 |
 |
|
|
|
Antidepressant drugs are associated with many gastrointestinal side effects (1). Depending on the severity of these effects, treatment may be initiated with medications that suppress the production of gastric acid, thereby adding to the cost and complexity of antidepressant therapy (2,3). There are biologically plausible reasons to suspect that antidepressants may be associated with an increased risk of gastric bleeding; mild disruptions in serotonin-mediated platelet aggregation induced by antidepressants could contribute to gastric bleeding (4,5). Consistent with this hypothesis, 2 large database studies have detected a modest increase in the risk of upper gastrointestinal bleeding among individuals taking SSRIs. These studies have also identified factors that may exacerbate the risk of gastric bleeding when combined with antidepressants, including type of antidepressant, concomitant medications, and age (6,7). Although the aforementioned studies strongly suggest an increased risk of gastric bleeding among individuals taking SSRIs, little information is available to describe the use of GARs in this population, particularly among users who are not elderly or who may not be captured in the previously studied databases. Since all antidepressants can potentially cause gastric side effects, the overall use of GARs may be more common than once thought. Because gastric pathology does not necessarily correlate with the severity of reported symptoms (8,9), a detailed survey of GAR use among individuals taking antidepressants may also provide valuable clues to the true impact of serotonin-mediated gastric side effects. The objectives of this study are to determine whether the prescription of GARs is more likely to occur among generally healthy users of antidepressants and to determine whether GAR use rates differ among antidepressant classes. We also aim to identify predictors of GAR prescribing among antidepressant users. For a controlled evaluation, we compared GAR use by antidepressant users to use by consumers of salbutamol. Salbutamol was used to define the control group for several reasons, including practicality: salbutamol is frequently used in our population (for example, for exercise-induced asthma), allowing us to obtain a control group of a reasonable size. Salbutamol is also usually inhaled, a method of administration not typically associated with significant gastrointestinal side effects (10,11), and is therefore not expected to unduly influence our results. MethodsStudy Design We performed a retrospective cohort study using pharmacy records. This study was submitted to, and received approval from, the Ottawa Hospital Research Ethics Board. Subjects and Database We created a database using data from both CF pharmacies and nonmilitary (that is, civilian) pharmacy sources between June 1997 and November 2002. The database includes information on all medications (both nonprescription and prescription) processed through a CF pharmacy, as well as those medications processed through nonmilitary pharmacies according to a CF-specific drug reimbursement program. The cohort population comprised CF members who received at least one prescription for an antidepressant. Antidepressants were classified as follows: SSRIs (that is, citalopram, fluoxetine, fluvoxamine, parotexine, or sertraline), tricyclic or heterocyclic antidepressants (that is, amitriptylline, clomipramine, desipramine, doxepin, imipramine, maprotiline, nortriptylline, or trimipramine), other nonspecific antidepressants (that is, bupropion, mirtazepine, nefazodone, or trazodone), irreversible monoamine oxidase inhibitors (that is, phenelzine or tranylcypromine), reversible inhibitor of monoamine oxidase (that is, moclobemide), or serotonin-norepinephrine reuptake inhibitors (that is, venlafaxine). Individuals who received bupropion for smoking cessation alone (that is, as the Zyban formulation) were excluded from the analysis. The control group comprised all CF members to whom one or more prescriptions for salbutamol were recorded and who were not also receiving an antidepressant. Definition of Case Exposure and Outcome A new period of exposure, either to an antidepressant or to salbutamol, was defined as a course of treatment lasting at least 29 consecutive days following an initial washout period of 90 days. A 30-day nonuse interval was used to distinguish between 2 separate exposures to the same antidepressant. Duration of exposure was defined from the initial dispensing date to 30 days following the last day of treatment as indicated on the prescription. Consecutive exposures that occurred within 30 days of each other were considered a single exposure period. Exposure periods were defined for salbutamol, for all antidepressants as a group, and for individual classes of antidepressants as listed in Table 1.
