Antidepressants: The scapegoat of poor outcome bipolar disorder?

Like so many constructs within psychiatry, antidepressants in the treatment of bipolar disorder have become the object of a split that renders them as being either ‘all good’ or ‘all bad’ in the eyes of many practitioners. Early case reports in the 1960s and 1970s noted that some depressed patients may develop manic or hypomanic symptoms soon after beginning a tricyclic antidepressant (TCA) or monoamine oxidase inhibitor (MAOI). Those observations in turn led investigators in the 1980s to pose the intriguing hypothesis that antidepressants might acutely nudge some depressed patients from depression into mania or hypomania, overshooting euthymia; longer term, they may increase the frequency of subsequent episodes of either polarity. Despite minimal systematic, prospective research to evaluate this hypothesis, the field seemingly overnight embraced this proposition as dogma. Contemporary guidelines either altogether shun the use of antidepressants for bipolar depression or else grudgingly acknowledge their role as a necessary evil, with the proviso that antimanic drugs precede antidepressant exposure to minimize the risk of destabilization.

What is actually known about mood destabilization from antidepressants in bipolar depression? Perhaps first and foremost, the scope of the alleged problem appears far less than one might think. In the largest meta-analysis of antidepressant-associated mania/hypomania to date, encompassing 114,521 bipolar subjects across 109 antidepressant trials, Tondo and colleagues (2010) found the overall incidence of antidepressant-induced mania or hypomania to be only 12.5%; remarkably, co-therapy with antimanic drugs did not appreciably lower overall risk. Rather than imagine antidepressants to wreak havoc on mood states indiscriminately for any and all depressed bipolar patients, it appears more realistic to consider mood destabilization as a relatively rare phenomenon that occurs with greater susceptibility in a definable high-risk subgroup. Characteristics that may help to identify such higher-vulnerability patients include:

• bipolar I > bipolar II sub-diagnoses (Altshuler et al., 2006);

• current depressive episodes that are preceded by manic/hypomanic (rather than euthymic) phases of illness (MacQueen et al., 2002);

• mixed episodes or sub-threshold mania symptoms during syndromal bipolar depression (Frye et al., 2009; Goldberg et al., 2007);

• a history of comorbid alcohol or substance use (Goldberg and Whiteside, 2002; Manwani et al., 2006);

• a history of past antidepressant-induced mania/hypomania (Truman et al., 2007);

• a genomic predisposition to SSRI-induced mania/hypomania, involving variants of candidate risk alleles such as the serotonin transporter gene polymorphism (Daray et al., 2010).

Additionally, one must account for several other practical considerations when contemplating whether antidepressants might be detrimental to the course of bipolar disorder. These may be summarized as follows:

(a) The field remains divided on the issue of whether mania/hypomania that emerges soon after antidepressant initiation in a hitherto unipolar depressed patient represents the unmasking of a bipolar diathesis (as was the view in DSM-IIIR) or, rather, an adverse drug effect (i.e. secondary mania/hypomania, as now described in DSM-IV, carrying no implications for an ‘underlying’ diagnosis of bipolar disorder or role for ongoing antimanic prophylactic therapy). If indeed an antidepressant serves only to catalyze a manic/hypomanic episode, removal of the catalyst alone should not avert the blossoming of an incipient episode; by contrast, mania symptoms that dissipate soon after stopping a presumptive offending agent likely reflect an adverse drug effect, rather than a core diagnosis. Notably, in Krauthammer and Klerman’s (1978) seminal description of secondary manias, drug-induced manias were found to be more common in people without rather than with pre-existing bipolar disorder; they also had a later onset than usually seen in bipolar disorder, and were seldom associated with a family history of bipolar disorder.

(b) It is inherently difficult to attribute mania symptoms with reasonable certainty to recent antidepressant exposure as opposed to the natural course of illness. Randomized, placebo-controlled data provide the only real means for reconciling that distinction. Notably, in the National Institute of Mental Health Systematic Treatment Enhancement Program for Bipolar Disorder (NIMH STEP-BD), risk for prospectively defined antidepressant-induced mania/hypomania was nearly identical among depressed bipolar patients taking a mood stabilizer plus antidepressant (10.1%) versus mood stabilizer plus placebo (10.7%) (Sachs et al., 2007). Other long-term randomized data within the STEP-BD found no difference in rates of new manias when an antidepressant was stopped versus continued for up to 3 years in conjunction with a mood stabilizer after initial response to their combination (Ghaemi et al., 2010).

(c) Most studies that link antidepressant use with multiple affective episodes (e.g. Schneck et al., 2008) draw upon observational, non-randomized data – making it impossible to know cause versus effect (that is, whether antidepressants cause multiple episodes or whether multiple episodes cause antidepressant prescribing).

