Content uploaded by J. G. Martins
Author content
All content in this area was uploaded by J. G. Martins on Nov 25, 2015
Content may be subject to copyright.
Review
EPA but Not DHA Appears To Be Responsible for the
Efficacy of Omega-3 Long Chain Polyunsaturated Fatty
Acid Supplementation in Depression: Evidence from a
Meta-Analysis of Randomized Controlled Trials
Julian G. Martins, MA, MBBS
Academy of Nutritional Medicine, Cambridge, UNITED KINGDOM
Background: Epidemiologic and case-control data suggest that increased dietary intake of omega-3 long-
chain polyunsaturated fatty acids (v3 LC-PUFAs) may be of benefit in depression. However, the results of
randomized controlled trials are mixed and controversy exists as to whether either eicosapentaenoic acid (EPA)
or docosahexaenoic acid (DHA) or both are responsible for the reported benefits.
Objective: The aim of the current study was to provide an updated meta-analysis of all double-blind,
placebo-controlled, randomized controlled trials examining the effect of v3 LC-PUFA supplementation in
which depressive symptoms were a reported outcome. The study also aimed to specifically test the differential
effectiveness of EPA versus DHA through meta-regression and subgroup analyses.
Design: Studies were selected using the PubMed database on the basis of the following criteria: (1)
randomized design; (2) placebo controlled; (3) use of an v3 LC-PUFA preparation containing DHA, EPA, or
both where the relative amounts of each fatty acid could be quantified; and (4) reporting sufficient statistics on
scores of a recognizable measure of depressive symptoms.
Results: Two hundred forty-one studies were identified, of which 28 met the above inclusion criteria and
were therefore included in the subsequent meta-analysis. Using a random effects model, overall standardized
mean depression scores were reduced in response to v3 LC-PUFA supplementation as compared with placebo
(standardized mean difference 520.291, 95% CI 520.463 to 20.120, z 523.327, p50.001). However,
significant heterogeneity and evidence of publication bias were present. Meta-regression studies showed a
significant effect of higher levels of baseline depression and lower supplement DHA:EPA ratio on therapeutic
efficacy. Subgroup analyses showed significant effects for: (1) diagnostic category (bipolar disorder and major
depression showing significant improvement with v3 LC-PUFA supplementation versus mild-to-moderate
depression, chronic fatigue and non-clinical populations not showing significant improvement); (2) therapeutic
as opposed to preventive intervention; (3) adjunctive treatment as opposed to monotherapy; and (4) supplement
type. Symptoms of depression were not significantly reduced in 3 studies using pure DHA (standardized mean
difference 0.001, 95% CI 20.330 to 0.332, z 50.004, p50.997) or in 4 studies using supplements containing
greater than 50% DHA (standardized mean difference 50.141, 95% CI 520.195 to 0.477, z 50.821, p5
0.417). In contrast, symptoms of depression were significantly reduced in 13 studies using supplements
containing greater than 50% EPA (standardized mean difference 520.446, 95% CI 520.753 to 20.138, z 5
22.843, p50.005) and in 8 studies using pure ethyl-EPA (standardized mean difference 520.396, 95% CI 5
20.650 to 20.141, z 523.051, p50.002). However, further meta-regression studies showed significant
inverse associations between efficacy and study methodological quality, study sample size, and duration, thus
limiting the confidence of these findings.
Address reprint requests to: Julian G. Martins, MA, MBBS, Academy of Nutritional Medicine, 80 Commercial End, Swaffham Bulbeck, Cambridge CB25 0NE, UNITED
KINGDOM. E-mail: julian.martins@aonm.org
Abbreviations: EPA 5eicosapentaenoic acid, DHA 5docosahexaenoic acid, AA 5arachidonic acid, LC-PUFA 5long-chain polyunsaturated fatty acid, v3 LC-PUFA
5omega-3, UREML 5unrestricted maximum likelihood method, HDRS 5Hamilton Rating Scale for Depression, HDRS-SF 5Hamilton Rating Scale for Depression-
short form, BDI 5Beck Depression Inventory, MADRS 5Montgomery-A
˚sberg Depression Rating Scale, EPDS 5Edinburgh Postnatal Depression Scale, SCID-IV 5
Structured Clinical Interview for DSM-IV, POMS 5Profile of Mood States, CDI 5Children’s Depression Inventory, CDRS 5Children’s Depression Rating Scale, IDS-
C5Inventory of Depressive Symptomatology, DASS 5Depression Anxiety and Stress Scales, LPS 5lipopolysaccharide, IL 5interleukin, TNF-a5tumor necrosis
factor-alpha, NF-kB5nuclear factor kappa-B.
Journal of the American College of Nutrition, Vol. 28, No. 5, 525–542 (2009)
Published by the American College of Nutrition
525
Conclusions: The current meta-analysis provides evidence that EPA may be more efficacious than DHA in
treating depression. However, owing to the identified limitations of the included studies, larger, well-designed,
randomized controlled trials of sufficient duration are needed to confirm these findings.
INTRODUCTION
Depression remains a serious public health problem with
significant associated morbidity, mortality, and economic cost.
Over the life course, major depression will affect 16.9% of
individuals [1]. In addition to the risk of suicide, the
commonest cause of death in individuals with depression,
mortality is also increased as a result of the association with
cardiovascular disease [2]. The estimated economic cost of
depression, both direct in terms of cost of treatment and
indirect through lost days at work and reduced productivity, is
substantial [3]. In comparison with other causes of disease, the
Global Burden of Disease Study has shown that, by 2030,
unipolar depressive disorder and ischemic heart disease will be
the major causes of disability in developed populations [4].
Given the increases in prevalence of depression [5] and
cardiovascular disease, the latter especially among socio-
economically disadvantaged groups [6] and young people with
obesity [7], it appears likely that a common underlying
environmental influence may account for these changes.
One theory that has been advanced to explain these
changes, at least in part, is the significant shift over the last
century in the dietary intake of long-chain polyunsaturated
fatty acids (v3 LC-PUFAs) toward an increase in saturated fat
and an increase in the ratio of omega-6 to omega-3 fatty acids
[8]. The pathophysiological basis of support for this theory
comes from evidence that inflammatory processes such as
excessive cytokine production [9] and glucocorticoid resis-
tance [10] underpin major depression, coupled with evidence
that v3 LC-PUFAs produce both anti-inflammatory eicosa-
noids [11] that reduce levels of pro-inflammatory cytokines in
depressed individuals [12] and have antidepressant protective
actions as identified from the epidemiologic association of
lower levels of depression with fish consumption [13–20], and
the physiological association between reduced plasma or red
blood cell (RBC) membrane v3 LC-PUFAs and depressive
symptoms [21–27] and suicide [28–30]. If true, this theory has
substantial implications for both the prevention and treatment
of depression, with the potential for large-scale impact through
dietary interventions. Whilst it is acknowledged that many
other factors may also be contributing to the rise in depression
and that effective treatments already exist, such as antidepres-
sant medication and cognitive behavioral therapy, even if the
effect of dietary intervention is very small at the individual
level, substantial benefits can result at the population level.
This is because a population-based intervention, if capable of
shifting the distribution curve of depressive symptoms in a
positive direction, even by a small degree, can remove large
numbers of individuals from the threshold of clinical disorder
[31]. Furthermore, established antidepressant therapy is not
effective in all cases, with less than 50% of patients showing
full remission of symptoms [32].
In order to answer this important question of dietary
influence in depression, Basant Puri’s group in London became
the first to demonstrate the efficacy of eicosapentaenoic acid
(EPA) in a therapeutic case study of depression [33].
Subsequently, a number of double-blind, placebo-controlled
trials have been conducted, some of which support the efficacy
of v3 LC-PUFAs in the adjunctive treatment of adult
depression [34–42] and in the sole treatment of childhood
depression [43] and adult depression [35,44], whereas other
trials have not demonstrated effectiveness [45–59]. Some of
these studies have been subjected to meta-analytic review by
various authors. Freeman et al. [60] and Lin and Su [61] report
overall benefit, whereas Appleton et al. [62] and, in an updated
analysis of the same study, Rogers et al. [52] report negligible
benefits. As a possible explanation for the variability in these
findings, all of these meta-analytic studies acknowledge
considerable between-study heterogeneity that may be ex-
plained by publication bias, severity of baseline depression,
diagnostic variation, and variability in the nature of the v3 LC-
PUFA regime employed. Given the negative findings of a trial
involving pure docosahexaenoic acid (DHA) [50], an addi-
tional meta-analysis conducted by Ross et al. [63] sought to
examine whether this heterogeneity could be explained by
variation in the v3 LC-PUFA regime employed using random
effects meta-regression. When this factor was taken into
account, the between-study variance was substantially reduced.
Specifically, the positive effect of v3 LC-PUFAs on
depressive symptoms appeared to be explained by the EPA
rather than DHA content of the regime.
Because several new randomized controlled trials of v3 LC-
PUFA preparations in depressive disorders have been published
in the last year, a further meta-analysis has been performed and
is reported here. The specific aims of the current meta-analysis
were to update the findings on overall efficacy and to provide
further meta-regression analyses to either confirm or refute the
original observation by Ross et al. [63] regarding the possibility
that EPA and not DHA is responsible for the therapeutic effect
of v3 LC-PUFAs in depression.
METHODS
The PubMed MeSH database was searched using the
following terms: ((‘‘Psychiatry and Psychology Category’’
EPA in Depression: A Meta-Analysis
526 VOL. 28, NO. 5
[Mesh] OR ‘‘Fatigue Syndrome, Chronic’’ [Mesh]) AND
(‘‘Fatty Acids, Essential’’ [Mesh] OR ‘‘Fatty Acids, Omega-
3’’ [Mesh] OR ‘‘Fish Oils’’ [Mesh] OR ‘‘Eicosapentaenoic
Acid’’ [Mesh] OR ‘‘Docosahexaenoic Acids’’ [Mesh])) AND
‘‘Randomized Controlled Trial’’ [Publication Type].
The overarching ‘‘Psychiatry and Psychology Category’’
was used to encompass not only the relevant terms for
depression (‘‘Depressive Disorder’’ [Mesh], ‘‘Depressive
Disorder, Major’’ [Mesh], ‘‘Depression’’ [Mesh], ‘‘Bipolar
Disorder’’ [Mesh], ‘‘Depression, Postpartum’’ [Mesh], ‘‘Dys-
thymic Disorder’’ [Mesh], ‘‘Seasonal Affective Disorder’’
[Mesh]) but also terms for all other mental disorders and mood
in normal subjects. This inclusive strategy was deemed
necessary as some studies examining, for example, patients
with schizophrenia also report changes in depressive symp-
toms. In addition, a number of studies have looked at the effect
of v3 LC-PUFA supplementation on mood in individuals
without evidence of mental disorder. The term for chronic
fatigue was also included, as there have been reports of the use
of v3 LC-PUFA supplementation in this disorder, which also
is associated with considerable comorbid depressive symp-
toms. As of May 4, 2009, this strategy identified 288 potential
studies for inclusion. Studies were then selected on the basis of
the following criteria: (1) randomized design; (2) placebo
controlled; (3) use of an v3 LC-PUFA preparation containing
DHA, EPA, or both where the relative amounts of each fatty
acid could be quantified; and (4) reporting sufficient statistics
on scores of a recognizable measure of depressive symptoms
(in certain circumstances, the authors of studies were contacted
directly for clarification of results).
