A Critical Look At The REDUCE-IT Trial

It’s finally here. The results of the REDUCE-IT trial were presented today at the American Heart Association conference and the paper is also up on The New England Journal of Medicine. I’ve been excited to see the results of this trial ever since the announcement in September (which I wrote about here) so let’s get right into it.

I’ll just copy and paste what I’d written before in September regarding the inclusion/criteria and statistical analyses so that I can get right into the results. You can skip this section and go to the results if you’d like or you can reread it to get reacquainted with the trial design.

September 2018:

That’s why I thought it would be a good time to review the preregistration protocol of the REDUCE-IT trial on ClinicalTrials.gov and a detailed paper that describes the rationale behind the study design. Let’s get into it.

So very clearly, the authors state what the study here is doing,

“AMR101 (icosapent ethyl [ethyl-EPA]) is a highly purified ethyl ester of eicosapentaenoic acid (EPA) being developed by Amarin Pharma Inc. for the treatment of hypertriglyceridemia.

The purpose of this study is to evaluate whether this drug, combined with a statin therapy, will be superior to the statin therapy alone, when used as a prevention in reducing long-term cardiovascular events in high-risk patients with mixed dyslipidemia.”

Comparing statins + highly purified, high doses of EPA to statins alone. Why are they using such high doses? The authors explain here,

“It is worth noting that the promising results from JELIS occurred with a high‐purity EPA preparation dosed at 1.8 g/d in a Japanese population, for whom the baseline EPA levels are higher than in western populations due to greater dietary intake of marine omega‐3 fatty acids.

Icosapent ethyl 12‐week dosing at 4 g/d in a high‐risk population similar to that within the Reduction of Cardiovascular Events with Icosapent Ethyl–Intervention Trial (REDUCE‐IT) who had persistent elevations of TG after treatment with statins resulted in significant reductions in TG and atherogenic lipoproteins, as well as comparable plasma EPA levels as the 1.8 g/d dosing group in JELIS.

Therefore, a dose of 4 g/d was selected as the dose for further study. In this context, REDUCE‐IT was designed to determine if treatment with icosapent ethyl 4 g/d vs placebo would reduce ischemic events in patients at increased CV risk already being treated with statins.”

Also “worth noting” that the JELIS trial was included in both meta-analyses mentioned at the beginning of the blog post.

The primary outcome is a composite endpoint that consists of CV death, MI, stroke, coronary revascularization, and hospitalization for unstable angina. Whenever I hear composite endpoint, I often become skeptical because I’ve seen composite endpoints abused or misused in several ways, resulting in very misleading results. There was a very nice post on composite endpoints by Hilda Bastian here.

The study (REDUCE-IT) enrolled 8,000 participants and began in 2011 and ended in 2018.

Why 8,000 participants?

The authors explain here,

“The sample‐size calculation was based on a hazard ratio assumption of 0.85. Therefore, 1612 events would be required to have approximately 90% power with a 1‐sided α‐level of 2.5% and with 2 interim analyses. This results in a total target sample size of 7990 patients.

Approximately 70% of randomized patients were to be in CV risk stratum 1 (established CVD) and approximately 30% of randomized patients were to be in CV risk stratum 2 (high‐risk primary prevention defined by diabetes mellitus and other risk factors). Randomization was stratified by CV risk strata, ezetimibe use, and by geographical region.”

The design power calculations seem like they were based on very optimistic expectations of the effect size, and apparently, justifiably so!

A bit unfortunate that there is no mention anywhere of what statistical analyses they planned on doing exactly.

In order for participants to be eligible for the trial:

  • they had to be over the age of 45
  • have high levels of triglycerides (hypertriglyceridemia) originally >150 mg/dL but then changed to >200 mg/dL via a protocol change
  • been on statins for at least a month
  • they had to have a history of heart disease
  • There’s a bunch of exclusion criteria too that I don’t really feel the need to go over, but you can see them listed out on the registry.

Also, worth mentioning of course that the trial is funded by Amarin, AKA the manufacturer of Vascepa.

November 2018


Quick Characteristics of Participants

  • Median Age: 64 years old
    • Females: 28.8%
  • Median LDL: 75.0 mg/dL
  • Median HDL: 40 mg/dL
  • Median Triglycerides: 216.0 mg/dL

Changes in Lipids:

The median change in triglycerides from baseline to 1 year was a decrease of 18.3% (−39.0 mg per deciliter) for the Vascepa group and an increase of 2.2% (4.5 mg per deciliter).  

A between-group comparison of change scores shows that the median group reduction was 19.7% higher (44.5 mg/dL, p<0.001) in the Vascepa group than it was in the placebo group.

LDL increased in both groups from baseline to year 1, however, the Vascepa group had a 6.6% (5.0 mg per deciliter, p<0.001) lower increase than the placebo group.

Differences in Clinical Endpoints (what we really care about):

A total of 1606 participants experienced one of the primary end-point events.

Approximately 17.2% of the participants in the Vascepa group experienced a primary end-point whereas 22.2% of the participants in the control group experienced a primary end-point.

Thus, the between-group difference is 4.8% (95% CI, 3.1 to 6.5).

The results are: (hazard ratio, 0.75; 95% confidence interval [CI], 0.68 to 0.83; P<0.001).

So very clearly we can see that there was a statistically significant difference and the confidence interval leans towards an effect.


The authors calculated a number needed to treat for this between-group difference (NNT: 21 (95% CI, 15 to 33)) and its confidence interval which slightly makes me laugh. I’ve written about the NNT before and its flaws and the reality is that it doesn’t do what it claims to do and has very bizarre statistical properties.

