October 19, 2012
Multivitamins May Slightly Decrease Cancer Risk in Older Men
My regular readers already know my deep skepticism about the benefit of vitamin supplements in well-nourished healthy people. My post last year summarized the available studies about many vitamins and ran under the headline “A Reminder to Dump Your Multivitamin”. Thus far, large randomized studies about vitamins have tended to study a single vitamin at a time usually in fairly high doses. That makes a lot of sense. If you want to figure out if vitamin E prevents heart attacks, design a trial in which lots of people get either vitamin E or placebo. (It doesn’t prevent heart attacks.)
This week the Journal of the American Medical Association published the results of the first large randomized trial to study the effects of a multivitamin. The trial began in 1997 and enrolled 14,641 male US physicians aged 50 years or older. It randomized the men to a daily multivitamin or a placebo. The multivitamin used was the same formulation as the formulation of Centrum Silver at the time, which has 30 different vitamins and minerals. The men were followed for an average of 11 years. New diagnoses of cancer (except non-melanoma skin cancer) and deaths were counted for both groups.
It’s important to note that this was part of a larger study to test the effects of a multivitamin on cancer, cardiovascular disease, eye disease, and cognitive decline. The current results publish only the cancer findings. The effects on cardiovascular disease, eye disease, and cognitive decline will be published separately.
Especially to a skeptic like me, the results were interesting. The multivitamin group had slightly fewer new cancer diagnoses than the placebo group. The media is reporting the decrease as an 8% lower risk, which is true, but I don’t think gives us a clear mental image of the magnitude of the benefit. Imagine two groups each of 769 older men. If one group took the multivitamin for a year and the other group took the placebo for a year, the vitamin group would have 13 new cases of cancer and the placebo group would have 14. So for every 769 men taking a multivitamin for a year one new diagnosis of cancer is prevented.
Total mortality and mortality due to cancer were the same in both groups. The study was too small to detect differences in occurrence rates of specific types of cancers (i.e. lung cancer, colon cancer, etc.).
Two objections demand to be raised. First, why would you test 30 different vitamins and minerals together? It’s conceivable that some of the ingredients have synergistic effects and work differently in combination than they do individually, but it seems that the scientific method suggests the opposite approach. First figure out the effects of each ingredient, then test some combination. The current results leave us completely mystified about which of the ingredients (or combination of ingredients) is responsible for the effect. The authors state that the trial studied the effects of multivitamins because so many people take them, but I am left wondering whether the trial was designed to raise our understanding or increase Centrum Silver sales. (Pfizer which sells Centrum Silver, supplied the placebo and the multivitamin for the trial, though it did not design or conduct the study. The study was funded by the NIH.)
The other objection is that the difference between the cancer incidences in the two groups barely rose to the level of statistical significance. Bear with me while I delve into the statistics briefly. Statisticians use a measure called a p value to decide whether a measured difference between two groups might have been due to chance alone and not due to the intervention being studied. The p value answers this question: If the intervention being studied is actually irrelevant (in this instance, if the multivitamin and the placebo have equivalent health effects) how likely is it that the two groups would show such different numbers (in cancer incidence, in this case) by chance alone? By convention, any p value less than 0.05 is considered statistically significant. That is an arbitrary choice agreed on by statisticians, but there is no math or science behind that number. We just need some cut-off point, so we decided that 0.05 is statistically significant and 0.06 is not. That still means that for all the randomized trials we do, if the null hypothesis is true (meaning, if the medicine is just like the placebo) 1 time in 20 we’ll get a result that shows that the two are different in a statistically significant way, meaning we’ll get a false result by chance. The p value in this trial for the difference in cancer incidence for the two groups was 0.04. Yes, that’s statistically significant, but just barely.
So what can we conclude?
First of all, this teaches us nothing about multivitamins in women or in men under 50.
Second, the design of the trial is very solid. It was randomized and blinded. So we should take the result at face value – that for men over 50 taking a multivitamin daily makes a small decrease in the risk of cancer (but not in dying of cancer). This effect is much smaller than quitting smoking. So it makes a lot of sense to quit smoking before you try to figure out whether or not to take a multivitamin.
My suggestion for now is to wait for the rest of the results from the trial. For example, if the cardiovascular results are even slightly negative, suggesting that the multivitamin increases stroke or heart attack risk by a small amount, that may overwhelm the cancer benefit.
For the time being I’m still not recommending a multivitamin for healthy adults, or taking one myself.
Multivitamin Cuts Cancer Risk, Large Study Finds (Wall Street Journal)
Important legal mumbo jumbo: