Antidepressants and Bipolar Disorder: The Plot Thickens

Bipolar depression is responsible for most of the disease burden imposed by bipolar disorder and predominates over the disease course (1, 2). Thus, there is a clear need for effective treatments.

The most appropriate use of antidepressants for bipolar depression is still debated. There is conflicting evidence for their efficacy, both as adjunctive therapy (3–6) and as monotherapy (7). Current guidelines do not endorse adjunctive antidepressants as a first-line treatment for acute bipolar depression and suggest avoiding monotherapy entirely in bipolar I disorder (1). This is in part because of the possible risk of a switch to mania but also because of the potential induction of rapid cycling.

The question of antidepressant-induced affective switch in patients with bipolar disorder has been a long-standing concern. In 1987, Wehr and Goodwin (8) highlighted the lack of randomized controlled trials (RCTs) examining this issue. Decades later, there is still great clinical uncertainty.

In this issue of the Journal, Rohde and colleagues (9) present the results of a “target trial emulation” designed to assess the risk of antidepressant-induced mania in bipolar depression. Although double-blind RCTs remain the gold standard in research, the execution of RCTs has many challenges. Especially with longer follow-up periods, RCTs can be prohibitively expensive, recruitment may be uncertain, retaining participants can be difficult, and generalizability of the results may be dubious.

A target trial emulation using longitudinal observational data may solve some of these challenges. For this study, the authors defined the target trial, created comprehensive eligibility criteria, and followed the appropriate methods to adjust for baseline confounding factors, all of which are key elements of a robust target trial emulation. However, target trial emulations are not without their own limitations. Data that are not collected for research purposes may include inconsistencies, such as ambiguous medical coding and a lack of standardized measurements. Additionally, they are unable to emulate RCTs with regard to placebo control or tight monitoring of treatment adherence (10).

Present Study

Using a combination of Danish registries, Rohde et al. were able to create a target trial emulation containing 979 individuals with bipolar disorder who were recently discharged from an inpatient admission for bipolar depression. The authors were able to achieve a power of 90% based on a true effect size of a hazard ratio of 1.77, taken from a previous meta-analysis (5). Although the trial was well powered, the authors acknowledge that the confidence in the causal relationship between antidepressants and hypomanic or manic episodes (internal validity) was below that of an ideally conducted double-blind RCT.

The target trial emulation included individuals ages 18 and over, who were discharged from their first (index) psychiatric admission with a diagnosis of bipolar depression, defined by ICD-10 codes (11). Individuals who had a diagnosis of bipolar depression or had used antidepressant medication in the 2 years preceding the index admission or had a previous diagnosis of schizophrenia or schizoaffective disorder were excluded. Individuals who were readmitted within 2 weeks after discharge were also excluded from the study.

Patients were described as allocated to the treatment-with-an-antidepressant arm if they redeemed a prescription for an antidepressant within the first 2 weeks of discharge. All others were considered allocated to the not-treated-with-an-antidepressant arm. Many baseline covariates, including (but not limited to) the use of sedatives, severity of bipolar disorder at index admission (defined by ICD-10 codes), psychiatric comorbidities, and year of admission, were taken into consideration.

The primary outcome was a psychiatric admission with a hypomanic or manic episode, with readmission for bipolar depression as a secondary outcome. Each participant’s record was followed until one of the following occurred: primary outcome (psychiatric admission with hypomania or mania), death, 1 year after discharge, or end of the study period.

Study Limitations

Although this target trial emulation provides an interesting alternative to an RCT, there are still several limitations. Target trial emulations using health registries are dependent on the quality and availability of health record data. In many countries, these may not be adequate.

The authors report investigating the level of treatment adherence, but it is not clear or reported how many individuals received acute treatment with antidepressants versus a prolonged maintenance treatment. Recent research (12) has found that continuing adjunctive antidepressant therapy for 52 weeks as a maintenance treatment for bipolar depression was not more effective than acute antidepressant treatment for 8 weeks in preventing relapse of depressive symptoms over a 1-year period. Likewise, that study did not find that prolonged maintenance treatment was more likely to induce a manic episode. Further investigation of treatment approaches in the emulated trial would be beneficial.

Furthermore, the lack of standardized mood measures means that the authors were unable to capture episodes of hypomania or mania or recurrent depression that occurred but were not severe enough for hospitalization or outpatient treatment. Additionally, using ICD-10 codes to measure severity allows for only three categories (mild or moderate, severe without psychotic features, and severe with psychotic features). Without more fine-grained symptom measures, it is difficult to make meaningful conclusions about severity, which was used as a covariate in the statistical analysis.

