Placebo nonresponders: An experimental investigation on their unreliability over time

In order to disentangle the effects of drugs from placebo responses, several approaches have been used, such as a placebo run-in phase in which only placebo nonresponders, or poor responders, are considered for further randomization to either placebo or active treatment. This study is aimed at investigating the variability of placebo nonresponders obtained through the classical placebo run-in paradigm (group RUN) and through mismatch conditioning (group MIS), as done in our previous study. To do this, we simulated a real clinical trial in the laboratory, in which the placebo responders of both groups were discarded and the remaining nonresponders of both groups RUN and MIS were randomized to either continuing on placebo (groups RUN-P and MIS-P, respectively) or receiving topical 0.5% lidocaine (groups RUN-L and MIS-L, respectively) applied to the skin. By measuring pain thresholds, we found that the placebo nonresponders selected on the first day of the experiment showed different responses on the following day in both group RUN and MIS. This led to no significant differences between placebo and lidocaine in both groups. Although this is an experimental laboratory situation far from the clinical trial setting, these findings show that placebo nonresponders are not necessarily constant over time, both when a placebo run-in protocol is used and when nonresponders are created in the laboratory. This questions the reliability of selecting placebo nonresponders as a methodological approach in clinical research. Therefore, we suggest reconsidering the validity and usefulness of placebo run-in protocols. PERSPECTIVE: Placebo nonresponders are sometime selected for further randomization to either placebo or active treatment. In this experimental study, which is a laboratory simulation of a clinical trial, we found that placebo nonresponders vary from day to day, thus questioning their validity as a methodological approach in clinical research.