trial involving observations in which some multiples may occur

[Thompson,Paul]

Folks:

I am having a discussion with a vascular surgeon who is doing a trial involving a vascular procedure. Most of the observations are on single individuals. In one case, we have a case where he did both sides, and wishes to retain both observations in the trial. I am not very enthused about this, since although the endpoint involves a physical evaluation of the artery, there are a number of aspects which are greatly complicated by multiple observations.

 

1)      Are there analysis approaches which would allow me to have (arguably) 48 independent observations from different persons and 2 from the same person?

2)      Thoughts on adverse events, etc?

3)      Any guidelines to discussions on this issue?

 

Paul A. Thompson

 

Director, Methodology and Data Analysis Center

Sanford Research/USD

Sanford Health

2301 E 60th St N

Sioux Falls, SD 57104

605-312-6462

 

----------------------------------------------------------------------- Confidentiality Notice: This e-mail message, including any attachments, is for the sole use of the intended recipient(s) and may contain privileged and confidential information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply e-mail and destroy all copies of the original message.

[Marc Schwartz]

On Apr 24, 2012, at 2:16 PM, Thompson,Paul wrote:

> Folks:
>
> I am having a discussion with a vascular surgeon who is doing a trial involving a vascular procedure. Most of the observations are on single individuals. In one case, we have a case where he did both sides, and wishes to retain both observations in the trial. I am not very enthused about this, since although the endpoint involves a physical evaluation of the artery, there are a number of aspects which are greatly complicated by multiple observations.
>
> 1) Are there analysis approaches which would allow me to have (arguably) 48 independent observations from different persons and 2 from the same person?
> 2) Thoughts on adverse events, etc?
> 3) Any guidelines to discussions on this issue?
>
> Paul A. Thompson

Paul,

My first question would be, what does the protocol specify regarding the surgical procedure?.

Does it specify that the procedure is only to be performed on one side, perhaps based upon certain inclusion/exclusion requirements or is it silent?

Were both sides done during the same surgical procedure or on separate dates/times?

If the protocol specifies that only one procedure is to be done, then the second is also a protocol deviation, which might result in the subject being excluded from a per protocol analysis if one is to be performed. The protocol might even define that the subject is to be withdrawn in that scenario.

If the procedures were done on different days/times, an argument can be made that the first procedure would be included in the study under ITT, however the second side procedure would be labelled as an adverse event, or even an SAE if the subject was admitted for it. From a DSMB/DMC perspective, that's more than likely how it would be adjudicated.

The second procedure on the subject is not independent of the first and presuming that for example, general anesthesia is used, the subject is at a greater risk for peri-operative AE's, more so if the time interval between the two procedures is relatively short and there are other relevant co-morbidities. Even if they were done during the same session under general, you have a greater risk for anesthesia and other complications since the subject is under longer.

That all can bias your safety profile for the procedure. It might only be one subject to date, but what would you do if a larger proportion of the subjects end up in a similar situation? Is there a likelihood that if you followed the subjects long enough, they would all end up with the same procedure on the opposite side?

Food for thought.

Regards,

Marc Schwartz

[Doug Altman]

These papers may be helpful. Of course it is best to consider this possibility when developing the protocol. (As it was a trial I assume there was one - but non trials studies may well not have one.) Also, I'm not sure whether the methods will work, or even make much difference, with only one within-patient pair. As there is only one I would choose one side at random. A more subtle point is whether all the other patient s had unilateral disease or whether some had bilateral disease but the surgeon chose which side to operate on.

Doug

How should randomised trials including multiple pregnancies be analysed?
Gates S, Brocklehurst P.
BJOG. 2004 Mar;111(3):213-9.

Eyes or patients? Traps for the unwary in the statistical analysis of ophthalmological studies.
Newcombe RG, Duff GR.
Br J Ophthalmol. 1987 Sep;71(9):645-6.

plus numerous subsequent papers in ophthalmology, most recently this one:

Karakosta A, Vassilaki M, Plainis S, Elfadl NH, Tsilimbaris M, Moschandreas J.
Choice of analytic approach for eye-specific outcomes: one eye or two?
Am J Ophthalmol 2012;153:571-579.

--
To post a new thread to MedStats, send email to MedS...@googlegroups.com .
MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules