Creating Define.xml v2 Using SAS® for FDA Submissions

Paper 1749-2014
Creating Define.xml v2 Using SAS® for FDA Submissions
Qinghua (Kathy) Chen, Exelixis, Inc. South San Francisco, CA
James Lenihan, Exelixis, Inc. South San Francisco, CA
ABSTRACT
When submitting clinical data to the Food and Drug Administration (FDA), we need to submit the trial results as well
as information that help the FDA understand the data. The FDA has requires the CDISC Case Report Tabulation
Data Definition Specification (Define-XML), which is based on the CDISC Operational Data Model (ODM), for
submissions using Study Data Tabulation Model (SDTM) for this purpose. Electronic submission to the FDA is
therefore a process of following the guidelines from CDISC and FDA. This paper illustrates how to create a CDISC
®
guidance compliant define.xml v2 from metadata using SAS .
INTRODUCTION
In 1999, The FDA standardized the submission of clinical data using the SAS version 5 Transport Format (XPT) and
the submission of metadata using Portable Document Format (PDF) as define.pdf, respectively. In 2005, the study
data specifications published by the FDA included the recommendation that data definitions (metadata) be provided
as a Define-XML file. The FDA has collaborated with the Clinical Data Interchange Standards Consortium (CDISC)
ever since its founding in order to standardize the content and structure of clinical trials data for regulatory
submission. The FDA has included the CDISC Case Report Tabulation Data Definition Specification (define.xml) in
the submission package using SDTM. In March 2013 the final version of define.xml 2.0.0 was released by the
CDISC XML Technologies team. The Define-XML specification has been improved from v1 with better clarifications
and some new features. The define.xml v2.0 has improved the:
•
•
•
•
Support for CDISC Controlled Terminology
Definition of Value Level metadata
The documentation of the data origin or source
The handling of comments.
The initial implementation of define.xml can be a long complicated process for statistical programmers in the
Pharmaceutical Industry. Because most SAS programmers may not have had any exposure to the XML language,
this may affect their production/submission time; since they will be learning the language as they are writing the SAS
code to generate the XML document.
What is XML? XML is an Extensible Markup Language, designed to transport or store data. It is self-descriptive as
you may make your own elements, also known as tags. XML also simplifies data sharing since the data are stored in
plain text format, and is software and hardware independent.
XML documents form a tree structure that starts at "the root" and branches to "the leaves.” An XML document
contains XML Elements and an XML element is everything from the element's start tag to the element's end tag. An
element can contain other elements, text, attributes or a mix of all of the above. Attributes provide additional
information about an element. From the example below you can see that for the root pet, type is an element and
name, nickname, color and age are the attributes of this element. If you have more than one element, a cat in this
example, you would have the cat block repeated with the correct information filled in.
<pet>
<type="CAT">
<name>Spot</name>
<nickname>Pig Monster</nickname>
<color>black</color>
<age>18</age>
</type>
</pet>
Creating Define.xml v2 Using SAS for FDA Submissions
In order to view the define.xml through a browser, we need the XSLT to transform an XML document into HTML.
XSLT is the recommended style sheet language of XML and it is far more sophisticated than CSS. Define.xml v2
recommend by FDA came with define2-0-0.xsl for us to use. So all we need to do is to put define.xml together and
the stylesheet will handle the display.
Here is what Define.XML looks like:
You can see define.xml v2 consists of different components. The first 20 lines should be included in each XML
header. The elements in the gray box should be implemented as needed. Annotated Case Report Form,
Supplemental document may include Reviewers Guide and Complex Algorithms are standalone documents.
Datasets tabulation which is part of Domain Level metadata, Variable Level Metadata, Value Level Metadata,
2
Creating Define.xml v2 Using SAS for FDA Submissions
WhereClause, Controlled Terminology, Computational Algorithms and Comments are metadata embedded. The
metadata for define.xml is kept in an Excel spreadsheet, partially derived from SDTM or ADaM data specifications
when possible, with some critical information added manually. The last three lines are the closing tags for this define
file.
An Annotated Case Report Form is a CRF with SDTM annotations on it. The Reviewers Guide is a document
provided to the FDA reviewer to help them understand the clinical data included and any other information that will
help speed up the review process. For better readability, complex algorithms are used when derivation of certain
variables gets too long to be displayed in the method cell and the entire explanation can be centralized in one place.
There are five steps to create a define file listed below and illustrated in the “Road Map”. They are:
1.
2.
3.
4.
5.
Create the metadata spreadsheet from the SDTM datasets and data spec
Create define.xml components using SAS
Construct all the define.xml supporting documents
Create .xpt files for submitted SAS datasets
Construct define.xml
Road map
5
Annotated
Case Report
Form
SDTM
Specifications
Define.XML
3
1
Reviewer Guide
3
Complex Algorithms
3
Datasets Tabulation
1
1
Value Level Metadata
1
Controlled Terminology
1
Computation Methods
2
Comments
2
Clinical Date in XPT
4
TOC Metadata
2
Variable Metadata
2
2
Enumerated
List/Code List
2
CREATING METADATA SPREADSHEET FROM SDTM SPECIFICATION
In order to improve the programming efficiency, it makes more sense to have one program generate the define.xml
rather than doing it manually. So it is very important to create the metadata spreadsheet for each of the sections
consistently across studies.