With regards to secondary efficacy end-point events, 11.2% of the participants in the Vascepa group experienced an event compared to 14.8% of the participants in the placebo group.

The absolute between-group difference was 3.6% (95% CI, 2.1 to 5.0).

Again, very clearly a difference.

They also ran prespecified subgroup analyses to see whether the effects would be influenced. What was interesting to me was when they claimed this:

“Baseline triglyceride levels (≥150 vs. <150 mg per deciliter or ≥200 or <200 mg per deciliter) had no influence on the primary or key secondary efficacy end points (Figures 2 and Figure 3).”

This is interesting to me because there were so few patients that were enrolled into the trial with triglyceride levels that were beneath 150 or hovering around 150. This is very clearly reflected by the wide confidence intervals in that subgroup which can be seen in the figure above. This is because the protocol changed early in the trial to restrict participant inclusion to people with higher triglycerides,

“The first protocol amendment in May 2013 changed the lower limit of the acceptable triglyceride level from 150 mg per deciliter to 200 mg per deciliter (2.26 mmol per liter), with no allowance for variability."

Whereas originally, it was,

“Eligible patients had a fasting triglyceride level of 150 to 499 mg per deciliter (1.69 to 5.63 mmol per liter)... because of the intraindividual variability of triglyceride levels, the initial protocol allowed for a 10% lower triglyceride level from the target lower limit, which permitted patients to be enrolled if they had a triglyceride level of at least 135 mg per deciliter (1.52 mmol per liter).”

Seems like more data is needed to say anything about this with confidence.

Hierarchical Tests

For their prespecified hierarchical tests, this is what they found:

  • Rates of Composite Ischemic Endpoints + Rates of Cardiovascular Death: (Vascepa: 4.3% vs. Placebo: 5.2%; hazard ratio, 0.80; 95% CI, 0.66 to 0.98; P=0.03)
  • Rate of Death From Any Cause: Vascepa: 6.7% vs. Placebo: 7.6% (hazard ratio, 0.87; 95% CI, 0.74 to 1.02).
    • Doesn’t seem to be statistically significant, but the coverage of the intervals leans towards an effect.
  • Rates of Adjudicated Sudden Cardiac Death: Vascepa: 1.5% vs. Placebo: 2.1% (hazard ratio, 0.69; 95% CI, 0.50 to 0.96)
  • Rates of Cardiac Arrest: Vascepa: 0.5% vs. Placebo: 1.0% (hazard ratio, 0.52; 95% CI, 0.31 to 0.86)

So most seem to favor the Vascepa group.


What about its safety? A look at the supplementary appendix gives us this:

There don’t seem to be any statistically significant differences for serious treatment-emergent adverse events.

However, among other treatment-emergent adverse events, Peripheral Edema, Constipation, and Atrial Fibrillation were significantly more frequent in the Vascepa group.

There you have it.


The results of this large trial clearly suggest that Vascepa had a notable effect on triglycerides and clinical endpoints. These data along with the results from the JELIS trial should be considered as supporting evidence for the hypothesis that high doses of EPA (not your over-the-counter normal fish oil) may have an effect on cardiovascular disease events (even though JELIS was an open-label study). Hopefully, future meta-analyses make high doses of EPA an inclusion criteria to avoid pooling null results from trials using smaller dosages, where it may be more difficult to see an effect. Currently, there are a few other ongoing trials that are also using high doses of EPA:

“Ongoing trials of moderate-to-high doses of pure EPA ethyl ester will provide further information on the effects of these agents. These trials include the Randomized Trial for Evaluation in Secondary Prevention Efficacy of Combination Therapy–Statin and EPA (RESPECT-EPA; UMIN Clinical Trials Registry number, UMIN000012069), a secondary prevention outcomes trial involving statin-treated patients in Japan, and the Effect of Vascepa on Improving Coronary Atherosclerosis in People with High Triglycerides Taking Statin Therapy (EVAPORATE; ClinicalTrials.gov number, NCT02926027), which is examining changes in coronary plaque over 9 to 18 months.”

Even though REDUCE-IT has been successful, it is only *one* trial, and it's always worth remembering what the history of fish is, even though this trial didn't look at fish oil but rather a pharmaceutical formulation of EPA.

A Brief Timeline of Fish Oil

1970s | A group of explorers noticed that people who consumed fish also had lower incidences of heart disease.

1980s-1990s | Early observational studies saw an inverse relationship between fish consumption and heart disease risk.

1990s | A few randomized trials were conducted to see the effects of fish oil supplementation and CVD outcomes and found benefits.

2002 | The American Heart Association puts out a statement suggesting that fish oil supplementation was beneficial for fatal outcomes.

2002 - 2016 | Several randomized trials and systematic reviews found inconsistent evidence for the effects of fish oil supplementation and CVD outcomes.

2016 | The American Heart Association creates a scientific advisory group to look at all the evidence and concludes that despite the controversy, there was a small benefit from fish oil and it could be worth taking for people who had a history of CVD events.

Early 2018 | Large meta-analysis published in JAMA with 77,000 + participants found no benefit from fish oil supplementation on real clinical endpoints.

Mid - 2018 | Most comprehensive meta-analysis and systematic review with 112,000 participants by Cochrane finds very small benefits to several CVD-related outcomes.

Late 2018 | Announcement by the manufacturer of Vascepa exclaiming that high-doses of fish oil in the REDUCE-IT trial had a large effect (25% reduction) on CVD outcomes.

Subscribe to the Blog

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 90 other subscribers