In this study, we found that antidepressant use was associated with about 10% decreased risk of hospitalization due to psychosis and 20% decreased risk of all-cause mortality. None of the studied drugs were related to an increased risk of psychosis or mortality. However, we found that the use of antidepressants was associated with a slightly increased risk of hospitalization due to non-psychiatric reasons.
We found an association between a lower risk of psychosis hospitalization with antidepressant use in general and with certain antidepressants. This is in line with the findings of Stroup et al. who found a lower risk of hospital admission due to psychiatric reasons for antidepressant initiators (aHR 0.84, 95% CI 0.80–0.88)12. However, they investigated add-on antidepressant use comparing it with the initiation of another antipsychotic as a reference, and included hospitalization due to any mental disorder for this outcome. These differences make direct comparisons between the studies difficult. Stroup and colleagues did not assess specific antidepressants for this outcome. Our study showed a lower risk of psychosis hospitalization associated with sertraline, fluoxetine, citalopram, venlafaxine, and escitalopram. We did not observe a lower risk associated with mirtazapine and mianserin use. One reason for this might be that they are widely used for insomnia in Finland, and usually, the dosage with this indication is relatively small compared to other indications, especially with mirtazapine27. We found an association with antidepressant polytherapy, referring to the concomitant use of two or more antidepressants, and increased risk of psychosis hospitalization in between-individual analysis. More severely ill patients are often treated with several medications, which might explain this finding.
We found a 20% lower risk of mortality associated with antidepressant use. Similar findings considering antidepressant use have been found in a previous study comparing different levels of antidepressant exposure with non-use (15–40% lower risk depending on the level of exposure)9. Another study comparing the levels of antidepressant exposure found 17% lower risk associated with low exposure and 9% lower risk with moderate exposure, but there was no difference in risk with the highest exposure10. Similar trend was also seen in a previous Finnish study, even though the results were not statistically significant (HR 0.57, 95% CI, 0.28–1.16)13. However, the study cohort was considerably smaller (N = 2588) than in our study. Stroup and colleagues also evaluated mortality outcomes and they did not find an association between antidepressant use and mortality (aHR 0.97, 95% CI 0.81–1.17)12. Considering specific drugs, we found a lower risk of mortality with sertraline, escitalopram, amitriptyline, mianserin, and citalopram compared to non-use of antidepressants, and none of the studied specific drugs was associated with increased risk of mortality.
There was a somewhat increased risk of non-psychiatric hospitalization associated with citalopram (5% increased), escitalopram (8%) use, and polytherapy (8%). Citalopram and escitalopram were so widely used altogether that this may have affected to results of all antidepressants observed as a whole. Similarly, as Stroup and colleagues12, we found no association with cardiovascular hospitalizations, and it seems we lacked the statistical power with this outcome (as indicated by wide confidence intervals). Antidepressant use has been associated with certain adverse events28, such as fractures, cardiovascular events, and hyponatremia. However, if these adverse effects would be very severe, contrary to what we found, it probably would have been seen as a higher risk of mortality. Another possible explanation for the small increase in the risk of non-psychiatric hospitalization is that reduction of negative or depressive symptoms, or reduction in diagnostic overshadowing led to improvement in detection of physical diseases and consequently, resulted in a necessary hospitalization.
The strengths of this study are that we used a large nationwide study cohort, including all persons diagnosed with schizophrenia in inpatient care in Finland during a moderately long time period. The inpatient care register has high specificity to identify schizophrenia diagnosis (i.e. no false positive cases)29. In addition, our study enabled a long-term follow-up of up to 22 years, which is required for relatively rare outcomes such as mortality. PRE2DUP method, which we used in this study to derive time periods when drugs were used vs. not used, has been evaluated to estimate antidepressant and antipsychotic use well26,30. For the hospitalization-based outcomes, we used within-individual design, which eliminates the impact of all time-invariant factors such as sex and needs to be adjusted for only time-varying factors. These were sequential order of treatments, time since cohort entry, and use of certain medications.
There are certain limitations in this study. We did not have information on depressive symptoms which can act as a confounder. It is likely that antidepressant use is initiated when the symptoms are present or worsening but the full efficacy of antidepressants is reached after a delay of 2–4 weeks. To overcome this, we ran analyses by omitting the first 30 days from all exposures and the results remained nearly the same as in the main analysis. However, as it was not possible to remove the impact of fluctuation or severity of depressive symptoms in time, our results on the effectiveness of antidepressants in preventing hospitalizations due to psychosis might be an underestimation. Although the diagnosis of schizophrenia is often made in inpatient care, some patients with a less severe course of illness who have been treated only in outpatient care may be missing from our study cohort. Adherence to medication is a common problem in persons with schizophrenia. From the register-based data, we only have information on the dispensed drugs, and it is possible that they are not actually taken. Another limitation is that Finnish registers do not record indications for prescribed medications, i.e., mirtazapine might have been used for the treatment of insomnia, often with relatively low doses, which might affect our results.
In conclusion, antidepressant use was associated with decreased risk of hospitalization due to psychosis and lower mortality among patients with schizophrenia. Although there was a somewhat increased risk of non-psychiatric hospitalization associated with citalopram and escitalopram, we observed a decreased risk of mortality with these substances and with antidepressant use in general, which suggests that antidepressants are relatively safe to use in persons with schizophrenia. However, more studies evaluating the safety and benefits of antidepressant use in persons with schizophrenia are needed.