Amsterdam and colleagues' findings

Amsterdam et al ^{Reference Amsterdam, Luo and Shults5} report that not only did patients with a history of rapid-cycling bipolar II disorder show no increased risk of treatment-emergent affective switches across 50 weeks of fluoxetine monotherapy, but they actually had a nominally better outcome than the patients who were randomly assigned to be switched from fluoxetine to lithium. Do these provocative findings indicate that our current practice guidelines need to be rethought or does a careful reading of this study reveal alternative explanations for such discrepant data?

Perhaps the most important methodological detail of this study concerns the way that Amsterdam and colleagues classified ‘rapid cycling’. ^{Reference Amsterdam, Luo and Shults5} Specifically, unlike the DSM-IV-TR criteria ⁶ for a rapid-cycling pattern, which specify that the individual must have experienced four or more distinct episodes during the past year, the patients in the current study were classified as ‘rapid cycling’ if they had an average of four or more affective episodes per year during the course of their illness. There is no doubt that this definition identified patients with a large number of lifetime episodes: their ‘rapid-cycling’ group had experienced more than three times the number of lifetime depressive episodes (22.8 v. 6.6) and five times more lifetime hypomanic episodes (45.0 v. 9.0) than the ‘non-rapid-cycling’ group. But, this definition did not ensure that the ‘rapid-cycling’ group actually met the DSM-IV-TR criteria for the past year, nor did it ensure that the ‘non-rapid-cycling’ group actually did not meet the DSM-IV-TR criteria. We do not know how many of the participants in Amsterdam et al's study would have met the DSM-IV-TR criteria for a rapid-cycling pattern. But we do know that the average length of the index depressive episode of their rapid-cycling group was reported to be of 13.3 months' duration, which essentially precludes the possibility that many of the patients in this group were actually in a rapid-cycling pattern of illness at the time they entered this study. Moreover, as the average index depressive episode of the non-rapid-cycling group was reported as only 4.0 months, their rapid-cycling-group was paradoxically more chronically and persistently depressed than the non-rapid-cycling group at study entry.

A second important methodological detail pertains to the design of the larger study from which these data were drawn ^{Reference Amsterdam and Shults7} : all of the patients included in this report had all been first treated with up to 80 mg/day of fluoxetine for 12 weeks before entering the 50-week, randomised, placebo-controlled discontinuation/substitution phase of this study. Moreover, only those patients who stably remitted on fluoxetine therapy, as defined by a Hamilton Rating Scale for Depression (HRSD) score of eight or less at week 12, entered the next phase of the study. As such, the longer-term study examined a patient group that was ‘enriched’ not only for response to fluoxetine, but also for the absence of treatment-emergent affective shifts during fluoxetine monotherapy.

The most parsimonious explanation of the results of Amsterdam et al ^{Reference Amsterdam, Luo and Shults5} is that there is a subset of patients who meet criteria for bipolar II depression who not only can remit on fluoxetine monotherapy but can also continue on this therapy for up to 1 year without an increased risk of treatment-emergent affective shifts. Further, their findings suggest that these patients can do well on fluoxetine monotherapy whether or not they have a past history of many past depressive or hypomanic episodes. As such, these results do not inform treatment decisions for patients who have experienced four or more episodes during the year before their decision to seek treatment.

Despite growing recognition that bipolar II disorder may be more prevalent than bipolar I disorder, there are few controlled studies of bipolar II depression and no definitive, placebo controlled studies of antidepressants (see, for example, Swartz & Thase ^{Reference Swartz and Thase8} ). The current study provides additional information about the utility of antidepressant monotherapy for at least some patients with bipolar II depression. In this respect, the results are fully consistent with a series of earlier reports by this research group, ^{Reference Amsterdam and Brunswick9} including prior studies of fluoxetine ^{Reference Amsterdam, Garcia-España, Fawcett, Quitkin, Reimherr and Rosenbaum10-Reference Amsterdam, Shults, Brunswick and Hundert13} and venlafaxine. ^{Reference Amsterdam and Garcia-España14-Reference Amsterdam, Wang and Shults17} As other investigators have observed venlafaxine to be associated with a high risk of treatment-emergent affective shifts in patients with bipolar depression taking concomitant mood stabilisers, ^{Reference Leverich, Altshuler, Frye, Suppes, McElroy and Keck18,Reference Vieta, Martinez-Arán, Goikolea, Torrent, Colom and Benabarre19} there is further reason to suspect that Amsterdam and colleagues have sampled into a group of patients with bipolar II disorder with a relatively low risk of treatment-emergent affective shifts. Nevertheless, the weight of the series of studies of Amsterdam and colleagues does suggest that there is little reason to systematically avoid the use of antidepressants for patients with bipolar II depression who have no prior history of antidepressant treatment or who have been treated successfully with antidepressants in the past.