Intended for healthcare professionals


Antidepressants and upper gastrointestinal bleeding

BMJ 1999; 319 doi: (Published 23 October 1999) Cite this as: BMJ 1999;319:1081

New results suggest a link

  1. Alain Li Wan Po, director (a.liwanpo{at}
  1. Centre for Evidence-Based Pharmacotherapy, School of Life and Health Sciences, Aston University, Birmingham B4 7ET

    General practice p 1106)

    Is there an increased risk of upper gastrointestinal bleeding associated with antidepressant therapy? Most clinicians probably think this unlikely Indeed, despite spontaneous case reports,13 most drug reference sources, including the British National Formulary and the Data Sheet Compendium, do not mention any such association. In this issue, however, de Abajo et al report that there is indeed an increased risk with selective serotonin reuptake inhibitors (p 1106).4 Moreover, they suggest a possible interaction with non-steroidal anti-inflammatory agents, leading to an increased risk far beyond a simple additive effect. How robust are these results and what are the implications for clinical practice?

    Retrospective observational studies are always subject to confounding, and the present case-control study is no exception. However, this study seems to be less prone to it. Upper gastrointestinal bleeding is not generally known to be associated with serotonin reuptake inhibitors, so channelling bias5 is less likely. The more obvious possible confounders were carefully adjusted for and the database used is respected for the quality of its data. Therefore, despite the possibility of confounding, the weight of evidence suggests that there is an increased risk of upper gastointestinal bleeding with some antidepressant compounds and that this risk is increased if the patient is also taking non-steroidal anti-inflammatory drugs.

    More detailed implications for practice are harder to tease out. The authors started off with a seemingly plausible biochemical hypothesis that, since serotonin is important in the haemostatic response to vascular injury, its depletion from platelets by the serotonin reuptake inhibitors may increase the risk of bleeding.4 6 The data they obtained are broadly consistent with this hypothesis. However, uncomfortable inconsistencies remain. The authors' classification of clomipramine as a serotonin reuptake inhibitor may cause some surprise but has a sound biochemical basis Clomipramine's serotonin selectivity of about 130, relative to norepinephrine, is close to those of fluoxetine and paroxetine but higher than that of trazodone (see table on Even imipramine's selectivity is not far off that of trazodone (27 versus 53). Indeed, except for nortriptyline and lofepramine, the tricyclic antidepressants shown in the table are serotonin reuptake inhibitors, based on the more recent radioligand binding assays using cloned human transporters.

    Therefore, it would be inappropriate to infer that serotonin reuptake inhibitors may increase the risk of upper gastrointestinal bleeding without first defining what we mean by a serotonin reuptake inhibitor. Structural classification (tricyclics) and biochemical classification (serotonin reuptake inhibitors) clearly clash. If a serotonin effect on haemostatic response is the proposed mechanism for the adverse effect, should we not focus on the size of the dissociation constants for the serotonin transporter rather than, or at least as well as, the selectivity? Doing so reveals other inconsistencies. For example, trazodone is associated with the highest risk of upper gastrointestinal bleeding. Yet it appears to be an outlier among the serotonin reuptake inhibitors with respect to the serotonin equilibrium constants, although the 95% confidence interval for the adjusted relative risk was wide. The authors obtained a pooled relative risk, associated with current use of a serotonin reuptake inhibitor only (fluoxetine, fluvoxamine, paroxetine, sertraline, and clomipramine), of 2.6 (95% confidence interval 1.7 to 3.8) with a corresponding figure of 3.7 (3.2 to 4.4) for those currently using non-steroidal agents only. Concurrent use of both types of drugs yielded a relative risk of 15.6 (6.6 to 36.6).

    Should prescribing practice be changed on the basis of these new data? Greater caution is probably warranted in co-administering non-steroidal anti-inflammatory drugs and serotonin reuptake inhibitors, including clomipramine, particularly for patients with risk factors for upper gastrointestinal bleeding. When both classes of drugs are judged necessary, there is better evidence on the choice of a non-steroidal agent8 than on the choice of a serotonin reuptake inhibitor, or indeed any antidepressant, for reducing the risk of bleeding. For example, changing from indomethacin to low dose ibuprofen is likely to reduce the risk more than changing from paroxetine to imipramine. Whether paroxetine was preferable to imipramine and indomethacin to ibuprofen in the first place is another debate. With greater clinical experience and validation, the newer COX-2 selective non-steroidal anti-inflammatory agents may make a contribution.

    There is an increasing tendency in drug evaluations to lump drugs together, often as part of meta-analyses, to come up with a prized “class effect.” Tatsumi et al7 and de Abajo et al4 remind us indirectly that great care is necessary when doing so. An antihistamine or a tricyclic drug may be a serotonin uptake inhibitor and vice versa. Just like patients, drugs act as individual agents, each with its own three dimensional and electronic structure to exert unique effects on three dimensional receptors.9 10

    Though the general practice database used by de Abajo et al is useful, it may fail to capture all events11 and is not fully representative of prescribing practice. It certainly does not capture self medication adequately. Over the counter antihistamines such as chlorpheniramine and diphenhydramine, which bind to the serotonin transporter and show selectivity towards it, and COX inhibitors, such as aspirin and ibuprofen, are widely used Such use may confound estimates of risk. Therefore, further studies using alternative methods are necessary to confirm these results.


    ALWP has consulted and received research funding from Boots Healthcare International and Novartis, who both manufacture non-steroidal anti-inflammatory drugs.


    • website extra The table appears on the BMJ's website


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