Designing ADaM Datasets: KISS the Complexity Goodbye PhUSE 2013

Designing ADaM Datasets: KISS the Complexity Goodbye PhUSE 2013
PhUSE 2013
Paper TS02
Designing ADaM Datasets: KISS the Complexity Goodbye
David Jordan, SAS Institute, Cary, NC USA
The CDISC SDTM standard provides many predefined domains, each providing a model for a specific purpose with
an appropriate set of columns. In the case of CDISC ADaM, a large number of possible variables are defined and it
is up to you to determine which variables are appropriate for your particular analysis modeling situation. Even though
the ADaM ADSL and ADAE dataset templates serve a particular purpose, you still need to select a subset of the
variables appropriate for your analysis needs. The BDS dataset template is a very general template that defines over
140 variables you can use, which can be overwhelming. Defining an ADaM analysis dataset is far more complex
than using a pre-defined SDTM domain.
Metadata describing both the dataset templates in the standard and existing datasets in your data environment can
be used when defining an ADaM dataset. When the user includes variables from existing datasets, this metadata
can automatically provide traceability information. Some ADaM variables require the management of xx, y, and zz
parameters within the ADaM variables. Software can assist in the management of this tedious and error-prone task.
The use of an underlying metadata framework and ADaM-specific algorithms to drive the dataset definition process
in a graphical environment eases and accelerates the use of the ADaM standard. The ADaM dataset creation tool
that has been incorporated into the SAS Clinical Data Integration® product is described in detail.
This paper describes the process for defining a CDISC ADaM dataset using the ADaM dataset wizard implemented
in SAS Clinical Data Integration release 2.4. Prior releases of SAS Clinical Data Integration supported the CDISC
SDTM standard. Defining an ADaM dataset involves some issues that do not arise when defining an SDTM domain.
The interface for defining an ADaM dataset in SAS Clinical Data Integration will be fully described, along with
aspects of ADaM dataset definition that do not exist in the creation of SDTM domains.
SAS Clinical Data Integration is an extension of the SAS Data Integration® product, providing management of
metadata and clinical trial data based on the CDISC standards. SAS Data Integration provides a nice user interface
environment for graphically defining extract, transform, and load (ETL) procedures for mapping and storing data.
Embedding metadata and CDISC-specific algorithms into this data integration platform eases the effort of mapping
one’s data from an existing data representation into a CDISC-compliant form.
When setting up your SAS Clinical Data Integration environment, a standards administrator imports metadata
associated with the clinical standards defined within the SAS Clinical Standards Toolkit®. The SDTM, ADaM and
SEND standards are supported. This metadata provides a complete description of all the datasets and associated
attributes and properties, as well as data and code used to validate the compliance of your datasets relative to the
standards. When the metadata is imported, it is placed within the SAS Metadata Server®.
A study administrator is responsible for defining characteristics of a study. This includes the selection of the relevant
standards, libraries in which to place datasets, and controlled terminologies to use with the study. The study
administrator can also select a folder template which initializes a folder hierarchy in which to place the data
associated with the study. A wizard walks you through the steps necessary to initialize the study.
Once you have completed these steps, you can begin creating datasets for the study. You will likely place the
datasets for SDTM and ADaM in separate folders within the study’s folder hierarchy. The Folders tab within the user
interface provides a graphical rendering of the folder hierarchy. The creation of an ADaM dataset involves the
invocation of a wizard that guides you through the necessary steps. You can select the folder within the study in
which you want to place your ADaM dataset, use the right-mouse-button context menu to select the New menu item,
and choose the Analysis Dataset menu entry. This brings up the wizard to create an ADaM dataset. Another means
PhUSE 2013
of selecting the New menu is a drop-down menu that resides directly below the File button in the upper left corner of
the SAS Data Integration Studio window.
In the following screen shot, we have selected the folder named ADaM within Study1 and selected the New menu
entry. There is an entry near the bottom of the new menu in the screen shot used to create an analysis dataset.
If you selected the New menu relative to the folder in which you want to place the ADaM dataset, the first page of the
wizard identifies the location and study in which you invoked the wizard. You do have the option of browsing to select
an alternative location, but normally you would press the Next button to proceed to the next page of the wizard. If the
study happens to be associated with more than one ADaM standard, the next wizard page lists the ADaM standards
and requires you to select one. Select the appropriate standard and press the Next button.
