I clicked on “Millions take fish oil for brain health. A two-year study found no benefit” braced to be let down, and came away asking a pettier question: what was this trial actually built to measure? The answer is spinal fluid. On that question, the one the study was sized and designed around, DHA won by a mile.
The trial, published in eBioMedicine by a Keck Medicine of USC team led by Hussein Naji Yassine, randomized 365 adults aged 55 to 80 who rarely ate fish and carried at least one dementia risk factor. Half took 2,000 mg a day of DHA, the omega-3 fat packed into the membranes of brain cells, for two years, while the other half took placebo. By the end the DHA group did no better on memory and thinking tests, and its scans showed no protection against shrinkage of the hippocampus, the small memory hub that tends to wither early in Alzheimer’s. Yassine told ScienceDaily that “our results do not support fish oil supplements as a preventive measure against Alzheimer’s.”
I take that sentence at face value. I’m much less willing to swallow the headline version, because the memory tests that produced “no benefit” were never the trial’s main event, and the paper says so in plain type.
Sized for a lumbar puncture
The study is called PreventE4 (NCT03613844), and according to the full results paper it had one primary outcome: the six-month change in the ratio of DHA to arachidonic acid, an omega-6 fat, in cerebrospinal fluid, measured by lumbar puncture in 181 of the volunteers. Two-year brain imaging, including hippocampal volume, was secondary. The RBANS cognitive battery, the source of every “no benefit” headline, was listed as exploratory, alongside blood fatty acids and plasma Alzheimer’s biomarkers.
Exploratory doesn’t mean meaningless, and a flat memory result still counts for something. It does mean the trial’s size was set with other questions in mind. The authors calculated that, assuming 30 percent dropout, the study could detect a cognitive effect of 0.35 standard deviations with 80 percent power, and they benchmarked that against an earlier DHA trial, MIDAS, which had reported an effect of 0.3 standard deviations on delayed verbal recall. So the effect a previous DHA study actually found sat just under what this one was sized to catch, and that was before attrition came in worse than planned.
Only 225 of the 365 participants finished, which means 38 percent dropped out, mainly because of COVID-19 pandemic disruptions. The people who left weren’t a random slice: they were more likely to be Hispanic and had less education, lower baseline RBANS scores and lower plasma DHA. The analysis kept every randomized participant and used mixed models that assume missing data are missing at random given what was observed, and the paper reports no imputation and no formal sensitivity analysis for the missing outcomes. Which way that cuts can’t be read off the published numbers, and that uncertainty belongs in any honest reading of “no benefit.”
| Measure | Share |
|---|---|
| of randomized participants did not finish | 38 percent |
The bottle matters too. Headlines say fish oil, but the baseline paper describes algal-derived triglyceride DHA, a single omega-3 grown from algae rather than the EPA-plus-DHA mix in most fish-oil bottles. It’s the same molecule your brain would pull from a piece of salmon, just not the product most readers keep in the cabinet, and everyone in both arms also took a daily vitamin B complex.
The fat showed up where it was sent
On the question it was built to answer, the trial delivered. At six months the CSF DHA-to-arachidonic acid ratio rose by 0.17 in the DHA group and dipped by 0.02 on placebo, a Cohen’s d of 1.64, which is a huge effect for a nutrition study. The rise didn’t detectably differ between carriers and non-carriers of APOE4, the strongest genetic risk factor for late-onset Alzheimer’s (interaction p=0.71). Over two years the omega-3 index in red blood cells climbed 6.18 points on DHA versus 0.21 on placebo. Three capsules with breakfast and two with dinner went down with food, got absorbed through the gut and turned up in the fluid that bathes the brain.
So why would a brain take delivery of that much DHA and show nothing for it on a memory test? That’s the question I kept circling, and the authors offer a hypothesis, labeled as one. An enzyme called calcium-dependent phospholipase A2, or cPLA2, clips fatty acids out of cell membranes, and if it’s pathologically switched on it may break down the incoming DHA and arachidonic acid before they can do their work at the synapse. The trial didn’t test this. I still find it a strange and compelling picture: the delivery truck arrives on schedule, and something at the loading dock shreds the cargo.
