reports for TSCs and unblinded and blinded sections of DMC meetings (clinical trials)

[statlimp]

Dear All,

I wonder if you can please advise me what the conventions are for reports for Trial Steering Committee/TSCs and unblinded and blinded sections of Data Monitoring Committees/DMC meetings during clinical trials.
I am particularly interested in what tables are produced and are they split by treatment arm.
There are various types of data collected in a clinical trial.
1) demographic eg gender, age, ethnic group
2) clinical data of various types
i) baseline data of clinical variables eg BMI, BP (if a stroke or CVD trial BP might be in iii) )
ii) follow up data of clinical variables
iii) baseline values of variables that are endpoints or closely related to endpoints (primary or secondary)
iv) post baseline data that comprise the trial endpoints - or are closely related to them
v) recruitment figures
vi) (S)AEs
vii) numbers of participants who die or withdraw from trial, or from treatment or other intercurrent events (these may be part of endpoints)
I may have missed some data types.

I am interested in what set of tables are produced for each of the 3 settings and are they split by treatment arm or not and whether comparisons are made eg p-value. I know in part its obvious - we don't present efficacy split by arm with a p-value to the TSC or the blinded TSC.
Some issues aren't clear to me - is there a difference in the tables for the TSC and blinded part of the DMC ?
eg can we show the treatment split of demographic demographic variables in the TSC and blinded part of the DMC or is treatment splits never allowed there ??
can we show the baseline values of efficacy type variables in the TSC and blinded part of the DMC (or in only 1 of those) eg baseline differences in fasting plasma glucose or PSA or tumour sizes ?
Are there clear guidelines on these ? I've not found them - I have found stuff on trial committee roles and types fo participant and general conduct but nothing specific on tables for each type
Does it depend on whether the trial arms are blinded to participants ie often "patients" (even if it is the arms may well not be blinded to trial investigators, statisticians or endpoint assessors)

many thanks in advance

 

[statlimp]

o dear - thanks everyone for viewing my query. I wonder if getting authorative guidelines on this are harder than I thought. I wonder if there is a more clinical trials oriented board or list ?

 

[hlsmith]

Hmm. Is this hypothetical - meaning you aren't currently conducting a study?

I have only registered with USClinicaltrials.gov. I haven't gone through this process, since my research is investigator initiated. I would imagine governing bodies have articulated general rules - partially dependent on risk. Which all should be written out a priori to commencing the study. I would be interested in what you find out. I will note that you list a lot of data/result points. I imagine the number of points wouldn't be that large and would be classified as primary and secondary, which would influence the level of detail needed for them.

 

[statlimp]

Hmm. Is this hypothetical - meaning you aren't currently conducting a study?

thanks for your thoughts - unfortunately its not hypothetical - I am independent stat (we distinguish independent and trial stat) for a number of clinical trials and will need to produce these reports over the next few years.
Yes we have a prmary endpoint but quite a few secondaries and sometimes even the secondaries have a lot of inputs in their calculation. The CRFs that capture information would have extra variables again that aren't part of endpoints but are presumably measured for a reason ie reporting at TSCs and DMCs. I forgot to mention QoL measures. we might have 6 or so questionaires each with 20 or so questions recorded as multiple post baseline visits so it would be 100s of variables (usually each on a 5 point likert scale so %s at each level might be wanted).

I think there might be stuff out there I've not looked at carefully enough - eg the DAMOCLES charter (for DMCs not TSCs) and this Harman paper t at seems to say TSCs never see data split by arm (and I guess the unblinded DMC section can see everything by arm - whcih jusy leaves the blinded DMC - maybe they can just see baseline and recruitment split by arm ???)

 

[fed2]

tables related to safety population use the actual treatment assignment, rather than the randomized assignment. hoepthat helps, fed2.

 

[hlsmith]

There is also the DMC not blinded to arms but blinded to which arm is the intervention under-investigation.

 

[statlimp]

There is also the DMC not blinded to arms but blinded to which arm is the intervention under-investigation.

ah yes that is a possibility thanks .. you mean members see info for arm A and arm B but won't know which is the novel treatment and which standard of care. I've heard of that but not sure when it is recommended to use that approach

 

[hlsmith]

When I have people provide data to me, I typically tell them not to inform me which treatment is novel. Triple blinded, subjects, investigators, and analyst.