The number of incident prescriptions of GAR was the primary outcome measure in this study. This was defined as a new prescription of a GAR lasting at least 30 consecutive days that occurred within 12 months of the initiation of either salbutamol (in the control group) or an antidepressant (in the case group). Thus GAR prescriptions that either predated drug initiation by more than 12 months or that occurred more than 12 months after drug initiation were not analyzed. Data Analysis We calculated ORs using numbers of new GAR prescriptions recorded pre- and postexposure for both antidepressant and salbutamol users overall. These ORs were compared to determine whether GAR use was more likely to occur among patients prescribed antidepressants as opposed to salbutamol. Incident rates of GAR prescriptions were calculated for salbutamol and antidepressant users as the number of new GAR prescriptions recorded during a period of interest, specifically preexposure (that is, prior to initiation of antidepressant or salbutamol) or postexposure (that is, following initiation of antidepressant or salbutamol). The postexposure period was further subdivided into a short-term exposure period, covering the first 4 weeks following processing of the prescription (that is, Days 1 to 28), and a long-term exposure period, covering the days following up to 12 months later (that is, Days 29 to 360). Day 0 was defined as the date on which the prescription was first dispensed. Further calculations were made by subgrouping all users of antidepressants, users of specific classes of antidepressants, and users of salbutamol. We estimated relative risk for incident GAR prescription using a ratio of incident prescription rates. We evaluated statistical significance and precision using 95%CI. To produce adjusted estimates of the relative risks, we performed multiple logistic regression analysis using subjects = age, sex, and concomitant gastric irritant medications as covariates. We analyzed concomitant medications, including antiplatelet agents, antithrombotic agents, oral corticosteroids, biphosphonates, and NSAIDs. NSAIDs were further subanalyzed as either nonselective agents or COXIBs. ResultsWe identified a total of 5588 antidepressant users (cases) and 3059 salbutamol users (control subjects). In the group exposed to antidepressants, the mean (SD) age of subjects was 33.6 (6.9) years; 76% were men, and most (72.2%) were aged between 27 and 41 years. Demographic characteristics were similar in both the exposed and control groups (data not shown). Of the antidepressant users, 1966 (35.2%) consumed at least one concomitant medication of interest. There were 8722 exposure periods identified for any antidepressants and 9974 exposure periods identified for separate categories of antidepressants (that is, 1102 exposures occurred in patients who were not antidepressant-naVve) among members of the exposed group. Most exposures (41%, n = 4089) were to SSRIs. This category also presented with the longest mean duration of exposure (Table 1). We identified 388 incident GAR prescriptions (4.45% of all exposures) in the post antidepressant exposure period, compared with 58 such prescriptions (1.4%) in the same period for salbutamol users (Figure 1). A negligible number of postexposure GAR prescriptions were initiated concurrently with either salbutamol or antidepressants. Most new GAR prescriptions, however, were initiated after 29 days of therapy (71%, 274 of 388 of antidepressant users, compared with 81%, 47 of 58 of salbutamol users).
Among both salbutamol and antidepressant users, the number of new (incident) GAR prescriptions decreased following initiation of drug therapy (Figure 1). The rate of decrease was greater, however, for indiduals who received salbutamol, compared with those who received antidepressants (OR 1.38; 95%CI, 1.12 to 1.71; P = 0.03; Table 2).
Further analysis of the incident rates of GAR prescription also demonstrates an increased risk for antidepressant users, with OR 4.93 (95%CI, 2.66 to 9.21; P < 0.001) in the short-term, and OR 2.83 (95%CI, 2.05 to 3.9; P = 0.001) following long-term use. The classes of SSRIs, tricyclic or heterocyclic, and other nonspecific antidepressants (that is, trazodone, nefazodone, bupropion, and mirtazepine) were associated with significantly increased risk of GAR use, both in the short- and long-term (ORs ranging from 3.53 to 5.58 and 2.25 to 3.10 for short- and long-term exposure, respectively). Logistic regression analysis (performed with incident GAR rates) also identified any antidepressant use as a predictor of GAR initiation. Adjusted ORs were significant both in the short- and long-term (2.53; 95%CI, 1.90 to 3.37; P = 0.001 and 2.12; 95%CI, 1.54 to 2.91; P = 0.001, respectively) for antidepressant use, compared with salbutamol. Logistic regression analysis for individual antidepressant classes identified SSRIs as the strongest predictor of incident GAR prescription, both in the short- and long-term (OR 2.61; 95%CI, 1.92 to 3.56; P = 0.001 and OR 2.32; 95%CI, 1.64 to 3.28; P = 0.001, respectively). Our logistic regression analysis did not identify sex or age as a predictor of GAR use. This analysis also identified the concomitant use of antiplatelet agents, bisphosphonates, oral corticosteroids, and NSAIDs as predictors of GAR use among antidepressant users, both in the short- and long-term, following antidepressant initiation (data not shown). Use of COXIBs was also a significant predictor of GAR prescription only, but among individuals who received more than 4 weeks of treatment. This risk was significant only in the short-term following antidepressant