Coprescription of antiulcer drugs with SSRIs is fairly common

EDITOR—The paper by de Abajo et al raises the possibility of an increased risk of upper gastrointestinal bleeding associated with selective serotonin reuptake inhibitors.1 The authors excluded patients with cancer, oesophageal varices, Mallory-Weiss disease, alcoholism, liver disease, and coagulopathies, but it is not clear if they excluded patients who were coprescribed antiulcer treatments. They do mention antecedents of upper gastrointestinal disorders as potential cofounders that were adjusted for.

To investigate this issue further we examined a large prescription database provided by the general medical services in the Republic of Ireland. For November 1998 we determined the odds ratios to estimate the relative risk for the coprescription of antiulcer drugs (including H₂ antagonists, proton pump inhibitors, and prostaglandin analogues but excluding antacids) with selective serotonin reuptake inhibitors and other antidepressants (table). We excluded people who were coprescribed aspirin, non-steroidal anti-inflammatory agents, or corticosteroids as this may have been cytoprotective coprescribing. This left a total of 95 929 patients for analysis, of whom 8921 received prescriptions for antiulcer drugs. There was a significant correlation between the prescribing of antidepressants of all types and antiulcer drugs.

In the past, tricyclic antidepressants were used to treat peptic ulcer disease because of their antimuscarinic effects, but because of poor efficacy they are now not used for this. The newer selective serotonin reuptake inhibitors may have fewer antimuscarinic side effects, although they have gastrointestinal side effects such as vomiting, dyspepsia, and abdominal pain. Paradoxically, antidepressants (in particular, selective serotonin reuptake inhibitors) have been recommended in functional dyspepsia.2

Coprescribing of antiulcer drugs with antidepressants may indicate patients with depression who have dyspeptic symptoms, an increased prevalence of peptic ulcer disease in patients with depression, or an increased prevalence of depression in patients with peptic ulcer disease. Whatever the true association between these conditions, an increased risk of bleeding from peptic ulcer disease would be expected on the basis of these data.

SSRIs are no more likely than other drugs to cause such bleeding

EDITOR—De Abajo et al suggest that selective serotonin reuptake inhibitors increase the risk of upper gastrointestinal bleeding, possibly by an effect on the haemostatic function of platelets.1 This conclusion, however, is based on a case-control study that included only cases that required referral or admission to hospital and excluded cases of Mallory-Weiss syndrome or oesophageal bleeding. It is not clear how controls were selected: they might have included patients with an upper gastrointestinal bleed but who had not been admitted to hospital.

Using prescription event monitoring, we have carried out observational studies on large cohorts of patients taking selective serotonin reuptake inhibitors, other antidepressants, and many other drugs.2 This technique uses exposure data from the Prescription Pricing Authority for general practitioners' prescriptions in England and outcome data on events, recorded on forms completed by the prescribing doctor. In our database we have data on 74 drugs, of different therapeutic classes.

We have reviewed our data on events coded as gastrointestinal bleeding during exposure to selective serotonin reuptake inhibitors and two of the other drugs on our database. These two were chosen to show the baseline rate of gastrointestinal bleeding in the population presenting to a general practitioner, where there is probably no confounding by indication or concurrent illnesses and drugs.

The table gives the results, which show a non-significant difference in rate between the two groups. This also applied when the patients were stratified by age (1-59 and ≥60 years): rate ratios (95% confidence intervals) were 1.27 (0.83 to 1.96) and 1.34 (0.83 to 2.18) respectively. If “haemorrhage gastrointestinal upper” was taken as the only outcome event the rate ratio was 1.30 (0.75 to 2.27); again, stratification by age made no significant difference. We also compared selective serotonin reuptake inhibitors with venlafaxine (an antidepressant with mixed serotoninergic and adrenergic effects); the rates were not significantly different (rate ratio 0.75 (0.50 to 1.17)).

