Mapping Considerations for Screen Failure, Not Assigned, and Not Treated Subjects

June 18, 2020

Our recent webinar Confusing Validation Rules Explained sparked lots of follow-up questions from you. We are addressing those questions in a series of posts. In this edition, we will clarify the best practices when mapping data for screen failure, not assigned, and not treated subjects. We will also help by describing the most effective ways to respond to these situations.

Contradictory Guidance for Mapping Planned Arm and Actual Arm

In SDTMIG up to and including v3.2, CDISC guidance was to populate ARMCD/ARM and ACTARMCD/ACTARM for screen failure, not assigned, and not treated subjects. CDISC provides specific values to use in these cases. However, FDA added guidance in their Study Data Technical Conformance Guide (TCG) that states, "Screen failures, when provided, should be included as a record in DM with the ARM, ARMCD, ACTARM, and ACTARMCD field left blank. For subjects who are randomized in treatment group but not treated, the planned arm variables (ARM and ARMCD) should be populated, but actual treatment arm variables (ACTARM and ACTARMCD) should be left blank." FDA acknowledged that their guidance conflicted with CDISC guidance. CDISC has since changed the guidance in SDTMIG v3.3 to match FDA guidance. However, this causes inconsistency in how screen failures are mapped across the industry.

So, Which Guidance Should be Followed for Planned Arm and Actual Arm?

With SDTMIG v3.3, the guidance between FDA and CDISC match, so there is no ambiguity. If you are using an SDTMIG version prior to v3.3, it is a tough question to answer. Since these variables are listed as required by SDTMIG, and leaving these variables null without listing a reason (screen failure vs not assigned, etc.) makes it unclear what happened to these subjects, perhaps following the SDTMIG is the best strategy. However, we caution that communication with the review division at the regulatory agency is advised, since they may have a preference on how you represent your data. You also need to consider that PMDA's requirements are different. PMDA considers these variables required as specified by SDTMIG, and missing variables will result in validation Errors which require explanation.

What the Changes Will Look Like in SDTMIG v3.3

In SDTMIG v3.3, CDISC changes the guidance to conform to FDA's expectations in the TCG, which states to leave planned arm/arm code (ARM/ARMCD) and actual arm/arm code (ACTARM/ACTARMCD) null for screen failure and not assigned subjects. ACTARM/ACTARMCD should also be left null if the subject has been randomized to a planned arm, but was not treated. In addition to leaving these variables null, you will need to populate the new Reason Arm and/or Actual Arm is Null (ARMNRS) variable in the Demographics (DM) domain.

Ambiguity in Planned Arm Definitions

It is not always straightforward which planned arm should be used for subjects. For example, if a subject has withdrawn consent between screening and randomization, but fulfilled all inclusion/exclusion criteria, is this regarded as a screening failure? For these types of cases, you would need to refer to the CDISC Definitions of the codelist terms. In the situation mentioned, SCREEN FAILURE is defined in CDISC SDTM CT as "The potential subject who does not meet eligibility criteria during the screening period." If a subject meets all inclusion/exclusion criteria, they would likely not be a screen failure. Perhaps NOT ASSIGNED is a better fit for these subjects.

Handling Missing Data

What should be done in the case that screen failure subjects are not being collected for your study? If you are not submitting screen failure subjects, explain it in the Reviewers Guide. There is a question in the Reviewers Guide (PhUSE template) that asks if screen failures are included in any datasets, which you can use to explain your situation.

In addition, if you have no records in the Inclusion/Exclusion Criteria Not Met dataset, you should not include an empty IE dataset in your submission. Use the Reviewers Guide to explain why this dataset is missing.

Handling Related Validation Findings

There are some validation rules that flag subjects where planned arm (or arm code) does not equal actual arm (or arm code). These rules typically fire for not treated subjects where planned arm (ARM) is the randomized arm but actual arm (ACTARM) is NOTTRT (or null), or where a subject receives a treatment different from their planned arm. While for these cases, it is correct that ACTARM should be different than ARM, the purpose of the rules is to identify these subjects where this important situation exists. Use these validation findings to provide a detailed explanation in Reviewers Guide about what happened in the study for these subjects.

Mapping Subjects that were Enrolled or Screened More Than Once

Pinnacle 21's Kristin Kelly wrote a paper on this subject, titled "Considerations When Representing Multiple Subject Enrollments in SDTM". Be sure to refer to this paper if your study has this situation, as the paper has examples of what to do today in order to comply with the current standards.


Blog Main Page

Want a demo?

Let’s Talk.

We're eager to share and ready to listen.

Cookie Policy

Pinnacle 21 uses cookies to make our site easier for you to use. By continuing to use this website, you agree to our use of cookies. For more info visit our Privacy Policy.