gustav bernard is an associate director at iqvia who has ...€¦ · gustav bernard is an associate...
TRANSCRIPT
Gustav Bernard is an Associate Director at IQVIA who has been with the company for 14 years. His work focuses on the implementation of CDISC Standards (SDTM, ADaM and Define-XML) within the IQVIA Global Biostatistics department. He is currently working on creating an ADaM specification creation/automation tool. He is also created the Define-XML 2.0 automation process within IQVIA. Gustav earned a bachelor of business in computer science from the University of the Orange Free State in South Africa.
Zach Dorman is a Senior Statistical Programmer at IQVIA. He has been in the CRO industry for 5 years, and has been a programming team lead on a number of studies for various clients, and has been involved in internal programming initiatives relating to CDISC and Define standards. He attended the University of Reading, United Kingdom for his BSc degree in Applied Statistics.
Copyright © 2017 IQVIA. All rights reserved.
PharmaSUG SDE RTP 2018
Gustav Bernard, Zach Dorman
Define XML
Expectations from
Various Clients, Tools
and Industry
Table of
Contents
2
+ Define XML Expectations from Various Clients, Tools and Industry
• Tools - Pinnacle 21
• Groups – CDISC
• Clients - Industry in General
3
Pinnacle 21
Define XML Expectations from Tools
• Define Datasets Sort Order
- Expected to be Sorted by Class then Alphabetically within Class.
- Newer versions of Pinnacle 21 sorts this list Alphabetically and not by Class
• Variable Level Origin should be missing if Value Level Metadata (VLM) for a Variable has multiple origins.
- Want to make sure everyone is aware of this Define-XML rule.
- Expectation from CDISC Define-XML and Pinnacle 21
- Not clear if this is expected from CDISC SDTM and ADaM Standards.
• WhereClause ID from WhereClauses tab within P21 Specification
- This is being created when latest version of P21 Define Generator is used to create the Define.xml file.
4
Dataset Order
Define XML Expectations from Tools
Example Define created with older version of Pinnacle 21, OpenCDISC 2.0.1 Dataset section ordered by Class then
Alphanumeric.
Example Define created with latest version of Pinnacle 21 2.2.0, Dataset section ordered Alphanumerically.
XML Code can be
manually modified
to correct dataset
sort order.
5
Dataset Order
Define XML Expectations from Tools
• The order for SDTM datasets is as follows:
- TRIAL DESIGN
- SPECIAL PURPOSE
- INTERVENTIONS
- EVENTS
- FINDINGS
- FINDINGS ABOUT
- RELATIONSHIP
• The Dataset order for ADaM is as follows:
- SUBJECT LEVEL ANALYSIS DATASET
- BASIC DATA STRUCTURE
- OCCURANCE DATA STRUCTURE
- OTHER
6
Quick Intro Into VLM
Define XML Expectations from Tools
• VLM Value Level Metadata for SDTM or PARAMETER Value Level Metadata for ADaM
• Test Level Variable Derivation
- If you want to break down how a variable was created on a Test \ Parameter Level.
- Pinnacle 21 Excel Specification
› ValueLevel Tab – Each row will represent a VLM Block to be created.
» Similar Structure to the Variables TAB, but Label column is replaced by a WhereClause Column.
▪ WhereClause Column Links to WhereClauses Tab
▪ Codelist Column Links to the Codelist Tab
▪ Method Column Links to the Methods Tab
▪ Value Level Comment Column Links to the Comments Tab
› WhereClauses Tab contains all the Where Clauses needed to select each of the VLM blocks.
» The Same WhereClause can be used for Multiple Blocks.
▪ So if the Same tests are grouped together for two different Target Variables you will only need one WhereClause
to be used for both.
7
Pinnacle 21: Variable Level Origin when Multiple VLM Origins
Define XML Expectations from Tools
• CDISC and P21 Expectation is to have Missing Variable Level Origin if Value Level Metadata has Multiple Origins.
• I would have expected to see the Variable Level Origin to be Derived if there are Multiple Origins within the VLM.
• For Example if you have a dataset where only one test is derived, but all other tests are assigned from SDTM I would
expect the variable to have an Origin of Derived.
Variable Level Missing Origin
Variable Level
Origin Populated
as Expected
8
Visual Illustration
Pinnacle 21: Variable Level Origin when Multiple VLM Origins
VLM for
ADVS.AVAL
with some
Tests having
Origin of
Derived and
Others with
Origin of
Assigned.VLM Block
VLM Block
9
WhereClause ID – Rewritten with Latest version of Pinnacle 21 Tool
Define XML Expectations from Tools
• ID Statement in the WhereClauses Tab in an P21 Excel Specification:
• XML code –Define Generator OpenCDISC 2.0.1: XML Code - Define Generator Pinnacle 21 2.2.0:
QID in Define.XML matches the ID in P21 Spec:
WC.ADVS.DTYPE.HEIGHT=WSTCIRS
QID in Define.XML does not match ID in P21 Spec:
WC.ADVS.PARAMCD.IN.HEIGHT-8f625ab5
I Believe this update was made because of a Lack of
unique ID statements in the P21 Specification.
Extract from P21 User Created Excel Spec
10
WhereClause ID – Example where this can be an issue:
Define XML Expectations from Tools
• Where This can be a issue:
When you have Multiple Where Clauses that start with
the SAME PARAMCD value, two similar WhereClause
QID’s Will be Created.
