| Questions ReceivedQuestion | Teams Collective Response | |
|---|---|---|
Historical Data Consideration in the SV Domain (under SDTMIG v3.4) | PHUSE Team Response: 09 December 2022 Pre-study findings, such as tests performed at the time the disease was diagnosed, can be assigned to the initial screening visit. In this case, the content of the visit variable represents the visit when the test result was recorded in the CRF. The date of the test (or sample collection date) will be stored in the –DTC variable of the applicable domain (e.g. MIDTC). In cases where historical data is stored as a finding, these historical test/sampling dates should not be taken into account when populating SVSTDTC for the particular visit. References: Assumptions 13 for the SV domain in SDTMIG v3.4: | |
How should the sex of transgender patients be collected and analysed in clinical trials? Should the sex at birth be collected only or should the gender preference also be collected? Which laboratory normal ranges should be assigned to transgender patients’ laboratory test results? How does hormone therapy affect data collection and/or analysis for transgender patients? | PHUSE Team Response: 30 June 2022 The CDISC CDASH team is currently working to publish either an updated guidance or white paper planned for 2025 on recommendations on capturing the sex for transgender patients. In the draft version, the recommendation would be to collect a two-stage question (note that the controlled terminology and collection text are a draft stage and not finalised): 1. “Sex at Birth” (Male | Female | Don’t know | Prefer not to answer) and 2. “Sexual Identity” (Male | Female | Intersex | Transgender | … | Don’t know | Prefer not to answer | Self-describe). In the interim, each sponsor should determine how the data should be collected. It is recommended to provide clarity on the definition of each question, perhaps within the CRF Completion Guidelines. For example, does Sex at Birth pertain to sex stated on the birth certificate, and how to complete the data entry if a patient does not have a birth certificate. The following articles may be reviewed to determine how hormone therapy affects laboratory results and, in general, analysis for transgender subjects:
| |
How do you proceed in providing the reason for the missing code? Do you collect the reason for the missing LOINC code or do you just provide a predetermined reason? The LOINC working group recommend providing a reason for missing code in the cSDRG. (See the extracted text from the Reference: https://www.fda.gov/media/109376/download)
The FDA TCG 4.6 recommends providing the LOINC code of the laboratory parameters for studies starting after March 2020, but nothing is mentioned in the case of missing code. | PHUSE Team Response: 07 February 2022 If the laboratory hasn’t sent the LOINC code, it is recommended to go back to the laboratory to obtain it. Per the team members’ experience, the FDA accepts if the laboratory hasn’t provided the LOINC code and it is missing. In the cSDRG, it notes the reason for it missing as “Lab did not provide the code”, or as noted in the LOINC working group’s screenshot. (Reference: https://www.fda.gov/media/109376/download) One solution would be to request the LOINC code from the lab at the study initiation phase, but it is expected that not all lab tests will have a corresponding LOINC code assigned. | |
There are a couple of papers which offer guidance for maintaining 1-1 mapps between AVAL and AVALC. Things like: https://www.lexjansen.com/wuss/2017/79_Final_Paper_PDF.pdf https://www.pharmasug.org/proceedings/2012/DS/PharmaSUG-2012-DS16.pdf However, neither of these papers explain how to consistently create derived records, where AVALC is a rounded version of AVAL, which satisfies the 1-1 criteria. For example, suppose (within a single PARAMCD) I need to compute an average and then present that in a list to 1 dp. For example, let's say AVAL=45.333333 so for the listing I want to show 45.3. I've computed an average for another subject where AVAL=45.26 which I also wan to show as 45.3 in a listing. If AVALC=45.3 for both records, then this is not a 1-1 mapping. I obviously can't round AVAL, because that would represent a loss of numerical precision in other calculations. One solution might be 'do not populate AVALC, do the rounding when producing the report'. However, this leaves a lot of work in the reporting program if many parameters are to be listed; the programmer would have to determine the rounding on a per-parameter basis. Ideally the 'heavy lifting' should already have been done at the dataset level. | PHUSE Team Response: 08 July 2020 Rounding values of AVAL for listing purpose - where to do the rounding and how/where to store the rounded value. Storing a rounded value in AVAL is not good practice as it typically results in a loss of precision for calculations in the tables. Storing rounded values in AVALC goes against the ADaM rule that there has to be a 1-1 mapping of AVAL to AVALC. Also, it is not the intent to store the character version of a numeric analysis value in AVALC. AVALC should be populated only when the character value is used for analysis. See ADaM IG v1.1, section 3.3.4, 'PARAM, AVAL, AVALC' paragraph 3. There is no ADaM guidance as to variable naming for variables used for listing purpose only. Rounding the analysis result can be done in the listings program, or alternatively, if one wants to store rounded value in the ADaM dataset, a custom variable can be added with an intuitive meaning, eg LISTVAL, to store the rounded value. | |
| Study treatment regimen will be A-B-C-D, therefore planned ARMCD can be ABCD. Most of the patients actual ARM ACTARMCD will also be ABCD. But a few patients may skip D or repeat ABC part which is ABC or ABCABCD. Shall we put UNPLANN in actors or put the real ABC or ABCABCD in the ACTARMCD? | PHUSE Team Response: 09 January 2020 The planned treatment should be reflected in ARMCD/ARM, while the actual regimen received should be reflected in ACTARMCD/ACTARM. In general, TA should reflect the protocol-specified treatment regimens to be administered. If the protocol specified the skipping of a treatment regimen by design, then it is acceptable to find inconsistencies between ARMCD and ACTARMCD. However, these should be noted in the cSDRG and explained in further detail. | |
| In SV domain, we search all the by-visit source data to get the min and max date of each CRF visit. if due to some reason, there are 1-2 days overlap among 2 consecutive CRF visits in SV domain, we can explain in SDRG or always make visits in SV without overlap which means we assign the overlapped days to 1 CRF visit in SV rather than keep the days in both visits as the source data shown? | PHUSE Team Response: 09 January 2020 Acceptable to have the overlap on the visits in SV domain. There will be no P21 consequences due to this, and as such it is not required to explain further in the DRG. The explain in the DERG would be left to the Sponsor's determination. | |
| For the Table like 'Summary of Common (>=X%) Adverse Events by Overall Frequency', should the flags for common AEs be created in the ADAE dataset? | PHUSE Team Response: 31 July 2019 5pct, 2pct custom flag variables can be added to ADAE Derivation of the flag variable depends on the definition in SAP/table Janssen - If derivation rule is complicated enough, include it in ADAE.
Other companies do not include in the ADAE and handle it in the table generating programs.
| |
| FDA impressed the wish to keep ETCD/ELEment to facilitate reviewer to review the data in 2011 CDER Common Data Standards Issues Document. However, in all later FDA published Study Data Technical Conformance Guide up to V4.1 published in 2018, only EPOCH is required. EPOCH by it's own should have been informative enough. FDA validator rules V1.2 published in DEC2017 still mentions that variables requested by FDA in Policy documents should be included in the dataset, e.g. EPOCH and ELEMENT. Do you know if FDA still require ELEMENT/ETCD in all domains? If yes, I would suggest to CDISC SDTM team to include those 2 variables in the parent domain and not the SUPP domain. | PHUSE Team Response: 04 July 2018
| |
| How should OTHER be represented for variables bound by non-extensible codelists? | PHUSE Team Response: 07 June 2017 Existing SDTMIGs (e.g., v3.1.2, v3.1.3, v3.2) do not explicitly define how "OTHER" should be implemented universally for all non-extensible codelists. Additional References: N/A | |
| How should MULTIPLE be used for variables bound by non-extensible codelists? | PHUSE Team Response: 07 June 2017 Existing SDTMIGs (e.g., v3.1.2, v3.1.3, v3.2) do not explicitly define how "MULTIPLE" should be implemented universally for all non-extensible codelists. Additional References: N/A | |
| What are best practices for creating CT for/representing questionnaire responses? | PHUSE Team Response: 07 June 2017 It is recommended to review SDTMIG (v3.1.2, v3.1.3, or v3.2) Section 4.1.3 Coding and Controlled Terminology Assumptions. Furthermore, please also review existing questionnaire CDISC Controlled Terminology (CT) and CDISC Questionnaires, Ratings & Scales (QRS) supplements and related details found on the QRS page – see reference below. | |
| What is the general recommendation/approach for generating/submitting custom domains (e.g. non-standard CDISC SDTM domains) to regulatory agencies? | PHUSE Team Response: 12 September 2017
The overall process for creating a custom domain are clearly explained in the SDTM IG and must always be based on one of the three SDTM general observation classes (interventions, events or findings). |