initiation (OR 2.42; 95%CI, 1.17 to 5.01; P = 0.017). DiscussionThis database study was designed to evaluate the patterns of use of gastric acid suppressants in a generally healthy population of antidepressant users. Despite a relatively short observation period, we detected an increased risk for GAR use among users of antidepressants, compared with users of a gastrically neutral control drug, salbutamol. Comparison With Existing Literature The results of this study are consistent with those previously reported. In the database studies of de Abajo and van Walraven, a modest increase in risk of gastric bleeding was identified among SSRI users (6,7). More recent studies evaluating the gastric effects of SSRIs, in particular, have also detected similar increases in risk (12,13). Owing to the nonspecificity of our marker of gastric bleeding (that is, GAR use), we obtained larger ORs that likely reflect the greater sensitivity of our marker for milder gastric disturbances, in addition to more serious gastrointestinal pathology. The participants in our study also differ substantially from those in the reports cited above. In contrast to the population-based databases used by the above authors, our database consists primarily of currently serving military personnel who are younger and healthier than the general population. The identification of an increased risk for GAR use in our study population strongly suggests that the gastric side effects of SSRIs may be more substantial than previously thought, particularly because this association remained significant even after adjusting for age, sex, and concomitant medication use. The results of our study remain robust when the characteristics of our control group are more closely scrutinized. Because asthma may present with symptoms similar to reflux (that is, retrosternal chest pain), a masking effect was possible. The finding of an increased risk of GAR use in the antidepressant consumers, despite similar potential for such increased use among the asthmatic control group, suggests that the increased risk is associated specifically with antidepressant use. In our study, the SSRIs had the highest adjusted association with increased risk for GAR use among all classes of antidepressants represented in our study population. However, the increase in absolute risk, although definite, was small; initiation of any antidepressant was associated with only 3.05% increased incidence rate for GAR use overall, compared with salbutamol. Although the differences in risk between the antidepressant categories were not significant, they are consistent with other reports. Of note, TCAs were also associated with an increased risk for GAR use in our study. These results are contradictory to the results of de Jong (12) and Dalton (13), although their results may be attributed to the small number of observations recorded. We were also unable to detect a predictive effect of age on risk of GAR use, likely owing to the relatively limited age range of our study population, compared with other populations studied to date. We also found that concomitant medications have an independent effect on risk of GAR prescription. Because we analyzed the COXIBS separately from the nonselective NSAIDs, we were unable to detect a uniformly increased risk of GAR prescription in this subgroup of antidepressant users. It is not clear whether this difference in risk profile is owing to a true gastroprotective effect of the COXIBs or to the relatively small number of users of such drugs. Consistent with the literature, we also identified an increased risk with other medications known to contribute to gastric bleeding, specifically, antiplatelet agents, corticosteroids, and NSAIDs. These observations warrant particular attention since they suggest that careful consideration should be given to concurrent treatment with antidepressants, even in younger patients who are free of preexisting gastric illness. Study Limitations Despite the use of multiple logistic regression models, which allow statistical adjustment for age, sex, and concomitant medications, this study possesses the same weaknesses as all database studies: we are unable to establish a definite causal relation between exposure and outcome. Although adjustment for concomitant medications can help minimize the potential effect of other agents on the risk of bleeding, it cannot serve as an adequate adjustment for morbidity. In our case, the increased risk detected may reflect an increased risk for gastric symptomatology among patients with psychiatric conditions (14,15), compared with patients with respiratory illness alone. Antidepressants are also frequently used to treat various nonpsychiatric illnesses, including chronic pain conditions (16,17) and migraine headaches (18). These conditions often provoke the use of certain analgesics (for example, opioids and anticonvulsants) that may themselves cause gastric disturbances. Although our salbutamol control group did not differ demographically from the group of antidepressant users (data not shown), we were unable to fully control for medical conditions and other factors that may predominate in one group. This explanation is consistent with the higher rate of GAR use prior to treatment initiation, which was observed in the antidepressant group. We were also unable to evaluate nonmedicinal factors that are known to contribute to gastric disturbances, such as alcohol use. Given that substance use disorders, including alcoholism, occur more frequently in those with psychiatric illness, the increased use of GARs in the antidepressant cohort is not unexpected. Our results may reflect