Abstract

In many medical research settings, such as the neonatal intensive care unit, the number of patients who are eligible for a randomised clinical trial is relatively small and recruiting a sufficient number of patients into trials is often difficult. Furthermore, some infants may have already been enrolled into a trial as a fetus. Sequential co-enrolment of patients into more than one trial may offer a solution, yet runs the risk of contaminated results. We consider the situation of two sequential trials and describe requirements for different possible treatments effects (‘estimands’) to be estimated in such situations. These estimands differ regarding the extent to which participation status and treatment status in the previous trial is accounted for. Because of differences in available information about previous trials, analyses may result in estimated effects which differ in terms of interpretation and generalisability, except when in the absence of an interaction between the studied treatments. If co-enrolment cannot be ruled out, researchers should collect information about co-enrolment and treatment status in a previous or concurrent trial and mitigate the trial analysis plan in order to estimate meaningful effects.