Download An Integrated Approach to Computer Systems for NDA Preparation and Presentation

AN INTEGRATED APPROACH TO COMPUTER SYSTEMS
FOR NDA PREPARATION AND PRESENTATION
Martin J. Rosenberg, PhD,
MAJARO INFOSYSTEMS,
INC.
ABSTRACT
Increasingly,
pharmaceutical
companies
are
attempting to enhance productivity by expanding the
scope of their clinical information systems to embrace
non-traditional users such as: investigators, in-house
clinical staff, and FDA reviewers. As these systems
become commonplace, companies are confronted
with the prospect of having up to five different
computer systems, each with incompatible data
structures: a remote data entry system; a database
management system for traditional date entry; a
clinical data review system for use by in-house clinical
staff; a statistical analysis system; and a computer
assisted NDA review system (CANDA).
in order to look for toxic effects in man. As evidence
of safety is accumulated, dosages are increased to
therapeutiC levels. In Phase II, the drug is tried for the
first time in human volunteers with the targeted
disease. Sample sizes are small, and information is
gathered concerning the appropriate dose and
regimen to use to obtain a therapeutic effect. Once
sufficient Information is gathered and hypotheses
about the drug's efficacy in various indications have
been formulated, large scale Phase III testing in human
volunteers begins.
The research process can take five or more years.
Throughout the process additionai information is
gathered concerning the safety of the drug in both
humans and animals; the stability 01 the drug (how
long it can remain on the sheff without degrading); the
pharmacokinetics of the drug, I.e., how it is
metabolized in humans; and the ability of the company
This papar describes the emerging features of SAS
software, including Version 6 enhancements such as
screen control language, multi-vendor architecture,
muhiple engine architecture, and PROC SOL; and
explores their role as a connectivity tool in constructing
remote data entry, CANDA, and clinical data review
systems in order to develop an integrated approach to
clinical information systems.
to manufacture the drug in production quantities.
When all the required research is completed, the drug
company submits its information to the FDA and
makes a formal request for permission to market the
drug In specified diseases. This registration process is
known as a New Drug Application or NDA in the
United States. An NDA Is frequently between 200 and
400 volumes long Of more than one hundred thousand
pages. Although information from many scientific
disciplines is presented, frequently the largest single
portion of the NDA is the clinical section. The FDA
then reviews the NDA and can take three actions:
approve the drug for all or some of the Indications;
not approve the drug; or request more Information be
gathered.
The review process frequently takes
another two or more years.
THE DRUG APPROVAL PROCESS
The pharmaceutical and biotechnology industries are
unique in that before they can bring a product to
market, they must first obtain government approval.
The process of obtaining government approval is a
long and complicated one. In the United States, the
company must first use the proposed drug in animals
to show that there are no gross toxic effects. With this
knowledge, the sponsoring company can obtain an
IND (Investigational New Drug) from the Food and
Drug Administration (FDA).
An IND permits the
sponsor to ship the drug across interstate lines and to
use It experimentaity in humans.
Although the details are different, the drug approval
process is conceptually the same in other countries.
Experimental information is gathered according to the
requirements of the particular country until sufficient
information Is gathered to submit the drug for approval
to a governmental regulatory agency.
Once an IND is obtained, the three pre-marketing
phases of clinical research begin. Phase I starts by
using small single doses in heahhy human volunteers
786
CLINICAL INFORMATION SYSTEMS
such systems Remote Data Entry and Monitoring
systems or REDEM.
As can be imagined from the magnHudeof information
that must be collected, computer systems have long
been a part of the drug development process. We call
a computer system for the collection and retrieval of
clinical trial information, a Clinical Information System
or CIS. The current concept of a Oinical Information
System is shown in Figure 1. Physicians who conduct
clinical trials of new drugs transmit the raw data to the
sponsor on case report forms (CRFs) which are then
entered into a database management system (DBMS).
The data from the DBMS is transferred to a data
analysls package for statistical analysis and computer
generation of tables and graphs which are then
incorporated into reports which become part of the
NDA. Although no atandard DBMS has emerged in
the U.S. pharmaceutical industry, SAS has long been
the standard software used for statistical analysis,
report generation, and production of summary tables
and graphs.
