multiple mechanistic theories [15, 16]. Theories that have been proposed include blockade of calcium mobilization [17], inhibition of nitric oxide synthase [18], decreased platelet secretion in response to collagen [19], and decreased platelet binding affinity [14] which may be caused by downregulation of 5-HT binding sites [20] and/or decreased 5-HT in platelets [21]. SSRIs also decrease expression of several platelet surface markers such as CD63 [22], ADP-PA, Coll-PA, CD 107a [23], CD9, GP 1b, VLA-2 [24], PECAM-1 [24–26], GP IIb/IIIa [23, 24], β-thromboglobulin [25, 26], E-selectin [26],p-selectin[23–26], and thromboxane B2 [25]. A decrease in platelet factor 4 may also contribute to antiplatelet activity [25, 27]. One in vivo study demonstrated a negative correlation between serum levels of sertraline and multiple platelet markers [25]. Of note, a clinical study of patients with reduced left ventricular ejection fraction reported a decrease in many of these biochemical markers when an SSRI is added to aspirin therapy, demonstrating that the antiplatelet effects of SSRIs may be additive to those of aspirin [28].
Case reports
Multiple case reports are published in the literature describing adverse bleeding events in patients on SSRIs. Reports have implicated paroxetine [29], sertraline [30], fluvoxamine [31], and fluoxetine [32], including eight cases in patients taking fluoxetine for obsessive–compulsive disorder [33]. Another report describes bleeding in a patient taking the combination of escitalopram, venlafaxine and mirtazapine [34]. Reports of bleeding events are not limited to the adult population as five cases in children taking sertraline have been reported [35]. The combination of in vitro platelet data and multiple case reports provided the hypotheses behind many of the following studies.
Observational studies
Increased bleeding risk (Table 2)
| Study (n) | Population | Findings | NSAIDs + SSRI (versus SSRI alone) | ASA + SSRI (versus SSRI alone) | Comments | Ref. |
|---|---|---|---|---|---|---|
| de Abajo, 1999 (11 899) | UGIB or ulcer perforation | Increased risk with SSRI exposure | Increased | Increased | Incidence of bleeding 0.013% per SSRI prescription | [36] |
| Dalton, 2003 (26 005) | Antidepressant users 1991–1995 | Increased UGIB in SSRI users | Increased | Increased | No increase in UGIB with antidepressants with no SRI activity | [37] |
| Tata, 2005 (64,417) | UGIB | Increased risk with SSRI or NSAID | Increased | N/A | Self-controlled analysis yielded similar results | [38] |
| Meijer, 2004 (64 647) | Antidepressant users 1992–2000 | Increased abnormal bleeding with high and intermediate SRI versus low SRI | N/A | N/A | Positive correlation between degree of SRI and odds of hospitalization for bleeding | [39] |
| Van Walraven, 2001 (317 821) | Antidepressant users 1992–1998; age >65 years | Rate of UGIB: low SRI < intermediate SRI < high SRI | N/A | N/A | Trend towards increased UGIB with increasing SRI not statistically significant | [40] |
| Movig, 2003 (520) | Orthopaedic surgery patients 1999–2000 | Increased risk of blood transfusion in SSRI users versus no antidepressants | N/A | N/A | No increase in transfusion requirements with non-SSRI antidepressants | [41] |
- ASA, aspirin; NSAIDs, nonsteroidal anti-inflammatory drugs; SSRI, selective serotonin reuptake inhibitor; SRI, serotonin reuptake inhibition; UGIB, upper gastrointestinal bleed; N/A, not analysed.
Several case–control and cohort studies have indicated an increased risk of bleeding is present with SSRI use. de Abajo et al. [36] conducted a population-based case–control trial in the UK examining SSRI use in patients with upper gastrointestinal bleeding or ulcer perforation when compared with controls without either event. There were 1651 cases of upper gastrointestinal bleeding and 248 cases of ulcer perforation analysed in comparison with 10 000 matched controls. SSRI exposure was identified through computerized prescription records recorded in the UK general practice research database. Cases and controls reported SSRI exposure rates of 3.1% and 1.0%, respectively, yielding a risk ratio of 3.0 (95% confidence interval 2.1–4.4). The authors reported one bleeding event for every 8000 SSRI prescriptions, an incidence of 0.013%. Aspirin in combination with an SSRI increased the risk ratio to 7.2 (95% CI 3.1–17.1). However, the highest risk was observed with nonsteroidal anti-inflammatory drugs (NSAIDs) in combination with an SSRI which resulted in a risk ratio of 15.6 (6.6–36.6). Of note, no correlation was found between SSRI therapy and ulcer perforation.
