marionkcampbell.bsky.social
Trialist; Statistician; Professor of Health Services Research; University of Aberdeen. Hebridean. Views my own.
https://www.abdn.ac.uk/people/m.k.campbell
81 posts
745 followers
155 following
Regular Contributor
Active Commenter
comment in response to
post
Whilst there remains no one-size-fits-all answer as to how to deal with multiple testing, what is clear is that your rationale must be justified and clearly reported. 7/7
comment in response to
post
Richard Hooper (@richardhooper.bsky.social) also recently published a commentary on the pros & cons of adjusting for multiple outcomes providing useful discussion of different rationales for adjusting/not and pointers for further reading 6/7
www.jclinepi.com/article/S089...
comment in response to
post
The main elements highlighted in the review include the need to pre-specify outcomes and analysis, methods of multiplicity adjustment, power and sample size implications, multiplicity & secondary outcomes, correlation between outcomes, and different applications 5/7
comment in response to
post
This week Hussein et al published a useful scoping review highlighting the range of guidance and methods available on the topic (including the regulatory expectations) but noted that agreed guidance is lacking in many areas 4/7
www.jclinepi.com/article/S089...
comment in response to
post
Several strategies have been put forward to deal with this - I discussed some of these in a previous MethodologyMonday post elsewhere 3/7
x.com/marionkcampb...
comment in response to
post
Multiple testing is when you want to carry out tests on multiple outcomes other than the primary outcome. The more tests that are carried out, however, the more the possibility of false-positives 2/7
comment in response to
post
Great to hear!
comment in response to
post
So sorry to hear this Richard. What an amazing talent to have - a gift that will no doubt leave lots of special memories. My condolences to you and your family.
comment in response to
post
So, as you have these (sometimes awkward) conversations about sample size, please do ensure that the clinical meaningfulness of the trial doesn’t get lost in the sample size samba beat!! 8/8
comment in response to
post
I sometimes wonder what we would find if we measured decisional regret at the end of such studies! 7/8
comment in response to
post
We often hear of trials that show apparently clinically meaningful differences but did not detect these statistically. This is often because their sample size calculations were unrealistic - effect sizes set too large in an attempt to minimise the sample size 6/8
comment in response to
post
Thus whilst adjustments can obviously be made, the trial must ultimately be deemed possible of detecting a clinically important change eg one should not change the target difference from say a desired 5% to an implausible 20% reduction in outcome just to drop the sample size! 5/8
comment in response to
post
Whilst, of course, the trial has to be able to recruit the target number of patients, a trial must fundamentally be designed to deliver a clinically meaningful answer 4/8
comment in response to
post
The term comes from the excellent Schulz & Grimes paper in the Lancet a number of years ago 3/8
www.thelancet.com/journals/lan...
comment in response to
post
Many of you have heard me talk about the “sample size samba” before - the description of the awkward “dance” between the effect size you want to detect and the sample size that you think is you think is achievable 2/8
comment in response to
post
Love it!!
comment in response to
post
It was a fabulous lecture Mirela!
comment in response to
post
6. The policy does not just affect funding going forward. All existing NIH grants will have their indirect rates cut to 15% as of today, the date of issuance.
For a large university, this creates a sudden and catastrophic shortfall of hundreds of millions of dollars against already budgeted funds.
comment in response to
post
Of course, even if a formal independent DMC is deemed not essential, all trials must still have robust monitoring in place to ensure good practice and research integrity (eg via a TSC and other standard monitoring mechanisms) 8/8
comment in response to
post
However, as Sydes et al note, should an independent DMC deemed not to be essential, the rationale for this must be fully justified in the protocol and agreed before the start of the trial. 7/8
journals.sagepub.com/doi/epdf/10....
comment in response to
post
These exceptional reasons included eg short trial duration where events for
patients would all accrue before the DMC could provide input, where there are only v minor hazards from the treatments eg administrative studies & trials aimed at demonstrating a biologic principle 6/8
comment in response to
post
However, there are a few scenarios where, exceptionally, it has been accepted in the literature that it may not be feasible/essential for a DMC to be set up. These were described in the DAMOCLES project which reviewed all aspects of DMCs 5/8
www.journalslibrary.nihr.ac.uk/hta/HTA9070
comment in response to
post
The independent DMC role is thus very important and the expectation rightly is that a DMC is routinely required for all trials 4/8
comment in response to
post
A DMC’s role is primarily in the recruitment phase of a trial in order to protect any future potential participants from being unnecessarily enrolled in the study (if sufficient evidence has been reached) or from exposure to potential harm (should a harm signal be identified) 3/8
comment in response to
post
The role of an independent DMC is to review accumulating data in a trial for evidence of any emerging safety issues or whether there is sufficient evidence that a trial should stop early for benefit or harm (or indeed futility) 2/8
comment in response to
post
Attard et al also highlighted the lack of wider stakeholders currently involved in setting the non-inferiority margin. There is neat ongoing work in this field led by @beatrizgoulao.bsky.social trying to involve patients and public in non-inferiority margin selection 8/8
comment in response to
post
Totton et al recently reviewed a cohort of non-inferiority trials. They highlighted that clinical considerations tend to dominate in margin setting. They also usefully summarised the current (mainly regulatory) guidance on the issue 7/8
trialsjournal.biomedcentral.com/articles/10....
comment in response to
post
They highlighted, however, that many were not confident in the applicability of the non-inferiority margin & highlighted different barriers & facilitators to identifying an appropriate margin 6/8
comment in response to
post
The two most commonly reported approaches to determine the margin were through a review of the evidence base (66%) and opinion seeking methods eg asking clinicians (59%) 5/8
comment in response to
post
Attard et al undertook a survey a couple of years ago to identify the most common methods currently being used to define a non-inferiority margin 4/8
trialsjournal.biomedcentral.com/articles/10....