Very little decline to prevent
The duller explanation may be the better one. Median baseline MMSE was 29 out of 30, mean age was 66.4, and the authors themselves describe a relatively young, highly educated cohort with preserved cognition and minimal hippocampal atrophy. On the RBANS, both arms improved over two years, up 2.76 points on DHA and 2.67 on placebo, a difference of 0.09. Scores that rise on a repeated test can reflect practice, though the paper can’t show that’s what happened here. Either way, the average volunteer handed the supplement very little slide to stop.
The hippocampus said the same thing. Annual shrinkage on the left side was 0.75 percent on DHA versus 0.63 percent on placebo; on the right it was 0.55 percent versus 0.67 percent. One side leaned toward placebo and the other toward DHA, and neither difference was significant. The authors list the healthy cohort, the single-center design and the dropout among their limitations, and they suggest future trials run 36 months or longer. Safety, at least, was clean: any adverse event occurred in 40.3 percent of the DHA group and 40.9 percent on placebo, and serious adverse events numbered 5 (2.8 percent) versus 13 (7.2 percent).
Subgroup hints that point in opposite directions
Earlier trials bracket the problem without solving it. In 2010, a JAMA trial gave the same 2 g a day of algal DHA for 18 months to 402 people who already had mild-to-moderate Alzheimer’s, and it didn’t slow decline: ADAS-cog scores worsened 7.98 points on DHA and 8.27 on placebo. In an exploratory look at APOE4 non-carriers, decline was 6.23 points on DHA versus 10.11 on placebo, a comparison the authors noted was not adjusted for multiple testing.
In 2024, a three-year trial gave 1.65 g a day of mixed EPA and DHA to 102 adults with an average age of 81. It missed its main white-matter endpoint, but among APOE4 carriers a measure of neuronal integrity declined significantly less on omega-3, a prespecified secondary finding.
Those hints don’t line up: one favored non-carriers who already had Alzheimer’s, the other favored carriers in their 80s, and neither proves anything alone. What stands out to me is who’s missing. Very sick patients sit at one end, sharp 66-year-olds at the other, and PreventE4 wasn’t designed to test the people in between who are just starting to slip.
The next step is a medication
USC’s public message ran further than the trial’s design. In the release, Yassine said “our findings showed that fish oil supplements do not appear to protect brain health,” a line resting on outcomes the protocol labeled exploratory. Then comes the next step: “We are working to develop medications that may help the brain better utilize these nutrients to preserve cognitive function.” That’s a legitimate research direction, and it’s also a sequence I’ve learned to watch for, where the off-the-shelf nutrient gets a public thumbs-down and the follow-up becomes a drug.
The paper’s declaration of interests opens with “The authors declare that they have no conflicts of interest,” then lists co-author Lon S. Schneider’s grants and contracts from NIH and industry partners including Biogen, Eisai, Eli Lilly and Roche/Genentech, plus consulting fees from organizations including AC Immune, Cortexyme, Lundbeck, Merck and Novo Nordisk, all outside the submitted work. The trial itself was funded by the National Institute on Aging and the Alzheimer’s Drug Discovery Foundation, which the paper says had no role in design or analysis, and nothing suggests Schneider’s ties shaped these results. A reader weighing a “fish oil doesn’t work” headline should still know who’s on the author list.
Even the release leans toward food in the end, presenting a Mediterranean-style diet, which supplies omega-3s alongside other nutrients and “has been associated with a lower risk of Alzheimer’s disease,” as a possibly better source than supplements. This trial didn’t test food at all, and it enrolled people precisely because they ate very little fish.
If you’re in your 60s, cognitively sharp, rarely eat fish and take 2 g of DHA a day to guard your memory, this is a solid two-year trial that found no difference on standard memory tests or hippocampal scans, and that deserves weight. It says nothing about five years, about people already starting to decline, or about omega-3s eaten as fish. I wouldn’t buy DHA capsules as Alzheimer’s insurance on this evidence, and I’m keeping the sardines on toast I eat twice a week, because that habit never depended on a capsule trial.
Sources
- eBioMedicine: Yassine et al., CNS target engagement of high-dose DHA supplementation in older adults at risk for dementia (2026)
- Europe PMC: full text of the PreventE4 results paper
- Journal of Prevention of Alzheimer’s Disease: Baseline findings of PreventE4
- JAMA: Quinn et al., DHA supplementation and cognitive decline in Alzheimer disease (2010)
- JAMA Network Open: Shinto et al., omega-3 PUFA for secondary prevention of white matter lesions in older adults (2024)
- ScienceDaily: Millions take fish oil for brain health. A two-year study found no benefit (2026)