A population-based retrospective cohort study in North Jutland, Denmark, retrospectively identified antidepressant users in a prescription database over a 5-year time period and then determined the incidence of hospitalization for upper gastrointestinal bleeding in this population [37]. This was then compared with the rate of hospitalization for upper gastrointestinal bleeding in patients who did not receive antidepressants. Based on population parameters the authors determined the expected number of upper gastrointestinal bleeds in a normal population, and then compared the observed number of upper gastrointestinal bleeds to the expected number, yielding an observed–expected ratio. Current users of SSRIs had an observed–expected ratio of 3.6 (95% CI 2.7–4.7), whilst non-SSRI antidepressant users demonstrated a lower ratio (2.3; 95% CI 1.5–3.4). Antidepressants with no serotonin receptor effect (nortriptyline, desipramine, trimipramine, maprotiline, amoxapine and minaserin) had no significant difference on bleeding rates (1.7; 95% CI 0.8–3.1). Of note, combining an SSRI with an NSAID (12.2; 95% CI 7.1–19.5) or aspirin (5.2; 95% CI 3.2–8.0) increased the risk of upper gastrointestinal bleeding above that observed with an SSRI alone. A subset analysis restricted the cohort to patients who had no history of exposure to medications with increased risk of gastrointestinal bleeding (NSAIDs, low-dose aspirin, high-dose aspirin, vitamin K antagonists, or oral corticosteroids). The observed–expected ratio for SSRI users was 3.4 (95% CI 2.1–5.2), indicating SSRIs may be an independent risk factor for gastrointestinal bleeding.
A population-based case–control study in the UK aimed to compare the risk of gastrointestinal bleeding amongst users of NSAIDs, SSRIs, and the combination of the two classes [38]. In the database 11 261 adult patients were identified as cases, defined as adult patients who had an upper gastrointestinal bleed. A total of 53 156 controls were matched based on age, gender, and medical practice. Both SSRIs (OR 2.38; 95% CI 2.08–2.72) and NSAIDs (OR 2.15; 95% CI 2.02–2.28) were associated with an increased risk of gastrointestinal bleeding. When examining the effect of the two agents concomitantly, a slight elevation in risk was found when comparing with either agent alone (OR 2.93; 95% CI 2.25–3.82). Of note, a self-controlled analysis to decrease interpatient variability was conducted in which periods of exposure to an antidepressant or NSAID were compared with control periods in 8130 patients who had at least one exposure to one of these agents. The results were similar to the overall results, with SSRIs [incidence rate ratio (IRR) 1.71; 95% CI 1.48–1.98] and NSAIDs (IRR 2.71; 95% CI 2.51–2.91) demonstrating increased risk of gastrointestinal bleeding, and a slightly higher risk when these agents are combined (IRR 3.25; 95% CI 1.95–5.42).
Another population-based study identified a cohort of 64 647 new antidepressant users in the Netherlands between 1992 and 2000 [39]. A nested case–control study was performed with hospitalization for abnormal bleeding as the case definition. Antidepressant exposure was identified by computerized pharmacy records and subgrouped according to the degree of serotonin inhibition (high, intermediate, or low). There were 196 admissions for abnormal bleeding reported, and 972 controls (approximately 1 : 5 ratio) were then matched for age and sex. Odds ratios were calculated to determine the risk of bleeding with each drug when compared with mirtazapine, the agent with least effect on serotonin reuptake, and odds ratios were also computed for each subgroup of antidepressants. Antidepressants categorized as having a high degree of serotonin inhibition (fluoxetine, sertraline, clomipramine and paroxetine) yielded an OR of 2.6 (95% CI 1.4–4.8) when compared to those with low affinity for the serotonin transporter (e.g. mirtazapine, trazodone, bupropion and nortriptyline). Intermediate affinity agents (venlafaxine, amitriptyline, fluvoxamine, imipramine, citalopram and dothiepin) had similar results, with an OR of 1.9 (95% CI 1.1–3.5). Logistic regression was also performed to analyse the relationship between inhibition of serotonin reuptake and odds ratio of hospitalization for bleeding. A correlation coefficient of 0.51 was reported, signifying an important relationship between these factors. Adjustments were made in the regression model to account for potential confounders (aspirin, NSAIDs, anticoagulants, glucocorticoids, oestrogens, progestins, histamine2-blockers, proton pump inhibitors, and prior hospitalization for bleeding). Some of the limitations of this study are that bleeding events not resulting in hospitalization or in death before hospitalization were undetected and possible misclassification of the diagnosis from the hospital records. Regardless, this study did show that higher degrees of serotonin inhibition increase the risk of bleeding.
