Abstract
Introduction
There are few areas in psychopharmacology that are more controversial or arouse stronger opinions than the use of antidepressants in bipolar disorder. Additionally, scholarly opinions in this area, which typically caution against the use of antidepressants, conflict greatly with the widespread use of these agents for bipolar depression (Baldessarini et al., 2007). At first glance, this would seem unexpected since the literature is not extensive, allowing all to read and assimilate all the relevant data. Yet, from reading individual papers or scholarly reviews, one might imagine that antidepressants for bipolar depression are either ineffective, dangerous and wildly overused treatments that should be prescribed only sparingly after all other potential options have been exhausted or helpful treatments that add meaningfully to our pharmacological options, with some potential risks but with more benefits. Of course, both of these views are caricatures. Surely, antidepressants can be exceedingly helpful for some bipolar patients while they can be destabilizing for others. The paucity of the data and the likelihood that different antidepressant classes confer different risks for bipolar patients both contribute to the often polarized tone of the discussion.
Methodological questions
The lack of an established methodology for studying the efficacy of antidepressant also contributes to the lack of consensus. As examples of the methodological questions:
Should all antidepressants be evaluated as add-ons to mood stabilizers?
If so, how long should the mood stabilizers be prescribed at a constant dose prior to antidepressant initiation?
Given the difference in both their natural history and in their switch potential in association with antidepressants, should there be different guidelines for the use of antidepressants in bipolar I vs bipolar II depression?
If antidepressants are prescribed for bipolar depression, how long should they be prescribed after symptomatic improvement (i.e. how long is the appropriate length of continuation treatment for bipolar depression a topic with no relevant controlled literature)?
Can we more reliably identify certain subgroups of bipolar patients with different risk/benefit ratios for antidepressants?
Summary of the controlled data
Over the last few years, a number of thoughtful overviews and meta-analyses have been published evaluating both the efficacy and risks (switch into mania/hypomania or cycle acceleration) of antidepressants for bipolar disorders (Ghaemi et al., 2008; Licht et al., 2008; Sidor and MacQueen, 2011; Tondo et al., 2010). As noted, their conclusions have varied. Nonetheless, it is worth a brief summary of the controlled data in this area.
For efficacy, the most efficient summary statement would be that the effectiveness of antidepressants for bipolar disorder is still unproven. In the best recent meta-analysis (Sidor and MacQueen, 2011), antidepressants showed a small but non-significant benefit compared to placebo, [relative risk (RR) = -1.18; 95% confidence interval (CI), 0.99–1.40; p= 0.06]. There was significant heterogeneity across studies due to the somewhat greater efficacy of placebo in the largest study. Comparable remission data also trended towards efficacy but were not significantly different (RR=1.20; 95% CI, 0.98–1.47; p= 0.09).
When the vast majority of patients are treated with concomitant mood stabilizers, meta-analysis shows no differences in affective switching in short-term studies compared to placebo (RR=0.97; 95%CI, 0.62–1.53; p= 0.90).
Predictors of antidepressant-induced switching
It would certainly be helpful if we had consistent predictors of which bipolar patients were most likely to switch into mania when antidepressants were added to their regimens. Many clinical features have been proposed (Salvi et al., 2008; Visser et al., 2005), such as mixed depressive features (Frye et al., 2009), hyperthymic temperament (Henry et al., 2001), bipolar I vs bipolar II disorder (Bond et al., 2008) and rapid cycling (Seretti et al., 2003), although some of these features may predict switching into mania regardless of antidepressant treatment (Perlis et al., 2010). Here too, however, consistent findings across studies are lacking. Selective serotonin re-uptake inhibitors (SSRIs) and bupropion are also associated with lower switch rates in most studies (Salvi et al., 2008; Sidor and MacQueen, 2011).
Is hypomania always bad?
However, we should also consider whether a switch into hypomania should always be considered a negative outcome. Will it hurt the patient’s function and always negatively impact their life? At least one author has suggested that mild hypomania should not always be considered such a bad thing and will not necessarily require vigorous active treatment or prevention (Parker, 2012). Since bipolar II patients are at lower risk for switching than bipolar I patients (Bond et al., 2008), and hypomania may not always be such a negative outcome (whereas mania is, by definition, a destructive state), our treatment recommendations may differ between bipolar I and II patients.
