Fifty-seven patients (65% female; mean age 42 ±14) were randomized to EPA (n=28) or Placebo (n=29). Treatments were assigned by the MGH Research Pharmacy upon successful completion of the screen visit. Six subjects were lost to follow up after the screen visit; seven subjects were disqualified at baseline because of significant clinical improvement or worsening; and three chose not to enter the study after completing the baseline visit (only one of them specified a reason, a lengthy commute). This resulted in 41 subjects (63% female; mean age 43 ± 13 yrs) entering double blind treatment (17 on EPA, and 24 on placebo). Thirty-five subjects (63% female; mean age 45 ± 13 yrs, 16 on EPA, 19 on Placebo) remained for at least one post-baseline visit and were evaluable in the intent to treat (ITT) analysis. Three evaluable subjects (1 on EPA, 2 on placebo) withdrew because of non-response; one EPA subject withdrew because of commuting difficulties; and one placebo subject chose to discontinue early because he was feeling better. The remaining early discontinuers were lost to follow-up and no reasons for their discontinuation were available. There was no significant difference in dropout rates between the two treatment groups (p>0.05). Twenty-four subjects (11 on EPA, 13 on placebo) completed the full 8 weeks of treatment.
Degree of Improvement in HAM-D-17 Scores, and Response and Remission Rates
Among the 24 study completers (63% female; 11 on EPA, 13 on Placebo), mean HAM-D-17 scores decreased from 21.3 ± 3.0 to 11.1 ± 8.1 for the EPA group (p=0.004) and from 20.5 ± 3.8 to 16.3 ± 6.9 for the placebo group (p=0.06), with a trend to significance in the difference between the two groups (U=42.00, z=-1.71, p=0.087) (); the effect size for EPA was 0.73. Completer response rates, based on 50% or greater decrease in HAM-D-17 score, were 45% (5/11) for the EPA group, and 23% (3/13) for the placebo group () (Fisher's p=0.39, OR=2.78, 95% CI, 0.48-16.03). Remission rates, based on a final HAM-D-17 score of 7 or less, were 36% (4/11) for the EPA group and 15% (2/13) for the placebo group () (Fisher's p=0.357, OR=3.14, 95% CI, 0.45-21.74).
Change in HAM-D-17 score over time for each treatment group (completers)
Response and Remission Rates for Each Treatment Group (completers)
In the ITT sample (63% female; 16 on EPA, 19 on Placebo), mean HAM-D-17 scores decreased from 21.6 ± 2.7 to 13.9 ± 8.9 for the EPA group (p=0.005) and from 20.5 ± 3.6 to 17.5 ± 7.5 for the placebo group (p=0.12), with a non-significant difference between the two groups (U=105.50, z=-1.54, p=0.123); the effect size for EPA was 0.55. ITT response rates were 38% (6/16) for the EPA group, and 21% (4/19) for the Placebo group (Fisher's p=0.45, OR=2.25, 95% CI, 0.50-10.10). Remission rates were 25% (4/16) for the EPA group and 16% (3/19) for the Placebo group (Fisher's p=0.677, OR=1.78, 95% CI, 0.33-9.43).
Only four EPA responders and three placebo responders entered the 8-week maintenance phase. This small sample did not allow for significant analyses or comparisons, but all these subjects maintained their response during this phase, with no relapses or significant depressive worsening.
Effect of EPA Administration on Plasma Fatty Acid Profiles
A total of 37 analyzable samples were available for baseline visit, 14 for the week 8 visit, and 9 for the week 16 visit (including the 7 double-blind responders, and two non-responders on escitalopram rescue). Lipid samples could not be obtained for certain patients due to unavailability for blood draws, and some samples were not analyzable due to damage during storage.