A Canadian retrospective cohort study examined a population of elderly antidepressant users to determine the risk of upper gastrointestinal bleeding [40]. The patients were stratified based on the affinity of his/her antidepressant for the serotonin transporter (low, intermediate, or high). Use of potential confounding drugs (NSAIDs, aspirin, glucocorticoids, histamine2-blockers, and proton pump inhibitors) was controlled for in the regression model. Risk was described as bleeds per 1000 patient years. The overall risk of upper gastrointestinal bleeding was 7.3 bleeds/1000 patient years. Subgroup analysis showed that the low-affinity antidepressants had the lowest rate of upper gastrointestinal bleeding, followed by intermediate and then high-affinity agents (6.6, 7.4 and 7.9, respectively). This trend did not reach statistical significance. Further subgroup analysis did show a statistically significant increase in upper gastrointestinal bleeding with increasing serotonin inhibition in patients >80 years of age and patients with a previous history of gastrointestinal bleeding. This indicates that degree of serotonin inhibition may play a role in bleeding effects, especially when combined with other risk factors.
Another retrospective cohort analysis identified all patients in one Dutch institution who underwent orthopaedic surgical procedures during a 2-year time frame to evaluate risk of blood transfusion as related to antidepressant use [41]. The patients were classified into three groups: SSRI users, non-SSRI antidepressant users, and non-antidepressant users. Of 520 patients, 26 were taking SSRIs and 14 reported non-SSRI antidepressants. There were 59 patients who required blood transfusions. When examining odds ratios, SSRI users were at increased risk of requiring a blood transfusion when compared with non-antidepressant users (OR 3.71; 95% CI 1.35–10.18). Patients taking non-SSRI antidepressants did not demonstrate an increased risk with an OR of 0.74 (95% CI 0.1–5.95). The authors also examined the use of other co-medications (aspirin, NSAIDs, glucocorticoids and vitamin K antagonists) and found no increased risk; however, this is only based on univariate analysis and does not incorporate the potential for additive effects with other agents (i.e. SSRIs).
No difference in bleeding risk (Table 3)
| Study (n) | Population | Findings | Comments | Ref. |
|---|---|---|---|---|
| de Abajo, 2000 (319) | Antidepressant users 1990–1997 | No increase in ICH amongst SSRI users | High doses of SSRIs did not increase risk over low–medium doses | [42] |
| Layton, 2001 (136 754) | Prescription event monitoring database 1986–1998 | No difference in bleeding risk with SSRIs versus other psychiatric drugs or nonpsychiatric drugs | Numerically more bleeds amongst SSRI users in month 1 of therapy versus subsequent months; not statistically significant | [43] |
| Bak, 2002 (44 765) | Cases = stroke discharge diagnosis 1994–1999, matched controls | No increased risk of haemorrhagic or ischaemic stroke amongst SSRI users | Risk adjusted for potential confounders (i.e. age, sex, NSAIDs, aspirin) | [44] |
- ICH, intracranial haemorrhage; NSAIDs, nonsteroidal anti-inflammatory drugs; SSRI, selective serotonin reuptake inhibitor.
Although the studies reviewed above demonstrate a difference in bleeding events with SSRI therapy, other studies with varying patient populations have failed to correlate SSRIs and abnormal bleeding. A case–control study reviewed a cohort of first-time antidepressant users in the UK general practice database between 1990 and 1997 [42]. Cases were defined as subjects with idiopathic intracranial haemorrhage, and up to four controls per case were matched on sex, age, calendar time, and practice. Subjects with a history of intracranial haemorrhage, coagulopathy, or receiving anticoagulants were excluded. There were 120 cases identified with 247 controls subsequently matched. The odds ratio for SSRI exposure was 0.8 (95% CI 0.2–2.3). Analysis of other, non-SSRI antidepressants yielded similar results (OR 0.7; 95% CI 0.3–1.6), signifying no increase in risk of intracranial haemorrhage with antidepressant use.
An observational cohort study of UK SSRI users compared rates of abnormal bleeding with patients on other psychiatric medications and with patients not on psychiatric drugs [43]. Data were obtained from a database of postmarketing surveillance of prescription medications between 1986 and 1998. Potential confounders such as NSAID or aspirin use were not incorporated in the analysis model. Although there was a slightly higher incidence of bleeding in the SSRI users it was not statistically significant when compared with users of other psychiatric medications (RR 1.14; 95% CI 0.77–1.72) or users of non-psychiatric drugs (RR 1.23, 95% CI 0.91–1.66). The incidence of bleeding was highest in the first month of SSRI therapy, leading the authors to theorize that a temporal relationship may exist with regard to bleeding events. However, no significant differences were observed when month 1 bleeding rates were compared amongst the groups.
