Protection against Incidences of Serious Cardiovascular Events Study with daily fish oil supplementation in dialysis patients (PISCES): protocol for a randomised controlled trial


By 2030, over 2.5 million people worldwide will have kidney failure with replacement therapy (KFRT,1 previously known as end-stage kidney disease).2–4 Haemodialysis (HD) patients suffer from premature mortality, with a median 5-year survival of approximately 42% in the USA.5 Compared with the general population, the estimated foreshortening of lifespan due to KFRT is more than 25 years in men and 30 years in women.5 Cardiovascular (CV) disease accounts for approximately 42% of deaths in these HD patients.5 The assessment of interventions to prevent CV events in patients with chronic kidney disease (CKD) and kidney failure is particularly important as the risk and outcomes of these events are significantly worse compared with individuals without CKD. For example, the relative risk for CV mortality is 3 times higher in patients with CKD than in those without CKD and almost 10 times higher in the setting of CKD with diabetes.6 7 Although approximately 50% of kidney failure patients have diabetes, the CV risk conveyed by CKD exceeds that conveyed by diabetes.8 While diabetes is a risk factor for CVD, with optimised medical care and risk reduction, the CV event rate has declined over time in the non-dialysis population.9 This is in stark contrast to the persistently high CV event rate in dialysis patients. Unlike the general population, few pharmacological interventions are known to improve patient-important outcomes in dialysis patients,10–13 highlighting the need for further innovative research.

One such approach leverages the natural supplementation with eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), the long-chain omega-3 polyunsaturated essential fatty acids (n-3 PUFAs) found in fish oil, premised on the additional risk factors and differences in mechanisms for CV events that burden dialysis patients compared with non-dialysis patients.14 15 Higher intake of n-3 PUFA is associated with a reduced risk of CV death and sudden cardiac death in non-dialysis patients16–18 and in patients with established CV disease.19 20 Although patients with the greatest reductions in kidney function may benefit most with n-3 PUFA21; studies to date have excluded individuals with KFRT. Available evidence on the effects of n-3 PUFA on CV events in dialysis patients is promising but limited.22 23 Our prior multicentre randomised placebo controlled Fish Oil Inhibition of Stenosis in Hemodialysis Grafts (FISH) study,22 which evaluated the effect of daily n-3 PUFA supplementation (4 g/day, ie, 1600 mg EPA and 800 mg DHA total/day) on HD synthetic graft vascular access outcomes had promising findings on key a priori defined secondary CV endpoints (CV events defined as a composite of myocardial infarction (MI), heart failure (HF), stroke, peripheral vascular disease (PVD) and cardiac-related death). We found a significant reduction in the rate of CV events in the n-3 PUFA supplemented group (142/1000 patient years (95% CI 73 to 256)) versus placebo (347/1000 patient years (95% CI 215 to 526, p=0.02)) and a >50% reduction in the risk of a first CV event in the n-3 PUFA supplemented group (HR 0.43; 95% CI 0.19 to 0.96; p=0.035). The number needed to supplement with n-3 PUFA to prevent 1 CV event was 12. However, the fact that the FISH study population required a synthetic graft vascular access relegated them to a select, high CV risk population whereby their vasculature was so poor (eg, burdened by arterio or atherosclerosis) that a native arteriovenous fistula (natural artery-vein anastomosis without the synthetic graft) could not be successfully created for vascular access. Further, the FISH study was limited by its small study size (n=201) and numbers of CV events (9/99 (9%) in the n-3 PUFA group and 17/97 (18%) in the placebo group). However, the consistency in the reduction in the rate and risk (HR) of CV events in the FISH study and others24 that include those with KFRT who have the greatest CV risk, merits confirmation in a larger clinical trial of representative HD patients.

Methods and analysis

Study design

PISCES: Protection against Incidences of Serious Cardiovascular Events Study with Daily Fish Oil Supplementation in Dialysis Patients, is a multicentre randomised, placebo controlled, parallel clinical trial to test the hypothesis that oral supplementation with n-3 PUFA, when added to usual care, leads to a reduction in the rate of serious CV events in HD patients when compared with usual care plus matching placebo. The study will recruit HD patients who have given informed written consent to participate, from 26 dialysis units in Canada and Australia (online supplemental file 1) with an expected treatment duration of 3.5 years and total study duration of 10 years. PISCES is a pragmatic trial; apart from randomisation to n-3 PUFA or matched placebo in a 1:1 ratio, patient care will be managed by usual care providers according to local policy and procedures. The study has been approved by each of the participating sites’ research ethics review boards and the study follows the Declaration of Helsinki. It is registered at (ISRCTN00691795). The study has completed enrolment and has been following participants since August 2019.

