Steven Kaas, 5th place

Abstract

Some research suggests selenium supplementation may reduce cancer risk. The evidence from this research is circumstantial and may well not represent a real effect. However imperfect the evidence of cancer risk reduction, it is stronger than the evidence for selenium supplementation being risky at the level of one supplement every day; in combination with the low cost (in terms of money and inconvenience) of taking at least the recommended daily allowance of selenium separately or as part of a multivitamin, this argues for supplementing with selenium.

Recommendations List

Patients (especially male or with reason to think their selenium intake is low) are advised to take 100 or 200 micrograms of selenium per day, separately or in a multivitamin.

Report Body

The human body requires a number of different minerals to function. Generally, outside some given range, clear deleterious effects appear both from having too much and too little of each mineral.

However, the majority of people obtain enough of each mineral from their diet to make it non-obvious whether their level is too high, too low, or just right. Risks for them either appear only statistically and in the longer term, like increased risk of cancer, heart disease, or stroke; or are difficult to measure, like minor effects on cognition.

Research addressing such benefits or risks requires large-scale studies. Many possible biases exist. There exists correspondingly little in the way of easily interpreted evidence.

This is the situation for each of a number of minerals; this report will focus on specifically on the element of selenium, which has some of the more compelling evidence of beneficial effects. Specifically, some studies have found an association between selenium exposure and reduced levels of cancer. Furthermore, the points made here may transfer to other minerals.

Selenium is a trace mineral that aids antioxidant functioning by being part of antioxidant proteins called selenoproteins. Making a rational decision whether to take selenium supplements requires a weighing of costs and benefits. The costs, in this case, include the cost of taking the supplement, as well as the risks of having too high levels of selenium in one’s body. The benefits come from whatever advantages there are from not having levels be too low.

The costs of taking selenium supplements are very low. The RDA for most adults is 55 micrograms. A year’s supply of selenium pills containing a few times this RDA can be bought for $10-20. Many multivitamins contain similar amounts of selenium; a year’s supply of multivitamins costs a few tens of dollars. The time costs of taking a pill each day (or at least most days) are also small.

To make a rough calculation putting this on the same footing as the costs of cancer, first consider the base rate of cancer deaths in the USA. According to 1, the probability of an average American dying from cancer in any given year is roughly 0.2%. A relative risk of 0.9, then, translates to 0.02% less chance per year of dying from cancer; a relative risk of 0.99 translates to 0.002%. There are different estimates of the value of a statistical life, but such estimates generally number in the millions of dollars. (These numbers come from preferences revealed by people’s risk-taking behavior, and there is of course nothing to stop anyone from valuing a chance of death at some different number.) For example, assuming $5,000,000 as the value of a life, achieving a relative risk of 0.9 gives an expected gain of $1000, and achieving a relative risk of 0.99 still gives an expected gain of $100.

Obviously, there are a great many caveats to such a calculation. But it does appear that, unless our best estimate (all things considered) of the relative risk from selenium supplementation is very close to 1, the cost of taking supplements is negligible relative to the health effects, whichever way these effects go on net.

In obtaining this best estimate, one should start out with a fairly wide range of uncertainty. While there is a large probability that any given intervention does nothing, there is some chance that the role selenium is known to play in human biochemistry translates into a genuinely beneficial or harmful effect.

One can then, at least in theory, update this prior uncertainty using the formalism of Bayesian probability theory. To do an exact calculation here would give a false suggestion of precision, but there are a number of qualitative points to be made.

Note that one can combine a weighted average of estimates of relative risk into a best estimate. Suppose one is 20% certain that a given line of evidence is valid; suppose that if the line of evidence is valid, the best estimate for the expected number of deaths in a population taking selenium supplements is 50; suppose that if the line of evidence is not valid, it is 100. Then one can combine this into a best estimate of .2 * 50 + .8 * 100 = 90.

Note also that we have some idea of how a confidence interval translates to a likelihood ratio. For example, for a normal distribution, a 95% confidence interval corresponds more or less to the range between -2 and 2 standard deviations from the mean. The normal distribution function at either end of this range is e^2 times lower than at the mean, so if the distribution is normal, the end points are disfavored by a likelihood ratio in the rough neighborhood of 7.

The available evidence on selenium is summarized in [2]. In particular, the risk estimate from observational studies that 2 gives for cancer incidence is 0.69 (95% confidence interval 0.53 to 0.91) and the risk estimate for cancer death is 0.55 (confidence interval 0.36 to 0.83). This represents the ratio of risk between the group with the highest selenium exposure and the risk with the lowest selenium exposure.

These are strong-looking effects, but since these are observational studies, there may be some common factor causing both selenium exposure and the absence of cancer and/or the absence of death from cancer. Other reasons for doubt are given in [2]. For example, studies of specific cancers, including randomized controlled trials, give far weaker effects, if any, though the estimates still tend to be on the lower side of 1 more than the higher side.

The key point, however, is that the lack of sufficiently clear data to come to a scientific conclusion does not imply that supplementation is worse than no supplementation on an individual level. Because the cost of supplementation is low, as long as good effects are more likely than bad, one might as well supplement just in case the effect turns out to be real.

References

[1] Cancer Statistics, 2011. Siegel R, Ward E, Brawley O, Jemal A. CA: a Cancer Journal for Clinicians. 2011

[2] Selenium for Preventing Cancer (Review). Dennert G, Zwahlen M, Brinkman M, Vinceti M, Zeegers MPA, Horneber M. Journal of Evidence‐Based Medicine. 2011