Effect of Fish Oil on Circulating Adiponectin: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Abstract

Context:

Seafood long-chain polyunsaturated omega-3 fatty acids (n-3 PUFAs) improve insulin sensitivity in animal experiments, but findings remain inconsistent in humans. Adiponectin is a robust marker for insulin sensitivity and adipocyte function. Whether n-3 PUFAs affect adiponectin in humans is unknown.

Objective:

Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, the objective of the study was to perform a systematic review and meta-analysis of randomized, placebo-controlled clinical trials (RCTs) to determine the effect of n-3 PUFA consumption on circulating adiponectin in humans.

Data Sources:

MEDLINE, EMBASE, CABI (CAB abstracts), Cochrane Central Registry of Controlled Trials, ClinicalTrials.gov, SIGLE, and Faculty of 1000 were searched through to June 2012, supplemented with author contact and reference list searches.

Study Selection:

RCTs of either fish oil supplementation or isocaloric fish meal feeding that evaluated adiponectin as an outcome were selected for the study.

Data Extraction:

Two investigators independently extracted the data. Effect estimates were pooled using inverse-variance weighted, random-effects meta-analysis. Heterogeneity was assessed by the I² and Q statistic. Prespecified sources of heterogeneity were investigated by meta-regression. Publication bias was assessed using funnel plots and Egger's test.

Data Synthesis:

Of 110 studies, 14 RCTs met inclusion criteria. Fourteen trial arms evaluated fish oil (fish oil, n = 682; placebo, n = 641). Fish oil increased adiponectin by 0.37 μg/mL [95% confidence interval (CI) 0.07; 0.67, P = .02]. Although effects in 11 of 14 trials were 0 or greater, statistical heterogeneity was evident (I² = 72.9%), unexplained by n-3 PUFA dose or duration, study quality score, study location, or baseline body mass index (meta-regression P > .05 each). The funnel plot was asymmetric in favor of smaller trials with greater effects (Egger's P = .11); the fill-and-trim method suggested a theoretical pooled effect of 0.18 μg/mL (95% CI −0.15; +0.52, P = .28). Only 2 trial arms evaluated fish feeding (n = 136 intervention and 68 control subjects), for which the pooled effect on adiponectin was not statistically significant (−0.01μg/mL, 95% CI −0.65; 0.64, P = 0.99), although CIs were broad due to the small number of subjects.

Conclusions:

In placebo-controlled RCTs, fish oil moderately increases circulating adiponectin, although with unexplained heterogeneity as well as potential publication bias. These findings provide no evidence for harm and support possible benefits of n-3 PUFA consumption on insulin sensitivity and adipocyte function.

Increased consumption of seafood long-chain polyunsaturated omega-3 fatty acids (n-3 PUFAs) improves insulin sensitivity in animal studies (1–3). However, evidence for the effects of n-3 PUFA on insulin sensitivity in human subjects remains limited and inconclusive (4). The potential molecular mechanisms that may link n-3 PUFA intake and altered insulin sensitivity in humans also remain unclear.

Adiponectin is a protein produced by adipocyte cells that possess potent insulin-sensitizing and antiinflammatory properties. Adiponectin protects against development of type 2 diabetes and atherosclerosis in animal models (5), and higher circulating levels of adiponectin are associated with a lower risk of type 2 diabetes and coronary heart disease in prospective cohort studies (6, 7). In vitro and animal studies demonstrate that n-3 PUFA consumption increases the levels of these fatty acids in adipocytes and protects adipose tissue against inflammation and insulin resistance (8). Animal experimental studies have also consistently found that n-3 PUFA intake increases circulating levels of adiponectin (1–3).

Whether consumption of n-3 PUFAs affects circulating adiponectin in humans is not established. Such an effect would elucidate the biological pathways of effects of n-3 PUFAs on adipocyte function and insulin sensitivity. To address this gap in knowledge, we performed a systematic review and meta-analysis of randomized placebo-controlled clinical trials to determine the effect of n-3 PUFA intake on circulating adiponectin in humans.

