We searched the following bibliographic databases from inception to Dec 6, 2013: PubMed/MEDLINE (National Library of Science), EMBASE (Ovid), CENTRAL (Cochrane library - Wiley), followed by forward searching for key articles in Scopus. To identify ongoing and unpublished trials, we searched the World Health Organization’s International Clinical Trials Registry Platform. We supplemented electronic searches by hand-searching the bibliographies of included trials, and relevant narrative and systematic reviews. Our search strategy used both controlled vocabulary and free text, and searches in MEDLINE and EMBASE were combined with a high-sensitivity filter for randomized controlled trials [28]. The detailed search strategy for MEDLINE is included in Table 1.

To identify a population at high risk for cardiovascular events, we included only randomized trials of adults aged 40 years or older with established peripheral arterial disease, which was diagnosed by the presence of stable intermittent claudication and/or an ankle-brachial index ≤ 0.9. The intervention period was required to be at least 12 weeks in duration to allow the intervention to impact cardiovascular function. The minimum intervention duration of 12 weeks was chosen based on studies of the appearance of dietary fatty acids in the plasma and blood cell plasma membranes (which are considered robust markers of omega-3 PUFA consumption), demonstrating that plasma membrane fatty acid composition changes are apparent within two weeks of a dietary fatty acid intervention [30]. Thus, it is feasible that downstream physiological effects of the diet could be demonstrated within 12 weeks. We included interventions with any omega-3 PUFA in diet or supplement form where the dose per day was reported and with appropriate comparators (i.e. placebo, omega-3 PUFA deficient diet or usual diet). We used a two-stage process for trial screening and selection employing standardized and piloted screening forms. Two reviewers independently screened the titles and abstracts of the search results to determine whether each citation met the inclusion criteria. Trials published in languages other than English were translated, and the full text of citations classified as include or unclear were independently reviewed with reference to the predetermined inclusion and exclusion criteria. Discrepancies between reviewers were resolved through consensus or by discussion with a third reviewer, as required.

From each trial, two reviewers independently abstracted population characteristics (including age, body mass index, and diagnosis of peripheral arterial disease), number of participants, trial duration and follow-up, intervention design (including type and dose of omega-3 PUFA, comparators, and co-interventions), clinical health outcomes (incidence of major adverse cardiac events, myocardial infarction, cardiovascular death, angina, and stroke, amputation, symptom-driven revascularization procedures (e.g. bypass surgery), maximum and pain-free walking distance, and quality of life), and adverse effects. Any disagreements were resolved through consensus. Where data were incompletely or imprecisely reported, we contacted study authors for clarification. We assessed the internal validity of the included trials using the Cochrane Collaboration Risk of Bias Tool [31]. This tool consists of six domains, each of which is rated “low risk,” “unclear risk,” or “high risk.” If one or more individual domains were assessed as having a high risk of bias, the overall rating was a high risk of bias. We considered the overall risk of bias low only if all components were rated as having a low risk of bias. We rated the risk of bias for all other trials as unclear. Information regarding methodological quality was used to guide subgroup analyses and to explore sources of heterogeneity.

We conducted meta-analyses of the data from included trials using Review Manager (Version 5.2, The Cochrane Collaboration, Copenhagen, Denmark). Pooled binary data were weighted using the Mantel-Haenszel method, and presented as risk ratios (or Peto odds ratios for rare events) with 95% confidence intervals (CI) [32]. Pooled continuous data were weighted by the inverse of variance and expressed as a weighted mean difference with 95% CI. We explored and quantified statistical heterogeneity of the data using the I-squared test [33]. We assessed publication bias by viewing the overlap of confidence intervals and using funnel plot techniques. For the primary outcome of major adverse cardiac events, we performed the following a priori subgroup analyses: short versus extended duration of intervention, high versus low omega-3 PUFA dose (g/day), type of omega-3 PUFA (EPA, DHA, ALA, or a combination of these), supplements or capsules vs. dietary sources of omega-3 PUFA, and the duration of follow-up. Final subgroup analyses were limited by the number of trials included and the availability of reported outcomes and covariates.