The authors acknowledge the inherent limitations of identifying individuals based on ICD-10 diagnoses, particularly the lack of differentiation between bipolar I and II subtypes. This distinction is of crucial interest because there is evidence that these groups may respond to antidepressants very differently (1).

Conclusions About Manic Switch

Previous evidence has been conflicting as to whether antidepressants can induce hypomanic or manic switch (5, 7, 13, 14). Furthermore, evidence for manic switch may vary across study types and settings, with higher rates of mania switch being seen in retrospective and cross-sectional studies compared with RCTs, bringing the causality of positive findings into question (15).

There is substantial evidence that switching with antidepressant treatment, either as adjunctive therapy (16) or as monotherapy (17), is much less likely to occur in bipolar II than bipolar I disorder. Bipolar II disorder is dominated more by depression (2), and it is likely that a substantial portion of the sample studied here comprised patients with bipolar II disorder.

The present study found that there were no significant associations between allocation to antidepressant treatment and recurrence of mania or hypomania, including in groups stratified by concurrent treatment with a mood-stabilizing agent or in post hoc sensitivity analyses that included mania or hypomania recorded from outpatient and psychiatric emergency department visits. The authors acknowledged that by excluding patients who were readmitted within 2 weeks, they are unable to comment on acute manic induction.

The lack of delineation between bipolar I and II disorders means that very little clinical guidance can be drawn from the analysis, as the lack of manic switch may reflect the lack of patients with bipolar I disorder in the sample. The authors also cite the statistical uncertainty of the results, which means that a mild protective or harmful effect of the antidepressants cannot be ruled out. All told, then, it seems clear that the current guidance against antidepressant monotherapy in bipolar I disorder should not be amended on the basis of these findings.

There was also no statistically significant difference in the risk of bipolar depression recurrence between the two groups, suggesting a lack of efficacy. However, because only inpatient admissions were measured, and standardized mood scores and outpatient treatment were not, the authors are unable to comment on antidepressant efficacy in a wider context.

Moving Forward—What’s Next?

Although two meta-analyses have shown a small positive effect of antidepressant use in bipolar depression (5, 6), other studies have brought their effectiveness into question. The EMBOLDEN-II trial, an RCT, showed no significant improvement in bipolar depression using paroxetine as a monotherapy, whereas treatment with the mood stabilizer quetiapine did show significant improvement (7). The Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD trial), one of the largest studies, also showed no differences between adjunctive antidepressant therapy and placebo over a treatment period of up to 6 months (4). A recent meta-analysis that included 19 studies has further identified a lack of clinically meaningful improvements when antidepressants are used as an acute, adjunctive treatment option (18).

Rohde and colleagues’ findings, along with recent research, emphasize the need for alternative treatments. Where does that leave us? Recent reviews have highlighted many of the emerging treatments in bipolar depression, such as new atypical antipsychotics (e.g., lumateperone and cariprazine) (19, 20).

Additionally, a recent systematic review and meta-analysis and a scoping review have shown ketamine or esketamine to be a promising treatment for both bipolar depression and treatment-resistant bipolar disorder (21, 22). Most of the studies included showed a positive effect; however, three of the studies reported a recurrence of depressive symptoms after 1–2 weeks, indicating a short-term positive effect. The incidence of manic or hypomanic symptoms was roughly 2.4% across all studies. Most importantly, the small sample sizes and the lack of RCTs indicate that these results should be interpreted with caution. Further research is warranted and is being conducted.

Although requiring more research, psilocybin has become a treatment of interest after its success in treatment-resistant depression. A survey of 541 individuals with bipolar disorder found that users were more likely to find psilocybin more helpful than harmful, even in the presence of adverse effects (23). One open-label, nonrandomized controlled trial found that a single dose of psilocybin was effective for treatment-resistant bipolar II depression and that the illness remained in remission for most participants by the end of the 12-week trial (24). Although these studies support further research of psilocybin as a potential treatment for bipolar depression, the lack of randomized trials, with appropriate control groups and sample sizes, makes it impossible to draw firm conclusions.

Conclusions

Rohde and colleagues utilized an alternative, albeit imperfect, research method to RCTs that may be suitable when the gold standard is not available or achievable.

Although the authors provided some evidence to suggest that the worry of antidepressant-induced manic switch may be overblown, the study is not sufficient to support a change in the current treatment recommendations given the inability to distinguish between patients with bipolar I and II disorders in this sample. It remains clear, however, that new efficacious treatments for bipolar depression are urgently needed.