Materials and Methods

We conducted this systematic review and meta-analysis according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines during all stages of design, implementation, analysis, and reporting (9). The prespecified study protocol is available from the authors upon request.

Search strategy and eligibility criteria

We searched multiple electronic databases from the earliest available online indexing year through June 2012, without language restrictions, including MEDLINE, EMBASE, CABI (CAB abstracts), Cochrane Central Registry of Controlled Trials, ClinicalTrials.gov, SIGLE, and Faculty of 1000. Key words included, among others, fatty acids, omega-3, eicosapentaenoic acid (EPA), docosahexaenoic acid (DHA), fish oils, adiponectin, and clinical trial; the full search terms are available on request. Additional studies were identified through searching of electronically linked related articles on MEDLINE and hand searching of citation lists of full-text articles selected for review.

Studies were included if they were randomized placebo-controlled trials of oral fish oil or isocaloric fish meal feeding that evaluated adiponectin as an outcome. Studies were excluded if in pregnant women or if they had a concomitant intervention for which effects could not be separated (ie, factorial design studies could be included); were of less than 2 weeks' duration; or were observational, commentaries, reviews, case reports, nonrandomized studies, or duplicate publication from the same study.

Study selection and data extraction

One investigator (J.W.) screened the title and abstracts of all identified articles for eligibility. The full-text of remaining articles was subsequently assessed independently in duplicates by 2 investigators (J.W. and L.C.) to determine inclusion or exclusion. Discrepancies were resolved by consensus.

For studies that met inclusion criteria, data were extracted independently in duplicate by 2 investigators (J.W. and L.C.), using a standardized electronic form. Extracted data included author, publication year, study location (North America, Europe, and others), population characteristics (age, gender, ethnicity, and comorbidities), and study design characteristics including randomization method, blinding of investigators and participants, parallel or crossover design, washout period if crossover, dose of n-3 PUFA, content of placebo, duration of intervention, percentage of dropout, method of adiponectin measurement, and type of adiponectin measured. We extracted the mean concentration and error (SD, SE, or confidence intervals) of circulating adiponectin at study baseline, posttreatment, and/or change between baseline and posttreatment. Study quality was assessed using the instrument of Jadad et al (10), which incorporated 0 or 1 point for each of the following 5 criteria: 1) presence of randomization, 2) appropriate method of randomization, 3) presence of double blinding, 4) appropriate method of double blinding, and 5) description of withdrawals and dropouts. Trials were considered high quality if they obtained a score of 3 or higher. We contacted study authors to obtain relevant missing information and received responses for 7 of 15 studies contacted.

Statistical analysis

The primary end point was change in circulating adiponectin due to fish oil supplementation (or fish feeding). For crossover studies, we compared end-treatment circulating adiponectin in the control vs fish oil period. For parallel-design trials, we compared end-treatment changes in adiponectin in the control vs fish oil groups. Within-person changes were used when possible; otherwise, we used group means. The SD was extracted, or if not reported, calculated from SEs. Within-person changes and SD were available and used for 10 of the studies. For the remaining 4 studies, group means and SDs were used to calculate the change in adiponectin (11), and the SD of the change was imputed (12, 13), using a within-person correlation of 0.93 (based on observed serial measured of adiponectin taken 1 year apart (personal communication, Rimm E., April 2012). We performed sensitivity analyses assuming a lower within-person correlation of 0.50. We calculated the variance of the difference in mean change in adiponectin between fish oil and the placebo group, using the following equation: (SD of change in adiponectin²_{fish oil group}/n_{fish oil group}) + (SD of change in adiponectin²_{control group}/n_{control group}). Details of the methods used for imputation of SD and calculation of treatment variance are given in Online Supplemental Materials, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org.

