Is combined use of SSRIs and NSAIDs associated with an increased risk of starting peptic ulcer treatment

De Jong et al. [1] reported that the combined use of nonsteroidal anti-inflammatory drugs (NSAIDs) with selective serotonin reuptake inhibitors (SSRIs) synergistically increased the risk of initiating a treatment course with peptic ulcer drugs as a proxy for gastrointestinal adverse effects. In an attempt to limit confounding by indication, tricyclic antidepressant (TCA) users were used as a control group. As the observed association was not adjusted for potential differences between TCA and SSRI users [1], unmeasured confounding may have affected the estimation of the adverse effects. We therefore addressed the association using another pharmacoepidemiological study design, a so-called prescription sequence symmetry design [2]. The main advantage of this design is that it controls for time-invariant (unmeasured) confounding, because the patient population serves as its own control group. We used data from the same pharmacy prescription database (IADB; http://www.IADB.nl) that de Jong et al. [1] used and applied similar inclusion and exclusion criteria. When assessing the association between initiating treatment A and B, the sequence ratio (SR) is calculated by dividing the number of individuals starting treatment A first and treatment B second by the number of individuals starting treatment B first and treatment A second. For the primary analysis, only patients with a time span between both incident prescriptions between 2 and 28 days were included to limit time-variant confounding. We adjusted for prescription time trends by dividing the crude SR by a null ratio, i.e. the SR obtained assuming no association between both drugs, based on overall prescription rates of both drugs in the total IADB population. We estimated 95% confidence intervals (CIs) of SRs using exact confidence intervals for binomial distributions using STATA 12 software (StataCorp LP, College Station, TX, USA). In total, 50 350 adult patients who initiated SSRI treatment were identified between July 1994 and December 2011. A peptic ulcer drug treatment course was started in 277 of them within a 4 week period of SSRI therapy initiation; 126 (45%) started SSRI therapy prior to peptic ulcer drug treatment, while 151 (55%) patients first started peptic ulcer drug treatment [adjusted (a)SR, 0.83; 95% CI, 0.65–1.06; Table ​Table1].1]. Concurrent use of SSRIs and NSAIDs was associated with a statistically nonsignificant increase in the risk of starting peptic ulcer drug treatment (aSR, 1.48; 95% CI, 0.90–2.49), which did not exceed the risk estimated for NSAID treatment alone (aSR, 2.50; 95% CI, 2.27–2.76). Similar results were obtained for TCA treatment alone and concurrent use of TCAs and NSAIDs (Table ​(Table11). Table 1 Symmetry analysis of selected drug therapy initiation within 4 weeks of peptic ulcer drug therapy initiation using a drug-free run-in period of half a year Given that we used the same database and similar inclusion and exclusion criteria as de Jong et al. [1], the main difference between their study and ours is that we used an alternative design that reduces time-invariant (unmeasured) confounding. In accordance with our results, Dall et al., who applied a case–control design in which they controlled for various potential confounders including alcohol abuse, did not observe an increased risk of uncomplicated peptic ulcers associated with combined use of SSRIs and NSAIDs [3]. As we restricted our analyses to relatively short time spans to reduce the potential influence of time-variant confounding, we could not detect a risk that develops later in the course of treatment. However, de Jong et al. found an increased risk for concomitant use of SSRIs and NSAIDs during an average follow-up of 21 days [1]. In addition, Dall et al. found that the risk of uncomplicated peptic ulcers is greatest during the first 30 days of SSRI treatment [3], indicating that our time span of 4 weeks should be long enough to capture potential drug-related increases in peptic ulcer drug prescribing. We could not capture peptic ulcer drug courses prescribed during hospitalization, measure gastrointestinal adverse effects that required hospitalization or measure over-the-counter NSAID and peptic ulcer drug use. However, de Jong et al. did previously find an increased risk in patients concurrently using SSRIs and NSAIDs, while coping with the same limitations [1]. Therefore, the lack of an increased risk associated with SSRI use, whether used concurrently with NSAIDs or not, is likely not to be related to these potential limitations. In conclusion, our results suggest that the observations reported by de Jong et al. [1] might, at least partly, be attributed to unmeasured confounding. Although comparison with a drug that is used to treat the same indication does limit confounding, it does not eliminate all potential confounders. Therefore, when limited data on potential confounders are available, a self-controlled design can have added value, because it eliminates all time-invariant confounders [4].