Handling Associated Persons Data in ADaM

Introduction

The SDTM Implementation Guide for Associated Persons (SDTMIG-AP) v1.0 provides guidance on how to include data collected about persons who are not the study subjects in SDTM. However, the ADaM Implementation Guide has not provided guidance regarding how to include data from SDTM AP-- domains in ADaM datasets if needed for analysis. 

Background on SDTM AP-- Domains

In SDTM, data on associated persons are stored in separate AP-- domains (e.g., APDM, APSC, APVS). The structure, format, and rules for these domains are similar to the respective SDTM domains for study subjects with the following main exceptions:

• The variable APID (Associated Person Identifier) is used instead of USUBJID (Unique Subject Identifier).
• The additional expected variable RSUBJID (Related Subject) includes the respective USUBJID value if there is a relationship to a study subject; otherwise, RSUBJID is missing.
• The additional required variable SREL describes the relationship of the associated person (APID) to the subject identified in RSUBJID
  or to the device identified in RDEVID or, if both RSUBJID and RDEVID are missing, to the study identified in STUDYID.
• Only the variables that make sense in the associated person context are included in the AP-- domains. The core designations from respective domains described in the SDTMIG are not applicable.
• Supplemental qualifiers for AP-- domains are stored in SQAP-- datasets (e.g., SQAPDM includes the supplemental qualifiers for APDM).

Advice/Instructions/Guidance

In ADaM, the appearance and structure of the data on associated persons generally depends on the analysis need.

The following scenarios describe best practices for handling associated persons data in ADaM.

• When specific associated persons data are not required for summarizing analyses or are required for listings only, there is no need to include them in any ADaM dataset.
• Whenever there is a clear relationship between study subjects and associated persons (i.e., RSUBJID is populated in the AP-- domains),
  <ADaM dataset>.USUBJID = <AP-- domain>.RSUBJID would be at least part of the join statement (see Examples 1-4) to combine study subject and associated persons data in ADaM.
• When associated persons data are required to be included as non-standard variables in a USUBJID based dataset like ADSL, a standard suffix “AP” can be used to distinguish the variables referring to associated persons from the variables referring to study subjects (see Example 1).
• When associated persons data would be required as parameters in a BDS dataset, which might contain study subject and associated persons related parameters, a standard suffix “AP” can be used for PARAMCD values referring to associated persons. Additionally or alternatively, PARCATy could be used to distinguish between parameters on study subjects vs. associated persons (see Example 2).
• When a given AP-- domain (e.g., APMH, APAE, APCE) is almost analysis ready, just lacking the core variables for analysis by treatment of the related study subject for example, the core variables could be added via the join AP--.RSUBJID = ADSL.USUBJID. Such an ADaM dataset would have the Class ADAM OTHER, since the variable USUBJID, a required variable in the ADaM standard data structures BDS and OCCDS, is not included but RSUBJID is present (see Example 4).
• BDS-like datasets, where all the parameters refer to associated persons, with APID + RSUBJID rather than USUBJID, are a possibility (see Example 3).
• ADaM datasets containing only associated persons analysis data should have names beginning with "ADAP" (see Examples 3-4).

It should be noted that there is not necessarily a one-to-one relationship between study subjects and associated persons. In pregnancy registries, for example, there may be terminated pregnancies or pregnancies resulting in multiple infants. Thus, the denominators within a given analysis population may be different for study subject and associated persons analyses. This may be an argument to keep these data in separate datasets. In more complex scenarios (e.g., unsolicited collection of medical history, adverse events or clinical events for the associated persons when there is not necessarily at least 1 event per associated person), a separate ADSL-like dataset for associated persons (e.g., ADAPL - Associated Person Level Analysis Data) may be called for. ADAPL would contain 1 record per associated person including basic information on the associated person as well as data merged from ADSL (e.g., the core variables for the related study subject RSUBJID). ADAPL could then serve to create tables on basic characteristics of the associated persons as well as to calculate the denominators for associated persons event summaries.   

Beyond adhering to dataset, variable, and parameter naming conventions, it is important to clearly describe the sources, derivations, and dataset design considerations in the metadata and the ADRG.

Examples

The following examples include only a subset of variables and records to focus on illustrating how associated persons data can be represented in ADaM datasets and combined with data on study subjects.

Example 1

In transplant studies, in addition to data on the recipient (study subject), data on the donor (associated person) is often of interest.

This example consists of 2 AP-- SDTM source datasets, an ADaM ADSL extract, and the corresponding analysis variable metadata. It illustrates the convention for using the prefix "AP" for variables relating to associated persons. Variables without the AP prefix refer to the study subject.

Traceability is provided via analysis variable metadata only.

Source SDTM AP -- Datasets

apdm.xpt

apsc.xpt

ADaM ADSL Dataset - Selected Variables, Class: SUBJECT LEVEL ANALYSIS DATASET

adsl.xpt

ADSL Variable Metadata

Example 2

Pregnancy registries, which study the effects of a particular medicine and gather health information during and after pregnancy, are another common use case for associated persons data. In this example the mothers are the study subjects and the infants the associated person. For a pregnancy outcome analysis, data collected for the mothers as well as data collected on the infants can be relevant. When these data should be presented in one table, the individual outcome parameters either referring to the mother or the infant can be included in one BDS dataset (here: ADPREOUT - Pregnancy Outcome Analysis). The RSUBJID from the source AP- - domain(s) is stored in USUBJID in this ADaM dataset. 

Parameters relating to study subjects or associated persons can be distinguished by the PARCAT1 values, the PARMCD values (with and without the "AP" prefix), and the datapoint traceability variables SCRDOM, SRCVAR, SRCSEQ, particularly SRCDOM. For the associated persons-related parameters, APID was added as additional traceability variable, since --SEQ in AP uniquely identifies a record within APID rather within RSUBJID. APID could also serve as additional key variable when a pregnancy results in more than 1 infant.

adprout.xpt

Example 3

Example 3 Example 3 also relates to a pregnancy registry. In contrast to Example 2, the infant outcomes are stored in a separate ADaM dataset (ADAPIOUT - Infant Outcome Analysis Dataset). The dataset name starts with "ADAP" as an indicator that this dataset contains only data related to associated persons. The structure of this dataset is BDS-like, but the class is ADAM OTHER as the key variables from the source dataset APVS were maintained. RSUBJID was not renamed to USUBJID, so the required variable USUBJID for a valid BDS dataset is missing. 

To allow for analysis by the treatment the mother received, TRTA was derived from ADSL, which includes 1 record per pregnant woman in this registry, merging APVS.RSUBJID with ADSL.USUBJID. Since the parameters in this dataset come from only 1 SDTM domain APVS, the SDTM sequence number variable VSSEQ was considered to be sufficient to provide datapoint traceability.

Example 4

Example 4 again stems from a pregnancy registry. The ADaM dataset is called ADAPCE (Infant Abnormal Conditions Analysis) and belongs to the Class: ADAM OTHER. 
In terms of conventions applied it is similar to Example 3 but follows an OCCDS-like structure in this case.

References and Resources 

In addition to the SDTMIG-AP v1.0, see the Kidney Transplant Therapeutic Area User Guide v1.0, Section 6.1.1, and the HIV Therapeutic Area User Guide v1.0, Section 8.3. (Although the latter is not an associated persons example, it does include both mother and infants are treated as study subjects).