With the introduction of these three new classes of
users Qnvestigators, in· house medical stafl, and FDA
reviewers) future clinical information systems will
resemble Figure 2. Not every study will require all the
facilities of the future CIS's. In particular, REDEM is
likely to be used in selected studies only. However,
pharmaceutical firms will increasingly expect to have
these capabilities at their disposal. As such systems
become commonplace, companies will be confronted
wHh the prospect of having up to five different
computer systems, each with incompatible data
structures: a remote data entry and monitoring sys·
tem; a database management system for traditional
data entry; a clinical data review system for use by in·
house clinical staff; a statistical analysis system; and a
computer assisted NDA review system (CANDA). The
introduction of separate, incompatible systems could
adversely impact productiWy.
Companies must
maintain duplicate databases for each structure (with
the attendant risk of inconsistencies between the
databases) and must either train the staff in the
operation of· mu~iple systems or maintain separate
stalls for each technology.
DBMS
I
STATISTICAL
ANALYSIS
REMOTE DATA ENTRY
AND MOIi(ro'UNG
I
Use by lrwestigatQrs
AUTOMATED
TABLE AND GRAPH
GENERATION
I
I
Figure 1: Current CIS
I
DBMS
T
I
I
STATISTICAL
In an effort to better manage the volume of information
and to reduce the length of time required to introduce
a drug in the UnHed States, the Food and Drug
Administration, in conjunction wHh the pharmaceutical
industry, has been experimenting with ways of using
computer· technology to .fac~Hate the NDA review
process.
Such a computer system is called a
Computer Assisted NDA Review system or CANDA.
Participants in the CANDA experiments have reported
that similar systems would be of use not only by FDA
reviewers, but by the medical stafl of pharmaceutical
corporations. We designate systems for use by
medical stafl to monttor ongoing clinical trials, Clinical
Data Review Systems. Simuttaneously, there has been
much interest in computer systems that permtt the
investigator to enter data and the sponsor to remotely
monHor the trial. The Intent of such systems is to
collect mare timely and accurate information. We call
MALYSIS
CLINICAL DATA
REVIEW
I
Use by in~house
AUTOMATEO
medical staff
TABLE AIID GRAPH
GENERATlCM
I
CA>OA
~
Use by fDA Reviewers
Figure 2: Future CIS
Consequently, it would be desirable to integrate these
fllle functions. Since SAS software is already in
common use throughout the pharmaceutical industry,
this paper explores the emerging capabilities of
SAS and the role they can play in producing the
integrated CIS of tomorrow. The remainder of this
787
paper explores three areas: the use of the current
release 01 SAS on mainframes; the new features of
Version 6.03 of SAS; and the features and architecture
of future Version 6 releases.
CuNAcCESS'S strength is the ease with which data can
be accessed and manipulated. For example, Figures
4-8 illustrate the process of creating a report. In
Figure 4, the user specifies the study and dataset for
the report. If the user is unsure of the dataset, he or
she can check a box and a list of available datasets
appears (Figure 5). In Figure 6, the user selects
whether all variables are to be displayed in the report
or only selected variables. In this case the user has
decided to select which variabies will be included in
the report. Note also that CUNAccESS has selected the
variables 10 and WEEK for automatic display. These
variables will be used to clarify the report In Figure 7,
the user selects which variables will appear in the
report by placing an X next to each variable to be
included. To print sums in the report, the user
additionally checks the sum column. Figure 8 shows
the finished report.
The data is automatically
displayed by identifying variables, in this case the
patient id and the week of the visit.
CUNAcCESS'M
CuNACCESS Version 5 is a Clinical Data Review System
designed for use by medical monitors, cfinical
research associates, managers and other nontradttionai users on the clinical staff. Written using
Version 5.18 SAS{AF software,. CUNACCESS has the
follOWing capabilities: single or double-key data entry;
viewing and querying of data; graphics; descriptive
statistics; and report generation. CUNAccESS features
an extensive context sensitive help system and an online tutorial 10 aid the new or infrequent user.
Figure 3 shows the main CUNAccESS menu. Users
select one of the options by entering its number on the
·Select Option" line and preSSing enter. The Database
Administrator is addttionally permitted to structure new
datasels, paint data entry screens, and manage study
libraries. A key feature of CUNAccESS is its· data
dictionary which is called the Clinical Questions
Catalog (CQC). The Clinical Questions Catalog
ensures uniformity of variables acrOSS studies, lacn;tating the pooling of information. The CQC also
simplifies the study definition process. To create a
new dataset, the Database Administrator simply places
an X in the selection field for any variable which is to
be included In the new dataset All the standard
variable attributes (format, inform at, label, type, and
length) are automatically defined. To provide f1exiblltty
while enforcing standardization, the label, informat,
and format may be customized to the study, while the
name, type, and length of the variable are fIXed.