This selection strategy resulted in 30 trials being identified
that could be subjected to meta-analytic review. It is
noteworthy that this selection strategy identified all of the
studies included in the meta-analysis by Appleton et al. [62],
who employed a rather more extensive search strategy of
multiple databases. Therefore, despite the simplicity of the
search strategy employed in the current study, it is unlikely
that significant studies have been missed. One of the
identified studies examined the effect of v3 LC-PUFA
supplementation in obsessive-compulsive disorder [64] and,
although depressive symptoms were measured in this study
using the Hamilton Depression Rating Scale, no outcome
values were reported. Therefore, that study was excluded
from further analysis, as it did not meet the selection criteria.
The 2 previous meta-analyses by Appleton et al. [62] and
Rogers et al. [52] included a study by Ness et al. [65] that was
not included in the meta-analyses by Freeman et al. [60] and
Lin and Su [61]. This study was based upon the assessment of
mood in a trial of men with angina who had been randomized
to receive advice to eat more fish or to receive no such advice,
with those not tolerating fish being offered Maxepa fish oil
supplementation instead. Given the absence of any possibility
of placebo control or the ability to accurately quantify the
relative amounts of DHA and EPA being consumed, it was
felt that inclusion of this trial was inappropriate, as again, it
did not meet the selection criteria.
Methodological quality was assessed using the Jadad score
[66] plus 6 additional components: (1) whether similarities in
baseline characteristics were adequately described; (2)
whether attempts were made to conceal the fish taste of the
active intervention; (3) whether the outcome assessors were
adequately blinded; (4) whether data were analyzed according
to intention-to-treat (ITT) methods; (5) whether compliance
was assessed through measurement of RBC or plasma fatty
acids; and (6) whether blinding success was evaluated. This
gave a maximum possible quality score of 11. The
characteristics of the final 28 included studies are shown in
Table 1.
Standardized mean differences in depression scores were
computed and analyzed using the program Comprehensive
Meta-Analysis, identified as one of the best tools for this
purpose in a recent systematic review [67]. One of the many
advantages of this program is that effect sizes can be computed
from a wide range of reporting methods, enabling studies to be
included in the current analysis that were excluded by
Appleton et al. [62]. The program also allows mean effect
sizes to be computed in studies that use multiple outcome
measures, for example, 2 questionnaire measures of depressive
symptoms plus a categorical measure of clinical improvement,
allowing all available data that relate to depressive symptoms
to be included in this analysis. Some studies reported
additional outcomes using generalized measures such as the
Global Assessment of Functioning Scale or, in subjects with
bipolar disorder, the Young Mania Rating Scale; these
outcomes were excluded from the analysis to allow an
exclusive focus on depressive symptoms.
The meta-analytic strategy employed was as follows: (1) to
use random effects rather than fixed effects analyses, as it was
evident that there was considerable variation in clinical
populations studied, methodologies employed, and outcome
measures used; (2) to examine overall effect sizes using forest
plots; (3) to examine for possible publication bias using funnel
plots with Duval and Tweedie’s trim and fill method; (4) to
assess heterogeneity using Cohen’s Q, I
2
, and t
2
, where I
2
represents the percentage of heterogeneity that can be
attributed to true underlying differences in effect sizes between
studies and t
2
represents the extent of variance between
studies; (5) on the basis of these findings, to conduct further
sensitivity analyses on subpopulations of studies using random
effects analysis of variance (ANOVA); and finally, (6) to
conduct random effects meta-regression studies on relevant
moderator variables using the unrestricted maximum like-
lihood (UREML) method, a method least likely to generate
spurious findings.
EPA in Depression: A Meta-Analysis
JOURNAL OF THE AMERICAN COLLEGE OF NUTRITION 527
Table 1. Characteristics of the 28 Included Studies Listed Chronologically according to Publication Date
Study [reference] Clinical Group
Number, Total (v3
LC-PUFA/placebo)
v3 LC-PUFA
Preparation Daily Dosage Regime(s) Treatment Status
Duration
(days)
Outcome
Measure(s) Quality
Behan et al. 1990 [44] Chronic fatigue 63 (39/24) Efamol Marine 0.136 g EPA +
0.088 g DHA
Therapeutic, Monotherapy 90 Likert scale 6
Stoll et al. 1999 [38] Bipolar disorder 30 (14/16) Menhaden fish body
oil concentrate
6.2 g EPA +3.4 g DHA Preventive, Adjunctive 112 HDRS-31 7
Warren et al. 1999 [55] Chronic fatigue 50 (24/26) Efamol Marine 0.136 g EPA +
0.088 g DHA
Therapeutic, Monotherapy 90 BDI 7
Fenton et al. 2001 [45] Schizophrenia 87 (43/44) Laxdale Ltd 3 g Ethyl-EPA Therapeutic, Adjunctive 112 MADRS 8
Nemets et al. 2002 [36] Major depression 20 (10/10) Laxdale Ltd 2 g Ethyl-EPA Therapeutic, Adjunctive 28 HDRS-24 9
Peet and Horrobin 2002
[37]
Major depression 70 (17, 18, 17/18) Laxdale Ltd 1 g Ethyl-EPA or
2 g Ethyl-EPA or
4 g Ethyl-EPA
Therapeutic, Adjunctive 84 BDI, HDRS-17,
MADRS
6
Zanarini and
Frankenburg 2003
[54]
Borderline
personality
disorder
30 (20/10) Laxdale Ltd 1 g Ethyl-EPA Therapeutic, Monotherapy 56 MADRS 4
Llorente et al. 2003 [49] Perinatal depression 89 (44/45) DHASCO (Martek
Biosciences
Corporation)
0.2 g DHA Preventive, Monotherapy 120 BDI, EPDS,
SCID-IV
6
Marangell et al. 2003
[50]
Major depression 36 (18/18) Pure DHA (no
manufacturer
stated)
2 g DHA Therapeutic, Monotherapy 42 HDRS-28,
MADRS
6
Su et al. 2003 [39] Major depression 28 (14/14) Menhaden fish body
oil concentrate
4.4 g EPA +2.2 g DHA Therapeutic, Adjunctive 56 HDRS-21 5
Hirashima et al. 2004
[69]
Bipolar disorder 21 (6, 6/9) Fish oil 5.0–5.2 g EPA +
3.0–3.4 g DHA or
1.3 g EPA +
0.7 g DHA
Therapeutic, Adjunctive 28 HDRS-23 2
Silvers et al. 2005 [53] Major depression 77 (40/37) DHA-enriched tuna
fish oil (Clover
Corporation PLC)
0.6 g EPA +2.4 g DHA Therapeutic, Adjunctive 84 BDI, HDRS-SF 10
Fontani et al. 2005 [68] Non-clinical 33 (33/33)
1
Fish oil 1.60 g EPA +0.80 g DHA Preventive, Monotherapy 70 POMS 5
Frangou et al. 2006 [34] Bipolar disorder 75 (24, 25/26) Laxdale Ltd. 1 g Ethyl-EPA or
2 g Ethyl-EPA
Therapeutic, Adjunctive 84 HDRS-17 9
Nemets et al. 2006 [43] Major depression in
childhood
28 (13/15) Ocean Nutrition 0.4 g EPA +0.2 g DHA Therapeutic, Monotherapy 112 CDI, CDRS 6
Keck et al. 2006 [48] Bipolar disorder 116 (59/57) Laxdale Ltd 6 g Ethyl-EPA Therapeutic, Adjunctive 120 IDS-C 6
Hallahan et al. 2007 [41] Recurrent self-harm 49 (22/27) EPAX 5500 1.22 g EPA +
0.908 g DHA
Therapeutic, Adjunctive 84 BDI, HDRS 8
Grenyer et al. 2007 [47] Major depression 83 (40/43) Tuna fish oil (Clover
Corporation PLC)
0.6 g EPA +2.2 g DHA Therapeutic, Adjunctive 112 BDI, HDRS 10
Frangou et al. 2007 [46] Bipolar disorder 14 (7/7) Laxdale Ltd 2 g Ethyl-EPA Therapeutic, Adjunctive 84 HDRS 6
Rogers et al. 2008 [52] Mild-to-moderate
depression
218 (109/109) Minami Nutrition 0.63 g EPA +0.85 g DHA Therapeutic, Monotherapy 84 BDI, DASS 11
Jazayeri et al. 2008 [35] Major depression 48 (32/16) Minami Nutrition 1 g Ethyl-EPA Therapeutic, Adjunctive 56 HDRS-17 6
EPA in Depression: A Meta-Analysis
528 VOL. 28, NO. 5
Study [reference] Clinical Group
Number, Total (v3
LC-PUFA/placebo)
v3 LC-PUFA
Preparation Daily Dosage Regime(s) Treatment Status
Duration
(days)
Outcome
Measure(s) Quality
Rees et al. 2008 [51] Perinatal depression 26 (13/13) Fish oil 0.414 g EPA +
1.638 g DHA
Therapeutic, Monotherapy 42 EPDS, HDRS-
17, MADRS
11
Su et al. 2008 [40] Perinatal depression 33 (17/16) Menhaden fish body
oil concentrate
2.2 g EPA +1.2 g DHA Therapeutic, Monotherapy 56 BDI, EPDS,
HDRS-21
9
Freeman et al. 2008 [57] Perinatal depression 59 (31/28) Not specified 1.1 g EPA +0.8 g DHA Therapeutic, Monotherapy
2
56 EPDS, HDRS 5
van de Rest et al. 2008
[59]
Non-clinical 302 (100, 96/106) Lipid Nutrition 0.226 g EPA +
0.176 g DHA or
1.093 g EPA
+0.847 g DHA
Preventive, Monotherapy 182 CES-D, GDS-
15, MADRS
10
da Silva et al. 2008 [42] Parkinson’s disease 29 (6, 8/7, 8) Not specified 0.720 g EPA
+0.480 g DHA as
monotherapy
or 0.720 g EPA +
0.480 g DHA +
antidepressant
Therapeutic, Monotherapy
& Adjunctive
84 BDI, MADRS 5
Lucas et al. 2009 [58] Mild-to-moderate
depression
120 (59/61) Isodis Natura 1.05 g Ethyl-EPA +
0.15 g DHA
Therapeutic, Monotherapy 56 HDRS, HSCL-
D-20
11
Doornbos et al. 2009 [56] Perinatal depression 119
3
(42, 41/36) Not specified 0.22 g DHA
4
or
0.22 g DHA
+0.22 g AA
Preventive, Monotherapy 252 EPDS 4
1
Crossover design.
2
This study used v3 LC-PUFA supplementation as monotherapy, but adjunctive supportive psychotherapy was provided.
3
182 women were initially recruited, but 57 dropped out by the 36th week of pregnancy; data on the remaining 125 women are presented.
4
Only data from the DHA group were entered into this meta-analysis.
BDI 5Beck Depression Inventory, CDI 5Children’s Depression Inventory, CDRS 5Children’s Depression Rating Scale, CES-D 5Center for Epidemiologic Studies of Depression Scale, DASS 5Depression Anxiety and Stress
Scales, EPDS 5Edinburgh Postnatal Depression Scale, GDS-15 5Geriatric Depression Rating Scale, HDRS 5Hamilton Rating Scale for Depression (SF refers to short form), HSCL-D-20 520-item Hopkins Symptom Checklist
Depression Scale, IDS-C 5Inventory of Depressive Symptomatology, MADRS 5Montgomery-A
˚sberg Depression Rating Scale, POMS 5Profile of Mood States, SCID-IV 5Structured Clinical Interview for DSM-IV.
Table 1. Continued.
EPA in Depression: A Meta-Analysis
JOURNAL OF THE AMERICAN COLLEGE OF NUTRITION 529
RESULTS
The overall effect using a random effects model showed a
significant fall in depressive symptoms with v3 LC-PUFA
supplementation in the 28 studies; standardized difference in
means 20.291 (95% CI 520.463 to 20.120, z 523.327, p
50.001). However, significant heterogeneity was present; Q
598.16, p,0.0001 with an I
2
of 67.4 and t
2
50.147,
indicating that the heterogeneity found represented an under-
lying true difference in effect sizes across studies. In addition,
there was considerable evidence of publication bias as shown
by the funnel plot in Fig. 1. Using Duval and Tweedie’s trim
and fill, 9 additional imputed studies (solid circles) are needed
to correct the asymmetry of the observed studies (open circles),
which reduces the estimate of standardized mean difference to
a nonsignificant level of 20.066 (95% CI 520.252 to 0.121).
Given that substantial heterogeneity was present, rather
than assume that v3 LC-PUFA supplementation in depression
is ineffective, further analyses were undertaken to attempt to
identify what supplement regimes and in what disorders v3
LC-PUFA supplementation might show efficacy. In order to
investigate possible sources of this heterogeneity, sensitivity
and meta-regression analyses were performed. The most
obvious source of possible heterogeneity relates to the
differing clinical populations studied who are likely to have
very different baseline levels of depression. For example, the
pathophysiological processes of depressive symptoms involved
in schizophrenia [45] are likely to be very different from those
in a community sample of individuals with mild-to-moderate
depression [52] or in healthy people with little or no depression
[68]. To confirm this, a random effects meta-regression
analysis was performed of standardized mean depression
scores on baseline depression scores using the UREML
method. Baseline scores were standardized according to
published norms for all of the rating scales used in the study.
However, 3 studies had to be excluded from this analysis
because no baseline scores were published [48,68,69]. The
analysis showed a significant relationship between baseline
depression scores and efficacy, with the greater the baseline
depression score, the more likely that v3 LC-PUFA supple-
mentation would reduce depressive symptoms (point estimate
for slope 521.692, 95% CI 522.969 to 20.415, z 5
22.597, p50.0094).
To gain an understanding of the difference in likely effect
size as a result of baseline depression levels and to assess
changes in heterogeneity, the sample was split into 2 groups
based upon insignificant and significant levels of baseline
depression according to the cut-off cores for each rating scale:
normal and mild as insignificant, and moderate and severe as
significant. As expected, the group with significant baseline
depression showed a substantial and highly significant
reduction in standardized mean depression scores with v3
Fig. 1. Funnel plot for all 28 studies of precision (1/standard error) by standardized difference in mean depression scores. Observed studies are shown
in open circles with the associated estimate in an open diamond. Imputed studies are shown in solid circles with the adjusted estimate in a
solid diamond.
EPA in Depression: A Meta-Analysis
530 VOL. 28, NO. 5
LC-PUFA supplementation (20.605, 95% CI 520.871 to
20.339, z 524.451, p,0.0001), whereas the groups with
insignificant depression showed no significant reduction
(20.074, 95% CI 520.317 to 0.169, z 520.598, p5
0.55). However, despite overall heterogeneity falling by
16.2%, this remained significant (Q 582.25, p,0.0001).
Hence, other factors apart from baseline depression need to be
examined to account for the observed levels of heterogeneity.
To further explore the influence of subcategories of
depression and other diagnostic groups, together with the
influence of preventive versus therapeutic intervention, and v3
LC-PUFA monotherapy versus v3 LC-PUFA supplementation
used as an adjunct to antidepressant therapy, further sensitivity
analyses were performed. The results of these analyses are
shown in Table 2. As some diagnostic categories contained
only one representative study (borderline personality disorder,
Parkinson’s disease, recurrent self-harm, and schizophrenia),
results are reported in these studies for comparative purposes
only. However, when considering groups that contained 2 or
more representative studies, these results would suggest that
v3 LC-PUFA supplementation may be ineffective for the
treatment of depressive symptoms in chronic fatigue, mild-to-
moderate depression, and in non-clinical populations. When
considering groups that contained 5 or more representative
studies, there was stronger evidence to suggest that v3 LC-
PUFA supplementation may be effective for the treatment of
depressive symptoms in bipolar disorder and major depression
but ineffective in perinatal depression. Overall, diagnostic
variability accounts for 22.4% of the observed heterogeneity
(Q 576.12), which nevertheless remained significant (p,
0.0001). Finally, studies examining therapeutic intervention (n
523) as opposed to preventive intervention (n 55), and
studies examining adjunctive therapy (n 513) as opposed to
monotherapy (n 516) showed significant benefit, with these
factors accounting for only small proportions of the observed
heterogeneity; Q 595.2 (3%) and 93.56 (1.6%), respectively.
Having identified that baseline depression, diagnostic
category, preventive versus therapeutic intervention, and
monotherapy versus adjunctive therapy account for some of
the observed heterogeneity, further random effects meta-
regression analyses were conducted to identify possible
sources of outstanding heterogeneity. As in the Ross et al.
[63] study, variables associated with the dosages used of DHA
and EPA and the DHA:EPA ratio were included alongside
various indices of study characteristics, notably methodologi-
cal quality as assessed by the modified Jadad score, study
duration, and sample size. The results of these analyses are
shown in Table 3.
These results showed that whilst the total dose of the v3
LC-PUFA preparation was unrelated to efficacy, the purity of
EPA within the preparation appeared to be influential. This
was illustrated by the significant negative intercept for dose of
DHA, suggesting that studies containing no DHA were more
likely to show a fall in depressive symptoms, and the
significant negative intercept and positive slope for the
DHA:EPA ratio, suggesting that as the purity of EPA increased
the more likely the studies were to show a fall in depressive
symptoms. With respect to study characteristics, the significant
negative intercept of the modified Jadad score suggested that
studies of the lowest methodological quality were more likely
Table 2. Model Statistics for the Random Effects Subgroup ANOVA Analyses Examining the Influence of Diagnostic Category,
Preventive Versus Therapeutic Intervention, and v3 LC-PUFA Monotherapy Versus v3 LC-PUFA as an Adjunct to
Antidepressant Therapy
Model
Number of
Studies Estimate 95% CI z Value pof z Q pof Q
Diagnostic category 76.10 ,0.0001
Bipolar disorder 5 20.364 20.682 to 20.045 22.239 0.0251 7.73 0.1720
Borderline personality disorder 1 20.288 21.169 to 0.593 20.641 0.5218 0.00 1.0000
Chronic fatigue 2 20.140 21.366 to 1.086 20.224 0.8229 10.14 0.0015
Major depression 8 20.551 21.059 to 20.043 22.125 0.0336 46.23 ,0.0001
Mild-to-moderate depression 2 20.044 20.257 to 0.170 20.403 0.6870 0.071 0.7900
Nonclinical 2 0.016 20.164 to 0.197 0.178 0.8587 1.24 0.5379
Parkinson’s disease 1 21.405 22.229 to 20.580 23.340 0.0008 0.28 0.5965
Perinatal depression 5 20.071 20.507 to 0.365 20.318 0.7503 10.42 0.0339
Recurrent self harm 1 20.954 21.677 to 20.232 22.588 0.0096 0.00 1.0000
Schizophrenia 1 0.000 20.420 to 0.420 0.000 1.0000 0.00 1.0000
Preventive vs therapeutic 95.24 ,0.0001
Preventive intervention 5 20.060 20.282 to 0.163 20.525 0.5996 8.25 0.1432
Therapeutic intervention 23 20.362 20.578 to 20.147 23.296 0.001 86.99 ,0.0001
Mono- vs adjunctive therapy 93.56 ,0.0001
Monotherapy 16 20.130 20.317 to 0.057 21.360 0.174 33.47 0.004
Adjunctive therapy 13 20.475 20.780 to 20.169 23.042 0.002 60.09 ,0.0001
EPA in Depression: A Meta-Analysis
JOURNAL OF THE AMERICAN COLLEGE OF NUTRITION 531
to show a fall in depressive symptoms; the significant negative
intercept for study duration suggested that studies of the
shortest duration were more likely to show a fall in depressive
symptoms; and the significant negative intercept and positive
slope for sample size suggested that the smallest studies were
more likely to report a fall in depressive symptoms but, as
study sample size increased, no change in depressive
symptoms was more likely to be reported.
On the basis of these meta-regression findings suggesting
that EPA purity was a significant factor, a random effects
subgroup ANOVA was performed as in the Ross et al. [63]
study. However, in this case, the current studies were classified
into 4 groups on the basis of whether they employed pure
DHA, mainly DHA (.50%), mainly EPA (.50%), or pure
ethyl-EPA. One of the included studies employed ethyl-EPA,
but as a small quantity of DHA was also present, this study was
classified as mainly EPA rather than pure ethyl-EPA [58]. The
results of this analysis are shown in a forest plot (Fig. 2) and
the associated Table 4.
These results appear to confirm the observation originally
made by Ross et al. [63], that only v3 LC-PUFA preparations
with a predominant or pure EPA content show efficacy in
treating depressive symptoms. In addition, the observed
heterogeneity was now insignificant in the pure DHA (Q 5
0.687, p50.7093), predominantly DHA (Q 56.98, p5
0.0726), and pure ethyl-EPA (Q 516.76, p50.0799) groups,
with low levels of between-study variance (t
2
50.000, 0.064,
and 0.069, respectively). However, the predominantly EPA
group of studies still showed significant heterogeneity (Q 5
62.78, p,0.0001) and between-study variance (t
2
50.248),
and overall heterogeneity fell by only 11.15% (Q 587.2).
Despite these findings suggesting EPA efficacy, a note of
caution is necessary here, both because the number of studies
within each group was small, especially in the pure DHA and
mainly DHA groups, and because meta-analytic and meta-
regression findings are observational, so it is important not to
assume causality on the basis of these findings. The reader will
recall that the earlier UREML analyses (Table 3) showed that
as the proportion of EPA in the v3 LC-PUFA preparation rose,
so did efficacy, but that in contrast, study sample size and
study duration were inversely associated with efficacy. There-
fore, the ‘‘effect’’ of EPA-containing preparations could
simply be confounded by the fact that studies with small
sample sizes and of short duration are more likely to show
efficacy than are studies with large sample sizes and long
duration. Against this possibility, however, is the finding of
nonsignificant correlations between DHA:EPA ratio and both
study duration (Kendall’s tau 50.127, z 51.0869, p5
0.2771), and sample size (tau 50.216, z 51.8063, p5
0.0709).
Although the possibility of confounding relationships
appeared to be minimal in the whole group of studies, within
the pure ethyl-EPA group of studies, a number of anomalous
and/or confounding relations were suggested by UREML
meta-regression analyses (Table 5). Despite an overall bene-
ficial effect of pure ethyl-EPA on depressive symptoms, a
paradoxical inverse relationship between EPA dose and
improvement in depressive symptoms was evident in this
group as shown by a significant negative value for intercept
and a significant positive value for slope. In addition, both
study duration and sample size were inversely associated with
improvement in depressive symptoms to a highly significant
extent. Moreover, in this group of studies, a significant
correlation was evident between the EPA dose itself and study
duration (Kendall’s tau 50.4, z 51.988, p50.0468), which
could therefore indicate that the inverse relationship between
EPA and efficacy is artefactual and simply a consequence of
the inverse relationship between study duration and efficacy.
In contrast to these observations in the pure ethyl-EPA
group of studies, when the mainly DHA and mainly EPA
groups are considered independently (n 517 studies),
UREML meta-regression analyses shown in Fig. 3 demon-
strated a significant dose-response relationship for EPA
efficacy in reducing depressive symptoms (EPA point estimate
for slope 20.195, 95% CI 520.375 to 20.015, z 522.123,
p50.0338; and intercept 20.052, 95% CI 520.398 to 0.294,
z520.293, p50.7692), whereas DHA dose was unrelated to
efficacy (DHA point estimate for slope 20.032, 95% CI 5
Table 3. UREML Random Effects Meta-Regression Analyses
for Standardized Difference in Means of Depressive Symptoms
Variable
Point
Estimate 95% CI z Value pValue
Total dose of v3 LC-PUFA
Slope 20.050 20.146 to 0.045 21.037 0.2997
Intercept 20.195 20.476 to 0.086 21.361 0.1736
Dose of DHA
Slope 0.008 20.206 to 0.221 0.070 0.9439
Intercept 20.314 20.555 to 20.074 22.564 0.0103
Dose of EPA
Slope 20.077 20.192 to 0.037 21.322 0.1861
Intercept 20.187 20.440 to 0.066 21.450 0.1471
DHA:EPA ratio
Slope 0.005 0.0001 to 0.011 1.977 0.0480
Intercept 20.495 20.759 to 20.230 23.669 0.0002
Modified Jadad score
Slope 0.049 20.031 to 0.129 1.190 0.2342
Intercept 20.665 21.288 to 20.043 22.094 0.0363
Study duration
Slope 0.003 20.001 to 0.007 1.682 0.0926
Intercept 20.603 21.000 to 20.206 22.981 0.0029
Sample size
Slope 0.009 0.003 to 0.014 2.961 0.0031
Intercept 20.617 20.892 to 20.341 24.387 ,0.0001
EPA in Depression: A Meta-Analysis
532 VOL. 28, NO. 5
20.357 to 0.294, z 520.191, p50.8488; and intercept
20.296, 95% CI 520.756 to 0.165, z 521.257, p5
0.2088). In addition, in these studies employing mixed v3 LC-
PUFA preparations, there appeared to be no evidence of
confounding with study duration (point estimate for slope
0.002, 95% CI 520.005 to 0.009, z 50.517, p50.6052; and
intercept 20.497, 95% CI 521.201 to 0.208, z 521.381, p
50.1672), although studies with small sample sizes were still
more likely to show a fall in depressive symptoms with
supplementation (point estimate for slope 0.009, 95% CI 5
Fig. 2. Forest plot examining the effect of the type of v3 LC-PUFA supplementation employed on the reduction in depressive symptoms.
EPA in Depression: A Meta-Analysis
JOURNAL OF THE AMERICAN COLLEGE OF NUTRITION 533
0.001 to 0.017, z 52.223, p50.0262; and intercept 20.696,
95% CI 521.130 to 20.261, z 523.140, p50.0017). It is
noteworthy that in these groups the v3 LC-PUFA preparations
were predominantly of natural origin, in contrast to artificially
purified DHA or synthetic ethyl-EPA, with 3 out of the top 6
studies showing the strongest effect sizes employing the
mainly EPA containing menhaden fish body oil.
DISCUSSION
The results of the current meta-analysis appear to confirm
the original observation made by Ross et al. [63] that EPA and
not DHA may be the responsible agent conferring benefit for
the treatment of depressive symptoms with v3 LC-PUFA
supplementation (see Fig. 2 and Table 4). These results also
demonstrate that it is inappropriate to assume that the effects of
these 2 v3 LC-PUFAs will be the same, either in randomized
controlled trials or in meta-analytic studies examining the
effect of v3 LC-PUFA supplementation in a variety of
disorders. It is noteworthy that the meta-analyses by Appleton
et al. [62] and Rogers et al. [52], which largely concluded that
v3 LC-PUFA supplementation was ineffective in depression,
did not consider the differential effects of EPA versus DHA.
Moreover, these findings demonstrate that there is a significant
relationship between baseline depression levels and efficacy,
indicating that future studies examining the effects of v3 LC-
PUFA supplementation in depression should ensure that the
population studied is actually suffering from clinically relevant
levels of depressive symptomatology.
For example, in the large, excellently designed study by
Rogers et al. [52] baseline Beck Depression Inventory (BDI)
scores were 13.9 in both study groups. According to the
published norms for the BDI, a score of 0–13 is regarded as
indicative of minimal or no significant depression, 14–19 as
indicative of mild depression, 20–28 as indicative of moderate
depression, and 29–63 as indicative of severe depression [70].
This would suggest that, despite the recruitment strategy
having been designed to select individuals with mild-to-
moderate depression, the population studied by Rogers et al.
[52] was not suffering from clinically relevant depression and,
notwithstanding the mainly DHA-containing regime used in
this study, these 2 factors could explain why no clinical effect
was demonstrated.
In addition, the subgroup analyses presented in Table 2
provide further evidence that v3 LC-PUFA supplementation
may be most efficacious in clinical populations. Notably, the 2
studies in non-clinical populations failed to demonstrate
efficacy [59,68] and v3 LC-PUFA supplementation was most
effective in bipolar disorder and major depression, as a
therapeutic as opposed to a preventive intervention, and as an
adjunctive treatment rather than as monotherapy. The in-
creased efficacy of v3 LC-PUFA supplementation as an
adjunctive treatment is consistent with the observational study
of Fe´art et al. [24], which found that higher plasma EPA
concentration was associated with lower depressive symptoms,
especially in patients taking antidepressants. Overall, these
analyses lend further support to the conclusion that v3 LC-
PUFA supplementation is most efficacious in clinical popula-
tions. However, the point prevalence of major depression is
low at 2.6% [71], meaning that in order to identify whether v3
LC-PUFA supplementation is effective in non-clinical popula-
tions for the prevention of new episodes of depression, much
larger sample sizes, studied over an extended period of time,
may be required to demonstrate efficacy.
Although the current meta-analysis identified that diag-
nostic variation, baseline depression severity, and v3 LC-
PUFA regime type accounted for some of the observed
heterogeneity, significant heterogeneity remained. A possible
Table 5. UREML Meta-Regression Statistics for Standardized
Difference in Means of Depressive Symptoms in the Ethyl-
EPA Group of Studies
Variable
Point
Estimate 95% CI z Value pValue
Dose of EPA
Slope 0.122 0.026 to 0.218 2.484 0.0130
Intercept 20.694 21.040 to 20.347 23.924 0.0001
Modified Jadad score
Slope 20.055 20.212 to 0.102 20.686 0.4928
Intercept 0.002 21.124 to 1.127 0.003 0.9979
Study duration
Slope 0.013 0.005 to 0.021 3.346 0.0008
Intercept 21.535 22.270 to 20.801 24.097 0.0000
Sample size
Slope 0.014 0.004 to 0.024 2.720 0.0065
Intercept 20.790 21.174 to 20.405 24.028 0.0001
Table 4. Model Statistics for the 4 Subgroup Random Effects ANOVA of v3 LC-PUFA Preparation Type
Group
Number of
Studies Estimate 95% CI z Value pof z Q pof Q I
2
t
2
Pure DHA 3 0.001 20.330 to 0.332 0.004 0.9965 0.687 0.7093 0.00 0.000
Mainly DHA 4 0.141 20.195 to 0.477 0.821 0.4116 6.98 0.0726 57.01 0.064
Mainly EPA 13 20.446 20.753 to 20.138 22.843 0.0045 62.80 0.0000 77.71 0.248
Pure EPA 8 20.396 20.650 to 20.141 23.051 0.0023 16.76 0.0799 40.33 0.069
EPA in Depression: A Meta-Analysis
534 VOL. 28, NO. 5
outstanding source of this heterogeneity is the influence of sex,
which, however, could not be assessed in the current meta-
analysis, as none of the included studies reported outcomes for
depressive symptoms stratified by sex. Despite documented
epidemiologic associations between fish intake and reduced
risk of depression [13–19], one study that analyzed results by
sex found that low consumption of fish was associated with
depression only in women [20]. In addition, metabolic studies
indicate that, despite identical dietary intake, women show
higher levels of plasma DHA than men [72]. This increase in
DHA in women appears to be sensitive to estrogen and may
therefore represent a biological preparedness for the large
demands of the fetus and neonate for DHA during pregnancy
and lactation [73]. This, in turn, may cause women to be more
vulnerable to dietary deficiencies of v3 LC-PUFAs and may
also explain, at least in part, why women are at greater risk for
the development of depression [74]. It may also explain why
studies examining perinatal depression in the current meta-
analysis failed to show efficacy, as supplemented v3 LC-
PUFAs may have been diverted to the demands of the
developing fetus rather than contributing to therapeutic
efficacy for depressive symptoms. Subsequent studies of v3
Fig. 3. UREML meta-regression of DHA and EPA dose on standardized difference in mean depression scores in the mainly DHA and mainly EPA
groups of studies. The size of the circles is proportional to the statistical weighting attached to each study.
EPA in Depression: A Meta-Analysis
JOURNAL OF THE AMERICAN COLLEGE OF NUTRITION 535
LC-PUFA supplementation in depression should therefore
report outcomes in both males and females.