The advantage of these analyses is that all gastrointestinal bleeds recorded by the general practitioner are included, not simply those requiring hospital referral. Our results may therefore give a more valid picture of the situation in clinical practice than those in de Abajo et al's study. Thus we can find no evidence to suggest that selective serotonin reuptake inhibitors are more likely to cause gastrointestinal bleeding than other drugs.

Self treatment with non-steroidal drugs may be confounding factor

EDITOR—De Abajo et al concluded that selective serotonin reuptake inhibitors are associated with an increased risk of upper gastrointestinal bleeding, particularly when combined with non-steroidal anti-inflammatory drugs.1 They seemed to show a convincing association between degree of selectivity and relative risk. We have concerns that important factors have been overlooked.

Although patients with alcoholism were excluded, heavy drinkers are generally poorly identified in general practice. The authors might therefore not have identified a number of gastrointestinal bleeds secondary to alcohol misuse. Furthermore, depression is associated with several painful conditions such as menstrual disorders,2 arthritis,3 and back pain.4 These conditions frequently lead to self treatment with over the counter analgesics, which is likely to include substantial use of non-steroidal anti-inflammatory drugs.

General practitioners lack awareness of the prevalence of self treatment with non-steroidal drugs,5 and de Abajo et al consider their prescription data to be complete. But the United Kingdom general practice research database would be unable to identify the use of over the counter non-steroidal drugs in a substantial group of patients—specifically those with depression associated with pain. Therefore undue bias may be placed on the role of the antidepressant in the aetiology of gastrointestinal bleeding. Prospective studies that take such confounding variables into account are required before a causal relation between selective serotonin reuptake inhibitors and gastrointestinal bleeding can be shown.

Authors' reply

EDITOR—Williams et al postulate that the increased risk of upper gastrointestinal bleeding found in association with the use of selective serotonin reuptake inhibitors might be confounded by an association of depression with peptic ulcer disease. We think that this potential association cannot explain our results for three main reasons.

Firstly, we controlled for antecedents of upper gastrointestinal disorders, including dyspepsia. Furthermore, if we restrict our analysis to patients without any upper gastrointestinal antecedent the results are essentially the same (relative risk of upper gastrointestinal bleeding for selective serotonin reuptake inhibitors 3.3 (95% confidence interval 2.0 to 5.5); for the non-selective group 1.8 (1.2 to 2.6)).

Secondly, the increased risk was not observed to the same extent with all classes of antidepressants but mainly with selective serotonin reuptake inhibitors.

Thirdly, if depression were secondary to peptic ulcer the risk of bleeding would revert to the baseline value after the first few months of treatment, but this was not the case, as we stated in the paper.

We have reanalysed our data and included the use of antiulcer drugs (antacids, omeprazole, and H₂ blockers) in the model: the results hardly change (relative risk of upper gastrointestinal bleeding for selective serotonin reuptake inhibitors 2.9 (2.0 to 4.2); for the non-selective group 1.3 (1.0 to 1.8)).

Dunn et al suggest that our results may not be valid because we included as controls patients with minor upper gastrointestinal haemorrhage. We did this because our source of information was a general practice database, not a hospital based database. We identified potential cases through general practitioners' computerised records, which obviously included cases of minor gastrointestinal bleeding. But to enhance the validity of case ascertainment we selected as cases only those with the more serious outcomes. All potential cases (including cases of minor bleeding) not selected were consequently not eligible to serve as controls. Dunn et al do not present any convincing reason why readers should consider their approach to be more valid than ours.

Finally, Dickinson et al raise the interesting hypothesis that use of over the counter non-steroidal anti-inflammatory drugs could partially account for the increased risk associated with selective serotonin reuptake inhibitors, because pain is a common feature of depressive episodes. Unfortunately, we cannot test such a hypothesis, but the lack of a homogeneous pattern with all antidepressants speaks against that bias. Our study was intended to test a hypothesis raised by spontaneous reports and supported by biological evidence. The results obtained are consistent with this hypothesis, but we are aware that further studies are necessary.