In EPA subjects, mean plasma EPA level increased significantly over the 8 weeks of treatment, from 7.62 ± 8.21 ug/ml to 22.13 ± 7.04 ug/ml (Z=-2.028, p=0.043), and the n-6/n-3 ratio decreased significantly, from 13.78 ± 3.92 to 9.05 ± 1.57 (Z= -2.197, p=0.028). Plasma DHA showed no significant change, from 40.93 ± 15.13 ug/ml to 44.83 ± 10.62 (p>0.05), nor did total n-3 and n-6 (p>0.05). No significant changes in any plasma lipid parameters were observed for placebo subjects (p>0.05). The change in plasma EPA level in EPA subjects was significantly higher than the one from 4.25 ± 3.17 ug/ml to 4.39 ± 5.29 ug/ml observed in the placebo group (U=5.00, Z=-2.03, p=0.042), but differences in changes in the other lipid parameters between the two treatment groups did not reach significance (p>0.05). Among subjects who entered the maintenance/rescue treatment phase, only one EPA responder and two placebo responders had analyzable plasma lipid samples, so no significant comparisons in lipid parameters could be made for the EPA group; for the placebo subjects, changes in lipid parameters were not significant (p>0.05).
Relationship Between Plasma Lipid Levels, Severity of Depression, and Response to Treatment
Linear regression was carried out to determine the association between baseline plasma EPA, DHA, total n-3, total n-6, and n-6/n-3 ratio (independent variables) and baseline HAM-D-17 score (dependent variable). For all subjects with available baseline lipid data and HAM-D-17 scores (n=37), we found no significant associations between any of the baseline lipid parameters and severity of depression (p>0.05).
In study completers, we similarly examined the relationship between baseline plasma lipid parameters (independent variables) and the change in HAM-D-17 score (dependent variable) with treatment. For EPA group completers who had baseline lipid data available (n=8), we found a significant Pearson correlation between the baseline n-6/n-3 ratio and the change in the HAM-D-17 score with treatment (r= -0.686, p=0.030). No significant associations were observed between any of the other lipid parameters and the change in the HAM-D-17 score (p>0.05). Among placebo group completers who had baseline lipid data available (n=8), we found a significant Pearson correlation between baseline DHA and the change in the HAM-D-17 score (r=0.677, p=0.033), and between baseline total n-3 and the change in the HAM-D-17 score (r=0.694, p=0.028).
We examined the relationship between changes in plasma lipid parameters (independent variables) and change in HAM-D-17 score (dependent variable) with treatment. For EPA subjects, we found a significant Pearson correlation between change in n-6/n-3 ratio and change in HAM-D-17 score (r=0.784, p=0.032). No significant associations were observed between the changes in the other lipid parameters and the change in the HAM-D-17 score (p>0.05). In placebo subjects, no significant associations were observed between any changes in lipid parameters and changes in HAM-D-17 score (p>0.05).
Logistic regression was carried out to examine the relationship between mean baseline plasma lipid parameters (independent variables) and treatment response (dependent variable). For completers in the EPA and placebo groups, we found no significant association between treatment response and any of the baseline lipid parameters (p>0.05). In the ITT sample, we found a significant association between baseline n-6/n-3 ratio and treatment response (p=0.032) in EPA subjects, but no significant associations among placebo subjects (p>0.05).
We similarly examined the relationship between treatment-related changes in mean plasma lipid parameters (independent variables) and response (dependent variable). We found no significant associations between treatment response and changes in any lipid parameters in either treatment group, both for completers and the ITT sample (p>0.05).
Effect of Dietary Omega-3 Intake on Depression Severity, Plasma Lipid Parameters, and Response to Treatment
25 subjects (11 from the EPA group, and 14 from the placebo group) consistently filled out their food diaries. Thirteen subjects (6 from EPA, 7 from PBO) met criteria for low dietary n-3 (0-1 serving/week), nine (4 from EPA, 5 from PBO) for medium n-3 (2-3 servings/week), and three (1 from EPA, 2 from PBO) for high n-3 (4 or more servings/week). To simplify the analysis, in some cases we pooled the medium and high dietary n-3 groups (n =12) (“adequate”) to compare against the low dietary n-3 group (n =13) (“inadequate”).