Despite these conclusions concerning EPA efficacy, a
number of concerns remain regarding the robustness of the
findings concerning the effectiveness of EPA in depressive
symptoms identified in the current meta-analysis. These
concerns relate to the potential influence of specific method-
ological characteristics of the included studies; publication
bias; statistical power; anomalous findings with respect to the
influence of the proportion of EPA versus the absolute
amounts of EPA between differing subgroups of studies; and
the quantitative assessment of study methodological quality,
sample size, and duration.
First, the studies were of mixed methodological quality
according to the Jadad system: in 12 out of the 28 studies, the
method of randomization was not adequately described [39–
43,45,48,50,54,57,68,69]; in 6 studies the nature of the placebo
intervention was not adequately described [42,50,54,56,57,69];
and in a further 4 studies dropouts were not adequately
described [44,46,68,69]. Furthermore, in 2 studies dropout
rates were very high. In the negative study by Doornbos et al.
[56], examining the effect of low dose DHA for the prevention
of perinatal depression, 57 out of the 182 individuals dropped
out of the study during pregnancy. In the negative study by
Keck et al. [48], examining pure ethyl-EPA for the adjunctive
treatment of bipolar disorder, 54% of the participants dropped
out prior to completion.
Regarding the use of ITT analyses, only 1 out of the 3 pure
DHA studies used ITT [50]; all 4 of the mainly DHA studies
used ITT [47,51–53]; only 2 out of the 13 mainly EPA studies
used ITT [40,41]; and 5 out of the 8 pure ethyl-EPA studies
used ITT [34,36,37,45,48]. Given that the mainly EPA group
of studies showed the largest effect size, the relative absence of
ITT analyses in this group may have biased the results in favor
of EPA supplementation.
Concerning whether the outcome assessors were blinded to
treatment group, this could not be ascertained with certainty
from descriptions given in 2 out of the 3 pure DHA studies
[49,50], in 4 out of the 13 mainly EPA studies [38,39,68,69],
and in 3 out of the 8 pure ethyl-EPA studies [37,48,54]. Of
note, all of the mainly DHA studies had clear descriptions of
outcome assessor blinding [47,51–53]. It is possible, therefore,
that a relative reduction in outcome assessor blinding observed
in the studies employing EPA could have biased the results in
favor of EPA supplementation.
Regarding baseline comparisons, 1 pure DHA study [56], 5
mainly EPA studies [42–44,68,69], and 3 pure ethyl-EPA
studies [35,37,48] did not provide adequate descriptions of
baseline characteristics between supplement and placebo
groups. In 2 of the pure ethyl-EPA studies, baseline
measurements were provided, but either no baseline variance
[37] or appropriate labeling [35] of this variance was indicated.
It is possible, therefore, that the relative absence of information
regarding differences in baseline characteristics among studies
employing EPA supplementation could have led to over- or
underestimation of the efficacy of EPA. With respect to the use
of pure DHA, in one of these studies, baseline Hamilton
Depression Rating Scale measures were significantly greater in
the placebo group compared with the DHA group [50],
potentially biasing the results against demonstrating efficacy
for DHA supplementation.
Concerning the attempt to conceal differences in the
perception of fish taste between v3 LC-PUFA supplementa-
tion and placebo groups, none of the pure DHA studies
employed this strategy; 2 out of the 4 mainly DHA studies
added orange oil [52] or peppermint oil [51] to both active and
placebo preparations; 5 out of the 13 mainly EPA studies either
stated that the capsules were indistinguishable with respect to
taste [68]; or added 0.2% fish oil [58], 1% fish oil [57], or 1%
EPA/DHA mixture [41] to the placebo capsules; and 1 out of
the 8 pure ethyl-EPA studies stated that the capsules were
indistinguishable by aftertaste [48]. It is possible, therefore,
that a greater number of individuals in the pure ethyl-EPA
group of studies were able to perceive that they were in receipt
of v3 LC-PUFA supplementation rather than placebo, biasing
the results in favor of efficacy for EPA-containing prepara-
tions. However, none of the pure DHA groups made any
attempt to conceal fish taste and, in one of these studies, a fish
aftertaste was reported in 14 out of 35 participants, which
might have disrupted blinding and biased the result in favor of
the DHA group [50]. Moreover, the need to conceal fish taste
may be more important for mixed preparations than for
purified ethyl-EPA. In the study by Stoll et al. [38] employing
mainly EPA without concealment, 86% of the v3 LC-PUFA
group correctly guessed their group allocation compared with
63% in the placebo group, whereas in the study by Nemets et
al. [36] employing pure ethyl-EPA without concealment,
neither patients nor clinicians were able to guess group
allocation correctly. In addition, studies employing mixed
preparations, which conducted both concealment and the
assessment of blinding success, confirmed that individuals
were unable to guess group allocation [51,58]. Therefore, it is
unlikely that this factor significantly biased the overall results
presented in this meta-analysis, as concealment was under-
taken with equal frequency between the mainly DHA and
mainly EPA groups of studies.
Regarding the assessment of compliance using RBC
membrane or plasma phospholipid analyses, all 3 pure DHA
studies conducted these analyses, showing significant changes
in RBC membrane DHA levels as a result of supplementation
[49,50,56]; all 4 mainly DHA studies conducted these
analyses, showing significant increases in RBC membrane
EPA and DHA [47,53], plasma EPA and DHA [52], and
plasma total v3 LC-PUFA levels [51]; 8 out of the 13 mainly
EPA in Depression: A Meta-Analysis
536 VOL. 28, NO. 5
EPA studies conducted these analyses, showing either a
significant increase in RBC membrane EPA and DHA
[42,44,58], a significant decrease in plasma arachidonic acid
(AA)/EPA ratio [68], a significant increase in RBC DHA but
not EPA [40], or no significant changes in RBC membrane
EPA or DHA [39,55]; and only 1 out of the 8 pure ethyl-EPA
studies conducted these analyses, which nevertheless showed a
significant increase in RBC membrane EPA content [45]. The
lack of assessment of compliance in all but one of the pure
ethyl-EPA group of studies suggests that the reported benefits
on depressive symptoms in this group of studies cannot
therefore be definitively attributed to the EPA content of the
supplementation regime. In addition, the studies by Su et al.
[39,40], which contributed substantially to the overall efficacy
results of EPA-containing preparations, showed no significant
change in RBC membrane EPA or DHA in the first study [39],
and a significant change in RBC membrane DHA but not EPA
in the second study [40]. Given that these studies employed
EPA doses of 4.4 g and 2.2 g respectively, these findings are
surprising, as other studies employing much lower loses of
EPA, ranging from 80 mg to 600 mg, have shown significant
increases in RBC membrane EPA [75–78]. This could suggest
that the reported benefits on depressive symptoms in the
studies by Su et al. [39,40] may have occurred as a result of
factors other than EPA supplementation.
Outstanding methodological issues related to the exclusion
of placebo responders during 1-week single-blind run-in phases
in the Su et al. [39,40] studies, where 4 out of 34 patients in the
first study [39] and 4 out of 40 in the second study [40] were
excluded on the basis of a .20% reduction in HDRS scores.
Whilst inclusion of this strategy undoubtedly compromises
study generalizability, it has not, in fact, been shown to result in
magnified treatment versus placebo group differences or to have
any discernable impact on differential response rates [79]. Thus,
in a meta-analysis of 34 trials examining 3047 patients receiving
a selective serotonin-reuptake inhibitor for depression versus
3740 patients receiving placebo, there was no statistically
significant difference in effect size between the clinical trials
that had a placebo run-in phase followed by withdrawal of
placebo responders and those trials that did not [80]. Therefore,
it is unlikely that inclusion of the Su et al. [39,40] studies in the
current meta-analysis has resulted in overestimation of v3 LC-
PUFA supplementation treatment effects.
The second concern regarding the robustness of these
findings of EPA efficacy relates to the considerable evidence
of publication bias evident from both the funnel plot (see
Fig. 1) and the imputation of 9 further studies required to
correct this bias from the Duval and Tweedie trim and fill
analysis. Therefore, it is possible that had these ‘‘missing’’
studies been available for inclusion in the current meta-
analysis, the preferential effect of EPA versus DHA might no
longer be evident.
Third, the number of studies included in the analysis of
subgroups was small, especially in the pure DHA and mainly
DHA groups, which contained only 3 and 4 studies,
respectively. Thus, only 7 studies examined DHA in contrast
with 21 studies examining mainly EPA or pure EPA. With a
larger number of studies in the DHA subgroups, it is possible
that the preferential effect of EPA might become less evident.
Fourth, although the overall meta-regression analyses (see
Table 3) suggested that, as the proportion of EPA increases
within the preparation so does efficacy, analysis of the ethyl-
EPA subgroup of studies suggested an inverse relationship
between absolute EPA dose and efficacy (see Table 5). This
latter finding is consistent with Peet and Horrobin’s [37] dose-
ranging study that identified 1 g/d of EPA as being more
effective than either 2 or 4 g/d. It is also consistent with the
same research group’s hypothesis that the balance between
EPA and AA within neuronal membranes may be important for
optimal neurologic functioning. For example, in a study of
EPA supplementation in schizophrenia, RBC membrane AA
but not EPA correlated with clinical improvement [81],
suggesting that excessive dosages of EPA may deplete
neuronal AA, which may not be beneficial. However, against
the conclusion that only low-dose EPA is efficacious is the
finding from the current meta-analysis that in the larger group
of 17 studies employing mixed EPA and DHA supplements, a
significant positive relationship between absolute EPA dose
and efficacy was demonstrated. Taken together with the
significant correlation between EPA dose and study duration in
the ethyl-EPA group of studies (Kendall’s tau 50.4), this
would suggest that the inverse dose-response relationship with
efficacy in the ethyl-EPA group of studies is a chance
artefactual finding. However, further research will be required
to finally resolve this question.
Fifth, the meta-regression analyses on all 28 studies showed
that studies with the lowest methodological quality score,
sample size, and duration were more likely to show efficacy,
whereas studies with large sample sizes were least likely to
show efficacy (see Table 3). These findings could indicate that
the positive efficacy of EPA in ameliorating depressive
symptoms found in the current meta-analysis might be
confounded by indices of study quality and therefore be a
chance finding rather than a real effect. These concerns
appeared to be most extreme in the ethyl-EPA group of studies
where a highly significant inverse relationship between
efficacy and study duration and sample size was demonstrated
(see Table 5). Whilst this inverse association could indicate
that EPA has only a temporary effect in depression that
diminishes over time, the inverse relationship between efficacy
and both absolute EPA dose (as discussed above) and sample
size (see Table 5) would again tend to support the notion that
both the purported preferential effect of low dose ethyl-EPA,
and the purported temporary effect of ethyl-EPA, are
EPA in Depression: A Meta-Analysis
JOURNAL OF THE AMERICAN COLLEGE OF NUTRITION 537
artifactual chance findings. It is noteworthy that there was no
association between methodological quality and efficacy in the
ethyl-EPA group of studies, as generally these studies were
conducted to a high standard. The relative cost of ethyl-EPA as
opposed to fish oil or enriched natural preparations may have
limited sample size and duration, and thus be a possible
explanation for the exacerbation of chance associations in this
group of studies. However, against the conclusion that the
efficacy of EPA in depression is a chance finding is the fact
that in the whole group of studies the correlation between
DHA:EPA ratio and study duration was insignificant; and, in
the mainly DHA and mainly EPA group of studies, the largest
grouping of studies in this analysis, a significant positive
association between absolute EPA dose and efficacy was found
that was not confounded by study duration (see Fig. 3).