Complex data, complex interpretations: The absence of functional ratings as outcome variables
Implied in the discussion above is that it is not always easy to describe outcomes as positive or negative. As an example of how differing interpretations can be derived from a single data set, consider the double-blind, controlled study by Amsterdam and Shults (2010) on longer-term treatment (1 year) in bipolar II depression in which short-term responders to fluoxetine were randomly assigned to 1 year of treatment with either continued fluoxetine, lithium or placebo. Those subjects who continued on fluoxetine had fewer depressive relapses. There were no significant differences in a priori defined hypomanic episodes or mean Young Mania Rating Scale (YMRS) scores across the three treatment groups. However, examining YMRS ratings, it is clear that there was more mood fluctuation/variability in those treated with fluoxetine compared to the other two groups. So, did this study show greater efficacy with no more switches in those patients treated with fluoxetine for 1 year or should we describe the findings as showing efficacy but at the price of greater mood instability? Of course, for different individuals, that risk/benefit ratio may differ. Additionally, neither this study nor the vast majority of the others in the area measure functional outcome. One could argue that some mood instability is a small price to pay to avoid depression. Adding functional outcome measures may help clarify the comparative risks vs benefits.
Cycle acceleration
Concerns about cycle acceleration – a destabilizing of the course of bipolar disorder – has also been a consistent feature since the tricyclic era derived primarily the evidence from small studies of highly selected patients (e.g. Wehr et al., 1988) about the inverse relationship between cycle length and antidepressant use in bipolar individuals. Yet, no comparable data have emerged with the newer antidepressants (SSRIs, bupropion, and so forth). Are the newer antidepressants mood destabilizing? In the Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD) study, an association was found between prescription of antidepressants and rapid cycling (Schneck et al., 2008). However, as noted by Coryell et al. (2003), since episodes of depression – which then trigger the prescription of the antidepressant– often precede rapid cycling, the relationship between rapid cycling may be associative, not causal.
Maintenance antidepressants in bipolar disorders
The last important area of concern and marked disagreement is in the use of longer-term antidepressants in combination with a mood stabilizer for depression-prone bipolar patients. A number of naturalistic studies (Altshuler et al., 2001; Altshuler et al., 2003; Joffe et al., 2005) have described subsets of bipolar patients who seem to do best on a long-term combination of mood stabilizers and antidepressants. In these reports, those on the combination treatment have fewer depressive episodes compared to those treated with mood stabilizers alone with no increased risk for mania. Other studies (Ghaemi et al., 2010), however, demonstrated that antidepressants add little to mood stabilizers to depressive relapse prevention in bipolar patients. As an example of the need for longer-term studies in the area, in the most provocative recent report, bipolar I depressed patients were given paroxetine monotherapy (McElroy et al., 2010). Over 8 weeks, switch rates between paroxetine and placebo did not differ. Would this be true over a longer time frame or would an unopposed antidepressant eventually lead to both more switches and greater mood instability?
Clearly, one size never fits all (at least not in psychopharmacology) but it is reasonable to assume that there is a subset of patients for whom long-term antidepressants as a part of their treatment regimen – which, presumably, would usually include a mood stabilizer – is optimal for depressive relapse prevention without increased risk for manic relapses. We need to identify the size of this subgroup and, if possible, predictors so that they can be identified more easily than by trial and error.
Conclusion
It is a cliché to urge the design, execution and publication of more studies on antidepressants for bipolar depression but, above all, that is indeed what is needed. These studies should draw from a relatively unselected group of bipolar patients, include bipolar I and II individuals, include rapid cyclers and other subtypes – to ascertain for which subgroups the risks associated with antidepressants are greater than for others – should extend beyond the 6–12 week short-term time frame so that effects on changes in cycling frequency and depressive relapse prevention may be examined and, in some studies, include active comparators such as lithium, lamotrigine and quetiapine. Until we have such studies, the contrast between the regularity of antidepressant prescriptions for bipolar patients on the one hand and the consistent warnings against their overuse on the other will continue, as will the heated opinions on the topic as opposed to data-driven consensus recommendations. May that day come soon.