Pooled effects were estimated using the inverse-variance weighted, random effects meta-analysis of DerSimonian and Laird (14). The I² statistic and Q statistic were used to assess heterogeneity (15). Prespecified potential sources of heterogeneity were evaluated in stratified meta-analyses, with meta-regression examining whether effect estimates were significantly different. These included study location (Europe/North America vs others), duration of treatment (less than median; or equal to or greater than median of all studies), blinding (double vs single), major fatty acid of the placebo (linoleic acid vs oleic acid vs others), dose of n-3 PUFA (EPA + DHA, less than median; or equal to or greater than median of all studies), study quality score (low, 0–2; or high, 3–5), age (less than median; or equal to or greater than median of all studies), body mass index (BMI; less than median; or equal to or greater than median of all studies), and population (adults vs children/adolescents). We also assessed the duration of treatment (weeks), the dose of n-3 PUFA (grams per day), the quality score, age (years), BMI (kilograms per square meter), and gender (percentage male) as continuous covariates in the meta-regression analysis. Publication bias was assessed by visual inspection of funnel plots and statistically using Egger's and Begg's tests. We used the nonparametric trim-and-fill method to assess the potential influence of publication bias in sensitivity analyses and provide a theoretical pooled estimate accounting for estimated missing studies (16). All analyses were performed using STATA 10.0 (StataCorp, College Station, Texas), with a 2-sided alpha = .05.

Results

Search Results

Of 110 identified reports, 83 were excluded based on initial review (Figure 1). The full-text articles of the remaining 27 reports were retrieved and reviewed. Of these, 12 were excluded because they lacked placebo control (n = 9) (17–25), had a concomitant intervention (n = 1) (26), were duplicate publications from the same study (n = 1) (27), or lacked randomization (n = 1) (28). One additional study, published only as an abstract, nominally met inclusion criteria but did not contain sufficient data for quantitative analysis (29), which could also not be obtained after repeated author requests, and was therefore excluded. After the final exclusions, 14 studies met the eligibility criteria and were included in this meta-analysis (29–43).

Figure 1.

Flow chart of the literature search, screening, and selection process for randomized, placebo-controlled trials that investigated the effects of fish oil or fish consumption on circulating adiponectin in humans. *, One appropriate trial [published as an abstract (29)] was not included in the quantitative synthesis due to insufficient information about changes in adiponectin. We included this study as part of our qualitative review of published trials.

Overview of trial characteristics

Each of the 14 trials provided an effect estimate for the effect of fish oil supplementation on circulating adiponectin (Table 1). In addition, 1 trial also evaluated how different types of fish meals (lean, fatty) influenced adiponectin (39). Thirteen were parallel design (3 of which included a 2 × 2 factorial design), and 1 used a crossover design. Most (n = 11) trials were double blind; in 3 trials, only the investigators were blinded. In total, the trial treated 682 subjects with fish oil (range 6–247 subjects per study),and 641 subjects with placebo (range 5–239 subjects per study). Eleven trials were in adults (mean age ranging from 25.4 to 70 years; median 44.7 years), and 3 were in children. Most trials were performed in Europe (n = 10) or North America (n = 1); 3 were from Asia, Central America, and Australia. Subjects were generally at least overweight (mean BMI > 25 kg/m² in 11 trials), and 1 trial evaluated subjects with type 2 diabetes.

Table 1.

Characteristics of the 14 trials (16 Treatment Groups) Included in the Meta-Analysis^a