You must then select which ADaM dataset type you wish to create. Each type is presented in a list and you must
select one of the entries.
PhUSE 2013
One, and only one, ADSL dataset can be defined for a given study and its identifier must be “ADSL”. If you have
already created an ADSL dataset for the study, its entry in the list is disabled.
Press the Next button and you will be taken to the next wizard page where you can provide a name and identifier for
the dataset.
The identifier for all ADaM datasets must begin with the prefix “AD” and the wizard automatically provides this prefix
in the identifier field. It also makes sure you provide a name that is compliant with the ADaM standard. If you are
creating an ADSL dataset, the identifier is automatically provided and cannot be changed.
The next page of the wizard allows you to specify values for properties of the dataset. Default values are provided for
these properties, but you can enter alternative values.
The next wizard page allows you to select a library to assign to the dataset. If any default libraries have been
specified at the study level, they are listed in the table. There is also a means to select a different library.
ADaM allows you to select variables from existing data sources within your study to use in your dataset. This can
include other ADaM datasets, SDTM domains, and ordinary SAS data sets. The next wizard page presents all of
these data sets in a manner that reflects the folder hierarchy within the study.
PhUSE 2013
Set the checkboxes on any datasets that have variables you want to include. Then press the Next button to advance
to the next wizard page. If you did not select any source datasets, the next wizard page which allows you to select
variables is skipped.
You can then select variables from each of the source datasets.
The source datasets are listed in the table positioned at the top left of the wizard page. When you select a specific
dataset, its variables are listed in the table on the top right-hand side of the wizard page, as is shown for the CE
SDTM domain. The name and description of each variable is provided. If a particular variable is not compliant with
the ADaM standard for some reason, the third column labeled “Compliant” will have an X mark in it. Select a noncompliant variable and press the button labeled “X : Non-Compliance” and a popup window will explain why the
variable is not compliant.
Once you have selected one or more variables in the top right-hand table, you can press the Add button and they will
be added to the table in the bottom left part of the window. It was decided that the user should be allowed to add
non-compliant variables, so you can still add them. If you added one or more variables and decide that you would
rather not include them in the dataset, you can select them and press the Remove button. Once you have selected
all the variables that you want to include from the source datasets, press the Next button to advance to the next
wizard page.
PhUSE 2013
The next wizard page allows you to select the ADaM variables for your dataset. You are presented with the variables
associated with the particular type of dataset that you are creating. They are organized into the same categories that
are provided in the ADaM specification. Any variables that are required are preselected and their names will be
contained in square brackets. It is not possible to remove any of these required variables. The following screen
dump displays this wizard page.
The set of available variables are displayed in a list on the left and those that have been added to the dataset are
placed in the list on the right. To add one or more variables, select them in the list on the left and then press the Add
button. This causes them to be moved to the list of selected variables on the right. If you had selected a variable to
include but decide that you do not want it in the dataset, select it in the list of selected variables on the right and
press the Remove button, which will place it back in the list of available variables.
There are cases where inclusion of one variable requires the inclusion or exclusion of another variable based on
constraints in the ADaM specification. These variable selection dependencies can be evaluated by pressing the
button labeled “Check Compliance”. The set of variables that you have selected will be examined and you will be told
whether any variable dependency constraints have been violated. Once you complete the process of selecting the
variables for your dataset, press the Next button to proceed to the next wizard page. It is also always possible in the
wizard to use the Back button to return to a previous wizard page if you need to make changes to any prior
selections you have made.
You may have selected variables that contain parameters denoted by a lower-case xx, y, or zz. These are
placeholders to be replaced by numeric digits. The xx refers to a specific period of the study and must be replaced
with a zero-padded two digit integer (01-99) denoting a specific period. A single digit value for the y parameter is
used to refer to a grouping or other categorization, analysis criterion, or analysis range. A zz parameter is an index
for the zz record selection algorithm. It should be replaced with a zero-padded two digit integer (01-99). You would
typically have multiple actual variables derived from the same variable template that has a parameter, with the
integer substitution for the parameter providing a unique name for each variable.
