Findings from other studies

Few double-blind, placebo-controlled trials have examined efficacy and safety of antidepressant v. mood stabiliser monotherapy in bipolar disorder. Leverich et al ^{Reference Leverich, Altshuler, Frye, Suppes, McElroy and Keck7} found only a 23% sustained response rate without mood conversion episodes in patients with bipolar I and II disorder maintained on antidepressants (plus mood stabilisers). Schneck et al ^{Reference Schneck, Miklowitz, Miyahara, Araga, Wisniewski and Gyulai26} followed 1191 individuals with bipolar disorder (356 rapid cycling) for 1 year and found a 34% recovery rate, with a 61% mood conversion rate during antidepressant therapy. A double-blind, placebo-controlled comparison of quetiapine v. paroxetine found no advantage for paroxetine v. placebo in bipolar depression, ^{Reference McElroy, Weisler, Chang, Olausson, Paulsson and Brecher27} whereas a recent meta-analysis of controlled studies found no advantage for antidepressants per se in treating bipolar depression. ^{Reference Sidor and McQueen's28}

In contrast, a 5-year naturalistic study of 54 people with bipolar disorder found that antidepressants plus lithium maintained a response for an average of 17.2 months longer than if taking lithium alone, with a mood conversion rate of 14%. ^{Reference Peselow, Fieve, Difiglia and Sanfilipo29} A 1-year study of individuals with bipolar disorder who had recovered on antidepressants plus mood stabilisers (n = 19) v. mood stabilisers alone (n = 25) found a 32% and 68% depressive relapse rate, respectively (P = 0.0065). Antidepressants produced a threefold lower risk of relapse. ^{Reference Altshuler, Kiriakos, Calcagno, Goodman, Gitlin and Frye30} A similar benefit from long-term

Fig. 1

Flow chart of participants.

Table 1

Characteristics of participants with rapid- v. non-rapid-cycling bipolar disorder at the start of double-blind therapy

Characteristic	Rapid-cycling group (n = 25)	Non-rapid-cycling group (n = 56)	P a
Male gender, % (n/N)	64.0 (16/25)	41.1 (23/56)	0.06
Age, years: mean (s.d.)	35.3 (10.0)	38.8 (12.8)	0.23
Illness duration, years: mean (s.d.)	17.2 (10.0)	20.0 (11.9)	0.32
Depressive episode duration, months: mean (s.d.)	13.3 (15.3)	4.0 (19.3)	0.87
Prior depressive episodes, mean (s.d.)	22.8 (43.7)	6.6 (6.4)	0.008
Prior hypomanic episodes, mean (s.d.)	45 (56.2)	9.0 (13.4)	<0.0005
Hamilton Rating Scale for Depression score baseline, mean (s.d.)	7.2(6.1)	5.4 (4.0)	0.13
Young Mania Rating Scale score baseline, mean (s.d.)	1.3(2.8)	0.8 (1.7)	0.27

a P-values are from Student's t-test except for gender, which is from χ².

Table 2

Proportion of participants in the rapid- v. non-rapid-cycling group with treatment-emergent hypomania ^a

	Rapid-cycling group (n = 25)		Non-rapid-cycling group (n = 56)
	n	% (95% exact CI)	n	% (95% exact CI)	P b
Hypomania	5	20.0 (6.8–40.7)	5	8.9 (3.0–19.6)	0.27
Type I subsyndromal	5	20.0 (6.8–40.7)	10	17.9 (8.9–30.4)	0.82
Type II subsyndromal	2	8.0 (1.0–26.0)	5	8.9 (3.0–19.6)	0.89
Type III subsyndromal	2	8.0 (1.0–26.0)	13	23.2 (13.0–36.4)	0.13
Hypomania or type I	10	40.0 (21.1–61.3)	15	26.8 (15.8–40.3)	0.30
Hypomania or type I or type II	11	44.0 (24.4–65.1)	18	32.1 (20.3–46.0)	0.33
Hypomania or type I or type II or type III	11	44.0 (24.4–65.1)	26	46.4 (33.0–60.3)	0.99

a Some patients had more than one subsyndromal episode. Thus, the number of patients experiencing an episode shown in Table 2 may be smaller than the total number of episodes shown in Table 3.

b Fisher's exact test for comparison of rapid- v. non-rapid-cycling groups.

Table 3

Duration of treatment-emergent hypomanic episodes (in days) in the rapid- v. non-rapid-cycling group ^a

	Rapid-cycling group		Non-rapid-cycling group
	Mean (95% CI)	Episodes, n	Mean (95% CI)	Episodes, n	P b
Hypomania	23.2 (2.0 to 44.4)	5	5.8 (0.71 to 10.9)	5	0.06
Type I subsyndromal	2.6 (0.7 to 4.5)	5	5.8 (1.8 to 9.9)	11	0.27
Type II subsyndromal	9.0 (–16.4 to 34.4)	2	15.4 (6.8 to 24.3)	5	0.29
Type III subsyndromal	7.0 (–69.2 to 82.2)	2	2.3 (1.1 to 3.4)	16	0.05

a Some patients had more than one subsyndromal episode. Thus, the number of patients experiencing an episode shown in Table 2 may be smaller than the total number of episodes shown in Table 3.

b P-value is for comparison of mean duration between groups.

