The purpose of this study was to further evaluate the hypothesis that antidepressants may increase the risk of hospitalization for aspiration pneumonia in the elderly [7
]. This signal arose in the context of a systematic process to screen for iatrogenic factors that may lead to depression, and for complications of depression treatment. The positive association that we observed between antidepressants and pneumonia hospitalization, the primary outcome, before adjusting for clinical variables, is consistent with, but somewhat lower than (1.6 vs. 3.0) the earlier signal that hospitalization for aspiration pneumonia in the elderly is more common after than before a hospitalization for depression. In other words, pneumonia hospitalization does indeed seem to be more common in antidepressant users. However, the earlier study [7
] screened administrative claims data with the goal of identifying a potential wide range of potential signals to be further tested, did not measure actual exposure to antidepressant drugs, and made no attempt to control for patient factors that may change over time. In contrast, the current study focused on examining a particular association, with an emphasis on controlling for confounding.
The results of this study are reassuring that, once comorbidity measures are taken into account, there is no evidence that antidepressants increase the risk of hospitalization for pneumonia in the elderly. This is true for the general class of antidepressants and specifically for cyclic agents and SSRIs/SSNRIs. This conclusion is reinforced by the observation that the occurrence of pneumonia in patients prescribed antidepressants did not deviate from the expected wintertime peak in incidence. In addition to following up a potential signal, and in the process providing reassurance about the safety of antidepressants with regard to pneumonia in the elderly, this study confirms earlier findings that increasing age and the presence of comorbidities are strongly associated with pneumonia [13
This study has limitations. Although the GPRD is derived from the primary medical record, and diagnoses recorded in the GPRD are generally regarded as being of high quality [8
], these study outcomes have not been specifically validated. However, misclassification of the outcome seems unlikely to be responsible for our negative findings because there was a positive association between antidepressants and pneumonia in the minimally adjusted analysis. In addition, we did confirm associations with pulmonary and neurologic conditions, and other factors that would be expected to be associated with pneumonia. Exposure misclassification could also have played a role, in that patients prescribed antidepressant drugs might not take them, which could have tended to produce bias toward the null. However, exposure misclassification would have resulted in bias toward the null in both the unadjusted and adjusted analyses. Because of the large number of cases of hospitalization for depression, and the narrow CIs for this effect estimate, type-2 error is unlikely to have played a prominent role in the primary outcome. However, as we recognized at the outset, type-2 error is more of a concern for the secondary outcome.
In conclusion, we performed a controlled epidemiologic study to follow up on a prior data-derived hypothesis that, if true, could have had important therapeutic implications for elderly patients with depression. Although there was indeed an unadjusted association between antidepressant medication use and pneumonia hospitalization, the observed association appeared to be due to confounding by measured co-morbidity measures. This reinforces the need to independently confirm data-derived hypotheses before acting on them.