A Dose-Ranging Study of the Effects of Ethyl-Eicosapentaenoate in Patients With Ongoing Depression Despite Apparently Adequate Treatment With Standard Drugs FREE

DEPRESSION REMAINS an illness in which existing treatments have limited efficacy. The most widely prescribed drug, fluoxetine hydrochloride, produces a 50% improvement in symptoms in only 38% of those who start treatment and in only 56% of those who complete a full course.¹ Other drugs are similar in their effects.² Tricyclic antidepressants and selective serotonin (SSRIs) and norepinephrine reuptake inhibitors are similar in their efficacy.^{2 - 3} The SSRIs are marginally better tolerated, but the differences are small. On average, for every 100 patients who start treatment, 30 patients receiving a tricyclic compound during a 6-week trial will stop treatment compared with 27 receiving an SSRI.^{3 - 4} Discontinuation rates in ordinary clinical practice are probably higher. Therefore, novel approaches to the management of depression are needed.

Lipids, most of which are phospholipids, constitute 60% of the solid mass of the brain and are absolutely required for normal brain structure and function.^{5 - 9} Each phospholipid consists of a 3-carbon glycerol backbone with a fatty acid, usually a highly unsaturated fatty acid, attached to the middle (Sn2) carbon.^{5 - 9} The Sn2 highly unsaturated fatty acids may be of 2 common types, n-6 (also known as ω-6) derived from linoleic acid or n-3 (also know as ω-3) derived from α-linolenic acid. In the brain, the main n-6 fatty acid is arachidonic acid, with much smaller amounts of dihomogammalinolenic and adrenic acids. The main n-3 fatty acid is docosahexaenoic acid (DHA), with much smaller amounts of its precursors, eicosapentaenoic acid (EPA) and docosapentaenoic acid. The metabolic pathways are shown in Figure 1.

The fatty acids at the Sn2 position have important roles in neuronal signal transduction processes.^{5 - 9} Activation of most neurotransmitter receptors leads, via a G-protein mechanism, to activation of 1 or more of a group of enzymes called phospholipase A₂, which releases the fatty acid from the Sn2 position. Depending on its specific structure, that fatty acid may exert 1 or more of many signal transduction effects, which regulate ion channels, calcium, cyclic nucleotides, protein kinases, and gene function. For normal neuronal functioning, the right balance of fatty acids must be present at the Sn2 position.^{5 - 9} The Sn2 position of phospholipids is one of the major points in the body where gene and environment interact. The relevant enzymes are genetically determined, but they must work with fatty acids provided by the environment.^{5 - 9}

Evidence has been found that the balance of n-3 and n-6 fatty acids at the Sn2 position may be disturbed in depression. Compared with healthy control subjects, plasma and red blood cells from depressed patients show absolutely low levels of n-3 fatty acids or low levels relative to the concentrations of n-6 fatty acids. Similar findings have been reported in Australia, Japan, Europe, and North America.^{10 - 15} Also, low levels of n-3 fatty acids have been found in several of the medical diseases associated with depression, including cardiovascular diseases.⁶ Epidemiological data are consistent with these findings. A strong inverse relationship exists between the consumption of n-3 fatty acids in a population and the prevalence of both major depression and postpartum depression.^{16 - 17} In individuals, we can construct hypotheses whereby depression-induced changes in diet could lead to biochemical changes in blood. However, it is difficult to see how depression in a proportion of the population could lead to an overall change in the consumption of n-3 fatty acids. If causation rather than simple correlation is involved, it is more likely that changes in fatty acid intake in the population influence depression prevalence than vice versa.

The n-3 fatty acids have been tested in the treatment of bipolar disorder and schizophrenia. A partially purified mix of ethyl-eicosapentaenoate and ethyl docosahexaenoate was effective in preventing relapse and improving mood in bipolar disorder.¹⁸ Eicosapentaenoic acid triglyceride but not docosahexaenoic acid triglyceride was effective in schizophrenia.^{19 - 20} When a range of ethyl-eicosapentaenoate dosages was tested in schizophrenia, lower dosages of 1 to 2 g/d were effective, but a higher dosage of 4 g/d was not.²¹ The loss of response was associated with depletion of the n-6 fatty acid, arachidonic acid, indicating that the balance between n-3 and n-6 fatty acids matters. Abnormally high or low levels of either fatty acid type may be associated with malfunction. Because of the biochemical and epidemiological data, we hypothesized that EPA might have beneficial effects in depression. Possible effective dosages were unknown, and we therefore conducted a dose-ranging study of the pure ethyl ester derivative of eicosapentaenoic acid in patients who had failed to respond satisfactorily to standard antidepressant therapy.

Patients were recruited by family physicians who had a special interest in depression and experience in conducting clinical trials. Approval for the study was granted by the local regional ethical committees. After full verbal and written explanation of the study, each patient gave written informed consent. Patients were of either sex, aged 18 to 70 years, and depressed as indicated by a score of 15 or more on the 17-item Hamilton Depression Rating Scale (HDRS)²² despite ongoing treatment with a standard antidepressant at an adequate dose. This score was chosen as a level of depression that in the view of the trial investigators caused important impairment of function and therefore justified further attempts at treatment.

The test drugs were liquid paraffin placebo or pure ethyl-eicosapentaenoate in 500-mg soft gelatin capsules. The placebo and ethyl-eicosapentaenoate capsules appeared identical and were packed and coded by PCI Clinical Services, Bolton, England, an independent clinical trials packing organization. On entry, patients were randomly allocated by PCI Clinical Services computer to receive, each morning and evening, 4 capsules of placebo, 3 capsules of placebo and 1 of ethyl-eicosapentaenoate, 2 capsules of placebo and 2 of ethyl-eicosapentaenoate, or 4 capsules of ethyl-eicosapentaenoate. The capsules were encased in blister packs and labeled as morning and evening doses. PCI Clinical Services had no involvement with the rest of the trial. The patients therefore received placebo or ethyl-eicosapentaenoate at dosages of 1, 2, or 4 g/d.

Patients were randomized on a double-blind basis to 1 of the 4 treatment groups. They underwent assessment by means of 3 rating scales at baseline and at 4, 8, and 12 weeks. We used the 17-item HDRS,²² the Montgomery-Asberg Depression Rating Scale (MADRS),²³ and the patient-completed Beck Depression Inventory (BDI).²⁴ At each visit, patients were also asked about any adverse events.

When the last patient had completed the trial, the data were verified and the databases were locked before breaking the code. The primary variable was the HDRS, with the MADRS and the BDI as secondary variables. Two populations were used. The intention-to-treat (ITT) population included all patients who were randomized, while the per-protocol (PP) population included all patients who completed 12 weeks of treatment. For each rating scale at each time point, we used analysis of covariance to assess whether the overall differences among the 4 treatment groups were significant. Covariates included in the model were baseline HDRS score, center, and antidepressant class (tricyclic, SSRI, or other drugs that were norepinephrine or mixed reuptake inhibitors). Analysis of variance was then used to compare each ethyl-eicosapentaenoate treatment group with the placebo group.

Seventy-four patients underwent screening; of these, 70 met the entry criteria, agreed to participate in the trial, and were randomized. Eighteen patients were assigned each to the placebo and 2-g/d groups, and 17 were assigned each to the 1- and 4-g/d groups. Sixty patients completed 12 weeks of treatment. Four (22%) of 18 dropped out of the placebo group (1 was lost to follow-up, 1 withdrew consent during the study, 1 violated the protocol, and 1 had an adverse event not thought to be related to treatment). Two patients dropped out of each of the ethyl-eicosapentaenoate groups, making a total withdrawal from the ethyl-eicosapentaenoate groups of 6 (12%) of 52 (3 withdrew consent, 1 withdrew because of lack of efficacy, 1 violated the protocol, and 1 had a gastrointestinal adverse event). Compliance was estimated by results of capsule counts and was greater than 90% in all treatment groups.

The ages, sexes, and background treatments of the 4 groups are listed in Table 1. The groups were well matched. Women predominated.

All reported adverse events are listed in Table 2. Eight of 18 patients in the placebo group and 20 of 52 patients in the ethyl-eicosapentaenoate groups reported no adverse events. The events were evenly distributed across the groups. The only events attributed to treatment by the physicians were the gastrointestinal events that affected 4 of 18 in the placebo group and 20 of 52 in the ethyl-eicosapentaenoate groups. These events were attributable to the intake of 4 g/d of an oily substance, rather than being specifically caused by the study treatment. All but 1 of the gastrointestinal events were mild and self-limited and did not require cessation of treatment. Diarrhea developed in 1 patient in the 1-g/d group. None of the usual adverse events associated with antidepressant therapy and no effect on any blood parameter or liver function test result were seen.

