The honest answer is the boring one, and the boring answer is well supported. A multivitamin taken by a well-nourished adult changes almost nothing anybody measures. One outcome moved, slightly, in the right direction. The rest are flat lines, and a few single-nutrient products carry a measurable downside that has nothing to do with the combination.
The largest review
In 2022 the US Preventive Services Task Force published an updated evidence report: 84 studies, 739,803 people, restricted to adults without cardiovascular disease, without cancer, and with no known vitamin or mineral deficiency. That restriction is the whole argument in one clause, and it is the same restriction in every good study of the kind.
The pooled result on multivitamins was a lower incidence of any cancer, odds ratio 0.93 (95% CI 0.87 to 0.99), from four trials with 48,859 people. Lung cancer was lower too: 0.75 (0.58 to 0.95), two trials, 36,052 people.
Now the part the headline leaves out. The report also states the absolute risk difference across the adequately powered trials, and it runs from -0.2% to -1.2%. Take a population of a thousand well-nourished adults, give a thousand of them a daily multivitamin for years, and somewhere between two and twelve of them avoid a cancer diagnosis they would otherwise have had. That is the effect, expressed in people.
Everything else was a null
Cardiovascular events, cardiovascular disease, and death: no difference. Vitamin D on its own across 27 trials and 117,082 people was not associated with all-cause mortality (0.96; 0.91 to 1.02), nor with composite cardiovascular events (1.00; 0.95 to 1.05), nor with any cancer incidence (0.98; 0.92 to 1.03). Vitamin E across nine trials and 107,772 people: all-cause mortality 1.02 (0.97 to 1.07). Flat lines all the way down.
An earlier meta-analysis reached the same place from a different direction. Pooling 21 reports of randomised trials of multivitamin-multimineral supplementation, 91,074 people and 8,794 deaths, with a mean age of 62 and a mean supplementation period of 43 months, all-cause mortality came out at a risk ratio of 0.98 (0.94 to 1.02). Vascular causes: 1.01 (0.93 to 1.09). Cancer: 0.96 (0.88 to 1.04). There was a trend towards lower mortality in the primary-prevention trials alone, at 0.94 (0.89 to 1.00), which is exactly the kind of result that reads as a finding and disappears under scrutiny.
The biggest modern trial
COSMOS is the largest randomized trial of a daily multivitamin in this file, and by some distance: 21,442 US adults, 12,666 women aged 65 and over and 8,776 men aged 60 and over, median follow-up 3.6 years. Invasive cancer occurred in 518 participants on the multivitamin and 535 on placebo: hazard ratio 0.97 (0.86 to 1.09), P = .57. All-cause mortality was 0.93 (0.81 to 1.08). The one sub-result that moved was lung cancer, 0.62 (0.42 to 0.92) — a single site sitting inside an overall null, which is exactly why it needs a second trial before anyone leans on it.
Note what that means. The USPSTF cancer signal came from four trials with 48,859 people; COSMOS is one trial with 21,442 and a median of 3.6 years of follow-up, so the two are not interchangeable. What COSMOS does show is an independent, modern, preregistered test of the same outcome in the same kind of population coming back at 0.97 with P = .57. The honest reading is that a small effect on cancer incidence, if it exists, is smaller than the earlier pooled estimate suggested.
The one signal that survived scrutiny
Cognition is the most interesting corner, and the only place where a modern trial found something. In the COSMOS clinic subcohort, 573 participants completed detailed neuropsychological testing at baseline and again after two years. On global cognition the multivitamin group was ahead by 0.06 standard deviations (95% CI -0.003 to 0.13) — that interval includes zero. On episodic memory it was 0.12 SD (0.002 to 0.23), which just clears it.
Pooled across the three COSMOS cognitive substudies, global cognition came out at 0.07 SD (0.03 to 0.11; P = .0009) and episodic memory at 0.06 SD (0.03 to 0.10; P = .0007). The authors’ own conversion is the most useful line in the paper: the magnitude on global cognition is equivalent to reducing cognitive ageing by two years.
Read that carefully, because it is a smaller claim than it sounds. It is not that two years of ageing were removed from the brain. It is that the two groups diverged at a rate equivalent to two years of divergence, over two years, in people already in the age band where decline is measurable.
Where the harm actually is
The downside in this literature is not the multivitamin. It is high-amount single nutrients, and the best-quality trials are the ones that find it.
Pooling 68 randomized trials with 232,606 participants, the overall effect on all-cause mortality was nothing: 1.02 (0.98 to 1.06). But the trials judged low risk of bias told a different story. In 47 of those trials with 180,938 people, antioxidant supplements were associated with higher mortality: 1.05 (1.02 to 1.08). Within that subset, beta carotene was 1.07 (1.02 to 1.11), vitamin A 1.16 (1.10 to 1.24), and vitamin E 1.04 (1.01 to 1.07). Vitamin C and selenium showed nothing.
The 2022 evidence report points the same way from the other end. Beta carotene, with or without vitamin A, was associated with a higher risk of lung cancer (1.20; 1.01 to 1.42) and of cardiovascular mortality (1.10; 1.02 to 1.19). It also listed limited evidence of serious harms attached to specific nutrients: hip fracture for vitamin A, haemorrhagic stroke for vitamin E, kidney stones for vitamin C and calcium. The mechanism behind this is worth respecting: reactive oxygen species do drive ageing and chronic disease, but a large uncontrolled amount of those molecules is a different intervention from a nutrient obtained from food.
When it is a different question
None of the above applies to someone who is actually short. A trial of a supplement in a replete population says nothing about a trial of the same supplement in a deficient one, and the literature on deficiency is a separate body of work with a much better record.
The clearest case is vitamin B12 on a plant-based diet. A 2024 systematic review screened 19 studies into the review and pooled 17 of them on the full panel of B12 biomarkers. Unsupplemented vegan adults had significantly lower serum B12 than omnivores and higher total homocysteine; methylmalonic acid (p = 0.06) and holotranscobalamin (p = 0.09) pointed the same way without clearing significance. The authors are careful that serum B12 on its own is an imperfect marker. Critically, the subgroup analysis found that vegan adults who did take a B12 supplement improved on every one of those biomarkers relative to those who did not.
The second case is vitamin D, covered at length in the northern-winter piece. The mechanism for the shortfall is straightforward and well documented, and the supplementation evidence above was gathered in people who were not short. How to read a batch document is the other half of the practical question, because the label tells you what is in the tablet, not whether you need it.
What the data do not show
- They do not show that a multivitamin extends life. The largest pooled analysis of mortality is a null with a confidence interval straddling 1.
- They do not show that the cancer signal is real. One pooled estimate at 0.93, one far larger trial at 0.97 with P = .57. The absolute difference, where the review could calculate it, was between 0.2% and 1.2%.
- They do not show anything about people who are deficient. Every large trial above excluded them by design.
- They do not show that two years of cognitive ageing were reversed. The COSMOS authors described the effect size as equivalent to a two-year divergence, not a two-year repair.
- They do not show that the single-nutrient harms apply to multivitamins. Beta carotene at the intakes those trials used is not the same as the trace amount in a daily pill, and nobody has run that comparison. A gap, not a reassurance.
The short version
For an adult who eats a varied diet and is not in a group at known risk of shortfall, the expected return on a daily multivitamin is close to nothing, and the trial evidence is unusually blunt about it. The two places worth real attention are a documented deficiency and the specific nutrients where extra intake carries a documented cost. Everything in between is a purchase decision, not a health decision.
