Rule SD2237 "ACTARM does not equal ARM" violates FDA Technical Conformance Guide March 2019

Hi, Also the definition in…

Hi,

Also the definition in SDTM IG 3.2 states below:

Description of actual Arm. When an Arm is not planned (not in Trial Arms), ACTARM
will be “Unplanned Treatment”. Randomized subjects who were not treated will be
given a value of “Not Treated”. Values should be “Screen Failure” for screen failures
and “Not Assigned” for subjects not assigned to treatment. Restricted to values in Trial
Arms in all other cases.

Rule SD2237 "ACTARM does not equal ARM" violates FDA Technical C

Thanks for completing Priyanka!
The FDA Technical Conformance Guide also states that their statement deviates from what is in the SDTM-IG: "Although this convention is inconsistent with the SDTMIG, FDA recommends its use so that ‘Screen Failure’, ‘Not assigned’, and ‘Not treated’ are not specified as a treatment arm". 
Rule SD2237 is an FDA-derived rule, not a CDISC rule. So ...

Thanks Jozef for clarifying…

Thanks Jozef for clarifying.

Even I am encountering this issue with Pinnacle validation for my current study. Should we then follow FDA guidance and set Actual Arm in these cases as null and then report in SDRG and explain the Pinnacle error with justification that we have followed FDA TCG. Also what about cases where subject is randomised to Treatment Arm A , however given Treatment Arm B ? Can we have ARM = A and ACTARM= B?

Rule SD2237 "ACTARM does not equal ARM" violates FDA TCG

What I would do ... (but who am I ...):
For a submission to the FDA, I would follow the FDA TCG, so set ACTARM and ACTARMCD to null. If that causes a P21 validation error, I would document it in the SDRG as a false positive. If it is not an FDA submission, I would follow the SDTM-IG.
For the second case (ARM = A and ACTARM= B), I am not sure - I am currently going through all rules for a customer and check with the TCG, but I am not ready yet.

Any reaction from Pinnacle21?

Trevor

on August 16, 2019

P21

Hi,While we recognize that…

Hi,

While we recognize that the FDA technical conformance guide is inconsistent with the SDTMIG, we also need to be mindful of the fact that the FDA business rules and FDA validator rules (found in Section 4 of the Study Data Standards Resources page linked below) still reflect the need for a rule.

FDA Business Rule ID FDAB025: Randomized subjects are expected to receive a study treatment.

FDA Validator Rule ID SD0071: Invalid ARM/ARMCD.  Note: This is linked to CDISC Publisher ID CD0119.  The description is: Data for screen failure subjects, if submitted, should be included in the Demographics dataset, with ARMCD = “SCRNFAIL' and ARM = “Screen Failure”. Subjects withdrawn from a trial before assignment to an Arm, if they are not screen failures, should have ARMCD ='NOTASSGN' and ARM = 'Not Assigned'.

Note that FDA Validator Rules are intended to define the criteria used to ensure compliance to a business rule, a standard, or the Study Data Technical Conformance Guide.

This may be due to the fact that we have different release dates for these documents:

• FDA Validator Rules, released October 2018
• Technical Conformance Guide, released March 2019
• FDA Business Rules, released June 2019

It's worth noting, from P21s perspective, that rule SD2237 firing in these cases is not a false positive.  It is firing as expected.  When this rule fires, it's the sponsors opportunity to explain the specific situation in the cSDRG that this subject (or subjects) were either not treated or did not receive their planned treatment.  The rule is a Warning.

https://www.fda.gov/industry/fda-resources-data-standards/study-data-standards-resources

Thanks,

Trevor
Product Manager

Rule SD2237 "ACTARM does not equal ARM"

Thanks Trevor,

From P21s perspective  ... From a users perspective however, this is experienced as a false positive. The problem is that P21s perspective of a "warning" is "something unusual" (so not something wrong), to be explained in the SDRG. If a policeman gives me a warning, it usually is not because something is "unusual" (e.g. me driving a hydrogen-powered car) but because I did something wrong. I should not have to explain the policeman nor anyone else why I am driving a hydrogen-powered car.
So I propose that P21 changes all "warning"s in "unusual" ;-)

Jozef
(also PM - meaning professor in medical informatics)

Trevor

on August 16, 2019

P21

Jozef, Thanks for the…

Jozef,

Thanks for the feedback.  I'm curious to ask, what's the alternative?  Not having a rule and bringing no attention to the "unusual" data anomaly at all?  We don't think this is a good approach which is why we have rules like this.  I hope this makes sense.

Thanks,

Trevor

Rule SD2237 "ACTARM does not equal ARM"

Well, you could make it an "info" ...
The interesting is that the FDA wants to have ACTARM empty AS AN INDICATION that the subject was not treated but was randomized. Why should one then have to explain that a second time in the SDRG. The situation that a subject retreats before having first treatment is surely not "unusual" (my wife once did).
Also, I don't think it is up to P21 to decide what pharma companies like Pfizer, Novartis, ... should explain what was "unusual" in their clinical trial.
Better no rule than a bad rule.

About the "business rule", it says "expected". Well, I am expected not to write forum entries after 8pm. What time is it? Ooooh: 20:11 ...

Best regards,
Jozef