Editorials

SSRIs and gastrointestinal bleeding

BMJ 2005; 331 doi: http://dx.doi.org/10.1136/bmj.331.7516.529 (Published 08 September 2005) Cite this as: BMJ 2005;331:529
  1. Carol Paton, chief pharmacist (Carol.Paton{at}oxleas.nhs.uk),
  2. I Nicol Ferrier, professor of psychiatry
  1. Oxleas NHS Trust, Dartford, Kent DA2 7WG
  2. School of Neurology, Neurobiology, and Psychiatry, University of Newcastle upon Tyne, Royal Victoria Infirmary, Newcastle upon Tyne NE1 4LP

    Gastroprotection may be justified in some patients

    There are theoretical reasons for believing that selective serotonin reuptake inhibitors (SSRIs), widely used to treat depression, might increase the risk of gastrointestinal bleeding. Gastroprotective drugs are advocated for high risk patients taking non-steroidal anti-inflammatory drugs, another class of drug that causes gastrointestinal bleeding. What is the evidence that this advice should be extended to patients receiving SSRIs?

    Serotonin is released from platelets in response to vascular injury and promotes vasoconstriction and a change in the shape of the platelets that leads to aggregation.1 Platelets cannot themselves synthesise serotonin. SSRIs inhibit the serotonin transporter, which is responsible for the uptake of serotonin into platelets. It could thus be predicted that SSRIs would deplete platelet serotonin, leading to a reduced ability to form clots and a subsequent increase in the risk of bleeding.

    We have reviewed the published database studies on the relation between SSRI use and gastrointestinal bleeding. Four of these studies compared the risk of an upper gastrointestinal bleed in those prescribed SSRIs with those who were not. The odds ratios of a bleed in an SSRI-treated patient ranged from 1.38 to 3.6: 3.0 (95% confidence interval 2.1 to 4.4),2 3.6 (2.7 to 4.7),3 2.1 (0.6 to 8.3),4 and 1.38 (0.82 to 2.34).5 This roughly threefold increase in risk may also hold for other types of bleeding. Movig et al reported that patients taking SSRIs were 3.71 (1.35 to 10.18) times more likely to require a blood transfusion during orthopaedic surgery than patients not taking them.6 Meijer el al reported that women taking SSRIs with a high affinity for the serotonin transporter were 3.0 (0.8 to 4.9) times more likely to experience abnormal uterine bleeding than women who took antidepressants with low affinity for this transporter.4

    An association between the risk of bleeding and increasing affinity for the serotonin transporter has been noted in several studies,2 4 7 although the confidence intervals around the quoted odds ratios overlap considerably. Clomipramine, fluoxetine, sertraline, and paroxetine have a high affinity for the serotonin transporter while citalopram, fluvoxamine, and venlafaxine have intermediate affinity. Low affinity drugs include doxepin, mirtazepine, moclobemide, and nortriptyline.

    Risk decreases to the same level as controls in past users of SSRIs, indicating that bleeding is likely to be associated with the drug rather than the illness it was prescribed for.3 The association also holds when age, gender, and the effects of other drugs such as aspirin and non-steroidal anti-inflammatory drugs are controlled for.

    The mechanisms by which non-steroidal antiinflammatory drugs and SSRIs are associated with gastrointestinal bleeding are different. Non-steroidal anti-inflammatory drugs directly damage the gastrointestinal mucosa, while SSRIs reduce the effectiveness of the normal clotting mechanism. Aspirin does both. The absolute additional risk of an upper gastrointestinal bleed (requiring admission to hospital) with an SSRI prescribed alone is about 1 in 300 patient years, but co-prescription of SSRIs with aspirin increases the risk to 1 in 200 and with non-steroidal anti-inflammatory drugs to 1 in 80.3 The risk with a non-steroidal drug alone is 1 in 200.8

    The well established association between nonsteroidal anti-inflammatory drugs and upper gastrointestinal bleeding is estimated to result in 700-2000 deaths/year in the UK.8 9 This has led to the recommendation that patients in high risk groups should receive gastroprotection in the form of an H2

    antagonist, proton pump inhibitor, or misoprostil.10 High risk groups are defined as patients older than 65 years, those with a history of peptic ulcer or gastrointestinal bleed, those who are debilitated, and those receiving other drugs that are associated with an increased risk of bleeding such as warfarin and corticosteroids. Yet only misoprostol has been proved to reduce the risk of serious bleeds. Proton pump inhibitors have been shown to reduce endoscopically diagnosed mucosal damage and heal ulcers induced by non-steroidal anti-inflammatory drugs but not to reduce the incidence of severe gastrointestinal bleeds.10

    SSRIs are widely prescribed in the general population and for elderly people. Almost 14 million prescriptions were dispensed in community pharmacies in England in 2003.11 They are recommended by the National Institute for Health and Clinical Excellence as first line treatments in patients with at least moderate depression.12

    Gastroprotection is unlikely to be justified in patients who receive SSRIs alone, but those who are also taking non-steroidal anti-inflammatory drugs or aspirin are clearly at increased risk. This increased risk may also apply to those who are very old or have a history of gastrointestinal bleeding. The use of antide-pressants with low affinity for the serotonin transporter should be considered in these patients. Gastroprotective agents have not been shown to reduce the risk of bleeds associated with SSRIs alone or in combination with non-steroidal drugs, but until such studies are conducted we recommend that SSRIs are added to the list of drugs that increase the risk of bleeding induced by non-steroidal anti-inflammatory drugs and suggest that gastroprotection should be considered in patients who are prescribed both SSRIs and non-steroidal anti-inflammatory drugs or aspirin, including those under the age of 65.

    Footnotes

    • Competing interests None declared.

    References

    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    6. 6.
    7. 7.
    8. 8.
    9. 9.
    10. 10.
    11. 11.
    12. 12.