(d) Most of the literature examining antidepressant-induced ‘switches’ to mania or hypomania provide no rigorous operational definition of mania or hypomania (such as fulfillment of DSM-IV symptom criteria for a manic or hypomanic episode) (Malhi et al., 2010). Many studies rely on a patient’s own subjective, retrospective recall to identify past ‘episodes’, running the risk for counting as ‘switch events’ a wide variety of phenomena that could include agitation, insomnia, akathisia, anxiety, or merely random mood fluctuations. Relatedly, the time course for attributing a plausible causal connection between antidepressant use and subsequent outcomes is often poorly identified. If a depressed patient became euthymic 2 years after starting a selective serotonin reuptake inhibitor (SSRI) it would be hard to directly link that outcome with drug exposure; yet, clinicians may readily attribute newly emerging mania or hypomania to a longstanding antidepressant regimen, rather than consider the antidepressant as incidental to the intrinsic vicissitudes of the disorder itself. The International Society for Bipolar Disorders (ISBD) proposed guidelines for operationally defining switch events, including acknowledgement that beyond 12–16 weeks after antidepressant initiation or dosage changes, it is difficult to attribute polarity changes to drug effects (Tohen et al., 2009).

(e) To the extent that affective instability may be common in bipolar disorder irrespective of full manic/hypomanic or depressive episodes (Gottschalk et al., 1995; Henry et al., 2001), one can easily misattribute the presence of daily mood fluctuations to adverse treatment effects, rather than to the natural course of illness unaltered by an existing treatment regimen. ‘Mood stabilizers’ are generally studied for their ability to treat or prevent full syndromes, rather than to minimize or ameliorate sub-threshold affective instability in non-syndromal patients. Consequently, it becomes hard to know whether apparent ‘breakthrough’ mood symptoms truly reflect adverse drug effects as opposed to lack of efficacy against a fundamental illness characteristic.

(g) In patients with rapid cycling bipolar disorder, it remains a matter of debate whether antidepressants hasten cycling frequency (as was historically proposed by Wehr and Goodwin (1979)), improve depressive symptoms (as described by Coryell et al. (2003)), or foster more depressive recurrences (as recently reported by Ghaemi et al. (2010)). Given the high prevalence of depressive features among patients with rapid cycling, clinicians should be especially alert to the variety of possible short- and long-term effects that have been described when antidepressants are added to mood stabilizers in this subgroup.

(h) While a small literature espouses a higher risk for antidepressant-induced mania with TCAs or mixed agonists (such as the serotonin/norepinephrine reuptake inhibitor venlafaxine; Post et al. (2006)) than SSRIs or bupropion, existing studies do not account for the aforementioned predisposing risk factors for patient-specific proneness to mood destabilization, limiting broad generalizations about differential antidepressant effects in isolation from the risk profile of a given individual. Small sample sizes further limit the ability to form generalizable conclusions about relative drug outcomes; for example, an oft-cited comparison of adjunctive bupropion or desipramine for bipolar depression that reported a higher risk for switch with the latter agent, involved only 9 and 10 subjects in each respective treatment arm (Sachs et al., 1994).

As debate and speculation continue over whether antidepressants worsen the course of bipolar disorder, there exists a much larger and ominous risk associated with their use: namely, that antidepressants have never been shown to be more effective than mood stabilizers alone for treating bipolar depression. With the sole exceptions of fluoxetine plus olanzapine in bipolar I depression (Tohen et al., 2003) and fluoxetine monotherapy for relapse prevention after initial open-label response in bipolar II depression (Amsterdam and Shults, 2010), the randomized trial literature is saturated with negative findings from placebo-controlled studies for bipolar depression involving SSRIs (e.g. McElroy et al., 2010; Nemeroff et al., 2001; Sachs et al., 2007), TCAs (e.g. Nemeroff et al., 2001) or bupropion (Sachs et al., 2007). That literature stands alongside findings from non-controlled trials with other antidepressants that do not examine antidepressant efficacy above and beyond that seen with a mood stabilizer alone, or else a complete lack of data with agents unstudied for bipolar depression (e.g. desvenlafaxine, duloxetine, and mirtazipine, among others). In general, it is preferable to prescribe treatments that have been studied and found to be efficacious for a disorder before prescribing treatments that have not been studied, or that have been studied but found to be no better than placebo.

Much as antidepressants may destabilize mood for only a minority of depressed bipolar patients, so too they may distinctly improve depression for only a minority of individuals, above and beyond the antidepressant efficacy of mood stabilizers alone. The field has not yet identified predictors of antidepressant-responsive bipolar depression, but empirical data suggest that a full initial response portends better long-term remission with continued antidepressant use for up to 1 year (Altshuler et al., 2009), and that abrupt cessation of antidepressants in a stable bipolar patient may paradoxically induce mania by disrupting homeostasis (Goldstein et al., 1999), together suggesting that if a patient with bipolar depression has the good fortune to improve acutely with an antidepressant in his or her regimen – particularly in non-bipolar I presentations – clinical wisdom dictates changing nothing in the treatment so long as signs of mania/hypomania remain absent. However, for most bipolar patients, more likely scenarios may involve the lack of any effect on mood with traditional antidepressants (either beneficial or adverse), pointing to a niche role for novel therapeutic agents that demonstrate antidepressant properties in bipolar disorder (such as quetiapine, ketamine, riluzole, pramipexole, modafinil, and armodafinil), but are fundamentally different from those medications traditionally regarded as ‘antidepressants’ (Mathew et al., 2008).