Supplemental material

Objectives and outcomes

Primary objective and outcome

The primary objective of this study is to compare the rate of all, not just the first, serious CV events between HD patients whose diets are supplemented daily with n-3 PUFA (4 g/day) and those whose diets are supplemented daily with matching placebo. A serious CV event is a composite outcome comprising either a CV-related death (sudden and non-sudden cardiac death, fatal MI, fatal stroke) or a non-fatal CV event (non-fatal MI, non-fatal stroke and PVD requiring amputation) (see online supplemental file 2 for definitions).

Supplemental material

Why is heart failure not part of the primary outcome?

While heart failure is an important outcome that may potentially be influenced by n-3 PUFA,25–28 it is often difficult to distinguish between true CV systolic or diastolic heart failure in HD patients versus volume overload induced by dietary indiscretion, inappropriate target weight assignment or non-compliance with the dialysis prescription. In a prior study of n-3 PUFA in HD patients, the number of HF events was identical between treatment groups.22 Thus, we will capture heart failure as a secondary outcome rather than as a component of the primary composite outcome.

Secondary objectives and outcomes

The secondary objectives will include comparisons of the proportion and time to event of the primary composite outcome. Further, we will compare between groups the rates, proportions and time to the first event of: (1) a composite outcome that includes all-cause mortality and the non-fatal CV events from the primary outcome; (2) individual components of the primary composite outcome (cardiac death, MI, stroke, PVD); (3) heart failure requiring hospitalisation; (4) all CV related hospitalisations and interventions and (5) all-cause mortality. N-3 PUFA may have beneficial pleiotropic effects in HD patients that may reduce overall mortality. Indeed, a study of 216 incident dialysis patients found that those who ingested n-3 PUFA were approximately 50% less likely to die in the 3-year follow-up period29 than those who did not; thus it is important to include all-cause mortality. All primary outcomes and secondary outcomes that include mortality will be adjudicated by an independent Events Adjudication Committee composed of one or more cardiologists, nephrologists, neurologists and vascular surgeons, with clinical trials expertise and blinded to treatment assignment.

Tertiary outcomes that will be compared between treatment groups include blood pressure (online supplemental file 2 for definitions) and laboratory measures, specifically, the mean serum levels of n-3 PUFA and their key subcomponents (AA, EPA, DHA, EPA/AA). It is well established that dialysis patients exhibit lower levels of EPA and DHA in serum phospholipids than people without kidney disease30–33; in North America, HD patients have been found to have among the lowest levels recorded in the medical literature.31 The inadequate n-3 PUFA levels of HD patients may be due to reduced dietary intake,32 34 increased breakdown of n-3 PUFA, inflammation, diarrhoea, malabsorption, disturbances of metabolism and incorporation in the uremic milieu or by the HD procedure30 that does not occur in non-HD patients. We hypothesise that there may be a relationship between n-3 PUFA serum levels and CV events. Finally, data will be collected for quality of life (QOL) and economic analyses: QOL questionnaires (eg, 36-item Short Form Health Survey (SF-36), Kidney Disease Quality of Life (KDQOL), EQ-5D), the number of hospitalisations, lengths of hospital stays, number and type of CV and surgical interventions, and other healthcare utilisation data will be collected for future relevant economic and QOL analyses.

Inclusion/exclusion criteria

The inclusion criteria are designed to be broad:

  • Men and women (age>18 years) with KFRT on chronic HD 3 or 4 times per week of duration less than or equal to 5 hours per HD session.

  • Clinically stable (no hospitalisation or emergency room visit during the 1 month before enrolment).

Exclusion criteria are listed in box 1.

Box 1

Exclusion criteria

  • Reversible, acute kidney failure, likely with recovery of kidney function.

  • Pregnancy.

  • Active malignancy.

