Hi Pinnacle Team,
I was hoping for clarification about SDTM validation rule SD1066, which implies IDVARVAL must be null when RELTYPE is populated in RELREC. While I understand this is the most common way of relating datasets in RELREC, I do see opportunities to use IDVARVAL for these dataset-level relationships where RELTYPE is populated.
For example, if I want to display a relationship between PR and MH when PRCAT = MHCAT, but only when PRCAT and MHCAT are both “BIOPSY HISTORY”, I think an appropriate way of displaying this would be to leave USUBJID null, have IDVARVAL = “BIOPSY HISTORY”, and also have RELTYPE populated.
I understand that another variable could be added to MH and PR, like --LNKID, to solve this, but I also think the way I described above would be appropriate and would help keep the parent domains in a minimum structure.
I am curious to hear what led to this validation rule being created, and if there are any arguments to my logic.
Thank you,
Kylar
Hi Kylar,
There are two types of relationships between datasets. 1. Record-to-record and 2. variable-to-variable. RELTYPE variable covers the second option, where a reference to specific record is not applicable.
SD1066 rule is created based on SDTMIG “Relating Datasets” section. For example, see SDTMIG 3.2.
"RELTYPE - CDISC Notes: Identifies the hierarchical level of the records in the relationship. Values should be either ONE or MANY.
Used only when identifying a relationship between datasets (as described in Section 8.3).
8.3 Relating Datasets
The Related Records (RELREC) special-purpose dataset can also be used to identify relationships between datasets (e.g., a one-to-many or parent-child relationship). The relationship is defined by including a single record for each related dataset that identifies the key(s) of the dataset that can be used to relate the respective records."
This type of relationship between datasets uses only single key variable in each dataset. There is no need to provide a reference to specific values in each record. Those values are already defined by key variable in each dataset. See SDTMIG for details and examples.
Kind Regards,
Sergiy