CREATING METADATA SPREADSHEET FROM THE SDTM DATA SPEC:
The following are the metadata we need to create define.xml:
1. Header definitions
2. Datasets (TOC) definitions
3. Dataset variable definitions
4. Value level definitions
3
Creating Define.xml v2 Using SAS for FDA Submissions
5.
6.
7.
Computational Method
Controlled terminology definitions and code list
Comments
Header Metadata:
Header meatadata contains the information about the study, SDTM version and the version of the style sheet used.
FILE
OID
STUDY
OID
STUDYN
AME
STUDYDESCR
IPTION
PROTOCOL
NAME
STAND
ARD
VERSI
ON
SCHEMALOC
ATION
STYLES
HEET
XYZ123
123
XYZ123
A PHASE IIB,
DOUBLE-BLIND,
MULTI-CENTER,
PLACEBO
CONTROLLED,
PARALLEL GROUP
TRIAL OF ANALGEZIA
HCL FOR THE
TREATMENT OF
CHRONIC PAIN
XYZ123
SDTM
3.1.2
http://www.cdisc.org/n
s/odm/v1.3 define2-00.xsd
define2-0-0.xsl
TOC metadata:
The TOC metadata should contain definitions of all the datasets sent to FDA. The ORDID variable is used to
determine the order of domains in the display.
NA
ME
REPE
ATIN
G
ISREFEREN
CEDATA
PURPO
SE
LA
BE
L
STRUC
TURE
DOMAIN
KEYS
CL
AS
S
ARCHIVELO
CATIONID
TA
Yes
Yes
Tabulation
Trial
Arm
s
One record
per planned
Element per
Arm
STUDYID,
ARMCD,
TAETORD
Trial
Desig
n
../export/ta
DOCUMEN
TATION
ORDID
1
Variable Level Metadata:
The variable level metadata is used to describe the variables within each domain. The KEYSEQUENCE variable is
used to provide a sequential number key variables so the stylesheet can use it to populate the DOMAINKEYS in the
TOC session in define file automatically.
D
O
M
AI
N
VA
RN
U
M
VA
RIA
BL
E
T
Y
P
E
LE
N
GT
H
L
A
B
E
L
TA
1
STU
DYID
tex
t
15
TA
3
ARM
CD
tex
t
8
Stu
dy
Ide
ntifi
er
Pla
nne
d
Ar
m
Co
de
SIGNI
FICAN
TDIGIT
S
O
RI
GI
N
KEY
SEQ
UEN
CE
Der
ive
d
1
DISPL
AYFO
RMAT
COMPUT
ATIONME
THODOID
Der
ive
d
COD
ELIS
TNA
ME
ARMCD
MA
NDA
TOR
Y
R
O
L
E
ROL
ECO
DELI
ST
Yes
Ide
ntif
ier
ROLEC
ODE
Yes
To
pic
ROLEC
ODE
VAL
UELI
STOI
D
Value level metadata:
The value level metadata in define.xml v2 has changed significantly from v1. With where clause included, it not only
allows us to show the condition(s) to use to subset the data, it also provides link(s) back to its source domain so you
can refer to it when needed. The VALUELISTOID shows you for which variable the value level metadata is created.
The ITEMOID is used to define the record within VALUELISTOID and WHERECLAUSEOID that is used to
display/link the subset condition(s). When a DESCRIPTION is needed for one particular parameter in SUPPXX, you
can populate it with the source variable label and then it will show after the variable in the display within parentheses.
VAL
UEL
IST
OID
VA
LU
EO
RD
ER
VAL
UEN
AME
ITEMOID
WHEREC
LAUSEOI
D
TY
P
E
L
E
N
G
T
H
4
ORIG
IN
COM
PUTA
TION
METH
ODOI
D
COD
ELIS
TNA
ME
SIGN
IFIC
ANT
DIGI
TS
DISPL
AYFO
RMAT
MA
ND
AT
OR
Y
DE
SC
RIP
TIO
N
Creating Define.xml v2 Using SAS for FDA Submissions
VL.LB.
LBOR
RES
1
LBORR
ES
IT.LB.LBORRES.B
ILI.LBCAT.CHEMI
STRY.LBSPEC.BL
OOD
WC.LB.LBTEST
CD.BILI.LBCAT.
CHEMISTRY.LB
SPEC.BLOOD
float
3
eDT
VL.SU
PPDM.
QVAL
1
QVAL
IT.SUPPDM.QVAL
.RACE1
WC.SUPPDM.Q
NAM.RACE1
text
5
CRF
Page 6
4.2
YES
RACE
NO
Race
1
Here is how the label (Race 1) is shown in Define.xml.
Type
Variable
Length /
Display
Format
5
Where
QNAM EQ
RACE1
(Race 1)
QVAL
text
Controlled
Terms or
Format
["BLACK OR
AFRICAN
AMERICAN",
"OTHER",
"WHITE"]
<RACE>
Origin
Derivation/Comment
CRF Page
6
Computational Methods:
The Computational Methods section is used to display all the derivation rules used together in the datasets
submitted. COMPUTATIONMETHODOID is used to link to the Derivation/Comment and Computational Method
section. When the derivation of a variable can’t fit into a table cell nicely it should be saved in an external file called
Complex Algorithms.pdf. You should note that this metadata is built on prototyping data therefore the
DESCRIPTION of the computational method contains the pseudo code. When you prepare for an actual filing, actual
text needs to be written.
