4.3 Limitations of the study
There are several limitations in our study. First, drug adherences were unmeasurable in claims data, which can result in misclassification of exposure levels. Although we defined SSRIs by at least two prescriptions to reduce the possibility of including patients who never took the drugs in the first prescription, accurate information on medication use should be further confirmed by patients themselves. Second, older adults with obscure GIB might be misclassified into the control group because an obscure GIB can only be verified by endoscopic and radiologic evaluation[39]. This could result in the underestimation of the association in our study. Third, to improve the likelihood of identifying patients who truly had UGIB, patients with UGIB treated only in outpatient clinics were not included in our analysis. Therefore, our findings may only apply to patients with severe UGIB requiring hospitalization. Fourth, the statistical power may be insufficient to detect mild or moderate associations in dose analysis of some classes of SSRIs such as fluvoxamine and escitalopram, as the number of patients in each category was small.
In conclusion, among the adults aged ≥65 years receiving SSRI treatments, fluoxetine therapy was associated with increased risk of UGIB in a dose-response manner. Moreover, recent users of fluoxetine (<1 month before UGIB occurrence) had higher risk of UGIB than those who used this medication one month ago. Further investigations are needed to clarify whether paroxetine may incur lower risk of UGIB than other classes of SSRIs and the possible underlying causes. As SSRI use increases with age, physicians should accurately evaluate the patients for risk of gastrointestinal bleeding and calculate the total doses of SSRIs in long-term users because UGIB risk may increase with cDDDs. Careful considerations in treatment dosage and close monitoring are needed when providing care to older people.