If we have a study comparing standard treatment to experimental treatment, is it plausible to assume that there will be no defiers who choose to take the opposite treatment of what they have been assigned, since the patients are not blinded to which treatment they are taking?
I agree with Alvaro. Additionally whether the participants are able to take the other treatment (ie switch treatment groups from control to intervention) will depend on whether they have access to the intervention. This may be possible with physical interventions (eg physiotherapy) or lifestyle modifications, but should be more difficult with most drug treatment trials. Clearly you need to minimize this as far as is possible, monitor what actually happens as opposed to what the results of randomization are and then at the very least analyse by per-protocol and intention to treat. There are some more advanced statistical methods that can be used to try and adjust for differences between 'switchers' and 'non-switchers' (other than just the fact that they switch).
In circumstances of an unusual or toxic drug that does not have wide availability this is generally uncommon eg in oncology trials of new medications. It is thus a reasonable assumption. As alluded to above the biggest issue there is patients withdrawing, or poor accrual to the trial itself.
In circumstances where the intervention is widely available, especially if the perceived wisdom is that it is valuable it can be a real problem, and mechanisms to examine this should be built into the protocol. Examples include the early HIV/AIDS trials where some subjects pooled their medications to try and ensure they all got active drug, and in some of the screening trials where the screen test (eg PSA) was readily available.
- Badges
- Science topic