The results of the analyses of covariance for change from baseline to the end of the study across all treatment groups are shown in Table 3. Center and background medication had no significant effects on any rating scale in the ITT or PP populations. Baseline score had no effect on the HDRS and MADRS outcomes in the ITT or PP population: it had an effect on outcome on the BDI score in the ITT but not the PP population. Treatment had a significant overall effect on all 3 scores for both populations and was marginal only for the HDRS in the ITT population (P = .056).

The analyses of variance comparing change from baseline to the end of the study in each active treatment group compared with the placebo group are shown in Table 4. For the 1-g/d group, all of the analyses on all 3 scales showed this dosage to be significantly better than placebo. For the 2-g/d group, none of the comparisons approached significance. For the 4-g/d group, comparisons approached significance in the PP population but not in the ITT population.

For the 1-g/d and placebo groups, the changes at 4, 8, and 12 weeks in the PP population are shown in Table 5. For the HDRS and the MADRS but not for the BDI, the difference was already significant at 4 weeks. On all 3 rating scales, the difference was significant, or approached significance, at 8 and 12 weeks.

To probe what depressive symptoms might respond to the 1-g/d dosage of ethyl-eicosapentaenoate, the 3 main components of the HDRS (items 1-3, depression; items 4-6, sleep; and items 9-11, anxiety) and the 10 items of the MADRS were compared for the placebo and 1-g/d groups in the PP population. All 3 components of the HDRS and 9 of 10 items of the MADRS showed the 1-g/d dosage to be significantly better than placebo. On all 20 items of the BDI, the 1-g/d dosage was better than placebo, with particularly large and significant differences for sadness, pessimism, inability to work, sleep disturbances, and libido. The effect of ethyl-eicosapentaenoate applies to all major components of the depressive syndrome and is seen equally in the patient and physician assessments.

At all dosages given, ethyl-eicosapentaenoate was well tolerated, as indicated by the reported adverse events and the low withdrawal rate. Only 12% of patients receiving ethyl-eicosapentaenoate failed to complete 12 weeks of treatment. This compares with an average withdrawal rate in 6-week trials of about 30% of subjects receiving tricyclic antidepressants and of about 27% of subjects receiving SSRIs.⁴ The only common adverse event was mild and self-limited gastrointestinal disturbance. All of the patients in this trial were given 4 g/d of oily material. Such disturbance is likely to be less if in the future only the optimal dosage of 1 g/d is administered.

Results of all 3 rating scales demonstrate clear efficacy at the 1-g/d dosage of ethyl-eicosapentaenoate. The consistently significant effects are surprising given that the trial was a small exploratory study. The trial was too small to draw firm conclusions about the other treatment dosages. Although there appeared to be a trend toward significant efficacy at the 4-g/d dosage, larger studies would be required to elucidate possible beneficial effects of the higher dosages. Because most of the participants in the trial were women, it is not possible to draw conclusions about men, although inspection of individual scores indicated that men also responded. The reality of the antidepressant effect of ethyl-eicosapentaenoate is supported by results of 2 additional completed studies. One reported surprising improvement in a single suicidal depressed male patient who had proved exceptionally refractory to treatment.²⁵ The other reported a highly significant beneficial effect in a placebo-controlled study of patients with depression who had initially responded to standard therapy but then relapsed while continuing such therapy.²⁶

Analysis of the individual components of the HDRS, the MADRS, and the BDI indicates that the effect is a broad-spectrum one involving all of the components of the depressive syndrome. Depression, anxiety, sleep, and lassitude all responded equally well.

Larger studies of various dosages of ethyl-eicosapentaenoate as add-on therapy and as sole therapy in men and women are now required. It is likely that different individual patients will require different dosages. The substantial effect of the 1-g/d dosage in the present study is consistent with the findings of the epidemiological studies.^{16 - 17} These studies show a sharp fall in the prevalence of both major depression and postpartum depression at fish/seafood intakes that translate to long-chain n-3 fatty acid intakes of 0.5 to 1.0 g/d. If these epidemiological studies indicate a causal relationship, then they suggest that in many depressed patients, a 1-g/d dosage of ethyl-eicosapentaenoate should be effective. Other patients may require higher dosages.