COMPUTATIONMETHODOID
COMPUTATIONMETHODNAME
COMPUTATIONMETHODTYPE
DESCRIPTION
MT.AESTDY
Algorithm to derive AESTDY
Computation
AESTDY = AESTDTC
- RFSTDTC+1 if
AESTDTC is on or
after RFSTDTC.
AESTDTC RFSTDTC if
AESTDTC precedes
RFSTDTC
Codelist:
Codelist is a unique subset of the SDTM controlled terminology. For CDISC codelists, aliases are included at both
the CodeListItem and EnumeratedItem levels. When a description is needed for a coded term, the DECODED
variable should be marked as ‘YES’ so both the coded term and the translated term will be in the display. Rank is
provided for each EnumeratedItem to indicate the temporal order in analysis computation.
CODEL
ISTNA
ME
CODELI
STLABE
L
AE
Domain
Abbreviation
(AE)
Domain
Abbreviation
(DM)
Action Taken
with Study
Treatment
Action Taken
with Study
Treatment
Action Taken
with Study
Treatment
DM
ACN
ACN
ACN
R
A
N
K
1
2
3
CODE
DVAL
UE
TRAN
SLAT
ED
T
Y
P
E
CODELIST
DICTIONA
RY
source
variab
le
sourc
evalu
e
sour
cety
pe
AE
Adverse
Events
text
YES
DM
Demograph
ics
text
YES
DOSE
INCREASE
D
DOSE NOT
CHANGED
DOSE
INCREASE
D
DOSE NOT
CHANGED
text
aeaction
3
number
text
aeaction
4
number
DOSE
REDUCED
DOSE
REDUCED
text
aeaction
2
number
5
CODELI
STVERSI
ON
DEC
ODE
D
Creating Define.xml v2 Using SAS for FDA Submissions
Comments:
When a comment or an external document is referenced, we should create a record with COMMENTOID and a
description for it. Here is how it looks:
COMMENTOID
DESCRIPTION
COM.LBREFID
Accession number
COM.RELTYPE
All values are null since this is used only when identifying a dataset-level
relationship.
As you can see, some part of this metadata specification can be populated programmatically depending on how your
SDTM/ADaM specifications are constructed. Some information has to be filled manually. You can find an example of
how we created this metadata specification in Appendix A. In this case, we used a macro to create variable level
metadata from SAS datasets and SDTM specification, and read in the metadata template for all of the other tabs.
The benefit of doing this is that when the template changes, we don’t have to go back and change the SAS program
and thus makes our programs easier to maintain.
CREATE DEFNE.XML COMPONENTS USING SAS
Once the metadata specification is complete, we can use it to populate the define file. In order to populate the define
file correctly, we need to know how each variable is used in it. The Define-XML-2-0-Specification section 5 describes
how XML file is constructed, and how each element(s) and/or attribute(s) is defined/constructed. Based on that
specification, we can map the metadata into an element or attribute of ODM to form the define file in the end.
READING IN METADATA
The creation of define.xml basically involves taking data from the metadata spreadsheet one tab at a time and writing
it out in .XML format as a text file. The following code is used to read in the DEFINE_HEADER_METADATA sheet
from SDTM_METADATA.xls and convert the metadata to a SAS dataset called define_header. We should repeat
this code for all other sheets until we read in all the information needed for define.xml.
%xls2sas( _datarow=2,
_indir=../&path,
_infile=SDTM_METADATA.xls,
_labels=1,
_outdata=define_header,
_sheet=DEFINE_HEADER_METADATA,
_subset=,
_vars=1);
CREATING .TXT FILE ONE; FOR EACH OF THE COMPONENTS
Define_header.txt contains the header information for the XML file as well as the links to each component of the
define file. Variables studyoid, studyname, studydescription, protocolname, standard, version, schemalocation read
in from METADATA tab were populated into Define.xml elements accordingly. In addition to that, the link to the
external documents blankcrf.pdf and Reviewer’s Guide are created here by def:AnnotatedCRF and
def:SupplementalDoc.
