Double-Blind, Randomized, Placebo-Controlled Trials
of Ethyl-Eicosapentanoate in the Treatment of Bipolar
Depression and Rapid Cycling Bipolar Disorder
Paul E. Keck, Jr., Jim Mintz, Susan L. McElroy, Marlene P. Freeman, Trisha Suppes Mark A. Frye,
Lori L. Altshuler, Ralph Kupka, Willem A. Nolen, Gabriele S. Leverich, Kirk D. Denicoff, Heinz Grunze,
Naihua Duan, and Robert M. Post
Background: The results of pilot trials suggest that omega-3 fatty acids may have efficacy in the treatment of mood symptoms in
Methods: We conducted a 4-month, randomized, placebo-controlled, adjunctive trial of ethyl-eicosapentanoate (EPA) 6 g/day in the
treatment of bipolar depression and rapid cycling bipolar disorder. Subjects were receiving mood-stabilizing medications at
therapeutic doses or plasma concentrations. The measures of efficacy were early study discontinuation, changes from baseline in
depressive symptoms (Inventory for Depressive Symptomology total score) and in manic symptoms (Young Mania Rating Scale total
score), and manic exacerbations (“switches”). We also measured side effects and bleeding time, a biomarker of drug action.
Results: Overall, there were no significant differences on any outcome measure between the EPA and placebo groups.
Conclusions: This study did not find overall evidence of efficacy for adjunctive treatment with EPA 6 g/day in outpatients with bipolar
depression or rapid cycling bipolar disorder.
Key Words: Bipolar disorder, ethyl-eicosapenantoate, omega-3,
fatty acid, clinical trial
2002a; Su et al 2003) of four trials (Marangell et al 2003) in major
depressive disorder found a therapeutic effect. Stoll et al (1999b)
compared combined ethyl-eicosapentanoate (EPA) 6.2 g/day
and docosahexaenoic acid (DHA) 3.4 g/day with placebo in a
4-month trial involving 44 patients with bipolar I and II. The
omega-3 fatty acid group experienced a significantly longer time
in remission, greater improvement from baseline to end point in
mean total scores for the Clinical Global Impression (CGI) Scale
and the 31-item Hamilton Depression Rating Scale (HDRS), but
no significant difference in scores for the total Young Mania
Rating Scale (YMRS). The relative contribution of EPA, compared
with DHA, to therapeutic response could not be examined.
Calabrese et al (1999) noted important methodologic limitations
of this study. Frangou et al (2002) reported the results of a
mega-3 fatty acids may have a therapeutic effect in
patients with mood disorders (reviewed in Stoll et al
1999a). Three (Nemets et al 2002; Peet and Horribin
12-week, placebo-controlled, double-blind, parallel-group trial
of two fixed doses of EPA, 1 g/day or 2 g/day, as adjunctive
treatment in 75 outpatients with bipolar depression. Patients in
both EPA treatment groups displayed greater reductions in HDRS
total scores, but the interpretation of this study is compromised
because changes in patients’ medication regimens were allowed
during the trial for symptom worsening.
Methods and Materials
Study Groups and Recruitment
Subjects were men and women, 20–73 years old, who met
DSM-IV criteria (American Psychiatric Association 1994) for
bipolar disorder (types I, II, or NOS [not otherwise specified]).
Diagnosis was established by clinical interview, review of all
prior records, and administration of the Structured Clinical
Interview for DSM-IV (SCID) (Spitzer 1994). Bipolar depressed
subjects met DSM-IV criteria for a major depressive episode;
rapid cycling subjects met DSM-IV criteria for mania/hypomania
or major depression and for rapid cycling. We excluded subjects
with serious unstable medical illnesses, diabetes mellitus, current
alcohol or substance use disorder requiring separate treatment,
clinically significant suicidal or homicidal ideation, pregnant or
lactating women, and fertile women not using acceptable meth-
ods of contraception.
Subjects were required to receive at least one mood-stabiliz-
ing medication at therapeutic doses or plasma concentrations;
dose increases resulted in study termination. The appropriate
institutional review boards or human subjects committees at each
site approved the study protocol. After complete description of
the study to the subjects, written informed consent was obtained.