The issue of the specificity of the action of ethyl-eicosapentaenoate requires further exploration. Docosahexaenoic acid was not effective in schizophrenia, whereas eicosapentaenoic acid was.²⁰ A high dosage of 4 g/d of docosahexaenoic acid was slightly less effective than placebo in depression, but it is not known whether this finding was due to the dosage or to a lack of efficacy of DHA specifically.²⁷ At least 2 known mechanisms could lead to EPA being more effective than DHA. In depression, production of prostaglandins from arachidonic acid by the cyclooxygenase system has consistently been reported to be elevated.^{28 - 32} Eicosapentaenoic acid but not DHA is an effective substrate for cyclooxygenase and can compete with arachidonic acid at this point. Also, in some phospholipase A₂ assays, EPA but not DHA has been reported to be an effective inhibitor.³³ These different effects of EPA and DHA, which may include synergism and antagonism, mean that the biological effects of fish oils, which contain both in highly variable proportions, will be uncertain and difficult to predict.

The mechanism of action of ethyl-eicosapentaenoate requires much further exploration. We think it unlikely that it can be explained by improved pharmacokinetics or pharmacodynamics of existing drugs. Although individual patients may benefit from increasing antidepressant dosages, no substantial studies of existing drugs have shown such large improvements in outcome as a consequence of increasing the dosage as the improvement that were seen in the 1-g/d group. No differences were seen in the effect of ethyl-eicosapentaenoate between the different classes of antidepressants. Limited numbers of patients not receiving any antidepressant who have been treated by us in clinical practice have shown improvements similar to those in this trial. Patients with schizophrenia not receiving any drug therapy have responded to EPA.^{19 ,21} Therefore, although modulation of background drug pharmacokinetics cannot entirely be ruled out, we think it more likely that the ethyl-eicosapentaenoate action is on cell membranes and signal transduction systems.

Ethyl-eicosapentaenoate has one side effect that is likely to be beneficial in depression. It lowers triglyceride levels, inhibits platelet aggregation, and inhibits cardiac arrhythmias.^{6 ,34 - 35} In 2 large trials, EPA-containing products (providing EPA at a dosage of less than 1 g/d) have been shown to reduce mortality related to heart disease.^{36 - 37} In view of the steadily increasing evidence of associations between various types of cardiovascular disease and depression, and that both disorders are associated with low blood EPA levels, ethyl-eicosapentaenoate may be of particular benefit in depressed patients who are also at risk for cardiovascular disease.⁶

Ethyl-eicosapentaenoate offers an approach to depression that is radically different from that of existing drugs. Its position in the treatment spectrum will be established only by further trials.

Submitted for publication April 27, 2001; final revision received October 19, 2001; accepted November 1, 2001.

Dr Peet received research funding from Laxdale Research, Ltd, Stirling, Scotland.

This study was presented as a poster at the Society of Biological Psychiatry Meeting in New Orleans, La, May 2001; at the Fourth International Conference on Bipolar Disorder, Pittsburgh, Pa, June 2001; and at the British Association of Psychopharmacology Meeting, Harrogate, England, July 2001.

We thank Sharon Maxwell, Westlakes Scientific Consulting, for data management; Alastair Sword, Pharmapart UK Ltd, for statistical analysis; and Anne MacKenzie of Kendle UK, for project management. Packaging and provision of drug and placebo capsules were performed by PCI Clinical Services. Adverse event monitoring was performed by ClinTrials Research Ltd, Glasgow, Scotland. The patients were seen by the following 2 groups of family practitioners in the United Kingdom with a specialist interest in depression and in clinical trials of antidepressant drugs: MGB Group: George Beaumont, MB, ChB, and Max Gringras, MB, ChB, Poynton, England; and W. R. C. Aitchison, MB, ChB, and G. I. McLaren, Bridge of Weir, Scotland. QED Group: S. Connolly, MB, ChB, Hamilton, Scotland; J. Thompson, MB, ChB, P. D. Flanigan, MB, ChB, and D. Kilgour, MB, ChB, Coatridge, Scotland; and W. Scullion, MB, ChB, Airdrie, Scotland.

Corresponding author and reprints: David Horrobin, DPhil, BM, BCh, Laxdale Research, Ltd, Kings Park House, Laurelhill Business Park, Stirling, Scotland FK7 9JQ (e-mail: agreen@laxdale.co.uk).