Author (Year)	Design	Blinding	EPA + DHA, nb,c	Control, nb	Mean Age, yd	Male, %	Mean BMI, kg/m2	Population, Country	EPA + DHA, g/d	Duration, wk	Control Oil	Method for Adiponectin Measuremente	Dropout, %	Quality Scoref
Andersen et al (2011) (30)	Parallel	Double	55	54	0.76	53	17.7	Healthy infants, Denmark	1.25	39	Sunflower	RIA	14	5
Dangardt et al (2012) (32)	Crossover	Double	25	25	15.6	44	34.1	Obese adolescents, Sweden	1.22	12	Medium-chain triglycerides	ELISA	19	4
Damsgaard et al (2008) (31)	Factorial	Double	31	33	25.4	100	22.8	Healthy men, Denmark	2.9	8	Olive	ELISA	3	4
Gammelmark et al (2012) (33)	Parallel	Double	25	24	56.7	48	30.2	Overweight/obese adults, Denmark	1.12	6	Olive	ELISA	2	3
Kabir et al (2007) (34)	Parallel	Double	12	14	55	0	30	Women with type 2 diabetes, France	1.8	8	Paraffin	RIA	10	3
Koh et al (2012) (35)	Parallel	Single	50	49	54.7	59	25.3	Adults with hypertriglyceridemia, Korea	2	8	NR	ELISA	1	2
Krebs et al (2006) (36)	Parallel	Double	35	32	44.7	0	35	Obese women, United Kingdom	4.2	24	Linoleic rich	RIA	13	3
Lopez-Alarcon et al (2011) (37)	Parallel	Single	49	27	13.4	53	30.2	Obese children /adolescents, Mexico	0.9	4.3	Sugar pill	RIA	0	2
Micallef and Garg (2009) (38)	Factorial	Double	30	30	55.4	45	26.9	Adults with hyperlipidemia, Australia	1.44	3	Sunflower	ELISA	0	4
Ramel et al (2008) (39)	Parallel	Single	70	68	31.4	43	30.1	Overweight/obese adults, Iceland	2.1g	8	Sunflower	RIA	14	2
			74						1.1
			66						0.26g
Rizza (2009) (40)	Parallel	Double	26	24	29.9	50	26.2	Healthy adults, Italy	0.85	12	Olive	ELISA	0	2
Sofi et al (2010) (41)	Parallel	Double	6	5	54.4	82	29.3	Adults with nonalcoholic fatty liver disease, Italy	0.71	52	Olive	Bioplex assay	0	1
Troseid et al (2009) (42)	Factorial	Double	247	239	70	100	26.5	Older men with hyperlipidemia, Norway	1.32	156	Linoleic rich	ELISA	14	2
Vargas et al (2011) (43)	Parallel	Double	17	17	30	0	34.8	Women with polycystic ovary syndrome, United States	3.6	6	Soybean	RIA	13	2

Abbreviation: NR, not reported.

^aOne additional study, published only as an abstract, nominally met inclusion criteria but did not contain sufficient data for quantitative analysis (29) and was therefore excluded from the meta-analysis. It was a randomized, double-blind, placebo-controlled, 2 × 2 factorial-design study. Subjects were randomized to receive either EPA (2 g/d) or vitamin E (0.4 g/d), or EPA (2 g/d) plus vitamin E (0.4 g/d), or placebo (corn oil) for 12 weeks. In the published abstract, the authors reported that both EPA and the EPA plus vitamin E treatment increased adiponectin (29).

^bNumber of subjects who completed the intervention.

^cThe study by Ramel et al (39) had a fish oil group (1.3 g/d EPA + DHA) and 2 fish meal feeding groups (cod, 0.26 g/d EPA + DHA; salmon, 2.1g/d EPA + DHA). All other studies used fish oil treatment.

^dWhen mean age was not specified, the median age was used.

^eThe RIA measured all multimeric forms of adiponectin (low molecular weight, medium molecular weight, and high molecular weight) but not monomeric forms, whereas the ELISA and the Bioplex assay measured all multimeric and monomeric forms of adiponectin, except for the study by Dangardt et al (32), which measured high-molecular-weight adiponectin.

^fThe components of the quality scores were: presence of randomization (yes = 1, no = 0), appropriate method of randomization (yes = 1, no = 0), presence of double blinding (yes = 1, no = 0), appropriate method of double blinding (yes = 1, no = 0), description of withdrawals and dropouts (yes = 1, no = 0); with an optimum possible score of 5.

