Proper assessment of the response to a test drug needs a treatment-free control group in the experimental design

In preclinical biomedical research, the response elicited by a test drug is commonly evaluated relative to the response produced by its vehicle, with the difference assumed to represent the magnitude of the drug's action. However, there are many reports that the administration of vehicle can provoke a measurable response. For vehicles that are presumed to be inert (e.g. saline), this likely arises from the stress of the administration procedure itself. However, for drugs that are poorly soluble in water, there is extensive evidence that their vehicles can have profound physiological and toxic effects (including neurotoxicity). More complex interactions are also possible, whereby the effect of vehicle administration modifies the profile of the response to the test drug. This article discusses these confounders and argues that proper interpretation of the response to a drug treatment, and successful translation of novel compounds into the clinic, rests on establishing whether the vehicle itself affects the response of interest or modifies the response of the test drug. To investigate these possibilities, the experimental design needs to include a context-matched, treatment-free control group, as well as a vehicle-treated control group. We argue that only if the response of interest in the treatment-free control group is confirmed to be statistically equivalent to that of the vehicle-treated control group, is it appropriate to proceed with further experiments without including treatment-free control samples.