filename dheader "define_header.txt";
data define_header;
set define_header;
file dheader notitles;
creationdate = compress(put(datetime(), IS8601DT.));
put @1 '<?xml version="1.0" encoding="UTF-8" ?>' /
@1 '<?xml-stylesheet type="text/xsl" href="' stylesheet +(-1) '"?>' /
6
Creating Define.xml v2 Using SAS for FDA Submissions
@1 '<!-******************************************************************************* -->' /
@1 '' /
@1 "" /
@1 '' /
@1 '<!-******************************************************************************* -->' /
@1 '<ODM' /
@3 'xmlns="http://www.cdisc.org/ns/odm/v1.3"' /
@3 'xmlns:def="http://www.cdisc.org/ns/def/v2.0"' /
@3 'xmlns:xlink="http://www.w3.org/1999/xlink"' /
@3 'xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"' /
@3 'xsi:schemaLocation="http://www.cdisc.org/ns/def/v2.0' /
'../schema/cdisc.org-define-2.0/define2-0-0.xsd"' /
%if "&standard" = "ADAM" %then
@3 'xmlns:adamref="http://www.cdisc.org/ns/ADaMRes/DRAFT"' /
;
@3 'ODMVersion="1.3.2"' /
@3 'FileType="Snapshot"' /
@3 'FileOID="StudyXYZ123-Define-XML_2.0.0"' /
@3 'CreationDateTime="' creationdate +(-1) '"' /
@3 'Originator="CDISC XML Technologies Team"' /
@3 'SourceSystem="MH-System">' /
@3 'SourceSystemVersion="2.0.1">' /
@1 '<Study OID="' studyoid +(-1) '">' /
@3 '<GlobalVariables>' /
@5 '<StudyName>' studyname +(-1) '</StudyName>' /
@5 '<StudyDescription>' studydescription +(-1) '</StudyDescription>' /
@5 '<ProtocolName>' protocolname +(-1) '</ProtocolName>' /
@3 '</GlobalVariables>' /
@3 '<MetaDataVersion OID="CDISC.' standard +(-1) '.' version +(-1) '"' /
@5 'Name="' studyname +(-1) ',Data Definitions"' /
@5 'Description="' studyname +(-1) ',Data Definitions"' /
@5 'def:DefineVersion="2.0.0"' /
@5 'def:StandardName="'standard +(-1) '-IG"' /
@5 'def:StandardVersion="' version +(-1) '">' /
%if "&standard" = "SDTM" %then
%do;
@5 '<def:AnnotatedCRF>' /
@7 '<def:DocumentRef leafID="LF.blankcrf"/>' /
@5 '</def:AnnotatedCRF>' /
@5 '<def:leaf ID="LF.blankcrf" xlink:href="blankcrf.pdf">' /
@7 '<def:title>Annotated Case Report Form</def:title>' /
@5 '</def:leaf>'/
%end;
@5 '<def:SupplementalDoc>' /
@7 '<def:DocumentRef leafID="LF.ReviewersGuide"/>' /
@7 '<def:DocumentRef leafID="LF.ComplexAlgorithms"/>' /
@5 '</def:SupplementalDoc>' /
@5 '<def:leaf ID="LF.ReviewersGuide" xlink:href="reviewersguide.pdf">' /
@7 '<def:title>Reviewers Guide</def:title>' /
@5 '</def:leaf>'/
@5 '<def:leaf ID="LF.ComplexAlgorithms"
xlink:href="complexalgorithms.pdf">' /
@7 '<def:title>Complex Algorithms</def:title>' /
@5 '</def:leaf>'/;
run;
7
Creating Define.xml v2 Using SAS for FDA Submissions
TOC OF DOMAINS
The Itemgroupdef.txt is used to create the TOC part of define.xml. ItemGroupDef is used to define the ItemGroup
and ItemRef is used to refer to the key variables specified in variable metadata and any comments needed. The
variables highlighted in bold black fonts below are read from TOC_METADATA AND VARIABLE_METADATA
sheets.
filename igdef "itemgroupdef.txt";
data itemgroupdef;
length label $ 40;
merge toc_metadata(rename=(label=label1)) VARIABLE_METADATA;
by oid;
file igdef notitles;
if first.oid then
do;
put @5 "" /
@5 "" /
@5 "" /
@5 '<ItemGroupDef OID="IG.' oid +(-1) '"' /
@7 'Domain="' name +(-1) '"' /
@7 'Name="' name +(-1) '"' /
@7 'Repeating="' repeating +(-1) '"' /
@7 'IsReferenceData="' isreferencedata +(-1) '"' /
@7 'SASDatasetName="' name +(-1) '"' /
@7 'Purpose="' purpose +(-1) '"' /
@7 'def:Structure="' structure +(-1) '"' /
@7 'def:Class="' class +(-1) '"' /
@7 'def:CommentOID="' documentation +(-1) '"';
%if &standard=ADAM %then
put @7 'def:ArchiveLocationID="LF.' archivelocationid +(-1) '"' /
@7 'Comment="' documentation +(-1) '">'
;
%else
put @7 'def:ArchiveLocationID="LF.' archivelocationid +(-1) '">' /
;
;
put @7 '<Description>' /
@7 '<TranslatedText xml:lang="en">' label1 +(-1) '</TranslatedText>' /
@7 '</Description>';
end;
put @7 '<ItemRef ItemOID="IT.' itemoid +(-1) '"' /
@9 'OrderNumber="' varnum +(-1) '"' / ;
if keysequence ne ' ' then
put @9 'KeySequence="' keysequence +(-1) '"' / ;
if computationmethodoid ne ' ' then
put @9 'MethodOID="' computationmethodoid +(-1) '"' / ;
;
put @9 'Mandatory="' mandatory +(-1) @;
if role ne '' and "&standard" = "SDTM" then
put '"' /
@9 'Role="' role +(-1) '"' /
@9 'RoleCodeListOID="CodeList.' rolecodelist +(-1) '"/>';
else
put '"/>';
if last.oid then
put @7 "" /
@7 "" /
8
Creating Define.xml v2 Using SAS for FDA Submissions
@7
@7
'.xpt">' /
@9
@7
@5
run;
"" /
'<def:leaf ID="LF.' oid +(-1) '" xlink:href="' archivelocationid +(-1)
'<def:title>' archivelocationid +(-1)
'</def:leaf>' /
'</ItemGroupDef>';
'.xpt </def:title>' /
VARIABLE LEVEL:
The Itemdef.txt is used to create the variable level part of define.xml. ItemDef is used to create the variable element.