^gSubjects in these intervention arms were given 150 g of cod (which provided 0.26 g EPA + DHA/d) or 150 g of salmon (which provided 2.1 g/d EPA + DHA) 3 times a week.

The median dose of n-3 PUFA (EPA + DHA) was 1.3 g/d (range 0.26–4.2 g/d), and the median duration of treatment was 8 weeks (range 3–156 weeks). Olive and sunflower oil were the most commonly used placebos. One trial measured high-molecular-weight adiponectin (32), whereas the others measured total (all multimeric forms) adiponectin; these measures are highly correlated (r = 0.89) (44). Dropout rates were low (median 10%; range 0%–19%) and generally similar between placebo and fish oil arms in each trial (not shown). Seven trials met at least 3 study quality criteria on the Jadad scale (10), whereas 7 satisfied 2 or less.

Effect of fish oil on circulating adiponectin

The individual trial results and the pooled estimate for the effect of fish oil on adiponectin are shown in Figure 2. Fish oil treatment increased adiponectin by 0.37 μg/mL (95% confidence interval [CI] 0.07, 0.67, P = .02). Between-study heterogeneity was evident (I² = 72.9%, Q-statistic P < .001), although 11 of 14 trials suggested effects of 0 or greater, and only 2 suggested effects substantially less than 0. Among prespecified characteristics evaluated as potential sources of heterogeneity, the effect of fish oil on adiponectin appeared potentially stronger in trials of higher quality and in trials with younger or with obese subjects (Table 2). However, none of these potential interactions were statistically significant (P > .05 each). Even stratified by these factors, statistical heterogeneity remained in each pooled subgroup estimate (I² = 48% to 81.3%, Q-statistic P < .001 to 0.1). In a sensitivity analysis assuming a within-person correlation in adiponectin levels of 0.50, rather than 0.93, results were very similar, with fish oil increasing adiponectin by 0.28 μg/mL (95% CI 0.01, 0.54, P = .04, I² = 56.7%).

Figure 2.

Change in circulating adiponectin resulting from fish oil consumption. There were 13 parallel-design trials and 1 crossover-design trial. Among the parallel trials, 3 were 2 × 2 factorial-design studies, none of which reported an interaction effect between fish oil and the second intervention (31, 38, 42). In the study by Troseid et al (42), the main effect of fish oil on adiponectin was obtained from the authors directly and used in the pooled meta-analysis. For the other 2 studies, this information was not available. To avoid duplication of statistics from these trials, we used a separate meta-analysis to initially obtain a pooled effect size and variance estimate within each of these studies, before using the trial-specific estimates in a second meta-analysis evaluating all studies. This approach was also applied to the study by Dangardt et al (32), which displayed results separately for males and females and did not report a significant interaction between gender and fish oil on adiponectin. The effect estimate (shaded squares) and 95% CI (lines extending from shaded squares) for each trial is shown. The DerSimonian and Laird method of the inverse-variance weighted, random effects meta-analysis was used to pool the results (14). The overall pooled estimate and 95% CI of the effect of fish oil on adiponectin is indicated by the dotted line and clear diamond, respectively. Interstudy heterogeneity was quantified y the I² statistic (15).

Table 2.

Effect of Fish Oil on Circulating Adiponectin According to Prespecified Potential Sources of Heterogeneity^a

	Number of Trials	Pooled Effect of Fish Oil on Adiponectin, μg/mL (95% CI)	Within Subgroup Heterogeneity		Across Subgroup Meta-Regression, P Value
			I2, %	Q-Statistic, P Value
Overall	14	0.37 (0.07, 0.67)	72.9	<.001
Location
Europe/North America	11	0.54 (0.07, 1.01)	75.6	<.001	.79
Others	3	0.18 (−0.24, 0.59)	63.2	.07
Duration of treatment, wk
<8	4	0.61 (0.05, 1.17)	51.3	.10	.66
≥8	10	0.30 (−0.08, 0.68)	77.3	<.001
Blinding
Double	11	0.49 (0.09, 0.89)	75.4	<.001	.73
Single	3	0.24 (−0.46, 0.94)	57.9	.09
Major fatty acid in placebo
Linoleic acid	6	0.55 (−0.09, 1.20)	68.3	.008	.90
Oleic acid	4	0.72 (−0.20, 1.64)	81.3	.001
Others/not reported	4	0.24 (−0.42, 0.91)	73.9	.009
Dose of n-3 PUFA, g/d
<1.29	7	0.47 (−0.10, 1.04)	76.7	<.001	.92
≥1.29	7	0.45 (−0.04, 0.93)	72.3	.001
Study quality
Low, ≤2	7	0.34 (−0.04, 0.73)	75.6	<.001	.76
High, ≥3	7	0.47 (−0.16, 1.09)	74.1	.001
Age, y
<38	7	0.77 (−0.26, 1.80)	78.3	<.001	.77
≥38	7	0.28 (0.01, 0.56)	70.2	.003
Adult population
No	3	0.44 (−1.14, 2.01)	72.7	.03	.74
Yes	11	0.41 (0.10, 0.72)	73.6	<.001
BMI, kg/m2b
<30	6	0.16 (−0.14, 0.47)	73.0	.002	.50
≥30	7	0.81 (0.05, 1.58)	78.5	<.001

^aDuration of treatment, dose of n-3 PUFA, age, BMI, and baseline adiponectin were stratified based on the median value for all studies. When analyzed as continuous covariates, none of these factors were significant sources of heterogeneity (meta-regression, P all > .05).

^bExcluded the study of infants by Andersen et al (30).

Visual inspection of a funnel plot identified slight asymmetry (Figure 3), indicative of potentially more small studies showing stronger effects. Neither Egger's (P = .11) nor Begg's (P = .08) tests achieved statistical significance, although nonsignificant trends were seen. The trim-and-fill correction analysis obtained a theoretical pooled estimate of 0.18 μg/mL (95% CI −0.15, +0.52, P = .28) after 3 theoretically unreported studies were added.

Figure 3.

Funnel plot of change in circulating adiponectin in randomized, placebo-controlled trials of fish oil consumption. The change in adiponectin due to fish oil (x-axis) is plotted against the SE of the change in adiponectin (y-axis) in each trial. The vertical line represents the overall pooled fixed-effect estimate, and the dashed lines indicate the expected 95% CI for a given SE.

Effect of fish meal feeding on circulating adiponectin

Only 2 trial arms in a single study evaluated fish meal feeding (39). Combining these arms, the pooled effect of fish meal feeding on adiponectin was not statistically significant (−0.01 μg/mL, 95% CI −0.65, +0.64, P = .99) although CIs were broad due to the small numbers of subjects.

Discussion

This systematic review and meta-analysis suggests that fish oil consumption increases circulating adiponectin. To our knowledge, this is the first demonstration that fish oil influences circulating adiponectin in randomized, placebo-controlled trials in humans. Our findings quantify the potential magnitude of the effect, an approximately 0.37 μg/mL increase due to fish oil consumption, as well as identify the presence of unexplained heterogeneity and evaluate and quantify potential effects of publication bias.

The overall pooled effect was modest: based on the observed association between adiponectin and type 2 diabetes (6), a 0.37-μg/mL increase in adiponectin would correspond to about 3% lower incidence of diabetes. Furthermore, between-study heterogeneity was evident, suggesting that fish oil supplementation may have stronger influence on adiponectin in some populations and weaker effects in others. In an exploratory analysis, we did not identify any factors that explained the heterogeneity between studies, although meta-regression analyses to confirm sources of heterogeneity can be limited by insufficient statistical power (45). We did not detect a dose-response relationship, but most trials (11 of 14) tested doses in a relatively narrow range of 0.7–2 g/d, limiting our ability to detect a dose-response across a wider range. Additionally, our findings cannot confirm that enhanced insulin sensitivity is a clinically relevant mechanism of cardiometabolic protection of fish or fish oil consumption.

Interestingly, fish consumption has been linked to modestly higher incidence of type 2 diabetes in some but not all studies (46) and to slight increases in fasting glucose concentrations in some randomized controlled trials (47). Experimental studies have suggested that such effects might relate to increased hepatic gluconeogenesis, possibly a result of decreased de novo lipogenesis and lower release of triglycerides as an energy substrate by the liver, rather than due to any adverse effects on insulin resistance (48, 49). Given the key role of adiponectin in pathways of insulin sensitivity and metabolic health, the present findings provide further support that fish oil consumption does not adversely influence adipocyte function or related metabolic pathways and in fact could favorably influence these pathways. Such effects are consistent with known beneficial effects of n-3 PUFAs on triglycerides as well as modest increases in high-density lipoprotein concentrations and low-density lipoprotein participle size (50–52), and our results highlight the need for additional studies to investigate the effect of fish oil supplementation on other metabolic outcomes such as inflammation, free fatty acids, and insulin resistance.

In cultured rat and human primary adipocytes, enrichment of EPA and DHA leads to increased synthesis and secretion of adiponectin (53, 54). In these experiments, antagonism of peroxisome proliferator-activated receptor-γ, a key nuclear receptor known to be bound by n-3 PUFA, blocked the adiponectin-increasing effects of n-3 PUFAs (53, 54). n-3 PUFAs may also enhance adiponectin synthesis by inhibiting transient receptor potential canonical calcium ion channels, which are expressed on mature adipocytes and regulate adiponectin production (55). These findings in cultured adipocytes are supported by in vivo animal models, in which higher intake of n-3 PUFAs increases tissue and circulating adiponectin (56). Interestingly, recent experimental data suggest that specific DHA-derived metabolites, resolvins D1 and D2, can also stimulate the production of adiponectin (57). Our findings provide support that these pathways identified in vitro and in vivo may have functional relevance to effects of n-3 PUFAs on adipocyte function and adiponectin levels in humans.

There were several strengths of our analysis. We systematically searched for all published and unpublished (ie, abstracts) studies with broad search terms across multiple databases and also contacted authors for clarification and additional data, which minimized potential for publication bias and misclassification. We limited our analysis to randomized, placebo-controlled trials, and most were also double blind, strengthening inference for a cause-and-effect relationship. Attrition bias was also unlikely, due to low dropout rates that were also similar for the placebo and fish oil treatment in each trial. Only 1 trial used a crossover design, and its 6-week washout period minimized carryover effects. We evaluated several prespecified sources of heterogeneity, including dose, duration, and study location.

Limitations should be considered. Despite the broad search criteria, publication bias remains a possibility. Although statistical tests for heterogeneity were not significant, asymmetry was apparent in the funnel plot, and a trim-and-fill sensitivity analysis suggested the effect of fish oil could be overestimated in our unadjusted pooled analysis. Conversely, funnel plot asymmetry could also be due to other factors (16), such as true biological heterogeneity (effect modification) between different identified trials. We also excluded 1 trial that reported increased adiponectin after fish oil treatment due to insufficient reported data (29); inclusion of this trial may have reduced asymmetry in the funnel plot. Several of the trials had lower study quality as assessed by the Jadad scale (10). Yet effects of fish oil on adiponectin were not significantly different in trials with higher or lower study quality scores. Thus, study quality may not have had a systematic influence on the pooled findings.

In conclusion, this systematic review and meta-analysis of randomized, placebo-controlled clinical trials suggests that fish oil supplementation moderately increases circulating adiponectin. These findings support potential beneficial effects of fish oil supplementation on pathways related to adipocyte health and adiponectin metabolism.