During the initial screening visit, we obtained psychiatric and
medical histories, a physical examination, laboratory assessment
and vital signs, and administered the YMRS (Young et al 1978),
Inventory of Depressive Symptomatology (IDS-C) (Rush et al
1996), and Clinical Global Impression Scale-Bipolar Disorder
(CGI-BP) (Spearing et al 1997). These symptom scales and
adverse events were queried at each of nine subsequent visits.
Bleeding times were performed at baseline and at visit four
From the Psychopharmacology Research Program, Department of Psychia-
try (PEK, SLM), University of Cincinnati College of Medicine and the
Mental Health Care Line; the General Clinical Research Center (PEK),
Cincinnati VAMC, Cincinnati, Ohio; the Department of Psychiatry (JM,
California; the Department of Psychiatry (MPF), University of Arizona
School of Medicine, Tucson, Arizona; the Department of Psychiatry (TS),
The Netherlands; the University Hospital Groningen (WAN), Groningen,
and Anxiety Disorders Research Program (KDD), National Institute of
Mental Health, National Institutes of Health, Department of Health and
Human Services, Bethesda, Maryland; and the Psychiatrische Klinik der
Ludwig-Maximilian Universitat (HG), Munich, Germany.
sity of Cincinnati College of Medicine, 231 Albert Sabin Way, ML 559,
Cincinnati, OH 45267-0559; E-mail: paul.keck@UC.edu.
BIOL PSYCHIATRY 2006;60:1020–1022
© 2006 Society of Biological Psychiatry
(week 8) because omega-3 fatty acids may prolong bleeding time
(Nordoy et al 1994).
Subjects were randomized to receive either EPA 6 g/day or
matching placebo capsules (liquid paraffin), in addition to stable
ongoing usual treatment. Laxdale, Ltd. (Stirling, Scotland) gener-
ously supplied the EPA and matching placebo. The 6 g/day EPA
dose was selected on the basis of the dose of EPA contained in
the formulation in the study of Stoll et al (1999b) (6.2 g/day). An
Investigational New Drug application was filed with the U.S.
Food and Drug Administration (FDA) to conduct these trials.
Randomization tables were centrally generated by a computer
program at the Stanley Foundation Data Coordinating Center.
The EPA and matching placebo capsules were indistinguishable
by size, shape, color, odor, and after taste. Compliance was
estimated by results of capsule counts and was ?90% in all
The primary outcome variables were study discontinuation,
change from baseline on the IDS-C and YMRS at each of nine
scheduled visits, side effects, and “switches” into mania, defined
as a two-point CGI Mania score increase from a score of 1 or 2.
Time to discontinuation was analyzed by using the proportional
hazard regression (SAS PHREG) procedure. Symptom outcomes
were analyzed by using data from postbaseline visits two
through nine with mixed effects regression models with repeated
measures (SAS MIXED) and baseline as a covariate. Side effects
were rare, so these data were aggregated into a summary
measure at the patient level, by using a count of the total number
of separate side effect reports. Because most patients (71%)
scored zero on this index, the primary analysis used the Tobit
regression model (Tobin 1958). A secondary logistic regression
analysis used a dichotomy (any vs. side effects). Both side effect
analyses included days in treatment as a covariate.
Bleeding time was analyzed with analysis of covariance
controlling for baseline. Bleeding time data were gathered only
at baseline and at the fourth visit (8 weeks) and, therefore, were
not available for early study dropouts, a total of 49 patients with
missing data (17 at baseline, 45 at follow-up). We therefore also
analyzed bleeding times using a propensity-weighted analysis
(Hirano et al 2003; Rosenbaum and Rubin 1983), a method that
weights cases using baseline characteristics to adjust for possible
effects of dropout.
In their prior report, Stoll et al (1999b) excluded from all
analyses 12 (27%) of 44 participants who left the trial during the
first 30 days. Our early termination rate was similar (26/121
[22%]), but our primary analyses were intent-to-treat with
n ? 116 patients with postbaseline data. To permit direct
comparison with Stoll et al (1999b), we also estimated our
models omitting the 26 patients who discontinued study partic-
ipation during the first 30 days. We evaluated interactions of
medication with diagnostic subtype in all domains. None of these
interaction effects was significant, and they are not reported.