ClinACCeS$
DATA LIST REPORT
5274
study-
X_ Check (I) here w press enter to see a
list ef datasets fer the above study.
+-----------------------------------------------------------+
HEloP
DATA L.IST
+------------------
2 View" dataset
6
wri tun reports
7
l.i'HU'n 'to use
4 Graphics
STOOY AND OATASET MAtlE -------~---------+
study-
Cl iMcceas
8 bit Cl inAccesll
Libname: S274
_KEYS
p"
HELP
m
CONTual£
~EPtlRT
5274
+----------------------------------------------------------+
levtl 1 (Main Menu)
CANCEL
Figure 4: Selecting a Study and Dataset
COImIand ==>
5. Descriptive statistics
".
PF1
('till)
1 Enter data
dataset
MASS
LAB
MV_RX
etHED
ENTER
to select
Figure 3: The ClinAccess Main Menu
Figure 5: A List of Datasets is Displayed
788
OATA UST REPOiI.T
nATA I.ISTING ,Oft SMY $274
DATASET IS LAB
CcmMnd ===>
+------------------------------------------------------------+
I
STU>Y ftNO DATmT flAME: 5274.I.AB
I
+------------- _VARiABLES AUTCMATICAlLY OISPLA't£O ._----------+
I
,....'"
ID
,.,
I
+I··a;;;.;;~·;,;;;;;··· OTH. . . . . . . .lES TO ""PUT
················+1
_ Display all other variable$ in the -dataset.
X_ select other variables to display from a list.
r···················- 01''''''''
TITLES ---- •••••••••••••••••
.
j
HELP
Pf4
03tasets
PFS
Variables
Pf6
PF3
CANcel
SUI'PIlT
81LIRUBIN
CALCIUM
0.'
0.8
9.S
9.S
10.1
10.'
0.7
0."
0
4
8
12
1.
20
a.•
0.'
0.8
0.'
001
~
+---.--------------------------------------------------------...
Pfl
\/EEK
102
Figure 6: The User Fills in Blanks on the
Report Program Screen
SGOT
SOOT
13
14
14
17
15
19
8
12
10
12
12
13
-------_ .. _.
'2
.... _------67
•
0
4
8
12
1.
20
28
1.0
0.7
1.0
0.9
OJ
9.0
9.'
9.'
9.4
9.2
0.0
0.8
9.'
17
12
11
1.
13
14
11
9.7
.--~--~.-~--
102
1.
10
13
11
15
17
14
88
.~
..
-----~".-~-~
7
April 6. -1989
Figure 8: The Completed Report
CUNACCESS is an example of the kind of applications
which can be developed. using Version 5 SAS
software. Version 5 is available on IBM mainframes
under both the MVS and VMlCMS operating systems;
and on DEC VAX. Prime. and Data General
minicomputers.
The next. section describes the
substantial enhancements . possible under Release
6.03 of SAS for MS'DOS and UNIX.
VERSION 6.03 CAPABIUTIES
Edit dataset: 5274.LA8
kreeo 1
COIIIDIInd =::=>
"'"
Version 6.03 of SAS is available on IBM compatible
PC's under the MS·DOS operating system and on
selected HP and Sun Microsystems computers under
UNIX. In Version 6.03, major steps were laken to
make applications more interactive and to increase the
database management functions available. This was
. accomplished by the addition of Screen Control
Language (SCL) to SAS/AF and SAS/FSP. SCL is a
true programming language specifically designed to
create interective full"screen applications. The syntax
of SCL is similar 10 that of the SAS DATA step so
experienced SAS .programmers will quickly feel
comfortable.
1
Type- an X in the DISPLAY coll.J1l'l or use 1, 2~ 3, _._ to specify
the order displayed. IJ$., X to indiCf:te which variables to SlA'!.
DISPLAY $tJ4
x
;;
x
x
x
......
TYPE
tAtCIUM
NUM
SGOT
HUM
SOPI
GlUCOSE
lR011
X
x
'""'
BUN
8lLIRUBN
~eATNnl
PF1
lIELP
OESCRl?TION
.UM
.UM
....
IO.O!
IO.O!
.UM
.".
Phosphorus
Bitirubin
Crefltinine
pf7
I>rev- Vars
pfR
Next Val's
PF3
SUBMIT
A number of tools were added to make applications
easier to use. For example, Figure 9 shows a pop·up
window which can be used to select variables. The
user positions the cursor on the variable(s) to be used
Figure 7: Selecting Variables to Display and Sum
789
and presses enter, The need .to know and to type the
variable names is eliminated,
VIBl DATA AS REelIJIGUlAR tABLE
CominBnd ..,,;r,.