Of 741 citations identified from electronic and hand-searches, we included 5 unique trials enrolling a total of 396 individuals (Figure 1; Table 2). All were single-centre trials published between 1990 and 2010. All trials were adjudicated to be of high (3/5) [34‐36] or unclear (2/5) [37, 38] risk of bias (Table 3). Due to the small number of trials included, formal assessment of reporting bias (e.g., by using a funnel plot) was not possible.

Four trials were conducted in Europe [34‐37] and one in Japan [38]. One trial recruited individuals who were diagnosed with peripheral arterial disease and were also hypercholesterolemic and on statin therapy [38]. One trial included individuals who previously had angina or a myocardial infarction [34], while the rest excluded individuals with a recent history of cardiovascular/cerebrovascular events or revascularization surgery. Three trials [34, 37, 38] studied capsules containing varying doses of EPA alone or EPA and DHA combined, compared to capsules of sunflower oil [34] or corn oil [37], or no treatment [38]. The intervention in the remaining two trials [35, 36] studied a specially developed dairy product enriched with EPA, DHA and ALA, as well as vitamins A, B₆, D and E, folate and oleic acid, compared to a control group who received a product identical in appearance enriched only with vitamins A and D. None of the trials had follow-up periods beyond the duration of the intervention.

While all trials had random sequence generation (5/5), fewer reported adequate allocation concealment (2/5) [34, 37] and blinding of outcome assessment (3/5) [34‐36], and all trials were subject to unclear or high risk of attrition bias due to incomplete outcome data following substantial loss to follow-up or poor reporting of outcomes (5/5) (Table 3). The presence of co-intervention nutrients in 3/5 trials [34‐36] did not allow proper controlling for omega-3 PUFA effect.

Two trials reported the incidence of major adverse cardiac events following an intervention with EPA only [34, 38]. The risk ratio for major adverse cardiac events was not significantly decreased with supplementation of EPA (risk ratio 0.73; 95% CI 0.22 to 2.41; I ² 75%; 2 trials [34, 38]; 288 individuals; Figure 2). High statistical heterogeneity observed may relate to differences in the trial populations (Japanese vs. British, hyper- vs. normo-cholesterolemic), the ten-fold difference in daily dose of EPA (1.8 g vs. 0.18 g), the comparator in the control group (no treatment vs. sunflower oil), the co-intervention (statins vs. gamma-linolenic acid), or the length of the intervention (five vs. two years). In one of the trials (213 individuals), the risk ratio for major adverse cardiac events associated with 1.8 g/day EPA compared to no treatment for 5 years was 0.41; 95% CI 0.19 to 0.87 [38]. In the second trial (75 individuals) comparing 0.18 g/day EPA to sunflower oil for 2 years, the risk ratio for major adverse cardiac events was 1.38; 95% CI 0.55 to 3.50 [34].

Two trials with a total of 288 individuals reported the incidence of myocardial infarction, cardiovascular death and revascularization procedures (Table 4) [34, 38]. The pooled Peto odds ratios for myocardial infarction (Peto odds ratio 0.64; 95% CI 0.22 to 1.88), cardiovascular death (Peto odds ratio 0.60, 95% CI 0.19 to 1.90) and the pooled risk ratio for revascularization surgery (risk ratio 0.81, 95% CI, 0.13 to 4.91) demonstrated no significant difference between the EPA intervention group and control (no treatment or sunflower oil).

Four trials [34‐37] with a total of 187 individuals reported pain-free walking distance (metres), an indicator of intermittent claudication (Figure 3). We observed considerable heterogeneity and no statistically significant differences in pain-free walking distance associated with omega-3 PUFA supplementation (mean difference 115.40 metres; 95% CI -42.24 to 273.05; I ² 89%). None of the trials reported on quality of life. The remaining secondary outcomes (stroke, angina, amputation and maximum walking distance) were each reported by only one trial and were not found to be statistically different with omega-3 PUFA intervention (Table 4). Gastrointestinal upset, reported in a single trial [34], was more frequent with EPA treatment (risk ratio 1.58; 95% CI 1.01 to 2.48; 60 individuals).

Limitations of the data (i.e. the number of trials included, and the availability of appropriate outcome data and covariates reported) precluded most subgroup analyses. Comparisons of the effects of omega-3 PUFA intervention length, dose and type of omega-3 PUFA on pain-free walking distance did not reach significance, and statistical heterogeneity remained high (Table 5).