Three hundred eighty-five patients were screened and re-
cruited from the larger cohort participating in the Stanley Foun-
dation Bipolar Network (Leverich et al 2001). Of these patients,
130 met entry criteria and agreed to participate; 121 were
subsequently randomized after all screening procedures; 5 were
never assessed after baseline, leaving a total of 116 patients
analyzed here. Table 1 summarizes the demographic and clinical
characteristics of the study subjects.
Sixty-five (54%) patients exited the protocol before the final
(ninth) visit, the most common reasons being poor treatment
response (77%), side effects (9%), and noncompliance (8%). The
survival regression test of the medication effect on early termi-
nation was ?2? .11, df ? 1, p ? .74. The same ?2result was
obtained when patients who discontinued before 30 days were
The medication effect on change in the IDS-C depression
scale was not significant (F ? .05, df ? 1, 113, p ? .82). The mean
medication–placebo IDS-C difference was –038, SE ? 1.7, 95%
confidence limit (CL) ? ?3.8 to ?3.0. The medication effect on
the YMRS was also not significant (F ? .98, df ? 1, 113, p ? .32).
The mean difference in the YMRS change was .61, SE ? .61, 95%
CL ? ?1.82 to –.60). YMRS scores were generally low. Eighty-
five percent of the visits had YMRS scores below 8, and scores
above 16 occurred in only 2% of the visits. A CGI-based “switch”
occurred in 11 (19%) of 57 patients on placebo, and 15 (25%) of
59 patients on EPA, a nonsignificant difference (?2? .62, df ? 1,
p ? .43). Omitting the early dropouts had no meaningful effect
on any of these results.
Ethyl-eicosapentanoate was generally well tolerated, and
most adverse events were mild. Adverse events reported by
study subjects are summarized in Table 2. The Tobit regression
analysis on the sum across visits indicated a nonsignificant higher
average number of side effects for those on EPA than those on
placebo (?2? 2.97, df ? 1, p ? .08). Logistic regression on a
dichotomous side effect outcome (any vs. none) yielded a similar
inconclusive result (?2? 2.38, df ? 1, p ? .12, odds ratio ? 1.92,
95% CL ? .84 to 4.37).
Table 1. Demographic and Clinical Characteristics
Bipolar DepressedRapid Cycling
EPA (n ? 28) Placebo (n ? 29)p ValueEPA (n ? 31)Placebo (n ? 28)p Value
Age (mean, SD)
Male (n, %)
BP I (n, %)
BP II (n, %)
BP NOS (n, %)
Age tested by t tests with df ? 55, 57; other variables by chi-square with df ? 1.
EPA, ethyl-eicosapentanoate; BP, bipolar disorder; NOS, not otherwise specified.
P.E. Keck, Jr., et al
BIOL PSYCHIATRY 2006;60:1020–1022 1021
Bleeding Times Download full-text
The analysis of covariance including only the medication
main effect yielded nonsignificant differences in both the un-
weighted (F ? .25, df ? 1, 64, p ? .62) and propensity-weighted
analyses (F ? 1.05, df ? 1, 64, p ? .31).
The results failed to confirm any meaningful outcome differ-
ence between medication and placebo in any of the outcome
domains. The lack of a medication–placebo difference was not
related to diagnostic subtype (cycling type vs. depressed). This
result is a failure to replicate the positive results reported by Stoll
et al (1999b). They used a combination of DHA (3.36 g) and EPA
(6.16 g), whereas we used only EPA 6 g/day. The doses of EPA
were virtually identical, but it is possible that the EPA dose of 6
g/day was too high. Subsequent dose-ranging studies in patients
with unipolar major depressive disorder and in schizophrenia
have suggested efficacy for EPA 1–3 g/day, but not for higher
doses (Emsley et al 2002; Nemets et al 2002; Peet and Horrobin
2002b; Peet and Horrobin 2002). To the extent that increases in
DHA are required, Woodman et al (2002) reported that admin-
istration of DHA 4 g/day, but not EPA 4 g/day, increased DHA
levels in patients with diabetes mellitus. Also, possible therapeu-
tic effects of EPA may have been masked by coadministration
with mood-stabilizing agents (Horrobin 2003). The results of this
study do not support the efficacy of EPA 6 g/day in the treatment
of depression or rapid cycling in patients with bipolar disorder.
