Humbled again by the second tail

In statistical testing of any intervention trial it is best to consider that either benefit or risk might result from the intervention.  With any set of possible outcomes from a trial, there is a distribution of possible values ranging from low-likelihood extreme benefits to low-likelihood extreme hazards, with the more likely outcomes of either no effect or only modest effects residing in the middle. This distribution of all possible theoretical outcomes is a “bell-shaped curve” that is shaped like the Liberty Bell, with tapering edges on either side representing the two “tails” of the distribution of extreme theoretical outcomes. 

When we then conduct the trial and observe a finding, the statistical likelihood of that finding can be measured relative to that finding’s place on the bell-shaped curve.  Sometimes when we carry out studies we cannot imagine that the outcomes could be anything other than beneficial.  In the statistical testing game we play, if an adverse effect is literally impossible, then we can conduct the statistical test on only one end (one “tail”) of the distribution of possibilities.  However, if it is even remotely possible that an adverse effect might be seen, we should do (and almost always in fact do) testing of both possibilities (“two tailed” testing).

There are many examples of human trials that produced surprising findings.  Some classic examples include betacarotene for lung cancer prevention, estrogen-progesterone replacement therapy for heart disease prevention, and folic acid for colon polyp prevention.   We now know that taking betacarotene pills actually causes lung cancer, taking estrogen-progesterone pills actually causes heart disease, and taking folic acid pills actually increases growth of polyps.   Woops.

In JAMA last week we were humbled once again by the release of findings from the long-term follow-up of the SELECT trial, a study of over 30,000 men who took vitamin E and/or selenium pills vs placebo pills for over 7 years to prevent prostate cancer. There were 17% more prostate cancers in the group taking vitamin E, a finding that was not likely to have occurred by chance alone (which is what the statistical test tells us). 

The tail of harm is wagging more than the tail of benefit across vitamin pill studies.  We now have more evidence of harm from nutritional supplements than for any benefit we might have hoped for.  It is time that the Congress take a hard look at questions about marketing and harm related to the vitamin pill industry.  Taking vitamin pills might not be such a great idea.  I am increasingly glad I never started.