Subjects consuming low dietary n-3 (n=13) had a mean baseline HAM-D-17 score of 21.8 ± 3.0; subjects consuming intermediate n-3 levels (n=9) had a mean baseline HAM-D-17 score of 20.8 ± 2.8; and subjects consuming high n-3 levels (n=3) had a mean baseline HAM-D-17 score of 19.33 ± 6.7. One-way between-groups ANOVA showed no significant difference between baseline HAM-D-17 scores across the three dietary groups (p>0.05). After pooling medium and high dietary n-3, the mean baseline HAM-D-17 score for the group receiving adequate dietary n-3 (n=12) was 20.4 ± 3.8. The Mann-Whitney U test showed no significant difference in severity of depression at baseline between low and adequate dietary n-3 groups (z = -0.91; U = 61.5; p = 0.37).
Plasma lipid data were available for 18 of the subjects who filled out food diaries. In subjects with adequate dietary n-3 consumption, baseline plasma EPA (9.56 ± 9.11 ug/ml), and DHA (44.58 ± 19.46 ug/ml) were higher than in subjects with low consumption (EPA = 3.10 ± 2.28 ug/ml, and DHA = 35.19 ± 8.31 ug/ml). The n-6/n-3 ratio was slightly lower in subjects with adequate dietary n-3 consumption (12.57 ± 2.85) than in those with inadequate n-3 consumption (15.04 ± 2.58). The Mann-Whitney U test showed a significant difference between low and adequate n-3 consumers for plasma EPA levels only (z = -2.845; U = 8.00; p = 0.004), and a trend to significance for n-6/n-3 (z=-1.77, U=20.00, p= 0.076).
As an exploratory investigation, depression severity and response to treatment were examined for subjects at each level of dietary n-3 consumption. Among EPA subjects in the ITT group, the ones with low dietary n-3 consumption had the most robust response rate of 33% (n=2/6), with progressively lower response rates in the groups with medium (25%; n=1/4) and high n-3 consumption (0%; n=0/1), and the pooled (adequate n-3) group having a response rate of 20% (n=1/5). The same trend was observed for completers, with the low dietary n-3 group responding at a rate of 50% (n=2/4), the medium n-3 group at 33% (n=1/3), and the high n-3 group at 0% (n=0/1). The pooled group of subjects with adequate dietary n-3 had a response rate of 25% (n=1/4). Among placebo subjects, only the medium n-3 group had a moderately robust response rate, with 40% (n=2/5) in the ITT group and 33% (n=1/3) in completers, compared to the other two dietary groups, among which completers and ITT subjects all had response rates of 20% or less. None of the differences in response and remission rates between dietary n-3 groups were significant by Fisher's Exact test (p>0.05). No comparisons of HAM-D-17 changes, using one-way between-groups ANOVA (for the 3 dietary groups) and the Mann-Whitney U Test (for adequate vs low n-3 groups) reached significance (p>0.05 for all comparisons).
Impact of Smoking Status on Outcomes
Because smoking is known to lower omega-3 levels, we examined tobacco use in our subjects to rule out any potential confounding effects. Only four of the 35 evaluable subjects reported smoking 10 or more cigarettes per day on a regular basis, which was unlikely to have a significant impact on our findings overall.
Tolerability and Side Effects
Seven subjects (2 on EPA, and 5 on placebo) reported mild side effects, all gastrointestinal. EPA subjects reported gas and an unspecified GI upset. Placebo subjects reported GI upset and increased bowel movements or diarrhea. Three of these subjects discontinued in the acute phase, and two in the maintenance/rescue phase. One cited non-response as the reason for his termination, and the rest were lost to follow-up. No subjects attributed their discontinuation to side effects.