Dllltaset to be vi ewed:
Datas9ls can also be linked with SCL. One thcrny
problem in clinical data management is to display
comments mada by investigators. A solution is to
place the comments in a separate dataset. When
viewing data, the user sees a note that a comment is
available (Figure 11). The user responds ''I' for yes to
indicate he wishes to view the comment and the
comment is displayed (Figure 12). When finished, a
single keyatroke returns the user to the dataset he was
browsing.
I
DEMO.D£MOC
study .dataset_1'tM'e
This pr«edu.re permits you to vi C<lllllliH'ld -"'=:>
are variables and rows are patie
OptionaUy select
~
to 3 variab Pleascnooke 3 selectim'l(s).
.-
(lD vadables remain visibte as
,
I.
•
x)OO(X)(
XX><XX><
.""""'
xxxx><x
Data $et: oaIO.tlEMOG
,
I.
I.
lOOOO(X
lOOOO(X
XX><l"'"
XlOOOO(
xxmx
XXXXXX
XXXXlO(
xioooo<
DATE
."'.
1
C"""
IlEEK
FSEDIT DEMO.DE
8
""" •
•
PHOS
NUll
NUll
CALCIUM
for dataset names
VA. for- variable names
",be'
.... ••
.... •
CllAR
NUM
PATIENT
SCREENS SEARCHES LETTERS
.ND
Fum, LOCATE, KAME, RFlNO, SEARCH, STRING
""".
ADD
•
HUM
SGOT
01.
''''1
TYPe
.EX
SIRTHOAY
PATIENT 10
-----'
Figure 9: Pop-up Window appears far selecting ID
STUDY •
Ii 151
Sex:
Birthday:
Age;
OEHO.DEMOG
DR(l'
fORlCAl
!'lOVE
088
SCROLL
SHO\.I
1
1
TREAllIENT: Placebo
Ka'.
"
Height:
111
Weight:
117.2:
CARO IOVASCUlAR:
Blood Pressure:
11.
sys
Pulse-;
Any (:()QIIlIfflts? YES
I
70
dies
View
cOImIents{Y/H) y
76
Figure 11: A screen contrallanguage program permits
conunents to be easily viewed.
(B)-------------,
JMFORMAT
Screen
01/28/66
PHYSICAL EXAM:
Version 6.03 also added a number of ways to display
information from several datasets simultaneously.
Customized screens· resembling case report forms
can be developed which display data from more than
one dataset. for example demographics and adverse
events.
In table view, mufiiple windows can be
opened. For example, in Figure 10 three datasets are
viewed simufianeously: demographics, adverse
events, and laboratory. The windows can be resized
or zoomed to fiU the entire screen,
FSPRINT:
Obo
DEMOGRAMlICS:
variables
COlOR
MORE
214
stDIT D f M O . C O M M E N T S - - - - - - - - - - - - - : - - - ,
Coomand ="''''".
Obs
OOSACK
Screen
SEX
PATIEUT
151
152
153
154
155
156
DATE
Male
Female
Kale
Femate
Male
female
11/26/85
04/22/66
03/14/86
05/12186
02103/86
021'25/86
CClIIIMI;NTS fOR STUDY 2.14
DIWG
Placebo
Cud tet
Curito{
Placebo
Placebo
euritot
19
23
32
72
19
24
110
102
156
102,
112
110
70
56
100
72
70
64
PATla.T:
FORK:
DEKO.ADV_R)( ( 8 ) - - ~SPRINT: DEMO. tAB ( 8 ) - - - ,
EDJTHlG HELP PRINTING HSCROll VSCROll N "'¥NAME GO
COlOR, DR.QP, fORMAT. INFORMAT, K Set the default- vertical scroll
151
KEY:
DEMOG
fORMIlATE:
FSPRINT:
Patient did not return for second ... isit.
()(SPLAY
on
PATIENT
t:OMPLNf1
PATtENT
JOINT ST I FfNESS
151
15. FREQUENCY OF URtNATl
'"
151
153 STREP THROAT
154
155 SORE SPOT
15. BURfUNG
,<1
8fLCM KNEf
INFWEHZA
,,8 ST(»\ACH ACIte
153
154
155
156
157
158
CALCIIJM
•
9,6
0.'
10.1
9.'
'.2
SOOT
SGPT
22
11
12
9
11
I.
17
342
11/16/86
Vas out of town
business trip.
