Can the Humble Multivitamin Prevent Dementia?

Probably not, but it made for a great headline.

I admit that I did not have on my bingo card for this Tuesday morning, “randomized controlled trial shows that multivitamin use reduced cognitive impairment by 60% over 3 years!” But here it was:

We’re used to tantalizingly-reported observational studies making all sorts of similar claims, of course. Yes, “people who eat chili peppers live longer,” but such assertions usually can be brushed off as plagued by confounding biases. (In the case of chili peppers, those of us who eat them are simply of stronger stock than those with delicate palates; that’s a scientific fact.)

Here, however, was an actual randomized, placebo-controlled trial, of decent size (over 2000 participants), over an adequate time period (3 years), with a fairly low drop-out rate (92% did at least one follow-up exam), and published in a peer-reviewed journal. Since the study was initially designed to investigate the effect of cocoa on cognition, and even had a bit of funding from Mars, Inc., it seems unlikely that researchers were in the back pocket of Pfizer, then still the maker of the Centrum Silver multivitamin used in the study. The study lead author seemed genuinely surprised — and more than a little excited — at the truly shocking results of their study:

It’s shocking because cognitive loss is right there with cancer in terms of “problems the pharmaceutical industry would like to solve,” since it would rightfully create the sort of lucrative drug market that would make Viagra seem like a dud in comparison. However, despite decades of trying, Big Pharma is yet to come up with a medication for which I have anything but the most tepid enthusiasm, including its most recent over-priced failure, Aduhelm.

So… a vitamin pill? 60%? It seems too good to be true — and I think it is.

For starters, this is not the first rodeo for multivitamins and cognition. A huge study of a cohort near and dear to my heart — male physicians — looked at this same question with similar techniques, and the net effect of a daily multi was zilch. Of course, male physicians might eat better, have different cognitive coping skills, etc. than the average person, so it was not unreasonable to test the hypothesis again on a more representative sample of American seniors, a decade later, with a slightly different formulation of Centrum Silver. Fine. But the odds of a “shocking” success would have to be considered quite low.

The crux of the matter likely boils down to how hard it is to accurately assess cognitive ability. For a quick office test, I favor the Montreal Cognitive Assessment, or MoCA. This is the cognitive test then-President Trump claimed to have “aced.” It tests several different modes of reasoning and recall, among them naming animals, drawing a cube, explaining word associations, and remembering a set of five words. This week I gave a patient his 4th MoCA in 3 years; the first scored a 25/30 (suggesting mild cognitive impairment), the second a 21/30 (possible mild Alzheimers), the third a 26/30 (considered a normal score), and now a 23/30 (back to mild cognitive impairment). It’s hard to measure cognition, and it can be a moving target since we all have good days and bad days.

For the study, they went with an hours long, telephone-based set of multiple tests. Outlined below, I think it is correctly termed a “battery”:

The authors opted not to provide raw scores — which generally makes me a little nervous — but rather to scale the “global cognition composite” from the above tests as a standard deviation from the mean, rather like an IQ test. The average composite score improved by a little over a tenth of a standard deviation in the multivitamin group at year 1 and year 2, and then plateaued for year 3. (To use our IQ test analogy, since every 15 points from the mean is one standard deviation, that would be about 4 IQ points after 2 years.) Since we are all presumed to be getting more scatter-brained each year, going up instead of down means the multivitamin really worked, right?

Well, there’s a catch. The placebo really worked, too.

It worked so well in the cocoa extract group that it obliterated any statistical significance from the trial findings. Mars, Inc. was no doubt displeased with these anti-aging placebo pills! However, the placebo cohort did quite well against Father Time in the multivitamin arms, as well. Yes, the difference between active and placebo groups was statistically significant, but barely so; the sad truth is that I need to squint to see the gap between the top red and the bottom blue lines of the confidence intervals in year 3 of the “positive” multivitamin trial.

Why would placebo participants improve their cognition for two years, and then stop improving, just like the vitamin group? And why did people who were already taking a daily vitamin pill prior to the study “benefit” just as much as those who were not?

Your guess is as good as mine, but the most likely explanation (also mentioned by the authors) is that there is some learning effect from taking these tests the second and third times that wanes by the fourth. It’s also possible that the timing of when participants dropped out of the trial, and how many of the year 3 test takers had missed a year or two, could have contributed to the plateau. One thing I can say with confidence: it’s extremely hard to dream up a mechanism by which a multivitamin could improve cognitive function for 2 years, then stop improving it for the next year, all the while exactly mirroring the performance of the placebo group, which would make more sense than a learning effect from repeating the same test year after year.

Really, though, the most disappointing aspect of the study discussion comes at the end; this is the source of the “slow decline by 60%” part of the headlines:

In case you think this is gobbledygook… you are correct. The authors apparently dropped a line from the initial composite scores of the youngest trial participants to the oldest, and found that the average score fell by 0.045 standard deviations with each year of aging. Note: this is not what actually happened in the study! The placebo group scores went up almost 0.1 standard deviations per year, especially in the cocoa extract placebo cohort. Based on the above graphs, it is criminal statistical negligence to assert in any way that the baseline differences between age groups have some sort of linear relationship to the differences between individuals once they start repeating the same cognitive test, year after year.

So, no, you most certainly cannot extrapolate from this study that your brain will age 60% slower if you start taking a daily multi. I wish neurologic health were this simple.

There is surely some modest proportion of folks, especially amongst the elderly, who don’t absorb B-vitamins in their gut very efficiently and might experience neurologic benefits from B vitamin supplementation, but quite possibly they would need more than what is found in a Centrum Silver. The many other vitamins and minerals that have been studied for brain health — from vitamin D to K to magnesium and many more — could have some theoretical value with the amounts packed into a vitamin pill. It’s just unlikely to have much of an effect size for a given individual, and especially not on the collective scale needed to produce robust results in a trial.

While there’s not much harm in popping a vitamin pill every day, the loss is in not doing things that could really matter. I’d much rather my patients prioritize getting 8 hours of sleep every night, eat a broad palate of unprocessed foods, avoid sweetened drinks, move regularly, take up challenging hobbies, and be active members of their community. Yeah, I’m aware the quality of the science demonstrating that any of those activities really stave off dementia is poor, too; but they’re going to feel better regardless of what’s happening to their cognitive health.

I’m 60% sure that I don’t need a randomized controlled trial to know that.