Workaround: Ensure that the First Value within a list of
tests used are unique, for the above example if the
values of HEIGHT and WEIGHT are swapped a unique
QID will be created for each entry.
11
QID Naming Convention
Define XML Expectations from Various Groups
• Not enough CDISC guidance on:
- Method and Comment QID Naming Convention, how to name duplicate comments and methods
- Naming convention for VLM non-NCI CT codelists
• List of Comments and Methods at end of a Define-XML File
- Most reviewers of an Define-XML package do not understand the rational for re-listing the comments and Methods in
this section.
- This is the main reason why there needs to be more direction with regards to Comment and Method Naming convention.
- Personally I would prefer if the Standard Stylesheet from CDISC be updated to not show the Comments and Methods
Sections when the define.xml is viewed within a browser.
12
QID Naming Convention
Define XML Expectations from Various Clients, Tools and Industry
SDTM and ADaM Sections. Should the CDISC
Stylesheet be updated to not show the Computational
Algorithms and Comments sections for SDTM Defines
and Analysis Derivations and Comments for ADaM?
Naming of Comments and
Methods are simple when there
is a 1:1 correlation to a
variable.
13
QID Naming Convention
Define XML Expectations from Various Groups
Naming of Comments and Methods are
not as simple when there are not a 1:1
correlation to a variable.
Example when the same comment is
used to derive multiple tests from multiple
domains how do you name that specific
comment?
One way to handle this is to find all
repeat comments or methods and then
assign an Numeric Identifier to this
Comment or Method. Since if you were to
include all datasets, PARAMCD’s and
Variable combinations the ID will be >
than 200 chars.Naming of VLM Non NCI CT Controlled
Terms has no predefined naming
convention. Codelist Name Created
using: DatasetName-Variable-CL-Unique
Identifier
14
QID Naming Convention – Define-XML-2-0-Specification
Define XML Expectations from Various Groups
15
QID Naming Convention - Specification for the Operational Data Model (ODM)
Define XML Expectations from Various Groups
16
CDISC Guidance - Supplemental Variable Naming Conventions
Define XML Expectations from Various Clients, Tools and Industry
• Conflicting advice about how to reference supplemental variables in Define submissions
• Three known ways that these are requested to be shown by various clients:
1. SUPPxx.QNAM.<variable-name>
2. SUPPxx.QNAM.<variable-name>.QVAL
3. SUPPxx.QVAL where SUPPDM.QNAM = “<variable-name>”
• We recommend the 3rd approach, for the following reasons:
- The first and second approaches are not factually correct as a file path (the variable name is not a valid variable in
supplemental datasets)
- The third approach shows the correct variable path, as well as information on how to populate the variable from the
supplemental dataset.
17
CDISC Guidance - Hyperlinks within Define
Define XML Expectations from Various Clients, Tools and Industry
• Discrepancies between various browsers in terms of displaying and using hyperlinks to the ADRG and eCRF
• Works in Google Chrome with the style sheet if you use bookmarks
- However, same bookmarks do not work in Internet Explorer
• Using the page number to perform hyperlinking works across both browsers
• Solution: Creation of a Define.html
- This combines XML and the stylesheet, and allows Define packages to be reviewed across any browser
- However, not currently accepted by FDA as acceptable format
18
Industry in General
Define XML Expectations from Various Clients, Tools and Industry
• In General, misunderstanding in Industry in relation to certain topics that go into a Define
- Makes if difficult for a CRO to apply standard Define processes across multiple clients, for a document that should have
a standard format
• Issues noticed as inconsistent are as follows:
- VLM Grouping within the Define XML
- Use of terminology between standards in Define
- Non-CDISC Defines
- Delivery of Define packages
- Purpose of ADRG
19
VLM Grouping and Format
Define XML Expectations from Various Clients, Tools and Industry
• Currently, not clear how VLM should be grouped
- Individual row per test OR Grouping of PARAMCD, PARCATx, by types, lengths and comments
• Character VLM Codelists
• For SDTM the Define-XML CDISC examples only use EQ and for ADaM it uses both EQ and IN.
- Not clear why there is a difference between SDTM and ADaM, due to lack of communication
20
Other Define Topics
Define XML Expectations from Various Clients, Tools and Industry
• Increased requests for creation of Defines for non-CDISC data
- Unsure of whether this can follow the M5 structure for Define packages
• Delivery of Defines across a study
- Vast number of different approaches to how to handle Define deliveries
- Recommendation: Deliver Define package version with each version of the datasets sent
• Creation of Define.pdfs
- Frequent customer request to have PDF files for Define
- Question the need for these – purpose of Define 2.0 is have a printable file, so no requirement for PDFs.
21
ADRG (Analysis Data Reviewers Guide)
Define XML Expectations from Various Clients, Tools and Industry
• Should the ADRG be the biggest document produced in a clinical trial?
- Clients are moving towards adding all information into the guide
- No considering whether information would be useful for a reviewer (derivations and external data clearly stated in
Define, also listed long-form in ADRG)
- Results in lack of clarity for what was done in the creation of ADaM datasets
• List of programs for submission in ADRG
- Should form part of the Reviewers Guide.
- Uncommon to have a complete list during the creation of the ADRG
- Better communication required between CRO, client and FDA upfront regarding what needs submitting
Questions
• Define XML Expectations from Various Clients,
Tools and Industry.