11
•
12
12
8
Figure 10: Multiple datasets can be viewed
simultaneously
Figure 12:
790
A comment for the patient from Figure 11
FUTURE ENHANCEMENTS
•
FutLKe releases of Version 6 are designed to further
increase the database management capabilities of
SAS and lis utility as an applicallons development
language. Release 6.06 is scheduled to include
features that will be particularly useful in developing
integrated
systems,
specifically:
multi-vendor
archttecture, multiple engine architecture, PROC SOL,
and indexing.
Proper Engine
Figure 14: Multiple Engine Architecture
Multiple engine architecture refers to the ab~ity to use
data stored in data structures other than a SAS
dataset, for example data stored in another database
management system (DBMS). Currently, interfaces
between SAS and a DBMS are constructed as in
Figure 13. A program is used to identify the data
available in the DBMS. The user selects the variables
and/or records desired and these are extracted from
the DBMS and restructured in the form of a SAS
dataset. This extra step of extracting data from a
database and creating a SAS dataset can be· a
considerable obstacle to the non-traditional user. In
addition, since the data is physically removed from the
DBMS, as new data are added or corrections are
made to the DBMS, the extracted data ceases to be
current and must be re-extracted.
Mufti-vendor architecture refers to the ability of a SAS
program written for one computer to be used on other
computers.
This is particularly important in the
pharmaceutical industry where companies may use
several types of computers, e.g. a corporate
mainframe, a departmental mini-computer such as a
VAX, and PC's.
PROC SOL and indexing bring to SAS two important
database management features for performing
queries. SOL is the emerging standard for relational
database query languages. The SAS implementation
will permit queries of SAS datasets and with multiple
engine architecture of other database structures.
Indexing is a common database management feature
that can speed on-line queries.
SUMMARY
As the role of the Clinical Information System expands
to encompass non-traditional users such as
investigators, medical monitors, CRA's, and FDA
reviewers, pharmaceutical companies are faced with
an increaSingly complex web of technology. In an
effort to simplify the process, we might look for ways
to integrate the various CIS processes. One way to do
this is to lOOk for existing standards and see if there
are ways to expand their use and thus leverage our
investment in the technology and personnel already
trained to use the technology.
User Specifies Variables
SAS Dataset Created
Data Used by SAS Application
Figure 13: Current Extract Process
SAS software is the de facto standard in the
pharmaceutical industry for performing statistical
analyses and presenting the resuHs in graphical or
tabular forms. Once data is in machine readable form,
most of the subsequent processing is typically
parforrned in SAS. Consequently, it is reasonable to
examine whether SAS can play an expanded role in
the CIS•
Under multiple engine architecture, SAS applications
will be able to run directly against other databases as
shown in Figure 14. When a SAS program requires
data, a request is sent to the engine supervisor, a
portion of the SAS system. The engine supervisor
.selects the appropriate software engine to read the
data, and the information is provided to the requesting
program. No SAS dataset needs to be created and
the entire process is transparent to the user.
Aithough initially a statistical. package, much of the
recent development of SAS has been to add features
commonly associated w~h relational database
791
management systems and fourth generation
languages. This paper has explored many of these
existing and emerging features and has shown that
SAS may be able to play a prominent role in
developing integrated, powertul, yet easy to use
systems.
REFERENCES
Rosenberg, Martin J. (1988). Using the SAS System
to Facilitate Clinical Trials Research and NDA
Approval. Proceedings of the Thirteenth Annual SAS
Users Group International Conference. SAS Institute
Inc., Cary, NC. pp. 550-556.
ACKNOWLEDGEMENTS
CU"ACCESS is a trademark of MAJARO INfOSVSTEMS,
INC., Mountain Vrew, CA.
All CUNACCESS screens shown are Copyrighted (C)
1988, 1989 by MAJARO INFOSYSTEMS, INC. and are
used by permission.
VNX is a trademark of Digital Equipment Corporation.
SAS, SAS/AF, and SAS/FSP are the registered
trademarks of SAS Institute Inc., Cary, NC, USA.
MAJARO INFoSVSTEMS INC. provides statistical and
infOrmatlon management services to the pharma·
ceutical, biotechnology, and food products
industries. and specializes in extending computer
technology to non·traditional users.
For further information regarding this paper, please
contact:
Martin J. Rosenberg. Ph.D.
MAJARO INFOSVSTEMS, INC.
99 East Middlefield Road
Su~e31
Mountain View, CA 94043
tel. (415)961·8432
(415)961·9260
792