  • Active major bleed within 1 month of enrolment (online supplemental file 2 details for definition of major bleed).

  • Blood pressure >180/120 (malignant level).66

  • Receiving >2 antiplatelet agents or anticoagulants, that is, use of aspirin and Coumadin is not an exclusion.

  • Life expectancy <6 months (ie, palliative dialysis patients).

  • Implanted ICD or planned ICD placement within the year.

  • Involvement in another drug trial.

  • N-3 PUFA supplementation at the time of randomisation.

  • Any known allergy to fish, soy, corn or their products.

  • Unable to provide written informed consent (self or legal representative).

Why were the inclusion and exclusion criteria chosen?

All incident and prevalent adult chronic HD patients with or without previous CV disease or events were considered to increase generalisability to a broad HD population. Despite the excessive CV morbidity and mortality, few medical interventions are known to prevent CV events in the high-risk HD population.35–37 A RCT that randomised 114 HD patients with dilated cardiomyopathy to a beta-blocker, carvedilol or placebo demonstrated improved 2-year survival in the carvedilol group.37 Other small RCTs have demonstrated benefits to CV mortality using angiotensin-receptor blockers38 39; however, a subsequent meta-analysis do not confirm these results.40 Finally, the SPACE trial of 196 patients showed reduced CV events in HD patients with pre-existing CVD who received vitamin E compared with placebo.35 These studies are limited by the select patient population studied (heart failure or those with pre-existing CVD) and by their small sample sizes (studies had fewer than 400 participants); thus, these studies did not change clinical practice. The broad inclusion criteria, large sample size, easy accessibility and relative low cost of the intervention in PISCES will allow for greater applicability into clinical practice should the study demonstrate benefit of n-3 PUFA.

Participants with implantable cardioverter defibrillators (ICDs) were excluded because while n-3 PUFAs may be antiarrhythmic under conditions that favour triggered activity (eg, dialysis), it is unclear whether they improve or facilitate reentrant arrhythmias in dialysis patients, leading to ICD related arrhythmias.41–44 At the time of protocol development and trial recruitment, no consistent data existed for HD patients with ICD, thus, until more data from large prospective studies on the risk to benefit ratio of n-3 PUFA in HD patients with ICD is established, such patients were excluded.


Randomisation will occur via an independent, central, web-based system to ensure allocation concealment ( Given the pragmatic study design, participants will only be stratified by study site and CV history (prior, known co-existing CVD (based on study outcome definitions) or no CVD and event naïve). The randomisation sequence will be generated by a computerised random number generator, using a random permuted block design with randomly chosen block sizes. After randomisation, the randomisation facility will inform the site-specific coordinator (blinded to treatment allocation) so that the appropriate study capsules can be allocated. Study capsules are contained in opaque bottles labelled by a bottle ID; bottles are only distributed to the participant after the coordinator verifies by a bottle verification number that the correct bottle has been retrieved. The study coordinators, investigators, outcome assessors and participants will remain blinded with respect to treatment allocation (n-3 PUFA or placebo) throughout the study.

Treatment regimens

The trial intervention is daily oral supplementation with either 4 grams (four 1 gram capsules) of n-3 PUFA or matching placebo to begin immediately after randomisation. The n-3 PUFA capsules are steam deodorised and citrus-flavoured. The placebo capsules contain citrus-flavoured corn oil and are packaged identically to the n-3 PUFA. Therefore, the placebo and treatment capsules resemble each other in colour, shape, odour, taste and consistency to maintain participant, investigator and study team blinding. In a prior RCT using similar study interventions (FISH Study), we determined that participants were adequately blinded, likely due to the careful use of steam deodorisation and flavouring.22 This is consistent with another RCT of n3-PUFA versus placebo in determining ‘how blind is double blind’, where despite a marked difference in taste experience, there was no significant difference in correct guesses between the two groups.45 Furthermore, the CV outcomes of the study are ‘hard’ outcomes that will be adjudicated by outcome assessors who will be blinded to treatment allocation. The study capsules will be self-administered with participants instructed to take with food (eg, snack or meal) to help with ease of ingestion, improve compliance and reduce any aftertaste that may occur with burping. The bioavailability is such that the participant may take all four capsules at once or two capsules two times per day. Participants will be instructed to continue taking their study capsules even if they are hospitalised (for any reason), as study outcomes may occur during hospitalisation. For example, 11%–18% of cardiac death occur in outpatient dialysis clinics46–48 while >70% occur in hospital.49 Participants will be reminded by the study coordinator to take their capsules during any hospitalisation. The allocated study treatment will continue until the end of the study period, regardless of whether or not a study outcome has occurred.

The n-3 PUFA supplement used in this study is similar to that used in the FISH study,22 which has established safety in HD patients. The principal potential safety concern about n3-PUFA is bleeding risk; however, in the FISH study, no increased bleeding risk was seen.22 Other studies have also shown no increased bleeding tendency with moderate amounts of n-3 PUFA supplementation (2–5 g/day).50–53 In the Shunt Occlusion Trial of n-3 PUFA to prevent coronary bypass graft stenosis, 610 patients received n-3 PUFA and either aspirin or warfarin and followed for 1 year. There were no differences in bleeding episodes between the n-3 PUFA and control groups.51 This was confirmed in an RCT specifically assessing the bleeding risk of n-3 PUFA supplementation in HD.52 N-3 PUFA is considered safe enough for the European best practice guidelines for HD to recommend fish oil to treat resistant hypertriglyceridaemia in dialysis patients.54 In a small percentage (5%–20%) of patients, an unpleasant aftertaste with burping or stomach upset may occur. If patients develop intolerance, they may split the dose or freeze the capsules prior to ingestion as this can help reduce gastrointestinal intolerance. An independent data safety and monitoring board will closely monitor study progress and any safety concerns.

Study visits and data collection

The study timeline is detailed in figure 1. Patients will be followed for a minimum of 3.5 years, regardless of the number of events experienced during follow-up (to capture the rate of the primary outcome). Participants will have monthly follow-up (either in-person or by telephone) to determine if outcomes have occurred and to capture any serious adverse events. Participants are given a toll-free phone number to inform the study coordinator if they have been hospitalised or had a CV event. The ‘captive’ nature of HD patients (who are treated in their dialysis unit three times a week) will simplify this procedure and minimise loss to follow-up. The first patient randomised into the study was on 28 November 2013 and the last patient follow-up was March 2023. Measurement of n-3 PUFA blood (0.5 mL) will be obtained during HD (within the first hour) from the patient’s blood circuit to determine n-3 PUFA compositions of total serum phospholipid to address some of the secondary objectives according to study timeline (figure 1). Samples will be labelled by study ID and date, kept frozen (−80°C), properly packaged, and shipped to the Department of Human Biology and Nutritional Services at the University of Guelph, Ontario, for analysis. Blinded results will be sent to the study coordinating centre at the Toronto General Hospital Research Institute.

Figure 1
Figure 1

Study timeline for PISCES. BP, blood pressure; KDQOL, Kidney Disease Quality of LIfe Instrument; PISCES, Protection against Incidences of Serious Cardiovascular Events Study; QOL, quality of life.

During the study, there will be two measures of compliance: (1) measurement of n-3 PUFA incorporation into cell membranes and (2) pill count. At randomly selected sites, the study coordinator will perform a pill count on a random sample of participants at 3, 6 and 12 months and then annually, at the discretion of the site investigator. EPA, DHA and other fatty acid fractions will be measured after extraction and derivation of fatty acid methyl esters, followed by capillary gas-liquid chromatography.55 This has been successfully done to assess compliance of oral n-3 PUFA supplementation in the FISH study and other placebo controlled RCTs.56 57

