Yes, having a formal and codified method for transfer of participant is a very good idea. In several trials that I was a part of, we had a formal transfer process. This ensured that the patient care was not interrupted, and that the patient would not withdraw from the study (and of course, patients can withdraw consent).
In all transfer processes, the process was initiated by the previous location. The initiation of the process would notify the DCC and the receiving or final location. Prior to the tranfer being complete, a response was required by the receiving location. This ensured that no one was unaware of the process.
For trials with short participation times, this is probably unnecessary. For trials with long interactions, I would think it quite necessary.
I would certainly in this case NOT RERANDOMIZE. The participant is in the clinical trial and not in the clinical trial at a location. The location is generally not a key factor, at least we hope not. Thus, transfer from one location to another should change nothing of the treatment. In particular, the ITT principle means that the participant is treated in the manner into which the randomization puts her. Under no circumstances would I rerandomize. In the matter of the medication, it depends on the status of the medication. Trials that I have participated in have a method of "returning allocated but undispensed medication" to inventory (that is, medication lots that have been assigned to the patient but not given physically can be returned to inventory). On the other hand, if the medication must be retained, there are many secure methods of sending the medication. Finally, unless there is a large cost or a problem of quantity, the medication at Location 1 can be simply discarded and new medication assigned and given at Location 2.
-------------------------------------------
Paul Thompson
Director, Methodology and Data Analysis Center
Sanford Research/USD
-------------------------------------------
Original Message:
Sent: 08-14-2013 15:01
From: Jon Shuster
Subject: randomization question
In the Pediatric Oncology Group (1980-2000 before it meged with the Children's Cancer Group) to become the Children's Oncology Group, we had a formal system for transfers. The fist institution is responsible for the subject initially. There was a formal transfer process where a formal agreement is made that the patient is released by the first institution and the second becomes responsible for the patient. Capitation is affected and Quality Assurance penalties and rewards for the patient go with the accepting institution, regardless of where the bad or good occurred. Refusal to accept the patient, while on active treatment, put the patient off study, as protocol mandated follow-up was not allowed outside the group umbrella.
Good statistical practice is that a patient is on their assigned treatment. Re-randomization under any circumstance should not be done.
Also, I strongly recommend randomizing only at the point of divergence of the treatments. This is especially true if the study is open label as our cancer studies were. If open label, it is unacceptable to start the clock at the time of divergence wen the randomization occurred earlier. One example would be 6 eweeks of chemo, then randomize to radiation (XRT) or not following chemo. Events during these 6 weeks may seem like noise, but they are not. Within the scope of the protocol, the chemo might be pushed harder for those ot destined to get XRT. That could make a difference in the outcomes of the first 6 weeks and beyond that are confounders caused by this poor design. Randomizing to XRT at the 6 week point cures this.
Jon Shuster
-------------------------------------------
Jon Shuster
University of Florida
-------------------------------------------
Original Message:
Sent: 08-14-2013 14:39
From: Lisa Spielman
Subject: randomization question
Got this from a client I'm working with on a multi-center clinical trial. I've removed identifying information but otherwise the exact question is below. I know what I think is the right way to go, but I'm curious to hear what the collective wisdom thinks:
"I'm writing with an interesting randomization question. We have a patient who entered treatment at [Site 1]. At the time he entered treatment, he was randomized to a certain arm of treatment. You may remember, though the randomization occurs right away, the treatment starts at session 3. [Lisa's note: the first 2 sessions are identical across treatment arms].
The patient completed 2 sessions of treatment and was then told he is being moved for work to [a new state, near Site 2]. The decision was made not to start treatment since he only had 1-2 weeks before moving. I met with him along with the clinician at [Site 1] via Skype so we could have a warm handoff and continue to see him in our study. The obvious question is, since he was randomized at [Site 1], do we keep the randomization and have them ship over the medication from their pharmacy to ours? Or do we re-randomize for our site [Site 2]?"
What do you all think?
There is the additional question, since each site has quotas, whether he counts for Site 1 or Site 2, but that's secondary.
Thanks,
Lisa
-------------------------------------------
Lisa A. Spielman, Ph.D.
Consultation Services
-------------------------------------------