Another population-based case–control study analysed the association of SSRIs with cerebrovascular accidents [44]. Discharge diagnoses for residents of Funen County, Denmark, were analysed between 1994 and 1999, and 4765 cerebrovascular accidents were identified. A total of 40 000 controls were then selected from the remaining population. Odds ratios were adjusted for confounders such as age, sex, index date, and use of NSAIDs, aspirin, or warfarin. Risk of intracerebral haemorrhage was not affected by current SSRI therapy (OR 1.0; 95% CI 0.6–1.6), SNRI therapy (OR 0.5; 95% CI 0.1–1.9), or other antidepressants (OR 1.5; 95% CI 0.8–2.8). However, it has been theorized that the antiplatelet effects of SSRIs may be protective against thromboembolic events such as myocardial infarction [45] or ischaemic stroke [46], but evaluation of the ischaemic stroke cases in current SSRI users yields an OR of 1.1 (95% CI 0.9–1.4), indicating there is no protective effect of SSRI therapy. Similar results were found with the SNRIs and other antidepressants.
A case–control study used database information to compare the rate of gastrointestinal bleeding during SSRI treatment to a group of patients receiving non-SSRI comparator drugs [47]. The comparator drugs chosen were moclobemide, an MAOI and salmeterol, a beta-agonist. There were 237 609 patient-months of exposure to SSRIs and 205 431 patient-months of exposure to the comparator drugs. The SSRI group had 0.43 gastrointestinal bleeds per 1000 patient months when compared with 0.35 in the comparator group, yielding a rate ratio of 1.24 (95% CI 0.91–1.70). The authors concluded there is no significant increase in the risk of bleeding with SSRI therapy. However, this study is not included in Table 3 as it has only been published in correspondence format and thus limited background information is available.
Discussion
Based on our review of the literature, SSRI therapy may increase risk of abnormal bleeding events. Higher incidence of upper gastrointestinal bleeding [36–38, 40], postoperative transfusions [41], and antiulcer prescribing [48] has been associated with SSRIs. It is also important to note that in several studies concomitant aspirin or NSAID therapy increased the risk of bleeding above that of SSRI therapy alone, and in some cases the risk exceeded the additive effect of either agent alone. Therefore the effects of SSRI and aspirin/NSAIDS could be cumulative on bleeding risks and caution should be exercised when using more than one of these drug classes in combination. Patient education with regard to these effects is essential.
Our review also found much variation in the degree of risk associated with SSRI therapy. This may be partially explained by interstudy differences in patient populations and serotonin reuptake inhibition (SRI). It is unclear if degree of serotonin reuptake inhibition affects adverse bleeding outcomes. It has been reported that antidepressants with a higher degree of SRI increase bleeding risk more than those with lower degrees of SRI [39]. One study [40] did not find a significant difference when comparing bleeding associated with different degrees of SRI, but did find a trend towards more bleeding events with higher dosage of SRI medications, although serum concentrations were not measured. This same study found patients with a history of gastrointestinal bleeding or who are >80 years of age are also more susceptible to the effects of higher SRI medications on bleeding risk [40]. Consideration should be given to these higher-risk patient populations. Also of note is the finding that the risk of a bleeding event on SSRI therapy is numerically higher in the first month of therapy than in subsequent months [43]. Future studies should incorporate temporal relationship into study design to further evaluate this risk.
Limitations and conclusion
Our review is inherently limited by currently available published studies. Most of the studies are retrospective in nature or case reports. However, it is unlikely that a randomized prospective study will be conducted on this topic given the required sample size and ethical considerations. Whilst the number of studies reporting harm are outnumber the number of studies that did not demonstrated bleeding risks, there is potential publication bias as negative studies (that did not find harm) are less likely to be published. With this in mind the studies demonstrating harm has showed consistent results when compared with other well-recognized agents that are associated with increased bleeding. The same studies also made a convincing case by examining dose-gradient and temporal relationship.
In conclusion, our compilation of the literature indicates that the platelet effects of SSRIs may have a clinically significant impact on bleeding risk, but by design, the studies currently are inherently limited by unmeasured factors. Currently, only two studies were self-controlled which adjusts for interpatient variability [37, 38]. Future studies that are more tightly controlled for confounding factors will be necessary to fully elucidate the true impact. Major confounding variables include duration of SSRI exposure, dosage, medication adherence, and baseline patient bleeding risk factors such as bleeding diathesis, alcohol consumption, and other pharmacodynamic and pharmacokinetic drug–drug interactions. Analysis of other antidepressants that affect serotonin but are not classified as SSRIs (i.e. tricyclics) is also needed [49]. Until more information is available, careful review of patients taking SSRIs is needed to identify other factors which may increase their risk of adverse bleeding events. In terms of clinic practice, both patients and providers need to appreciate the potential risk of SSRI-related bleeding adverse events.
Conflict of interest statement
No conflict of interest was declared.