Table 4

Proportion of patients (with 95% exact CI) with a major or minor depressive episode

	Rapid-cycling group (n = 25)		Non-rapid-cycling group (n = 56)
	n	% (95% exact CI)	n	% (95% exact CI)	P a
Major depressive episode	1	4.2 (0.1–20.4)	0	0 (0.0–6.4)	0.13
Type I minor depression	8	32.0 (14.9–53.5)	18	32.1 (20.3–46.0)	0.99
Type II minor depression	6	24.0 (9.4–45.1)	23	41.1 (28.1–55.0)	0.14
Type III minor depression	1	4.2 (0.1–20.4)	0	0 (0.0–6.4)	0.13
Major or type I minor depression	17	68.0 (46.5–85.1)	41	73.2 (59.7–84.2)	0.63
Major, type I or type II minor depression b	16	64.0 (42.5–82.0)	46	82.1 (69.6–91.1)	0.08
Major, type I, type II or type III minor depression b	16	64.0 (42.5–82.0)	46	82.1 (69.6–91.1)	0.08

a Fisher's exact test for comparison of rapid- v. non-rapid-cycling groups.

b The same patients who experienced major, type I, type II or type III minor depressive episodes also experienced a type III minor depressive episode, resulting in similar values for the last two rows in Table 4.

Table 5

Duration of treatment-emergent depressive episodes (in days) in the rapid- v. non-rapid-cycling group

	Rapid-cycling group		Non-rapid-cycling group
Minor depression	Mean (95% CI)	Episodes, n	Mean (95% CI)	Episodes, n	P a
Type I	10.4 (3.9 to 17.0)	9	14.6 (4.3 to 24.9)	18	0.47
Type II	97.7 (–27.8 to 223.0)	12	14.7 (8.8 to 20.6)	27	0.10
Type III	3.0	1	0.0	0	–

a P-value is for comparison of mean duration between groups.

antidepressant therapy was reported by Kupfer et al, ^{Reference Kupfer, Chengappa, Gelenberg, Hirschfeld, Goldberg and Sachs15} whereas a double-blind, placebo-controlled study of maintenance fluoxetine in individuals with bipolar II disorder who had recovered showed a clinically meaningful trend for fewer depressive relapses (P = 0.08). ^{Reference Amsterdam, Shults, Brunswick and Hundert19,Reference Amsterdam and Shults20} Finally, the primary analysis from the current study showed the estimated hazard for depressive relapse as 2.5 times greater during lithium v. fluoxetine monotherapy (P = 0.04). ^{Reference Amsterdam and Shults21}

Limitations

Several caveats should be considered when interpreting the current findings. Results of this exploratory analysis are not definitive. The study was not powered to detect significant differences in efficacy or mania ratings between the rapid- v. non-rapid-cycling groups. The failure to identify significant differences in efficacy and safety in the current analysis does not mean that differences do not exist. We note the limited sample size of the rapid- v. non-rapid-cycling group within each treatment condition. Larger samples would be needed to detect small differences in mood conversion rates between groups.

In the current study, patients were stabilised on fluoxetine prior to randomisation. This design methodology may have influenced the long-term efficacy and safety ratings in favour of fluoxetine, whereby patients randomised to fluoxetine were more likely to stay well and less likely to experience mood conversion episodes.

Our definition of rapid cycling differed from the DSM-IV definition and was based on an average of ≥4 affective episodes per year over the course of the illness (rather than ≥4 affective episodes in the preceding year). This difference may have resulted in a rapid-cycling cohort with fewer affective episodes. ^{Reference Schneck, Miklowitz, Miyahara, Araga, Wisniewski and Gyulai26} Analysis of rapid cycling by DSM-IV criteria may have produced different results.

Finally, we limited our YMRS analysis to participants who experienced a change in YMRS scores over baseline in order to avoid averaging zero values from individuals with no change in YMRS scores. The frequency and severity of mood conversion episodes may have differed between groups had we used different threshold criteria for subsyndromal episodes, or if we employed a longer treatment duration. For example, Schneck et al ^{Reference Schneck, Miklowitz, Miyahara, Araga, Wisniewski and Gyulai26} found a 61% mood conversion rate in patients with stabilised bipolar disorder during 1 year of antidepressant therapy, whereas Koukopoulos et al ^{Reference Koukopoulos, Sani, Koukopoulos, Minnai, Girardi and Pani3} reported rapid cycling in 88% of patients taking antidepressants in a 36-year naturalistic study. In contrast, the current study found no difference in the proportion of the patients in the rapid- v. non-rapid-cycling group in any treatment condition with increases in YMRS scores. Although it is possible that this low mood conversion rate resulted from the inclusion of more patients who were mildly ill with bipolar II disorder with a lower propensity for developing manic symptoms, the illness severity of the current patient cohort was similar to that of prior cohorts with bipolar II disorder in studies by us and others. ^{Reference Leverich, Altshuler, Frye, Suppes, McElroy and Keck7,Reference Ghaemi, Klara, Ko, Baldassano, Kontos and Baldessarini10,Reference Goldberg and Ghaemi11,Reference Kupfer, Chengappa, Gelenberg, Hirschfeld, Goldberg and Sachs15-Reference Baldassano, Datto, Littman and Lipari17}

Implications

Although not definitive, these findings suggest that maintenance lithium or fluoxetine monotherapy are similar to placebo in preventing depressive relapse and treatment-emergent mood conversion episodes during long-term relapse-prevention therapy of rapid- and non-rapid-cycling bipolar II disorder. The findings call into question practice guideline recommendations to avoid maintenance antidepressants in patients with rapid- and non-rapid-cycling bipolar II disorder. ^{Reference Bauer, Callahan, Jampala, Petty, Sajatovic and Schaefer31-Reference Sachs, Printz, Kahn, Carpenter and Docherty34}