High-Dose EPA Linked to 32% Lower Death Hazard After Colorectal Liver Surgery, but Trial Missed Primary Endpoint

Patients who took 4 grams of purified EPA, an omega-3 fatty acid, every day around surgery for colorectal cancer that had spread to the liver were more likely to be alive at five years than patients given a placebo. The trial's primary endpoint, time until the cancer progressed, showed no significant difference between the two groups. Both facts come from the same study, the phase 3 EMT2 trial, presented at the ESMO Congress 2026.

What is actually known about this setting is limited. Surgery to remove liver metastases offers a chance of long-term survival, but the cancer often returns. Preclinical and observational work had suggested EPA might act against colorectal cancer, and an earlier phase 2 trial hinted at a survival benefit. That phase 2 trial missed its own primary endpoint, tumor cell proliferation, and the hint came from a prespecified exploratory analysis.

A trial built around progression, not survival

EMT2 enrolled 418 patients at 13 hepatobiliary surgery centers in the UK, all scheduled for liver resection with curative intent. Patients were randomly assigned to 4 grams a day of icosapent ethyl, a prescription form of EPA, or to a mineral oil placebo. Treatment began at least one day before surgery and was planned to run two to four years, including after any progression. About 20% of each group received adjuvant chemotherapy.

The design was published in BMJ Open in 2023. It named progression-free survival as the primary endpoint and overall survival as the key secondary one. The team planned 448 patients to detect a hazard ratio of 0.7 for progression, and assumed the placebo group would have a median of 21 months without progression. The observed medians were 13.5 months with EPA and 12.9 months with placebo (hazard ratio 0.89, 95% confidence interval 0.70 to 1.13, p = 0.33). The control arm progressed sooner than the plan expected, and the EPA arm did not pull away from it.

The survival numbers

Estimated five-year overall survival was 62.0% with EPA and 45.1% with placebo. The hazard of death was 32% lower with EPA (hazard ratio 0.68, 95% confidence interval 0.48 to 0.96, p = 0.03). Median overall survival had not been reached with EPA, against 59.0 months with placebo. A per-protocol analysis gave a hazard ratio of 0.60 (0.36 to 0.99).

Diabetes was twice as common at baseline in the EPA group (16% versus 8%), and the survival result held after adjusting for it. The interval's upper bound, 0.96, sits close to no effect. EPA caused no excess adverse events, bleeding or atrial fibrillation compared with placebo, and quality of life did not worsen.

Why a split result needs care

Professor Mark Hull of the Leeds Institute of Medical Research, the lead author, called the survival benefit clinically meaningful if other studies confirm it. He said the effect might reflect how the body responds to cancer, including cachexia, the weight loss and fatigue that can accompany it. He also noted that longer survival without a progression gain has been seen in cancer immunotherapy trials, and that more research is needed on whether EPA works through the immune system.

“The results are definitely intriguing, from both a biological and a clinical point of view.”

Cremolini added that the mismatch between the two endpoints requires careful interpretation. Progression-free survival is a less established stand-in for survival after liver metastasis surgery than after surgery for early-stage bowel cancer, she said. She wanted to see what treatments patients received after their cancer returned, and where it spread.

The placebo is a second open question. The 2023 protocol notes that in the REDUCE-IT heart trial, 4 grams a day of mineral oil was associated with small rises in inflammatory and lipid markers, which its authors estimated at no more than 3% added cardiovascular risk. Whether that could widen the gap in a cancer trial is not addressed in the material reviewed here.

This account rests on the ESMO press summary as republished by News-Medical, the 2023 protocol, and an Oncodaily preview of the ESMO program (abstract LBA30). The abstract itself was not reachable, and no journal paper was found. Base-rate context is also missing: none of these sources says how often a survival gain without a progression gain holds up in later trials.

The bottom line

The trial tested one purified prescription product in one surgical population, not fish oil supplements and not other cancers. Whether the survival signal is real will depend on replication in a second trial, and on the post-recurrence treatment data Cremolini asked for.