The Crossover of p-Values into the Results of Placebos
*Ben Locwin, Applied Pharmaceutical Intelligence
Keywords: Healthcare, Pharma, Pharmaceutical, Clinical Trials, Methodology, NNT, Placebo, NHST, Patient
Because all pharmaceutical cinical trials are conducted by NHST, they all must pre-list primary (and subordinate) endpoints against which they test efficacy. The placebo effect is well-documented and very powerful in these trials across a vast array of drug treatment categories. But when hypothesis testing to, say, the alpha = 5% level, if 20 individuals take a particular treatment, one of them is likely to show a positive outcome (that is, one whose outcome rejects the null that it is no better than placebo) just based on random chance. When summed across the entire United States and world, that is a tremendous number of people treated by drugs whose positive outcomes are related only to chance. This is closely related to the concept of Number Needed to Treat (NNT). If the NNT is 20 (that is, the number of patients needed to be treated for one positive outcome), that NNT result could have been solely the result of chance from a trial whose alpha was set at 5%. This is an incredible discussion about our society, drug-taking practices, and large numbers.
Important Dates & Deadlines
- October 9 - 11, 2013