P values and Multiple Endpoints II: Noxious Placebos in the Population

Abstract

The previous chapter motivated complicated work. Because it has the advantage of fiscal efficiency and buttreses a causality argument, the clinical trial with multiple treatment arms and/or multiple endpoints is here to stay. It is the physician-scientist’s responsibility to incorporate these dual concerns of efficiency and epidemiology within the paradigm of significance testing. The investigator’s community protection responsibilities clearly require the unambiguous interpretation of type I errors in multiple endpoint clinical experiments. The p value is the probability that the therapy efficacy contained in the sample, through the play of chance, does not reflect the truth of no efficacy in the population. The occurrence of this (type I) error leads to the community of patients being exposed to a medication which is not effective but produces side effects. Thus, we may view a p value in therapy trials as the probability that a noxious or poison placebo is deemed (through the play of chance) to be effective. This noxious placebo may spread through the community as the false result is picked up and disseminated through the printed and, visual media, as well as through regulatory channel of the F. D. A. Thus, in a concordantly executed experiment, the smaller the p value, the smaller the probability that a noxious placebo will be deemed effective.