The CodeListRef and def:ValueListRef are used for codelist display and value level data drill down respectively.
The variables used here are from VARIABLE_METADATA sheet.
**** CREATE "ITEMDEF" SECTION;
filename idef "itemdef.txt";
data itemdef;
set VARIABLE_METADATA end=eof;
by oid;
file idef notitles;
if _n_ =
put @5
@5
@5
1 then
"" /
"" /
"" ;
put @5 '<ItemDef OID="IT.' itemoid +(-1) '"' /
@7 'Name="' variable +(-1) '"' /
@7 'DataType="' type +(-1) '"' /
@7 'Length="' length +(-1) '"' /
@7 'SASFieldName="' variable +(-1) '"' ;
if significantdigits ne '' then
put @7 'SignificantDigitis="' significantdigits +(-1) '"';
if displayformat ne '' then
put @7 'def:DisplayFormat="' displayformat +(-1) '"';
if computationmethodoid ne '' then
put @7 'def:MethodOID="' computationmethodoid +(-1) '"';
put %if "&standard" = "SDTM" %then
@7 'Origin="' origin +(-1) '"' / ;
@7 'Comment="' comment +(-1) '"' /
@7 'def:Label="' label +(-1) '">';
if codelistname ne '' then
put @7 '<CodeListRef CodeListOID="CodeList.' codelistname +(-1) '"/>';
if valuelistoid ne '' then
put @7 '<def:ValueListRef ValueListOID="' valuelistoid +(-1) '"/>';
put @7 '<Description>' /
@7 '<TranslatedText xml:lang="en">' label +(-1) '</TranslatedText>' /
@7 '</Description>';
put @5 '</ItemDef>';
run;
VALUE LEVEL:
Value Level Metadata should be provided when there is a need to describe different metadata attributes for
subsetting the data within a column. It can also be used on other types of SDTM domains to provide information
9
Creating Define.xml v2 Using SAS for FDA Submissions
useful for data interpretation but, the value level metadata approach is very complicated so it is recommended to use
it only when it is absolutely necessary. The value level metadata consists of three parts: the itemdef_value, value list
and the whereclause list. The Itemdef_value defines how the value level metadata is displayed, and the where
clause shows the subset condition in define.xml and the value list is used to link to the whereclause definition. The
ItemDef is used to create the valuelist item, and def:WhereClauseDef and def:ValueListDef are used to create the
whereclause and valuelist. Conditional Description is used for variables in SUPPXX domains that need a label. The
variables read from value level metadata are used here. As you can see from the code below, we need to parse the
WhereClauseId so the stylesheet can get the information it needs and display it correctly in the browser. Also, it is
worthwhile to mention that this code is for prototyping, and cannot handle all real-world situations as yet. You may
find it fun to dig through the stylesheet and try to figure out how to make it work.
filename idefvl "itemdef_value.txt";
data itemdefvalue;
set valuelevel end=eof;
by valuelistoid;
file idefvl notitles;
if _n_ =
put @5
@5
@5
1 then
"" /
"" /
"" ;
put @5 '<ItemDef OID="' itemoid +(-1) '"' /
@7 'Name="' valuename +(-1) '"' /
@7 'DataType="' type +(-1) '"' /
@7 'Length="' length +(-1) '"';
if significantdigits ne '' then
put @7 'SignificantDigitis="' significantdigits +(-1) '"';
if displayformat ne '' then
put @7 'def:DisplayFormat="' displayformat +(-1) '"';
if computationmethodoid ne '' then
put @7 'def:MethodOID="' computationmethodoid +(-1) '"';
put @5 '>';
if description ne ' ' then
put @7 '<Description>' /
@7 '<TranslatedText xml:lang="en">' description +(-1) '</TranslatedText>' /
@7 '</Description>';
if codelistname ne '' then
put @7 '<CodeListRef CodeListOID="CL.' codelistname +(-1) '"/>';
put @7 '<def:Origin Type="' origint +(-1) '"' /;
if origint ^= "CRF" then
put @7 '/>' ;
else
put @7 '>' /
@9 '<def:DocumentRef leafID="LF.blankcrf">' /
@11 '<def:PDFPageRef PageRefs=" ' originpn +(-1)'" Type="PhysicalRef"/>'
/
@9 '</def:DocumentRef>' /
@7 '</def:Origin>' /;
put @5 '</ItemDef>';
run;
filename whrlist "wherelist.txt";
data valuelevel;
set valuelevel;
by valuelistoid itemoid;
10
Creating Define.xml v2 Using SAS for FDA Submissions
file whrlist notitles;
if _n_ =
put @5
@5
@5
1 then
"" /
"" /
"";
if first.itemoid then;
put @5 '<def:WhereClauseDef OID="' whereclauseoid +(-1) '">';
do i=3 to countw(whereclauseoid,'.')-1 by 2;
varcat=compress('IT.' || scan(whereclauseoid,2,'.') || '.' ||
scan(whereclauseoid,i,'.'))
