Site Level Enrollment Caps in a Multi-Center Study

[Marc Schwartz]

Greetings all,

ICH-E9 contains language, in the section on multi-center studies, regarding variations in site level sample sizes. This largely relates to site/treatment effect interactions and assessments of site level bias affecting findings from the aggregate data set. This context, of course, is largely in the domain of powered, randomized, confirmatory trials.

I am aware of a "rule of thumb", which is to take the mean number of enrollments per site (total study sample size divided by the number of sites) and multiply that figure by three (3) to five (5) to get a possible site level cap.

I have not been able to find any language elsewhere, in study design books that I have or online, that provides a clear framework for determining reasonable site level enrollment caps that would constrain the influence of high volume sites on the study and the differential between high and low volume sites.

Conceivably, the same concept would apply to regional level influences in a multi-national study.

In regulatory submissions, the notion would seem to be that if the FDA (or other relevant body) approves the study design, you are good to go. If not, adjust as needed and then obtain approvals.

For late phase and observational studies, there does not seem to be a clear framework to influence study designs.

As we all know, study enrollment progression can be one of the most significant challenges for a successful study. While site level enrollment caps can constrain the high end, the low end enrolling sites are still a challenge and I believe, largely gets to site selection characteristics. It is not uncommon to increase the number of sites when enrollment is materially behind projections. In some cases, I have seen enrollment caps increased to enable sites that can, bring in more patients to meet powered sample size requirements, rather than ending up with an underpowered study.

I found one recent paper in JACC:

Relationship Between Clinical Trial Site Enrollment With Participant Characteristics, Protocol Completion, and Outcomes
Insights From the EVEREST (Efficacy of Vasopressin Antagonism in Heart Failure: Outcome Study with Tolvaptan) Trial
Butler et al
J Am Coll Cardiol. 2013;61(5):571-579. doi:10.1016/j.jacc.2012.10.025
http://content.onlinejacc.org/article.aspx?articleid=1484450

which on a post hoc basis and within a narrow sample, assessed a trial for such influences. The study raises many questions regarding site and regional level influences on aggregate findings. Some of the points raised in the paper may be correlated to variances in protocol compliance and perhaps even patient selection bias, rather than representing actual characteristics and outcomes differences that may be associated to site volume alone.

Any guidelines, references and thoughts the group members have on these issues would be most appreciated.

Thanks and regards,

Marc Schwartz

[John Whittington]

At 07:54 28/01/2015 -0600, Marc Schwartz wrote:
>In regulatory submissions, the notion would seem to be that if the FDA (or
>other relevant body) approves the study design, you are good to go. If

>not, adjust as needed and then obtain approvals. ... For late phase and

>observational studies, there does not seem to be a clear framework to
>influence study designs.

As I'm sure you know, even for regulatory submissions, there are
potentially significant transatlantic differences. FDA scientific advice
is largely 'binding', in as much as if they 'approve' a study design, they
are more-or-less bound to 'accept' it as a satisfactory part of a
regulatory submission. In Europe, that is generally not the case (they will
usually not commit themselves as to the regulatory adequacy of a study
design), so the fact that a trial has been 'approved' (in the sense of
being allowed to happen) in Europe does not guarantee that it will be
regarded as satisfactory as part of a subsequent regulatory submission.

>As we all know, study enrollment progression can be one of the most
>significant challenges for a successful study. While site level enrollment
>caps can constrain the high end, the low end enrolling sites are still a
>challenge and I believe, largely gets to site selection characteristics.
>It is not uncommon to increase the number of sites when enrollment is
>materially behind projections. In some cases, I have seen enrollment caps
>increased to enable sites that can, bring in more patients to meet powered
>sample size requirements, rather than ending up with an underpowered study.

As you imply, this often does really come down to practicalities and
pragmatism - since an adequately powered study with markedly unequal
distribution of subjects between sites is probably 'better' than an
underpowered study!

Kind Regards,


John

----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: Joh...@mediscience.co.uk
Buckingham MK18 4EL, UK
----------------------------------------------------------------

[Marc Schwartz]

> On Jan 28, 2015, at 8:24 AM, John Whittington <Joh...@mediscience.co.uk> wrote:
>
> At 07:54 28/01/2015 -0600, Marc Schwartz wrote:
>> In regulatory submissions, the notion would seem to be that if the FDA (or other relevant body) approves the study design, you are good to go. If not, adjust as needed and then obtain approvals. ... For late phase and observational studies, there does not seem to be a clear framework to influence study designs.
>
> As I'm sure you know, even for regulatory submissions, there are potentially significant transatlantic differences. FDA scientific advice is largely 'binding', in as much as if they 'approve' a study design, they are more-or-less bound to 'accept' it as a satisfactory part of a regulatory submission. In Europe, that is generally not the case (they will usually not commit themselves as to the regulatory adequacy of a study design), so the fact that a trial has been 'approved' (in the sense of being allowed to happen) in Europe does not guarantee that it will be regarded as satisfactory as part of a subsequent regulatory submission.
>
>> As we all know, study enrollment progression can be one of the most significant challenges for a successful study. While site level enrollment caps can constrain the high end, the low end enrolling sites are still a challenge and I believe, largely gets to site selection characteristics. It is not uncommon to increase the number of sites when enrollment is materially behind projections. In some cases, I have seen enrollment caps increased to enable sites that can, bring in more patients to meet powered sample size requirements, rather than ending up with an underpowered study.
>
> As you imply, this often does really come down to practicalities and pragmatism - since an adequately powered study with markedly unequal distribution of subjects between sites is probably 'better' than an underpowered study!
>
> Kind Regards,
>
>
> John

Thanks for taking the time to reply John and I appreciate your comments.

Best regards,

Marc