Much of what is confusing and controversial in the treatment of bipolar disorder revolves around the role of antidepressants. Are they dangerous or safe? Are they effective or ineffective? Does it matter whether the patient has Bipolar I or II?

Do antidepressants cause manic switching?

In some circles, even asking this question is tantamount to asking if the earth is round. Manic switching seems to be something we see frequently in our clinical practice. The problem is that appearances can be deceiving. Bipolar disorder is, by definition, a cycling condition. If a bipolar patient becomes manic after being put on an antidepressant, a cause-effect relationship may be inferred when, in fact, this may be a part of the natural course of the patient’s illness, unrelated to the antidepressant.

An older, chart review study, though limited methodologically, makes this basic point clearly (Lewis and Winokur, Arch Gen Psychiatry 1982; 39(3):306). The authors reviewed medical charts of 27 bipolar patients who had received no treatment, versus 26 patients who had received treatment with tricyclics. Patients who did not receive treatment had a 41% switch rate, vs. a 28% switch rate in treated patients. While we can’t read too much into the numbers of this observational study, it is a good reminder that the natural history of bipolar disorder is one of shifting mood states, with or without antidepressants on board.

In order to validly answer the question of whether ADs cause manic switching, it is necessary to compare switch rates on ADs to switch rates on placebo. The placebo group provides an estimate of the spontaneous switch rate, allowing us to interpret the significance of switch rates on antidepressants.

In 2004, a meta-analysis was published reviewing 12 placebo-controlled clinical trials in which patients with bipolar depression were randomly assigned to antidepressants vs. placebo or other active treatments (Gijsman et al., Am J Psychiatry 2004;161:1537-1547). Of the 12 studies, 5 reported manic switch rates, and in these studies the overall manic switch rate for patients on antidepressants was 3.8% (11 out of 287 patients) and for patients on placebo, it was 4.7% (23 out of 492). This suggests that antidepressants do not cause manic switching. Of note, most patients were on mood stabilizers as well as an antidepressant.

The researchers wondered whether the low overall switch rate might mask differences between types of antidepressant. They found three trials which compared the rates of manic switches for patients taking tricyclics vs. those taking SSRIs. The switch rate was 8% for tricyclics and 0% for SSRIs (not statistically significant due to low sample sizes). Overall, this meta-analysis suggests that SSRIs do not cause manic switching when patients are on concomitant mood stabilizers. The data is not as clear for tricyclics.

In the largest placebo-controlled trial of antidepressants for bipolar depression, 366 patients with bipolar depression were randomly assigned to mood stabilizer plus antidepressant, or mood stabilizer plus placebo. The researchers chose the two antidepressants widely considered the least likely to lead to manic switches, namely, paroxetine and bupropion. After 26 weeks of treatment, the manic switch rates were nearly identical in the two groups: 10.7% in the placebo group and 10.1% in the antidepressant group (Sachs GS, et al., NEJM 2007;356:1711-1722). There may be some problems with the generalizability of these findings, however, which we’ll discuss in more detail when we talk about the surprising efficacy results of this study.

In another large clinical trial, 159 patients with bipolar depression (all of whom were taking mood stabilizers) were randomized in a double-blind protocol to receive adjunctive treatment with Wellbutrin (bupropion, mean dose, 286 mg/day), Zoloft (sertraline, 192 mg/day), or Effexor (venlafaxine, 195 mg/day). The patients were followed for 10 weeks, and some entered into a continuation treatment arm that went on for up to a year. There were no significant differences in the switch rates to mania or hypomania among any of the three antidepressants, with the overall switch rate being 19.3% at 10 weeks (Leverich et al., Am J Psychiatry 2006;163:232-239). Of course, since this trials was not placebo-controlled, it is hard to understand how to interpret this percentage. If a placebo arm had been included, patients taking placebo may have had a comparable spontaneous switch rate.

In a post-hoc analysis, the researchers reported that venlafaxine had a significantly higher threshold to subthreshold switch ratio than bupropion or sertraline. Unfortunately, the authors do not explain what the clinical meaning of this ratio might be. Nonetheless, they conclude that venlafaxine is more likely to lead to manic switching than bupropion or sertraline.

In sum, the placebo controlled data indicates that antidepressants (when added to a mood stabilizer) do not cause manic switching, and that we may be fooled by our preconceptions in our clinical practices. As always, of course, patients who enroll in randomized studies are not necessarily representative of the patients we see in the office, who may have more severe illness and more comorbid conditions.
Antidepressant monotherapy for bipolar disorder II

What about treating bipolar disorder with antidepressant monotherapy? Most of us would avoid this practice, in part because most treatment guidelines discourage it, but is it as dangerous as feared?

The Gijsman meta-analysis cited earlier reported the results of one study comparing tranylcypromine (Parnate) with placebo in patients with either bipolar I (n = 10) or bipolar II (n = 19) depression; most patients improved and there were no manic switches in either the active or placebo arms (Himmelhoch et al., J Nerv Ment Dis 1982;170:628-634).

More recently, a few studies have examined antidepressant monotherapy of patients specifically with bipolar II depression — intuitively, a safer therapeutic maneuver since any switch would presumably lead to a less severe hypomanic episode. Two studies have already endorsed the efficacy and safety of this approach (Amsterdam et al., Bipol Disord 2004;6:75-81 and Parker et al., J Affect Disord 2006 Jun;92:205-14), and hot off the press is another one. In this new open label study, 83 patients with bipolar II depression were randomized to receive either Effexor XR (mean dose, 186 mg/day) or lithium (mean dose, 966 mg/day). After 12 weeks, patients in the Effexor group improved more both on the Ham-D scale and on response rate (58% for Effexor vs. 20% for lithium). Only one patient in each treatment group experienced a hypomanic switch (Amsterdam and Shults, J Clin Psychopharm 2008;28:171-181).

The bottom line is that controlled trials are increasingly showing that antidepressants are safe and effective, either as monotherapy or as adjunctive therapy in bipolar II depression. Should we be convinced enough by these trials to treat our own bipolar II patients with antidepressants? Clinicians will have to answer that question for themselves.

Do antidepressants improve the short-term course of bipolar disorder?

Researchers break down the treatment of bipolar disorder into three phases: 1) the acute phase, lasting one to two months, and defined as treatment oriented toward treating the most dangerous and life-threatening symptoms; 2) the continuation phase, lasting 6-12 months, defined as continuing to improve symptoms until the patient is really stable, and 3) the maintenance phase, lasting one year or longer.

Most of the studies looking at antidepressants for bipolar disorder have been short-term studies, usually no longer than 10 weeks. The Gijsman meta-analysis found that patients with bipolar depression who are put on antidepressants do better over a 4 to 10 week time frame than patients put on placebo, in terms of both response rates and remission rates (Am J Psychiatry 2004; 161: 1537 – 1547). However, their efficacy analysis was based on only 5 studies, and 75% of patients were taking either a mood stabilizer or an atypical antipsychotic in addition to an antidepressant. Nonetheless, these results appeared to endorse the common practice of putting patients on antidepressants to pull them out of an acute depression.

But then, a large study seemed to dash these hopes. This was the Sachs study mentioned above in which 366 patients with bipolar depression (all on a mood stabilizer) were randomly assigned to adjunctive treatment with an antidepressant (either bupropion, median dose 300 mg/day or paroxetine, median dose 30 mg/day) or placebo. The primary outcome variable was “durable recovery,” defined as being euthymic for 8 consecutive weeks. After 26 weeks of treatment, durable recovery was achieved in 27.3% of patients taking mood stabilizer plus placebo vs. 23.5% of patients taking mood stabilizer plus antidepressant. Basically, antidepressants added nothing of value over and above placebo (Sachs GS, et al., NEJM 2007;356:1711-1722).

Before you reject antidepressants in bipolar patients, though, you should know that the study’s generalizability has been questioned. In an editorial accompanying the paper, Belmaker pointed out that the study initially recruited 4360 patients with bipolar disorder. Of these, 2689 patients had a depressive episode at some point and were therefore eligible to be recruited into the double blind study. But of these, only 366, or 14%, were actually enrolled (Belmaker RH, NEJM 2007;356:1771-1772).