Statistical analysis

The primary analysis will compare the rate of CV events (potentially recurrent non-fatal CV events and CV-related deaths) between the n-3 PUFA group and the placebo group using the Prentice, Williams and Peterson (PWP) gap time model, an extension of the Cox proportional hazards model for recurrent events.58 Unlike a simpler Poisson-regression based model, this model allows the baseline rate of events to vary over time, an important consideration in the group of patients newly starting HD, where early rates are expected to be higher than later ones.59 In addition, the PWP model allows baseline rates to vary by the previous number of events and the gap time model resets the time origin to zero at the occurrence of each event, in a sense matching on time since last event when comparing intervention groups. Participants will be followed from the point of randomisation and all recurrent events counted until the end of follow-up. If end of follow-up is a CV related death, this will also count as an event. If the participant dies a non-CV death, withdraws or reaches the end of the study alive, the end of follow-up is a censoring time. As a sensitivity analysis of the primary approach described above, we will use a joint frailty model that distinguishes recurrent CV events from terminal CV events, and allows separate effects of treatment on each, as well as a global test of the effect of treatment.60 The secondary outcome of time to first event will be compared between treatment groups using the Cox proportional hazards model, with baseline hazard stratified by new or prevalent HD status at randomisation. Even in the presence of competing risks (here, non-CV death), the estimation of cause-specific HRs is the recommended approach for determining the effect of an intervention in a clinical trial with a rate outcome.61 Sensitivity to the competing risk of non-CV death will be assessed by estimating the effect of intervention on the subdistribution hazard of a first CV event or CV related death using the Fine and Gray model.62 All of these analyses will include covariates for prior CVD, age, sex and type of access (catheter or arteriovenous access i.e. fistula or graft) at the time of randomisation; the baseline hazard will be allowed to vary by study site and dialysis vintage. The effects of treatment will be presented as adjusted HRs and rates of CV events will be reported as numbers per 100 patient years in each study group; incidence of first CV events over time will be estimated using the cumulative incidence function. A single interim analysis of the primary outcome will occur when half the anticipated number of events has occurred, using the Haybittle-Peto boundary α=0.001; this does not change α for the final analysis. Analyses of continuous outcomes collected at the study visits in figure 1 will use linear mixed effects models, with adjustment for the baseline value of the outcome and other covariates listed above. All analyses will use a p value of less than 0.05 to indicate statistical significance. The full statistical analysis plan describing all primary, secondary and exploratory analyses, and approaches to missing data and potential non-proportional hazards can be found in online supplemental file 3.

Supplemental material

Sample size

The overall sample size for the study will be 1100 participants total, accrued over 6 months at eash site and followed for at least 3.5 years. Data from administrative data bases in Ontario, Canada (Institute for Clinical Evaluative Sciences – Kidney, Dialysis & Transplantation Research Program (ICES-KDT)) for a HD population not exposed to n-3 PUFA was used to estimate the rate of recurrent CV events at 0.270 (270/1000 patient-years) and the rates of CV and non-CV related death to be 0.044 and 0.098. We assumed the HR for treatment with n-3 PUFA to be 0.825,22 a relative rate reduction of 17.5% that is clinically meaningful. The 95% CI for the HR in the FISH study22 (95% CI 0.19 to 0.96) suggests that the true benefit of n3 PUFA is plausibly larger than this. Finally, a yearly drop-out rate of 10% (higher than the FISH study in a similarly ‘captive’ population) and a drop-in rate of 3% based on baseline data from the FAVOURED study.63 64A simple two-group comparison in a Poisson regression over an average follow-up time of 2.5 years (to account for yearly loss to follow-up of 10% and yearly non-CV death of 9.8%) indicates 580 per group to attain 80% power when the rate ratio is 0.825. To assess the performance of the analysis of the recurrent events described above with real (not idealised) time to event data, a simulation study was carried out within a retrospective cohort of HD patients assembled using Ontario administrative data. Repeated samples of 1100 patients were selected from the cohort and randomised to n-3 PUFA or placebo; the actual times to study-defined events were left unchanged in the placebo group and rescaled (lengthened) in the intervention group by an amount consistent with a HR of 0.825. Allowing 3.5 years of potential follow-up power was estimated to be 82% using an α of 0.05 for the PWP gap time model. This multicentre study will recruit participants from academic and community affiliated dialysis units, including associated satellite units to achieve the target sample size.

Future studies and analysis

PISCES may inform future physiological and clinically oriented studies that investigate mechanisms that affect CV integrity impacting patient morbidity and mortality. Studies evaluating the effects of n-3 PUFA on QOL and its cost-effectiveness for prevention of CV events will be done if clinical benefit is found. The data to support these analyses will be collected in this study and fully executed in future work.

Patient or public involvement

Patients were not involved in the design and develop ment of the protocol. We have informed key stakeholders, including international patient associations about our study prior to its initiation and throughout the course of the study. The findings of our study will be discussed with patients, healthcare professionals, policy-makers and the public during the course and at the end of our study.

This post was originally published on