this comorbidity and not an increased risk of gastric bleeding or reflux from antidepressant therapy per se. Owing to the nonspecificity of GAR prescribing as a marker for gastric bleeding, from our study results, it is impossible to determine whether antidepressants are truly associated with gastric bleeding. However, given the significantly increased use of GARs in the antidepressant user cohort, it is possible that the gastrointestinal side effects from antidepressants (which can contribute to the use of GARs) far outweigh the risks of gastric bleeding, particularly in our young and relatively healthy population. ConclusionUse of an antidepressants was associated with an increased rate for new GAR prescription, which may reflect an increased risk for gastrointestinal side effects. This association remained significant after adjusting for age, sex, and use of concomitant gastric irritants. These results suggest that prescription of antidepressants should be undertaken with due consideration of the added risk of gastric bleeding and significant gastroesophageal reflux disease, even among young, generally healthy adults without significant comorbidity. Funding and SupportThis study was funded entirely by research funds from the Department of National Defence. Dr Boddam has received speaker fees from Solvay Pharma for lectures on antidepressants. None of the other authors received sponsorship fees or honoraria from pharmaceutical manufacturers or advocacy groups relating to the use or prescribing of antidepressants or gastric acid suppressants. AcknowledgementsA preliminary investigation, which formed the basis for this study, was performed by Dr L Maria Gutschi, BScPhm, PharmD. References1. McManis PG, Talley NJ. Nausea and vomiting associated with selective serotonin reuptake inhibitors: incidence, mechanisms and management. CNS Drugs 1997;8:394–401. 2. Stahl SM. Placebo-controlled comparison of the selective serotonin reuptake citalopram and sertraline. Biol Psychiatry 2000;48:894–901. 3. Spigset O. Adverse reactions of selective serotonin reuptake inhibitors: reports from a spontaneous reporting system. Drug Saf 1999;20:277–87. 4. Li N, Wallen NH, Ladjevardi M, Hjemdahl P. Effects of serotonin on platelet activation in whole blood. Blood Coagul Fibrinolysis 1997;8:517–23. 5. Javors MA, Houston JP, Tekell JL, Brannan SK, Frazer A. Reduction of platelet serotonin content in depressed patients treated with either paroxetine or desipramine. Int J Neuropsychopharmacol 2000;3:229–35. 6. De Abajo R, Garcia-Rodriguez LA, Montero D. Association between selective serotonin reuptake inhibitors and upper gastrointestinal bleeding: population based case-control study. BMJ 1999;319:1106–9. 7. Van Walraven C, Mamdani MM, Wells PS, Williams JI. Inhibition of serotonin reuptake by antidepressants and upper gastrointestinal bleeding in elderly patients: retrospective cohort study. BMJ 2001;323:655–8. 8. Okamoto K, Iwakiri R, Mori M, Hara M, Oda K, Danjo A, and others. Clinical symptoms in endoscopic reflux esophagitis: evaluation in 8031 adult subjects. Dig Dis Sci 2003;48:2237–41. 9. Frezza M, Gorji N, Melato M. The histopathology of nonsteroidal anti-inflammatory drug induced gastroduodenal damage: correlation with Helicobactor pylori, ulcers, and haemorrhagic events. J Clin Pathol 2001;54:521–5. 10. Schindlbeck NE, Heinrich C, Huber RM, Muller-Lissner SA. Effects of albuterol (salbutamol) on esophageal motility and gastroesophageal reflux in healthy volunteers. JAMA 1998;260:3156–8. 11. Albuterol monograph. In: American Society of Health-System Pharmacists. American Hospital Formulary Service 2004. Bethesda (MD): American Society of Health-System Pharmacists; 2004. p 1233. 12. De Jong JC, van den Berg PB, Tobi H, de Long-van den Berg LT. Combined use of SSRIs and NSAIDs increase the risk of gastrointestinal adverse effects. Br J Clin Pharmacol 2003;55:591–5. 13. Dalton SO, Johansen C, Mellemkjaer L, Norgard B, Sorensen HT, Olsen JH. Use of selective serotonin reuptake inhibitors and risk of upper gastrointestinal tract bleeding: a population-based cohort study. Arch Intern Med 2003;163:59–64. 14. Naliboff BD, Mayer M, Fass R, Fitzgerald LZ, Chang L, Bolus R, and others. The effect of life stress on symptoms of heartburn. Psychosom Med 2004;66:426–34. 15. Avidan B, Sonnenberg A, Giblovich H, Sontag SJ. Reflux symptoms are associated with psychiatric disease. Aliment Pharmacol Ther 2001;15:1907–12. 16. Salerno SM, Browning R, Jackson JL. The effect of antidepressant treatment on chronic back pain: a meta-analysis. Arch Intern Med 2002;162:19–24. 17. Max MB. Treatment of post-herpetic neuralgia: antidepressants. Ann Neurol 1994;35 Suppl:S50-3. 18. Punay NC, Couch JR. Antidepressants in the treatment of migraine headache. Curr Pain Headache Rep 2003;7:51–4. Author(s)Manuscript received April 2005, revised, and accepted August 2005. 1. Outcomes Research Pharmacist, Canadian Forces Health Services, Ottawa, Ontario. 2. Formerly Pharmacy Branch Advisor, Canadian Forces Health Services; Currently Director of Pharmacy, Children’s Hospital of Eastern Ontario, Ottawa, Ontario. 3. Practice Leader, Psychiatry and Mental Health, Canadian Forces Health Services, Ottawa, Ontario. 4. Biostatistician, JSS Medical Research Inc, Westmount, Quebec. 5. Associate Professor, Department of Surgery, McGill University, Montreal, Quebec; Associate Professor, Department of Epidemiology and Biostatistics, McGill University, Montreal, Quebec. Address for correspondence: Dr J Ma, 1745 Alta Vista Drive, Room 214A, Ottawa, ON K1A 0K6 e-mail: ma.j@forces.gc.ca
1 | 2
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||