This study was sponsored by the Stanley Medical Research
Institute, a National Alliance for Research on Schizophrenia and
Depression (NARSAD)Young Investigator Award (to MPF) and
an American Psychiatric Association (APA)/Lilly Psychiatry Re-
search Fellowship (to MPF), and Laxdale, Ltd., who provided the
This article was presented in part at the European Stanley
Foundation Bipolar Conference, Freiburg, Germany, September
14, 2002, and the Fifth International Conference on Bipolar
Disorder, Pittsburgh, Pennsylvania, June 12–14, 2003.
Mental Disorders, 4th ed. Washington, DC: American Psychiatric Press.
Calabrese JR, Rapport DJ, Shelton MD (1999): Fish oils and bipolar disorder.
Emsley R, Myburgh C, Oosthuizen P, van Rensburg SJ (2002): Randomized,
placebo-controlled study of ethyl-ecosapentaenoic acid as supplemen-
tal treatment in schizophrenia. Am J Psychiatry 159:1596–1598.
Frangou S, Lewis M, McCrone P (2002): Efficacy of ethyl-eicosapentaenoate
in bipolardepression: a randomized, double-blind, placebo-controlled
study [abstract]. Bipolar Disorders 4:123.
ment effects using the estimated propensity score. Econometrica 71:
Horrobin DF (2003): Omega-3 fatty acids for schizophrenia. Am J Psychiatry
dation Bipolar Treatment Outcome Network: I. Longitudinal methodol-
sahexaenoic acid in the treatment of major depression. Am J Psychiatry
Nemets B, Stahl Z, Belmaker RH (2002): Addition of omega-3 fatty acid to
maintenance medication treatment for recurrent unipolar depressive
Nordoy A, Hatcher L, Goodnight S, Fitzgerald DA, Conner WE (1994): Effects
of dietary fat content, saturated fatty acids, and fish oil on eicosanoid
production and homeostatic parameters in normal men. J Lab Clin Med
Peet M, Horrobin DF (2002a): A dose-ranging study of the effects of ethyl-
eicosapentanoate in patients with ongoing depression despite appar-
ently adequate treatment with standard drugs. Arch Gen Psychiatry 59:
Peet M, Horrobin DF (2002b): A dose-ranging exploratory study of the ef-
fects of ethyl-eicosapentanoate in patients with persistent schizo-
Rush AJ, Gullion CM, Basco MR, Jarrett RB, Trivedi MH (1996): The Inventory
of Depressive Symptomatology (IDS): psychometric properties. Psychol
Rosenbaum PR, Rubin DB (1983): The central role of the propensity score in
observational studies for causal effects. Biometrika 70:41–55.
Spearing MK, Post RM, Leverich GS, Brandt D, Nolen W (1997): Modification
of the Clinical Global Impressions (CGI) Scale for use in bipolar illness
(BP): the CGI-BP. Psychiatr Res 73:159–171.
Spitzer RL (1994): Structured Clinical Interview for DSM-IV. Washington, DC:
American Psychiatric Press.
Stoll AL, Locke CA, Marangell LB, Severus WE (1999a): Omega-3 fatty acids
and bipolar disorder: a review. Prostaglandins Leukot Essent Fatty Acids
Stoll AL, Severus WE, Freeman MP, Rueter S, Zboyan HA, Diamond E, Cress
KK, Marangell LB (1999b): Omega 3 fatty acids in bipolar disorder. A
preliminary double-blind,placebo-controlled trial. Arch Gen Psychiatry
Su KP, Huang SY, Chiu CC, Shen WW (2003): Omega-3 fatty acids in major
depressivedisorder. A preliminary double-blind, placebo-controlled
trial. Eur Neuropsychopharmacol 13:267–271.
Tobin J (1958): Estimation of relationships for limited dependent variables.
of purified eicosapentaenoic acid and docosahexaenoic acid on glyce-
Young RC, Biggs JT, Ziegler VE, Meyer DA (1978): A rating scale for mania:
Reliability, validity and sensitivity. Br J Psychiatry 133:429–435.
Table 2. Adverse Events Reported by Study Subjects
EPA (n ? 61) Placebo
Central nervous system
Serious adverse event
1022 BIOL PSYCHIATRY 2006;60:1020–1022
P.E. Keck, Jr., et al