A separate wizard page is provided for the setting of each parameter type (xx, y, zz). A wizard page is only shown for
the parameter types included in the variables that you chose. In the following screen dumps, variables with y and zz
parameters were selected for inclusion in the dataset.
PhUSE 2013
Each wizard page has a similar layout. A table is provided on the left that lists the variables that contain the particular
parameter type. Each variable’s name and description are provided and the third column lists the values that you
have assigned for the parameter. On the right-side of the wizard page is a list of values that can be associated with
the parameter. You can select one or multiple values from this list. Once you have selected one or more variables
and one or more values for their parameters, you press the Assign button which places the values you selected into
the third column of the table labeled Values.
For the y parameter type, the value list only contains the values 1-9. For the xx and zz parameter types, a list is
provided of all 99 values from 1-99. You can enter a maximum value in the field above the values to reduce the
number of entries in this list of values. Once you have assigned at least one value for each variable with an xx, y, or
zz parameter, you can advance to the next wizard page.
For a given variable, if you provide N values, you will get N corresponding variables added to the dataset. There are
a few variables that include both an xx and y parameter. In that case you will get N*M variables, where N is the
number of x values and M is the number of y values you have assigned.
Some ADaM variables include a wildcard character prefix (_) in the name and wildcard suffix (…) in the description,
allowing you to provide your own substitution characters. If you selected any of these variables, a wizard page is
presented so you can provide the text to substitute for the wildcard character. The wizard page presents a list of the
selected wildcard variable templates.
You should select a variable, specify a name prefix and description suffix and then press the Add button. This will
add the variable to the list of generated variables. You can add additional prefix and suffix values for the same
PhUSE 2013
variable, resulting in additional variables in the dataset. If you add a variable and decide you no longer want to
include it, select the variable in the list of generated variables and press the Remove button.
You can change the order of the variables in the dataset on the next wizard page. Select one or more contiguous
entries in the table and press either the Move Up or Move Down buttons to rearrange the order of variables in the
There is also a button provided to order the keys for the dataset. Pressing the button labeled “Order Keys” brings up
a separate dialog listing the key variables, allowing you to rearrange them in a similar manner.
The next wizard page displays a complete list of all the variables in your dataset. This includes all the substitutions
for wildcard characters as well as all values you provided for the xx, y, and zz parameters. For variables you directly
included from a source dataset, the folder, dataset, and variable columns will have the appropriate values.
PhUSE 2013
You can enter values for each of these columns to provide traceability information about the derivation of your
variables. The source derivation column on the far right has an associated multi-line editor you can use to provide a
detailed description or pseudo-code describing the algorithm to be used in the derivation. This editor is invoked by
pressing the button labeled “Edit Source Derivation.”
A final wizard page is displayed with some summary information about the dataset you are creating and a Finish
button. When you press the Finish button, the ADaM dataset and associated metadata is created.
Users of SAS Clinical Data Integration may want to define a set of their own fully-defined custom templates for their
ADaM datasets that can be shared across studies. This allows you to establish organization-wide consistency in your
dataset definitions. This is possible in SAS Clinical Data Integration by promoting a dataset to be a user-defined
template, which adds it to the list of datasets associated with the standard.
Once you have defined a dataset and promoted it to the standard, you can directly create an instance of one of
these datasets without going through the process of defining all the variables to be included. Select the folder in
which you want to place the new dataset and under the New menu, a menu entry labeled “User-defined Analysis
Dataset” is used to create an instance of the dataset.
The new ADaM dataset creation wizard in SAS Clinical Data Integration allows you to easily generate an ADaMcompliant dataset. It is driven by metadata describing the standard’s dataset templates. The user interface allows
you to quickly identify data from other datasets to be used as sources for variables in the ADaM dataset. It also
provides an interface for defining the values for the xx, y, and zz parameters found in some ADaM variables. It also
provides a means for generating variables that include wildcard characters. The wizard provides a quick and easy
means of generating your ADaM datasets.
Your comments and questions are valued and encouraged. Contact the author at:
David Jordan
SAS Institute
SAS Campus Drive
Cary, NC 27513
Work Phone: 919-531-1233
Email: [email protected]
Brand and product names are trademarks of their respective companies.
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF