cdm 1 sas in pharmaceutical industry 30 july 2009 arjun roy & madan gopal kundu clinical data...
TRANSCRIPT
CDMCDM 1
SAS in Pharmaceutical Industry
30 July 2009
Arjun Roy&
Madan Gopal KunduClinical Data Management & Biostatistics
MACR
CDMCDM 2
Statistical software
SAS – Advantage, History, Definition, Windows
Basic Programming
SAS Macro, Examples
Validation, Compliance
Contents
CDMCDM 3
Statistical Software – why??
Solution is Statistical software!!
Manual computation is error prone and time consuming
Clinical trials produces huge volume of data
CDMCDM 4
Statistical Software
CDMCDM 5
SAS – why??
Advanced statistical analysis is much more accessible
Non standard analyses can be programmed
Comparatively faster when working with very large dataset.
Better reporting tool
Also offers data warehousing capability
Popularity
CDMCDM 6
History
• Statistical Analysis System
• Developed by Jim Goodnight and John Shall in1970 at N.C. University
• Initially Developed for Agricultural Research
• SAS Institute founded in 1976
• 98 of world’s top 100 company in Fortune 500 use SAS
• CFR part 11 compliant
CDMCDM 7
SAS solutions for life sciences
SAS for Clinical Data Integration
SAS for Life Sci. Sales & Marketing
SAS Drug Development
SAS Patient Safety
CDMCDM 8
What is SAS ??
Tool for Stat. Analyses
Reporting tool
Programming Language
Data Warehouse
Database
CDMCDM 9
SAS as a Data Warehouse
CDMCDM 10
SAS as a Database
Import/ Export facilities – Can read or export data from a variety of formats
Performing query, merging or data manipulation is possible
Data transformation, derivation of new variables
CDMCDM 11
SAS as a Programming Language
Macro facility
Matrix manipulation
Possible to write routines for new methods
CDMCDM 12
SAS for Statistical Analyses
Descriptive statistics
Contingency Tables
Correlation / Regression
t-test
Wilcoxon test
General Linear Model (ANOVA, ANCOVA)
Logistic regression
Chi-square/ Fisher’s exact test
Trend test
Dunnett Multiple comparison
Logrank test/ Kaplan Meier
CDMCDM 13
SAS as a Reporting Tool
Almost any kind of tables for CSR can be programmed that meets the Clinician’s and Regulatory requirement.
Reporting procedures in SAS
PROC REPORTPROC PRINTPROC TABULATEDATA STEP
CDMCDM
Learning SAS!!
CDMCDM 15
• Editor • Log• Output• Result• Explorer• Graph
SAS Windows
CDMCDM 16
EDITORTo write/ modify SAS program code
LOG • To check execution of the program.• Helps in identify the error in SAS code
• Tells about details such as amount of time it taken to execute the code
EXPLORER
It displays the list of libraries (containing dataset, formats, compiled macros and graphs)
CDMCDM 17
OUTPUT
It displays the output generated upon execution of SAS code
RESULTS
It displays index of the output
CDMCDM 18
Libraries & Datasets SAS stores Datasets in Libraries.
Libraries are just a referred location in Hard-drive. (e.g., “F:\MADANKU\Ragacin\”)
Datasets in Libraries can be generated using Data steps.
Can be imported from other formats (e.g., Excel, Oracle Clinical etc.)
CDMCDM 19
Procedures in SAS SAS procedures analyze data in SAS data sets
to produce summary statistics to produce tables, listings & graphs to perform SQL queries to perform Statistical analyses to manage and print SAS files.
SAS Procedures come in modules (e.g., SAS/BASE, SAS/STAT, SAS/SQL, SAS/IML, SAS/GRAPH)
Commonly used procedures:
PROC PRINT PROC REPORT PROC UNIVARIATE
PROC MEANS PROC MIXED PROC LOGISTIC
PROC TTEST PROC NLIN PROC GPLOT
PROC FREQ PROC SQL PROC IML
CDMCDM 20
Example (Canada Guieline, 1992)Subject Sequence Treatment Period AuCt ln(AUCt)
A TR T 1 365 5.89990
B RT T 2 405 6.00389
C RT T 2 703 6.55536
E TR T 1 233 5.45104
F RT T 2 247 5.50939
G TR T 1 178 5.18178
H RT T 2 246 5.50533
I TR T 1 408 6.01127
K RT T 2 315 5.75257
L TR T 1 140 4.94164
M TR T 1 165 5.10595
N RT T 2 88 4.47734
O RT T 2 183 5.20949
P TR T 1 122 4.80402
Q RT T 2 68 4.21951
R TR T 1 275 5.61677
CDMCDM 21
Subject Sequence Treatment Period AuCt ln(AUCt)
A TR R 2 375 5.92693
B RT R 1 595 6.38856
C RT R 1 471 6.15486
E TR R 2 190 5.24702
F RT R 1 257 5.54908
G TR R 2 175 5.16479
H RT R 1 382 5.94542
I TR R 2 361 5.88888
K RT R 1 218 5.38450
L TR R 2 92 4.52179
M TR R 2 269 5.59471
N RT R 1 106 4.66344
O RT R 1 290 5.66988
P TR R 2 230 5.43808
Q RT R 1 144 4.96981
R TR R 2 344 5.84064
Example (Canada Guideline, 1992)
CDMCDM 22
Procedures in SAS
CDMCDM 23
Type 3 Tests of Fixed Effects
Effect Num DF Den DF F Value Pr > F
Sequence 1 14 0.09 0.7675
Period 1 14 0.33 0.5734
Treatment 1 14 1.89 0.1909
Estimates
Label Estimate Standard Error DF t Value Pr > |t| Alpha Lower Upper
T VS R -0.1314 0.09563 14 -1.37 0.1909 0.1 -0.2999 0.03699
LabelAUCt Ratio
(T/R)Lower Limit of
AUCt RatioUpper Limit of
AUCt Ratio
T VS R 87.68% 74.09% 103.77%
MACRMACR
Procedures in SAS
CDMCDM 24
Macros in SAS
Collection of SAS statements which can be used repeatedly
Why macro?
- Same program can be used repetitively
- Makes program simpler
- Data driven programs can be made, letting SAS decide what to do based on actual data values
Macros are complicated, but makes the work lot easier
CDMCDM 25
Macros in SAS
Defining of a macro
Calling of a macro
Specifying the analysis, algorithm etc.
Name of the macro
Key-parameter
CDMCDM 26
SAS in CDM Clinical Trial of all phases
- Sample size estimation- Randomization schedule- Tables, Listings & Figures (TLFs)
Pre-clinical Data Analyses
Pharmacovigilance signal generation
Pharmacokinetic (PK) analyses
Pharmacodynamic (PD) analyses
Non-standard- Repeated Measure- Nonlinear Mixed Model- Bayesian
CDMCDM 27
In-house Developed Macro
Pre-clinical Data Analyses
Pharmacovigilance
Sample Size
Randomization Schedule
CDMCDM 28
CDMCDM 29
Body weight – Change and % change
Clinical Chemistry parameters (n=20)
Hematology parameters (n=21)
Urine parameters (n=4)
Organ weights (n=8-9)- Absolute- Relative to body weight- Relative to brain weight
Parameters
Analysis is done for both Main and Recovery part of the study
For male and female separately
CDMCDM 30
Flow of Stat Analyses
Verifying Normality Assumption
• ANCOVA • Dunnett pair-wise
comparison
• K-W test• Wilcoxon pair-wise
comparison
• Log transform• Inverse transform• Square root
Norm
al
Non N
ormal
CDMCDM 31
CDMCDM 32
Process flow
Excel data
SAS data Tables and
graphs
%normtest%toxico%toxico_comb%toxico_rec
CDMCDM 33
CDMCDM 34
Haematology of M ale - I: M ean Platelet Volume (micro-meter cube) : Day29
Krishna M ohan, 12M A Y 09 09:09 Output: F:\KRISHN A M O\S tudy\Toxicolog y\GTF-069-08 \Output\tab5_1.rtf
Vehicle control 30 mg /kg 60 mg /kg 120 mg /kg
7.0
7.5
8 .0
8 .5
9.0
9.5
Mea
n Pl
atel
et V
olum
e (m
icro
-met
er c
ube)
Dose
N 9.00
CDMCDM 35
CDMCDM 36
NProportional Reporting RatioRelative Reporting RationChi- square
Stat task…
CDMCDM 37
Process flow
Excel data
SAS data Tables and
graphs
SAS programs
CDMCDM 38
CDMCDM 39
CDMCDM 40
CDMCDM 41
Validation
Program validation
Dataset validation
Output validation
Macro validation
CDMCDM 42
Program Validation
“Documented evidence that program performs as expected”
Log inspection Log enhancement Intermediate results checking Style
SimplicityReadability (use of comments)Re-usability
Syntax checkingLogical Dead endsInfinite loopsCode never executed
CDMCDM 43
Program Validation
“Documented evidence that program performs as expected”
Log inspection Log enhancement Intermediate results checking Style
SimplicityReadability (use of comments)Re-usability
Syntax checkingLogical Dead endsInfinite loopsCode never executed
“Act in haste and repent in leisure, Code too soon and debug forever”
- Raymond Kennington
CDMCDM 44
Program Validation
Cross verification with requirement/ algorithm Documentation (History and Version)
CDMCDM 45
Output Validation
Matching of exact values Layout Format of the values Consistency with the SOPs/ SAP/ Specification
document
CDMCDM 46
SAS Macro ValidationWhite Box testing
- Takes account internal mechanism of macro
- Testing with known, provided data and known results
- Check for the correct results
- Only legal parameters should be specified for its arguments
Black Box testing
- Ignores the internal mechanism of macro
- Testing with unknown data and unknown results
- Check for plausible results
- Any kind of parameters should be specified for its arguments
- Focuses only on the output
CDMCDM 47
ComplianceSAS System Installation Mgt.
- All installations are documented to the <SASROOT>\core\sasinst\hotfix directory
- Testing of installation done by SAS Institute supplied installation test kit located in <SASROOT>\core\sastest.
Version Control
- Important for a regulated environment to track changes in program file, log file and output file.
- SAS does not provide these feature.
- It can be attained through use of version control packages such as Microsoft Visual SourceSafe.
CDMCDM 48
Compliance
Security of SAS Datasets
- Controlled access to the contents of SAS datasets can be administered through password protection of the dataset
Retrieval of Electronic Records- Compliance is straightforward
- Printing audit trails can be done by setting the TYPE option to TYPE=Audit in PROC PRINT
SAS Coexistence with FDA 21 CFR Part 11, How Far Can We Get? – Available at www.lexjansen.com/pharmasug/2002/proceed/fdacomp/fda05.pdf
Audit Trails for SAS Datasets - With PROC DATASETS, it is possible to initiate
SAS dataset specific audit trails, that log dataset updates, modification and deletions.
CDMCDM 49
SAS for CDISC
Data standards are critical component in quest to improve global public health.
Varying data standards
CDISC attempts to define an industry standard for clinical data formatting
SDTM, ODM, LAB and ADaM can be effectively implemented in SAS Drug Development
ODM SDTM
PROC CDISC
SAS XML LIBNAME ENGINE
SAS Dataset
CDMCDM 50
Data Flow in e-Submission
CDMCDM 51
Any Questions
CDMCDM 52
Thank You…!