While the study’s authors did not report why patients chose not to be enrolled in the double blind study, we can surmise a range of typical reasons. Some may have responded well to antidepressants in the past, and didn’t want to be randomized into a study where they had a 50/50 chance of getting a placebo. Others may have had poor experiences with antidepressants in the past – or at least to bupropion or paroxetine, which were the ADs used in the study. Others may have been advised by their doctors not to volunteer for other reasons. The bottom line is that the patients who actually got randomized may have particular characteristics that make them less likely to respond to antidepressants than the typical, unselected patient that you and I see in our practice. On the other hand, this argument can cut both ways, because some studies purporting to show a benefit of antidepressant treatment have generalizability problems as well.

Should we avoid maintenance antidepressants in bipolar disorder?

But what about maintenance treatment? We all have patients with bipolar disorder who improved on an antidepressant, and 5 years later, they are still taking it. Often, they will energetically resist any of our well-meaning attempts to discontinue it.

How important is it to discontinue ADs in bipolar disorder? How solid is the evidence that they make things worse over the long term?

To begin, it’s pretty clear that long term studies have failed to demonstrate that keeping patients on antidepressants for years helps patients. In one study, for example, researchers located 7 blinded, controlled clinical trials in which patients were randomized to antidepressants vs. other treatments, usually lithium.

More recently, the STEP-BD study group followed 1,191 patients with bipolar disorder, about 1/3 of whom were rapid cyclers (at least 4 mood episodes over the prior 12 months). They found that patients who received antidepressants over one year of follow-up were 3.8 times more likely to experience rapid cycling, and 1.7 times more likely to experience a single mood episode, than patients not receiving ADs (Schneck CD et al., Am J Psychiatry 2008;165:370-377).

Sounds pretty damning of antidepressants—until you realize that, since this was an observational study, we cannot know whether the antidepressants caused the cycling, or, alternatively, if the cycling caused doctors to prescribe more antidepressants. Let’s face it, when a bipolar patient comes into your office profoundly depressed, it’s very hard to resist prescribing an antidepressant, even when experts are warning that you might be making a bad situation worse.

The authors of the STEP-BD study interpret their findings as implying that antidepressants are “cycle-promoters,” and an accompanying editorial by Nassir Ghaemi wonders if antidepressants are “mood destabilizers,” saying that these results “may be one more nail in the coffin of antidepressant use in bipolar disorder.” (Ghaemi SN, Am J Psychiatry 2008;165:300-301).

I wouldn’t buy a cemetery plot yet. A study by Altshuler, published back in 2003, is still quite influential, and argues in favor of long-term AD use. In this naturalistic study, 84 patients who had a remission to an antidepressant were observed over the next year. The risk of depressive relapse among 43 subjects who stopped antidepressants within 6 months after remission was compared with the risk among 41 subjects who continued taking antidepressants beyond 6 months. Patients who discontinued had a 70% rate of depressive relapse, while those who continued had only a 36% relapse rate (Altshuler et al, Am J Psychiatry 2003; 160:1252-1262). But this study, like others, is flawed in various ways, including the fact that it was not a randomized clinical trial, and that the 84 patients included were selected out of an initial pool of about 1000, limiting generalizability.

So what can we say about long-term use of antidepressants in bipolar disorder? Not much. The data are sparse and of low quality. At this point, the consensus seems to be that it’s best to discontinue ADs after you’ve stabilized your patient, but this is definitely advice in progress.
Bottom-line

We’ve besieged you with data, and we owe you some bottom line recommendations. Here they are:

1. Antidepressants are likely safe and effective for an acute episode of depression in bipolar disorder, particular if the patient is already on a mood stabilizer.

2. The newer antidepressants probably are not more likely than placebo to induce manic switching, even in patients with bipolar I disorder. Tricyclics are probably more dangerous in this regard.
3. Antidepressant monotherapy (i.e., without a mood stabilizer) is probably safe and effective for patients with bipolar II depression.

4. Maintenance treatment with antidepressants may be ineffective for preventing depressive relapses, and, depending on how you interpret the studies, may actually worsen the course of bipolar disorder for patients who are rapid cyclers. So try to wean patients off these meds 6 to 12 months after remission.

5. All this advice is based on the relatively little high quality data available, so our treatment guidelines are likely to change drastically over time, as larger studies are reported.