Please consider how you would respond to the following situation as the statistician on a randomized controlled trial.
Participants are randomly assigned to treatment or control group. The baseline assessment occurs just prior to randomization, determines eligibility, and will be used in ANCOVA. Follow-up assessments, including measurement of the primary outcome, are planned in a series of visits starting 3 months after randomization and go on for one year.
Shortly after randomization, a participant says something came up in their personal life and itâs not convenient to be in the study right now, but they would like to participate one year from now. The study will still be recruiting a year from now.
Do you allow them to receive the treatment in one year, postpone the series of follow-up assessments, and just consider the shift in the timeline to be a protocol deviation? How do the following 3 scenarios affect your answer?
The investigators and participants are blinded to group assignment.
Due to the nature of the treatment, it is impossible to blind the investigators and participants. The participant was already informed that they were assigned to the active treatment when they said it is not convenient right now but they would like to participate one year from now.
Now suppose that the control arm is actually a one-year delay in treatment. Participants are randomized either to treatment start now, or treatment start in one year. So eventually all participants will be offered the treatment, but the follow-up assessments are planned to be during the first year for both groups. So essentially, the participant has decided they would rather be in the control group that receives treatment one year from now.
If you do not accommodate the shift in the treatment and follow-up assessment timeline, how do you handle the participant in the analysis? If they are not willing to undergo treatment until next year, but they are willing to undergo the follow-up assessments the first year, do you collect data on the planned timeline and include in ITT analysis (in the treatment group even though they didnât receive the treatment)? Or exclude them and call it modified ITT because they havenât received a single dose?
I am facing scenario #3, but I thought it would be useful to think through simpler scenarios first. Is the shift in the follow-up timeline unacceptable under any circumstances? Or is the lack of blinding the problem? Or that the participant decided to delay treatment, which is essentially the control group?
This is time sensitive, and I would greatly appreciate input. Thank you!
We donât have the ability to do that, because the participants are already on a waitlist to get the treatment and their turn will come up in roughly a year. The study doesnât have the authority to deny them the treatment at that point. So by participating in the study they are getting a 50% chance of skipping to the front of the line and getting the treatment early. Unfortunately a couple of the participants who were randomized to get the treatment early are declining, and this may continue to happen periodically. We need to decide how to handle this.
a) Collect their data on the planned timeline even though they declined the opportunity to get the treatment early, and include in the analysis in the spirit of ITT?
b) Do not bother to collect their data because they arenât getting a single dose on the planned timeline, in the spirit of modified ITT?
c) Allow them to postpone the follow-up assessments until the next year when they get the treatment, and just call it a protocol deviation?
or �
I realize the particulars of this trial are unusual, but I would think that even in a âtypicalâ pharma trial of drug vs placebo, a participant could have a personal issue after randomization that would delay their participation, so as a starting point for thinking this through I was wondering how that would be handled.
I donât see the problem. The RCT is testing the strategy of providing a treatment one year earlier and quantifying whether this strategy results in benefits on the outcome measure. As with any clinical strategy, there will never be 100% adherence. Thus, including these participants in the intention to treat analysis seems prudent. You may then exclude them from a per protocol analysis. Is my thinking overly simplistic?
I figured that was the textbook ITT approach. However the FDA has recognized the complexities with handling intercurrent events. Although FDA guidance does not directly apply to this trial, I would like to better understand this.
We are ultimately interested in estimating the effect of the treatment, not just receiving the treatment one year earlier. We used this design for practical/logistical reasons. To be clear, the plan was to do follow-up assessments on everyone the first year and compare those randomized to treatment vs control. The fact that controls are getting the treatment the following year is not part of the study and not something we have any control over.
What is the possible clinical significance of delaying the treatment 2 years - one year from now plus an additional year if this patient is randomized to the delayed-treatment study group?
Does delaying treatment for 2 years increase the risk to the participant? Does the outcome from this participant bias the analysis of the study results?
The FDA recognizing the problem but hasnât recommended how to solve it. To me the best solution is to abandon the phrase âintercurrent eventâ in favor of âeventâ, leading to analyzing such events as bad outcomes, but perhaps not as bad as the main event efficacy is targeted to.