;
varval=compress(scan(whereclauseoid,i+1));
put @7 '<RangeCheck SoftHard="Soft" def:ItemOID="' varcat +(-1) '"
Comparator="EQ">' /
@11 '<CheckValue>' varval +(-1) '</CheckValue>' /
@7 '</RangeCheck>';
end;
if last.itemoid then
put @5 '</def:WhereClauseDef>';
run;
filename vallist "valuelist.txt";
data valuelevel;
set valuelevel;
by valuelistoid;
file vallist notitles;
if _n_ =
put @5
@5
@5
1 then
"" /
"" /
"";
if first.valuelistoid then
put @5 '<def:ValueListDef OID="' valuelistoid +(-1) '">';
put @7
@9
@9
if
'<ItemRef ItemOID="' itemoid +(-1) '"' /
'OrderNumber="' valueorder +(-1) '"' /
'Mandatory="' mandatory +(-1) '"' ;
computationmethodoid ne '' then
put @9 'MethodOID="' computationmethodoid +(-1) '"';
put @9 '>' /
@9 '<def:WhereClauseRef WhereClauseOID="' whereclauseoid +(-1) '"/>' /
@7 '</ItemRef>';
if last.valuelistoid then
put @5 '</def:ValueListDef>';
run;
It is worthwhile to mention that the way the stylesheet handles the variable label in the whereclause is different for the
SUPPXX domains than others. For the value level metadata in the case of supplemental qualifiers, the stylesheet will
display the label description. This is conditional on the availability of a description. (Such as RACE1, RACE2, ..,
RACEOTH will have "Race 1", "Race 2", ..., "Race, Other"). For other types of whereclauses the stylesheet will try to
display the decoded value if controlled terminology is attached to the variable that is part of the whereclause.
COMPUTATIONAL METHODS:
11
Creating Define.xml v2 Using SAS for FDA Submissions
MethodDef is used to display the computation method used.
filename comp "compmethod.txt";
data compmethods;
set compmethod;
file comp notitles;
if _n_ = 1 then
put @5 "" /
@5 "" /
@5 "";
put @5 '<MethodDef OID="' computationmethodoid +(-1) '"'/
@5 'Name="' computationmethodname +(-1) '"' /
@5 'Type="' computationmethodtype +(-1)'">' /
@7 '<Description>' /
@7 '<TranslatedText xml:lang="en">' description +(-1) '</TranslatedText>' /
@7 '</Description>' /
@5 '</MethodDef>';
run;
COMMENTS:
def:CommentDef is used to display the comments.
**** CREATE COMMENTS SECTION;
filename comp "comments.txt";
data comments;
set comments;
file comp notitles;
if _n_ = 1 then
put @5 "" /
@5 "" /
@5 "";
put @5 '<def:CommentDef OID="' commentoid +(-1) '">'/
@7 '<Description>' /
@7 '<TranslatedText xml:lang="en">' description +(-1) '</TranslatedText>' /
@7 '</Description>' /
@5 '</def:CommentDef>';
run;
CODELIST:
The Codelist section has changed dramatically from v1 as well. Not only did the Codelist CODE get added to the
display, codelist has been separated into EnumeratedItem and CodeListItem. The Alias Context attribute
refereneces “nci:ExtCodeID” to indicate the NCI/CDISC SDTM Terminology standards. When the coded value is the
same as the translated value, only code value is shown. If the codelist is saved in an external dictionary, a separate
section will be created just for that.
Please note, I used the CODELIST as the last section in the define file as the order of elements for each section does
not matter. If you choose to do it differently, you can move the last 4 lines of code to a different place, where
appropriate. You can also create a separate section if you like.
12
Creating Define.xml v2 Using SAS for FDA Submissions
filename comp "codelist.txt";
data codelists;
set codelists end=eof;
by codelistname rank;
length codelistname $200;
file codes notitles;
if _n_ =
put @5
@5
@5
1 then
"" /
"" /
"" ;
if first.codelistname then do;
if index(upcase(ncipreferredterm),'DOMAIN') then
put @5 '<CodeList OID="CL.' codelistname +(-1) '.DOMAIN"';
else put @5 '<CodeList OID="CL.' codelistname +(-1) '"';
put @7 'Name="' codelistlabel +(-1) '"' /
@7 'DataType="text">';
end;
**** output codelists that are not external dictionaries;
if codelistdictionary = '' then do;
if decoded ne 'YES' then do;
put @7 '<EnumeratedItem CodedValue="' codedvalue +(-1) '"' @;
if rank ne . then
put ' OrderNumber="' rank +(-1) '">';
else
put '>';
if code ne ' ' then
put @7 '<Alias Name="' code +(-1) '" Context="nci:ExtCodeID"/>';
put @7 '</EnumeratedItem>';
end;
else do;
put @7 '<CodeListItem CodedValue="' codedvalue +(-1) '"' @;
if rank ne . then
put ' OrderNumber="' rank +(-1) '">';
else
put '>';
put @9 '<Decode>' /
@11 '<TranslatedText>' translated +(-1) '</TranslatedText>' /
@9 '</Decode>' ;
if code ne ' ' then
put @7 '<Alias Name="' code +(-1) '" Context="nci:ExtCodeID"/>';
put @7 '</CodeListItem>';
end;
if codelist ne ' ' then
put @7 '<Alias Name="' codelist +(-1) '" Context="nci:ExtCodeID"/>';
end;
**** output codelists that are pointers to external codelists;
else if codelistdictionary ne '' then
put @7 '<ExternalCodeList Dictionary="' codelistdictionary +(-1)
'" Version="' codelistversion +(-1) '"/>';
if last.codelistname then
put @5 '</CodeList>';
if eof then
put @3 '</MetaDataVersion>' /
@1 '</Study>' /
@1 '</ODM>';
run;
13
Creating Define.xml v2 Using SAS for FDA Submissions
CONSTRUCT ALL THE DEFINE.XML SUPPORTING DOCUMENTS
Create Annotated CRF based on SDTM specifications and put lengthy data derivation rules, if any, in Complex
Algorithms. For anything else that will help the FDA to understand our data and to speed up the review process, we
should put that information in the Reviewer’s Guide.
CREATING .XPT FILES FOR CLINICAL DATA
Since FDA standardized the submission data format in 1999, converting SAS datasets to .xpt files in version 5 is a
must. There are different ways to convert SAS datasets to .xpt file, and the following SAS code provides one
example:
Libname source ‘SAS-data-library’;
Libname xportout xport ‘transport-file’;
proc copy in=source
out=xportout memtype=data;
run;
CONSTRUCT DEFINE.XML
Here is the command that was used in Unix to concatenate the text files together to form the define.xml.
cat define_header.txt compmethod.txt valuelist.txt itemgroupdef.txt itemdef.txt itemdef_value.txt comments.txt
codelist.txt > define.xml
With the Stylesheet’s help, Define.xml v2 can be viewed via IE or Firefox. The define.xml file is listed on the right for
your information.
CONCLUSIONS
Creating define.xml is difficult especially if you don’t have any knowledge about XML at the beginning. However, with
the approach I have outlined above it is readily achievable.
14
Creating Define.xml v2 Using SAS for FDA Submissions
APPENDIX A:
/***************************************************
*H*
*H* PROGRAM:
mk_varmeta.sas
*H*
*H* USAGE:
create the SDTM variable_metadata sheet from the STDM
specs
*H* REQUIRES anadata:
*H* REQUIRES rawdata:
*H* REQUIRES sdtmdata:
*H* REQUIRES export:
XL184XXX_SDTM_Speciifcatins_V8.0.xls
*H* REQUIRES macros:
%xls2sas %sas2xls
*H* PRODUCES:
*H*
*H*
*H* REVISION HISTORY:
*H*
20131216 jkl Created.
*H*
*H* $Id:$
***************************************************/
options mlogic mprint symbolgen validvarname=upcase ;
*PN-----------------------------------------------------------------*;
*PN-----------------------------------------------------------------*;
*PN macro that
1) converts an individual sheet from an xls file to SAS
2) select and label variables
3) creates the dataset "final" by concatinating the datasets
*PN-----------------------------------------------------------------*;
*PN-----------------------------------------------------------------*;
*PN------------------------------------------------*;
*PN macro to convert xls sdtmdata sheet to sas data*;
*PN------------------------------------------------*;
%xls2sas(_datarow = 2,
_indir = ../import,
_infile = Def_head.xls,
_labels =1,
_outdata = DEFINE_HEADER_METADATA,
_sheet =sheet1 ,
_vars = )
;
*PN --------------------------------------------------------------------*;
*PN --BRING IN SDTM VARIABLES FROM SPEC SPREADSHEET -------------------*;
*PN --------------------------------------------------------------------*;
%macro doit(dsn=,sort=);
%xls2sas(_datarow
_indir
_infile
=
=
=
7,
,
XL184307_SDTM_Specifications_V8.0.xls,
15
Creating Define.xml v2 Using SAS for FDA Submissions
_labels =
_outdata =
_sheet
=
);
6,
,
&dsn
data &dsn;
retain domain var16 var8 var17 var15 var12 var14 var18 var19 ;
set &dsn(drop=VAR1-VAR7 );
*PN------------------------------------*;
*PN crete the DOMAIN and SORT variables*;
*PN------------------------------------*;
domain = "&dsn";
sort = "&sort";
label
var8
var12
var13
var14
var15
var16
var17
var18
var19
domain
=
=
=
=
=
=
=
=
=
=
'VARIABLE'
'DISPLAYFORMAT'
'FLAG'
'COMPUTATIONALMETHODOID'
'COMMENTS'
'VARNUM'
'ORIGIN'
'MANDATORY'
'ROLECODELIST'
'DOMAIN'
;
if var8= '' or var13 ne ''
sort_var=var16*1.0;
then delete;
drop var13 var8;
run;
*PN-------------------------*;
*PN concatenate the datasets*;
*PN-------------------------*;
proc append base=specs data=&dsn force;
run;
proc datasets library=work;
delete &dsn;
quit;
%mend;
*PN--------------------------*;
*PN import from the SDTM spec*;
*PN--------------------------*;
%doit(dsn=AE,sort=A);
%doit(dsn=CD,sort=C);
%doit(dsn=CM,sort=D);
%doit(dsn=DA,sort=E);
%doit(dsn=DD,sort=F);
%doit(dsn=DM,sort=G);
16
Creating Define.xml v2 Using SAS for FDA Submissions
*PN---------------------------------------------------------------------*;
*PN------------ THIS SECTION EXTRACTS INFORMATION FROM SDTM DATA -------*;
*PN---------------------------------------------------------------------*;
%macro doit2(dsn=,sort=);
*PN---------------------------*;
*PN get the contents datasets *;
*PN---------------------------*;
proc contents data=sdtmdata.&dsn out=&dsn noprint;
run;
proc sort data=&dsn;
by varnum;
run;
proc print data=&dsn;
title "supp &dsn";
run;
data &dsn;
length mandatory $3 type $4;
set &dsn(drop = type rename=(memname=domain name=variable));
by varnum;
*PN------------------------------------------------------*;
*PN populate the Mandtory column in the variable_metadata*;
*PN------------------------------------------------------*;
if varnum in('4','5','10') then mandatory='No';
else mandatory='Yes';
*PN----------------------------------------------*;
*PN create the type column and the sort variable *;
*PN----------------------------------------------*;
type='text';
sort = "&sort";
keep domain varnum variable type length label mandatory sort;
label domain='DOMAIN'
;
run;
data &dsn(rename=(temp11=var11 temp16=var16));
retain domain var16 var8 var10 var11 var9 var18
length temp11 $3 temp16 $3;
set &dsn(rename=(varnum
= var16
variable = var8
type
= var10
length
= var11
label
= var9
mandatory= var18))
;
*PN------------------------*;
*PN make length and varnum *;
*PN------------------------*;
17
;
Creating Define.xml v2 Using SAS for FDA Submissions
temp11 = left(put(var11,4.0));
temp16 = left(put(var16,2.0));
drop var11 var16;
label
var8
var9
var10
temp11
temp16
var18
=
=
=
=
=
=
'VARIABLE'
'LABEL'
'TYPE'
'LENGTH'
'VARNUM'
'MANDATORY'
;
run;
*PN --------------------*;
*PN append the datasets *;
*PN---------------------*;
proc append base=final data=&dsn force;
run;
/*
proc print data=&dsn label;
title "&dsn";
run;
*/
*PN------------------------------*;
*PN clean out the work directory *;
*PN------------------------------*;
proc datasets library=work;
delete &dsn;
quit;
%mend;
*PN---------------------------------------------*;
*PN import datasets from SDTM data including SPP*;
*PN---------------------------------------------*;
%doit2(dsn=AE,sort=A);
%doit2(dsn=CD,sort=C);
%doit2(dsn=CM,sort=D);
%doit2(dsn=DA,sort=E);
%doit2(dsn=DD,sort=F);
%doit2(dsn=DM,sort=G);
%doit2(dsn=SUPPAE,sort=B);
%doit2(dsn=SUPPDM,sort=H);
data final;
length domain $6;
set final;
*PN----------------------------------------------------*;
*PN need numeric to sort the variable numbers correctly*;
*PN----------------------------------------------------*;
sort_var = var16*1.0;
run;
18
Creating Define.xml v2 Using SAS for FDA Submissions
proc sort data=final out=VARIABLE_METADATA/*(drop=sort sort_var)*/;
by sort sort_var ;
run;
proc sort data=specs/*(drop=sort sort_var)*/;
by sort sort_var ;
run;
*PN-----------------------------------------------------------------------*;
*PN merge the vars from the specc SS onto data imported from SDTM datasets*;
*PN-----------------------------------------------------------------------*;
data VARIABLE_METADATA;
merge VARIABLE_METADATA(in=a) specs;
by sort sort_var ;
if a;
drop sort sort_var;
run;
*PN--------------------------------------------------------------*;
*PN----------------- export to a spread sheet -------------------*;
*PN--------------------------------------------------------------*;
%sas2xls(_inlib=work,
_first=domain var16 ,
_header=L,
_files = DEFINE_HEADER_METADATA VARIABLE_METADATA ,
_outfile = SDTM_METADATA);
19
Creating Define.xml v2 Using SAS for FDA Submissions
REFERENCES
•
•
•
•
•
•
http://support.sas.com/publishing/authors/holland_chris.html
http://support.sas.com/resources/papers/proceedings13/201-2013.pdf
http://support.sas.com/resources/papers/proceedings10/157-2010.pdf
Define-XML v2 – Whatis New by Lex Jansen, SAS Institute Inc., Cary, NC, USA
Implementing CDISC Using SAS: by Chris Holland and Jack Shostak
Define-XML-2-0-Specification
SAS and all other SAS Institute Inc. products or service names are registered trademarks of SAS Institute Inc. in the
USA and other countries. ® indicates USA registration.
Other brand and product names are registered trademarks or trademarks of their respective companies.
ACKNOWLEDGEMENTS
Lex Jassen, SAS Institute.
Amos Shu, Medimmune Inc.
CONTACT INFORMATION
Your comments and questions are valued and encouraged. Contact the author at:
Qinghua (Kathy) Chen
Exelixis Inc
210 East Grand Ave,
South San Francisco
CA, 94080
Email: [email protected]
James Lenihan
Exelixis Inc.
210 East Grand Ave,
South San Francisco
CA, 94080
Email: [email protected]
20

Thank you for your participation!

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