Despite the methodological problems outlined in the
previous sections, the evidence from 17 studies indicating a
dose-response relationship between EPA dose and efficacy for
depressive symptoms would suggest that EPA in the form of
natural or enriched preparations may be more beneficial than
pure ethyl-EPA. It is noteworthy that, in the 6 studies showing
the greatest efficacy, 3 of them employed the mainly EPA
containing menhaden fish body oil with an EPA:DHA ratio of
approximately 65%. Further studies are required of sufficient
sample size, duration, and methodological quality to compare
pure ethyl-EPA with natural EPA preparations containing
EPA:DHA ratios ranging from 60% through to 100% to
resolve the question of what constitutes an optimal v3 LC-
PUFA formulation in the treatment of depression.
DHA is the most abundant LC-PUFA present in brain cell
membranes in contrast to EPA, which is present at levels
several hundred-fold lower than DHA [82,83]. Consequently,
the rationale for supplementation with DHA has historically
rested upon the assumption that increasing the nutritional
availability of a major structural component of neuronal
membranes would have beneficial effects on brain function,
including the amelioration of depression. However, rather than
providing a structural substrate, evidence is accumulating that
v3 LC-PUFAs may instead exert their effects through cell
signaling mechanisms as outlined below. This may explain
why EPA, present at very low levels in the brain as compared
with DHA, may have beneficial effects in depression.
Regarding the possible reasons why EPA and not DHA
may be more effective in depression, there is increasing
evidence that DHA supplementation may have damaging
effects on the nervous system. DHA, the most highly
unsaturated v3 LC-PUFA in the body, is very susceptible to
lipid peroxidation and can damage DNA [84], increase
production of reactive oxygen species in glial cells [85], and
worsen neurologic state in rat perfusion-injury models [86].
Highly reactive A
4
/J
4
neuroprostanes produced from DHA in
vivo under conditions of oxidative stress [87] may explain
some of the above negative findings. In addition, although
retro-conversion of DHA to EPA can occur to a certain extent,
DHA is at the end of the biosynthetic pathway of v3 LC-
PUFAs and therefore supplementation may boost DHA to
levels that cannot be adequately handled by metabolic
pathways, which in turn, may further exacerbate production
of damaging reactive derivatives. It is noteworthy that under
normal circumstances, the rate of conversion of dietary a-
linolenic acid to DHA is about 1% [88] and daily turnover of
DHA in the adult human brain is only 4.6 mg [89], suggesting
that it may not be desirable to boost DHA levels to such an
extent as occurs during supplementation [82].
In contrast to the above findings with respect to DHA, there
are many lines of evidence to indicate why EPA might be
beneficial in depression. First, EPA has neuroprotective
actions on lipopolysaccharide (LPS)-induced hippocampal
dysfunction via the prevention of LPS-induced phosphoryla-
tion of c-Jun N-terminal kinase, c-Jun and Bcl-2, which in turn
prevents the secretion of interleukin 1b(IL-1b), prevents
increases in mitochondrial membrane permeability, prevents
release of cytochrome C, and prevents neuronal apoptosis [90].
Second, oxidized derivatives of EPA, unlike the A
4
/J
4
neuroprostanes derived from DHA as described above, have
beneficial anti-inflammatory effects [91], in addition to the
well-documented anti-inflammatory effects of EPA-derived
eicosanoids [11]. Third, with respect to the inflammatory
hypothesis of depression, EPA appears to have the following
effects: (1) EPA is more effective than DHA at reducing the
inflammatory cytokines tumor necrosis factor-a(TNF-a), IL-6,
and IL-1b[92], an action that occurs via the mechanism of
EPA inhibition of the activity of nuclear factor kappa-B (NF-
kB), an important nuclear regulator of the inflammatory
response [93]; (2) although dietary EPA and DHA incorporate
into cell membranes equally as DHA, dietary EPA is more
effective at reducing inflammation in vivo [94]; and (3) dietary
DHA may not be beneficial in depression owing to the fact that
it appears to induce a T helper cell type 1–like immune
response with a raised interferon-cto IL-10 ratio, whereas
EPA does not induce this effect [95]. For this reason, the
authors of this latter study argue that highly purified EPA, free
of any DHA, should be used in the treatment of depression.
Given that the available evidence discussed above suggests
that pure EPA, in contrast to DHA, may be beneficial for
depressive symptoms, this might appear at odds with the
epidemiologic evidence linking fish consumption, a source of
mixed fatty acids, with reduced risk of depression. However,
the fatty acid content of fish varies substantially by species,
method of farming, and season. For example, the percentage
wet weight of EPA versus DHA is, respectively, 15.7 versus
0.7 for anchovy oil, 11.0 versus 9.1 for Atlantic menhaden oil,
and 6.2 versus 9.1 for Atlantic salmon oil (see Table 2.2 on
page 25 of reference 96). In addition, seasonal variation in
EPA in Depression: A Meta-Analysis
538 VOL. 28, NO. 5
fatty acid content is largely determined by spawning. Thus,
fish accumulate EPA and DHA in muscle tissue before the
reproductive season and transfer these fatty acids to the gonads
during spawning when levels of LC-PUFA in hard roe can
reach 3–4 times higher than those in muscle tissue [97]. These
natural variations in fatty acid levels may explain, in part, the
mixed findings obtained from epidemiologic studies of fish
consumption and intervention studies employing fish oils for
therapeutic benefit in depression. If possible, further studies
should examine whether individuals consuming predominantly
EPA-containing fish are at lower risk for the development of
depression compared with those consuming fish with a higher
DHA content. If confirmed, this might have important
implications with respect to dietary advice regarding the
prevention of depression.
In conclusion, there is substantive evidence both from the
current meta-analysis, and from the studies outlined above, that
EPA and not DHA may be effective in depressive disorders.
However, further studies are required of sufficient methodo-
logical quality, duration, and sample size to confirm these
findings. A direct comparative trial of EPA versus DHA should
be conducted, preferably also comparing these fatty acids in
triglyceride versus ethyl ester forms to identify whether the
method of fatty acid delivery influences efficacy. If EPA
superiority is confirmed, further studies are needed to answer
the following outstanding issues: (1) Are pure EPA prepara-
tions better than those containing a maximum of 40% DHA?
(2) If DHA-free EPA preparations are found to be more
effective, is highly purified ethyl-EPA better than enriched
natural products containing triglycerides? and (3) If a U-
shaped dose-response curve and short-term effect of EPA in
depression is replicated, could this be due to increasing AA
depletion as supplementation progresses? If so, could the
addition of c-linolenic acid to the supplement regime prolong
the effect of EPA and allow larger doses of EPA to be tolerated
by both increasing the synthesis of AA and by increasing the
production of dihomo-c-linolenic acid derived anti-inflamma-
tory eicosanoids?
ACKNOWLEDGMENT
The author wishes to thank Mr Peter Avakian, Director of
the Academy of Nutritional Medicine, for his unfailing support
and encouragement during the preparation of this manuscript.
REFERENCES
1. Andrade L, Caraveo-Anduaga JJ, Berglund P, Bijl RV, De Graaf
R, Vollebergh W, Dragomirecka E, Kohn R, Keller M, Kessler
RC, Kawakami N, Kilic C, Offord D, Ustun TB, Wittchen HU:
The epidemiology of major depressive episodes: results from the
International Consortium of Psychiatric Epidemiology (ICPE)
Surveys. Int J Methods Psychiatr Res 12:3–21, 2003.
2. Penninx BW, Beekman AT, Honig A, Deeg DJ, Schoevers RA,
van Eijk JT, van Tilburg W: Depression and cardiac mortality:
results from a community-based longitudinal study. Arch Gen
Psychiatry 58:221–227, 2001.
3. Sobocki P, Jo¨nsson B, Angst J, Rehnberg C: Cost of depression in
Europe. J Ment Health Policy Econ 9:87–98, 2006.
4. Mathers CD, Loncar D: Projections of global mortality and burden
of disease from 2002 to 2030. PLoS Med 3:E442, 2006.
5. Klerman GL, Weissman MM: Increasing rates of depression.
JAMA 261:2229–2235, 1989.
6. Jemal A, Ward E, Anderson RN, Murray T, Thun MJ: Widening of
socioeconomic inequalities in U.S. death rates, 1993–2001. PLoS
One 3:E2181, 2008.
7. O’Flaherty M, Ford E, Allender S, Scarborough P, Capewell S:
Coronary heart disease trends in England and Wales from 1984 to
2004: concealed levelling of mortality rates among young adults.
Heart 94:178–181, 2008.
8. Puri BK: Cardiovascular disease and depression: the PUFA
connection. Int J Clin Pract 62:355–357, 2008.
9. Raison CL, Capuron L, Miller AH: Cytokines sing the blues:
inflammation and the pathogenesis of depression. Trends Immunol
27:24–31, 2006.
10. Pace TW, Hu F, Miller AH: Cytokine-effects on glucocorticoid
receptor function: relevance to glucocorticoid resistance and the
pathophysiology and treatment of major depression. Brain Behav
Immun 21:9–19, 2007.
11. Calder PC: n-3 Polyunsaturated fatty acids, inflammation, and
inflammatory diseases. Am J Clin Nutr 83:1505S–1519S, 2006.
12. Kiecolt-Glaser JK, Belury MA, Porter K, Beversdorf DQ,
Lemeshow S, Glaser R: Depressive symptoms, omega-6:omega-
3 fatty acids, and inflammation in older adults. Psychosom Med
69:217–224, 2007.
13. Appleton KM, Peters TJ, Hayward RC, Heatherley SV,
McNaughton SA, Rogers PJ, Gunnell D, Ness AR, Kessler D:
Depressed mood and n-3 polyunsaturated fatty acid intake from
fish: non-linear or confounded association? Soc Psychiatry
Psychiatr Epidemiol 42:100–104, 2007.
14. Appleton KM, Woodside JV, Yarnell JW, Arveiler D, Haas B,
Amouyel P, Montaye M, Ferrieres J, Ruidavets JB, Ducimetiere P,
Bingham A, Evans A: Depressed mood and dietary fish intake:
direct relationship or indirect relationship as a result of diet and
lifestyle? J Affect Disord 104:217–223, 2007.
15. Hibbeln JR: Fish consumption and major depression. Lancet
351:1213, 1998.
16. Hibbeln JR: Seafood consumption, the DHA content of mothers’
milk and prevalence rates of postpartum depression: a cross-
national, ecological analysis. J Affect Disord 69:15–29, 2002.
17. Noaghiul S, Hibbeln JR: Cross-national comparisons of seafood
consumption and rates of bipolar disorders. Am J Psychiatry
160:2222–2227, 2003.
18. Silvers KM, Scott KM: Fish consumption and self-reported physical
and mental health status. Public Health Nutr 5:427–431, 2002.
19. Tanskanen A, Hibbeln JR, Tuomilehto J, Uutela A, Haukkala A,
Viinamaki H, Lehtonen J, Vartiainen E: Fish consumption and
depressive symptoms in the general population in Finland.
Psychiatr Serv 52:529–531, 2001.
EPA in Depression: A Meta-Analysis
JOURNAL OF THE AMERICAN COLLEGE OF NUTRITION 539
20. Timonen M, Horrobin D, Jokelainen J, Laitinen J, Herva A,
Ra¨sa¨ nen P: Fish consumption and depression: the Northern
Finland 1966 birth cohort study. J Affect Disord 82:447–452,
2004.
21. Adams PB, Lawson S, Sanigorski A, Sinclair AJ: Arachidonic acid
to eicosapentaenoic acid ratio in blood correlates positively with
clinical symptoms of depression. Lipids 31 Suppl:S157–S161,
1996.
22. Conklin SM, Harris JI, Manuck SB, Yao JK, Hibbeln JR, Muldoon
MF: Serum omega-3 fatty acids are associated with variation in
mood, personality and behavior in hypercholesterolemic commu-
nity volunteers. Psychiatry Res 152:1–10, 2007.
23. Edwards R, Peet M, Shay J, Horrobin D: Omega-3 polyunsaturated
fatty acid levels in the diet and in red blood cell membranes of
depressed patients. J Affect Disord 48:149–155, 1998.
24. Fe´art C, Peuchant E, Letenneur L, Samieri C, Montagnier D,
Fourrier-Reglat A, Barberger-Gateau P: Plasma eicosapentaenoic
acid is inversely associated with severity of depressive symptom-
atology in the elderly: data from the Bordeaux sample of the
Three-City Study. Am J Clin Nutr 87:1156–1162, 2008.
25. Maes M, Christophe A, Delanghe J, Altamura C, Neels H, Meltzer
HY: Lowered omega3 polyunsaturated fatty acids in serum
phospholipids and cholesteryl esters of depressed patients.
Psychiatry Res 85:275–291, 1999.
26. Maes M, Smith R, Christophe A, Cosyns P, Desnyder R, Meltzer
H: Fatty acid composition in major depression: decreased omega 3
fractions in cholesteryl esters and increased C20: 4 omega 6/C20:5
omega 3 ratio in cholesteryl esters and phospholipids. J Affect
Disord 38:35–46, 1996.
27. Peet M, Murphy B, Shay J, Horrobin D: Depletion of omega-3
fatty acid levels in red blood cell membranes of depressive
patients. Biol Psychiatry 43:315–319, 1998.
28. De Vriese SR, Christophe AB, Maes M: In humans, the seasonal
variation in poly-unsaturated fatty acids is related to the seasonal
variation in violent suicide and serotonergic markers of violent
suicide. Prostaglandins Leukot Essent Fatty Acids 71:13–18, 2004.
29. Huan M, Hamazaki K, Sun Y, Itomura M, Liu H, Kang W,
Watanabe S, Terasawa K, Hamazaki T: Suicide attempt and n-3
fatty acid levels in red blood cells: a case control study in China.
Biol Psychiatry 56:490–496, 2004.
30. Sublette ME, Hibbeln JR, Galfalvy H, Oquendo MA, Mann JJ:
Omega-3 polyunsaturated essential fatty acid status as a predictor
of future suicide risk. Am J Psychiatry 163:1100–1102, 2006.
31. Rose G: Sick individuals and sick populations. Int J Epidemiol
14:32–38, 1985.
32. Berton O, Nestler EJ: New approaches to antidepressant drug
discovery: beyond monoamines. Nat Rev Neurosci 7:137–151,
2006.
33. Puri BK, Counsell SJ, Hamilton G, Richardson AJ, Horrobin DF:
Eicosapentaenoic acid in treatment-resistant depression associated
with symptom remission, structural brain changes and reduced
neuronal phospholipid turnover. Int J Clin Pract 55:560–563, 2001.
34. Frangou S, Lewis M, McCrone P: Efficacy of ethyl-eicosapentae-
noic acid in bipolar depression: randomised double-blind placebo-
controlled study. Br J Psychiatry 188:46–50, 2006.
35. Jazayeri S, Tehrani-Doost M, Keshavarz SA, Hosseini M,
Djazayery A, Amini H, Jalali M, Peet M: Comparison of
therapeutic effects of omega-3 fatty acid eicosapentaenoic acid
and fluoxetine, separately and in combination, in major depressive
disorder. Aust N Z J Psychiatry 42:192–198, 2008.
36. Nemets B, Stahl Z, Belmaker RH: Addition of omega-3 fatty acid
to maintenance medication treatment for recurrent unipolar
depressive disorder. Am J Psychiatry 159:477–479, 2002.
37. Peet M, Horrobin DF: A dose-ranging study of the effects of ethyl-
eicosapentaenoate in patients with ongoing depression despite
apparently adequate treatment with standard drugs. Arch Gen
Psychiatry 59:913–919, 2002.
38. Stoll AL, Severus WE, Freeman MP, Rueter S, Zboyan HA,
Diamond E, Cress KK, Marangell LB: Omega 3 fatty acids in
bipolar disorder: a preliminary double-blind, placebo-controlled
trial. Arch Gen Psychiatry 56:407–412, 1999.
39. Su KP, Huang SY, Chiu CC, Shen WW: Omega-3 fatty acids in
major depressive disorder. A preliminary double-blind, placebo-
controlled trial. Eur Neuropsychopharmacol 13:267–271, 2003.
40. Su KP, Huang SY, Chiu TH, Huang KC, Huang CL, Chang HC,
Pariante CM: Omega-3 fatty acids for major depressive disorder
during pregnancy: results from a randomized, double-blind,
placebo-controlled trial. J Clin Psychiatry 69:644–651, 2008.
41. Hallahan B, Hibbeln JR, Davis JM, Garland MR: Omega-3 fatty
acid supplementation in patients with recurrent self-harm. Single-
centre double-blind randomised controlled trial. Br J Psychiatry
190:118–122, 2007.
42. da Silva TM, Munhoz RP, Alvarez C, Naliwaiko K, Kiss A,
Andreatini R, Ferraz AC: Depression in Parkinson’s disease: a
double-blind, randomized, placebo-controlled pilot study of omega-
3 fatty-acid supplementation. J Affect Disord 111:351–359, 2008.
43. Nemets H, Nemets B, Apter A, Bracha Z, Belmaker RH: Omega-3
treatment of childhood depression: a controlled, double-blind pilot
study. Am J Psychiatry 163:1098–1100, 2006.
44. Behan PO, Behan WM, Horrobin D: Effect of high doses of
essential fatty acids on the postviral fatigue syndrome. Acta
Neurol Scand 82:209–216, 1990.
45. Fenton WS, Dickerson F, Boronow J, Hibbeln JR, Knable M: A
placebo-controlled trial of omega-3 fatty acid (ethyl eicosapentae-
noic acid) supplementation for residual symptoms and cognitive
impairment in schizophrenia. Am J Psychiatry 158:2071–2074,
2001.
46. Frangou S, Lewis M, Wollard J, Simmons A: Preliminary in vivo
evidence of increased N-acetyl-aspartate following eicosapenta-
noic acid treatment in patients with bipolar disorder. J Psycho-
pharmacol 21:435–439, 2007.
47. Grenyer BF, Crowe T, Meyer B, Owen AJ, Grigonis-Deane EM,
Caputi P, Howe PR: Fish oil supplementation in the treatment of
major depression: a randomised double-blind placebo-controlled
trial. Prog Neuropsychopharmacol Biol Psychiatry 31:1393–1396,
2007.
48. Keck PE Jr, Mintz J, McElroy SL, Freeman MP, Suppes T, Frye
MA, Altshuler LL, Kupka R, Nolen WA, Leverich GS, Denicoff
KD, Grunze H, Duan N, Post RM: Double-blind, randomized,
placebo-controlled trials of ethyl-eicosapentanoate in the treatment
of bipolar depression and rapid cycling bipolar disorder. Biol
Psychiatry 60:1020–1022, 2006.
49. Llorente AM, Jensen CL, Voigt RG, Fraley JK, Berretta MC,
Heird WC: Effect of maternal docosahexaenoic acid supplementa-
tion on postpartum depression and information processing. Am J
Obstet Gynecol 188:1348–1353, 2003.
EPA in Depression: A Meta-Analysis
540 VOL. 28, NO. 5
50. Marangell LB, Martinez JM, Zboyan HA, Kertz B, Kim HF,
Puryear LJ: A double-blind, placebo-controlled study of the
omega-3 fatty acid docosahexaenoic acid in the treatment of major
depression. Am J Psychiatry 160:996–998, 2003.
51. Rees AM, Austin MP, Parker GB: Omega-3 fatty acids as a
treatment for perinatal depression: randomized double-blind
placebo-controlled trial. Aust N Z J Psychiatry 42:199–205, 2008.
52. Rogers PJ, Appleton KM, Kessler D, Peters TJ, Gunnell D,
Hayward RC, Heatherley SV, Christian LM, McNaughton SA,
Ness AR: No effect of n-3 long-chain polyunsaturated fatty acid
(EPA and DHA) supplementation on depressed mood and
cognitive function: a randomised controlled trial. Br J Nutr
99:421–431, 2008.
53. Silvers KM, Woolley CC, Hamilton FC, Watts PM, Watson RA:
Randomised double-blind placebo-controlled trial of fish oil in the
treatment of depression. Prostaglandins Leukot Essent Fatty Acids
72:211–218, 2005.
54. Zanarini MC, Frankenburg FR: omega-3 Fatty acid treatment of
women with borderline personality disorder: a double-blind,
placebo-controlled pilot study. Am J Psychiatry 160:167–169, 2003.
55. Warren G, McKendrick M, Peet M: The role of essential fatty
acids in chronic fatigue syndrome. A case-controlled study of red-
cell membrane essential fatty acids (EFA) and a placebo-
controlled treatment study with high dose of EFA. Acta Neurol
Scand 99:112–116, 1999.
56. Doornbos B, van Goor SA, Dijck-Brouwer DA, Schaafsma A,
Korf J, Muskiet FA: Supplementation of a low dose of DHA or
DHA+AA does not prevent peripartum depressive symptoms in a
small population based sample. Prog Neuropsychopharmacol Biol
Psychiatry 33:49–52, 2009.
57. Freeman MP, Davis M, Sinha P, Wisner KL, Hibbeln JR,
Gelenberg AJ: Omega-3 fatty acids and supportive psychotherapy
for perinatal depression: a randomized placebo-controlled study. J
Affect Disord 110:142–148, 2008.
58. Lucas M, Asselin G, Merette C, Poulin MJ, Dodin S: Ethyl-
eicosapentaenoic acid for the treatment of psychological distress
and depressive symptoms in middle-aged women: a double-blind,
placebo-controlled, randomized clinical trial. Am J Clin Nutr
89:641–651, 2009.
59. van de Rest O, Geleijnse JM, Kok FJ, van Staveren WA,
Hoefnagels WH, Beekman AT, de Groot LC: Effect of fish-oil
supplementation on mental well-being in older subjects: a
randomized, double-blind, placebo-controlled trial. Am J Clin
Nutr 88:706–713, 2008.
60. Freeman MP, Hibbeln JR, Wisner KL, Davis JM, Mischoulon D,
Peet M, Keck PE Jr, Marangell LB, Richardson AJ, Lake J, Stoll
AL: Omega-3 fatty acids: evidence basis for treatment and future
research in psychiatry. J Clin Psychiatry 67:1954–1967, 2006.
61. Lin PY, Su KP: A meta-analytic review of double-blind, placebo-
controlled trials of antidepressant efficacy of omega-3 fatty acids.
J Clin Psychiatry 68:1056–1061, 2007.
62. Appleton KM, Hayward RC, Gunnell D, Peters TJ, Rogers PJ,
Kessler D, Ness AR: Effects of n-3 long-chain polyunsaturated
fatty acids on depressed mood: systematic review of published
trials. Am J Clin Nutr 84:1308–1316, 2006.
63. Ross BM, Seguin J, Sieswerda LE: Omega-3 fatty acids as
treatments for mental illness: which disorder and which fatty acid?
Lipids Health Dis 6:21, 2007.
64. Fux M, Benjamin J, Nemets B: A placebo-controlled cross-over
trial of adjunctive EPA in OCD. J Psychiatr Res 38:323–325,
2004.
65. Ness AR, Gallacher JE, Bennett PD, Gunnell DJ, Rogers PJ,
Kessler D, Burr ML: Advice to eat fish and mood: a randomised
controlled trial in men with angina. Nutr Neurosci 6:63–65, 2003.
66. Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJ,
Gavaghan DJ, McQuay HJ: Assessing the quality of reports of
randomized clinical trials: is blinding necessary? Control Clin
Trials 17:1–12, 1996.
67. Bax L, Yu LM, Ikeda N, Moons KG: A systematic comparison of
software dedicated to meta-analysis of causal studies. BMC Med
Res Methodol 7:40, 2007.
68. Fontani G, Corradeschi F, Felici A, Alfatti F, Bugarini R, Fiaschi
AI, Cerretani D, Montorfano G, Rizzo AM, Berra B: Blood
profiles, body fat and mood state in healthy subjects on different
diets supplemented with omega-3 polyunsaturated fatty acids. Eur
J Clin Invest 35:499–507, 2005.
69. Hirashima F, Parow AM, Stoll AL, Demopulos CM, Damico KE,
Rohan ML, Eskesen JG, Zuo CS, Cohen BM, Renshaw PF:
Omega-3 fatty acid treatment and T(2) whole brain relaxation
times in bipolar disorder. Am J Psychiatry 161:1922–1924,
2004.
70. Rush JA, First MB, Blacker D: ‘‘Handbook of Psychiatric
Measures.’’ 2nd ed. Washington, DC: American Psychiatric
Publishing, Inc., 2008.
71. Singleton N, Bumpstead R, O’Brien M, Lee A, Meltzer H.
Psychiatric Morbidity Among Adults Living in Private House-
holds. London: The Stationery Office; 2001. Accessed at: http://
www.statistics.gov.uk/downloads/theme_health/psychmorb.pdf
72. Giltay EJ, Gooren LJ, Toorians AW, Katan MB, Zock PL:
Docosahexaenoic acid concentrations are higher in women than in
men because of estrogenic effects. Am J Clin Nutr 80:1167–1174,
2004.
73. Burdge GC, Wootton SA: Conversion of alpha-linolenic acid to
eicosapentaenoic, docosapentaenoic and docosahexaenoic acids in
young women. Br J Nutr 88:411–420, 2002.
74. Blazer DG, Kessler RC, McGonagle KA, Swartz MS: The
prevalence and distribution of major depression in a national
community sample: the National Comorbidity Survey. Am J
Psychiatry 151:979–986, 1994.
75. Clayton EH, Hanstock TL, Hirneth SJ, Kable CJ, Garg ML, Hazell
PL: Reduced mania and depression in juvenile bipolar disorder
associated with long-chain omega-3 polyunsaturated fatty acid
supplementation. Eur J Clin Nutr 63:1037–1040, 2009.
76. Fujioka S, Hamazaki K, Itomura M, Huan M, Nishizawa H,
Sawazaki S, Kitajima I, Hamazaki T: The effects of eicosapentae-
noic acid–fortified food on inflammatory markers in healthy
subjects—a randomized, placebo-controlled, double-blind study. J
Nutr Sci Vitaminol (Tokyo) 52:261–265, 2006.
77. Itomura M, Hamazaki K, Sawazaki S, Kobayashi M, Terasawa K,
Watanabe S, Hamazaki T: The effect of fish oil on physical
aggression in schoolchildren—a randomized, double-blind, pla-
cebo-controlled trial. J Nutr Biochem 16:163–171, 2005.
78. Stevens L, Zhang W, Peck L, Kuczek T, Grevstad N, Mahon A,
Zentall SS, Arnold LE, Burgess JR: EFA supplementation in
children with inattention, hyperactivity, and other disruptive
behaviors. Lipids 38:1007–1021, 2003.
EPA in Depression: A Meta-Analysis
JOURNAL OF THE AMERICAN COLLEGE OF NUTRITION 541
79. Posternak MA, Zimmerman M, Keitner GI, Miller IW: A
reevaluation of the exclusion criteria used in antidepressant
efficacy trials. Am J Psychiatry 159:191–200, 2002.
80. Lee S, Walker JR, Jakul L, Sexton K: Does elimination of placebo
responders in a placebo run-in increase the treatment effect in
randomized clinical trials? A meta-analytic evaluation. Depress
Anxiety 19:10–19, 2004.
81. Horrobin DF, Jenkins K, Bennett CN, Christie WW: Eicosapen-
taenoic acid and arachidonic acid: collaboration and not antago-
nism is the key to biological understanding. Prostaglandins Leukot
Essent Fatty Acids 66:83–90, 2002.
82. Arterburn LM, Hall EB, Oken H: Distribution, interconversion,
and dose response of n-3 fatty acids in humans. Am J Clin Nutr
83:1467S–1476S, 2006.
83. Carver JD, Benford VJ, Han B, Cantor AB: The relationship
between age and the fatty acid composition of cerebral cortex and
erythrocytes in human subjects. Brain Res Bull 56:79–85, 2001.
84. Umegaki K, Hashimoto M, Yamasaki H, Fujii Y, Yoshimura M,
Sugisawa A, Shinozuka K: Docosahexaenoic acid supplementa-
tion–increased oxidative damage in bone marrow DNA in aged
rats and its relation to antioxidant vitamins. Free Radic Res
34:427–435, 2001.
85. Leonardi F, Attorri L, Di Benedetto R, Di Biase A, Sanchez M,
Nardini M, Salvati S: Effect of arachidonic, eicosapentaenoic and
docosahexaenoic acids on the oxidative status of C6 glioma cells.
Free Radic Res 39:865–874, 2005.
86. Yang DY, Pan HC, Yen YJ, Wang CC, Chuang YH, Chen SY, Lin
SY, Liao SL, Raung SL, Wu CW, Chou MC, Chiang AN, Chen
CJ: Detrimental effects of post-treatment with fatty acids on brain
injury in ischemic rats. Neurotoxicology 28:1220–1229, 2007.
87. Fam SS, Murphey LJ, Terry ES, Zackert WE, Chen Y, Gao L,
Pandalai S, Milne GL, Roberts LJ, Porter NA, Montine TJ,
Morrow JD: Formation of highly reactive A-ring and J-ring
isoprostane-like compounds (A4/J4-neuroprostanes) in vivo from
docosahexaenoic acid. J Biol Chem 277:36076–36084, 2002.
88. Goyens PL, Spilker ME, Zock PL, Katan MB, Mensink RP:
Compartmental modeling to quantify alpha-linolenic acid conver-
sion after longer term intake of multiple tracer boluses. J Lipid Res
46:1474–1483, 2005.
89. Rapoport SI, Rao JS, Igarashi M: Brain metabolism of nutrition-
ally essential polyunsaturated fatty acids depends on both the diet
and the liver. Prostaglandins Leukot Essent Fatty Acids 77:251–
261, 2007.
90. Lonergan PE, Martin DS, Horrobin DF, Lynch MA: Neuropro-
tective actions of eicosapentaenoic acid on lipopolysaccharide-
induced dysfunction in rat hippocampus. J Neurochem 91:20–29,
2004.
91. Brooks JD, Milne GL, Yin H, Sanchez SC, Porter NA, Morrow
JD: Formation of highly reactive cyclopentenone isoprostane
compounds (A3/J3-isoprostanes) in vivo from eicosapentaenoic
acid. J Biol Chem 283:12043–12055, 2008.
92. Bhattacharya A, Sun D, Rahman M, Fernandes G: Different ratios
of eicosapentaenoic and docosahexaenoic omega-3 fatty acids in
commercial fish oils differentially alter pro-inflammatory cyto-
kines in peritoneal macrophages from C57BL/6 female mice. J
Nutr Biochem 18:23–30, 2007.
93. Zhao Y, Joshi-Barve S, Barve S, Chen LH: Eicosapentaenoic acid
prevents LPS-induced TNF-alpha expression by preventing NF-
kappaB activation. J Am Coll Nutr 23:71–78, 2004.
94. Sierra S, Lara-Villoslada F, Comalada M, Olivares M, Xaus J:
Dietary eicosapentaenoic acid and docosahexaenoic acid equally
incorporate as decosahexaenoic acid but differ in inflammatory
effects. Nutrition 24:245–254, 2008.
95. Maes M, Mihaylova I, Kubera M, Bosmans E: Why fish oils may
not always be adequate treatments for depression or other
inflammatory illnesses: docosahexaenoic acid, an omega-3 poly-
unsaturated fatty acid, induces a Th-1–like immune response.
Neuro Endocrinol Lett 28:875–880, 2007.
96. Barrow CJ, Shahidi F: ‘‘Marine Nutraceuticals and Functional
Foods.’’ Boca Raton, FL: CRC Press, 2007.
97. Nadezhda N, Sushchik MI, Kalachova G: Seasonal dynamics of
fatty acid content of a common food fish from the Yenisei river,
Siberian grayling, Thymallus arcticus. Food Chem 104:1353–
1358, 2007.
Received September 19, 2008; revision accepted July 13, 2009.
EPA in Depression: A Meta-Analysis
542 VOL. 28, NO. 5
