Are we really doing Independent Validation of Advanced Statistical

49
Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion Are we really doing Independent Validation of Advanced Statistical Procedures in SAS? Sameer Bamnote Cytel Statistical Software & Services Pvt. Ltd., Pune, India 15 October, 2013 Sameer Bamnote PhUSE 2013 Cytel 1/20 Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Transcript of Are we really doing Independent Validation of Advanced Statistical

Page 1: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Are we really doing Independent Validation ofAdvanced Statistical Procedures in SAS?

Sameer Bamnote

Cytel Statistical Software & Services Pvt. Ltd., Pune, India

15 October, 2013

Sameer Bamnote PhUSE 2013 Cytel 1/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 2: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Disclaimer

Any views or opinions presented in this presentation are solelythose of the author and do not necessarily represent those of thecompany.

Sameer Bamnote PhUSE 2013 Cytel 2/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 3: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Agenda

1 Introduction

2 Alternative Approach For Validation

3 Why R for validation?

4 Example

5 Summary

6 Conclusion

Sameer Bamnote PhUSE 2013 Cytel 3/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 4: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction

Statistics plays a crucial role in clinical trial reports

SAS is widely used to perform the statistical analysis

Validation - ensure if the results produced are accurate,consistent and reliable

Present scenario:

Source programmer: performs statistical analysis to producethe required results in SASValidator (QC Programmer): independent code in SASHowever, statistical PROC used by both remains same

Sameer Bamnote PhUSE 2013 Cytel 4/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 5: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction

Statistics plays a crucial role in clinical trial reports

SAS is widely used to perform the statistical analysis

Validation - ensure if the results produced are accurate,consistent and reliable

Present scenario:

Source programmer: performs statistical analysis to producethe required results in SASValidator (QC Programmer): independent code in SASHowever, statistical PROC used by both remains same

Sameer Bamnote PhUSE 2013 Cytel 4/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 6: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction

Statistics plays a crucial role in clinical trial reports

SAS is widely used to perform the statistical analysis

Validation - ensure if the results produced are accurate,consistent and reliable

Present scenario:

Source programmer: performs statistical analysis to producethe required results in SASValidator (QC Programmer): independent code in SASHowever, statistical PROC used by both remains same

Sameer Bamnote PhUSE 2013 Cytel 4/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 7: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction

Statistics plays a crucial role in clinical trial reports

SAS is widely used to perform the statistical analysis

Validation - ensure if the results produced are accurate,consistent and reliable

Present scenario:

Source programmer: performs statistical analysis to producethe required results in SASValidator (QC Programmer): independent code in SASHowever, statistical PROC used by both remains same

Sameer Bamnote PhUSE 2013 Cytel 4/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 8: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction

Statistics plays a crucial role in clinical trial reports

SAS is widely used to perform the statistical analysis

Validation - ensure if the results produced are accurate,consistent and reliable

Present scenario:

Source programmer: performs statistical analysis to producethe required results in SAS

Validator (QC Programmer): independent code in SASHowever, statistical PROC used by both remains same

Sameer Bamnote PhUSE 2013 Cytel 4/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 9: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction

Statistics plays a crucial role in clinical trial reports

SAS is widely used to perform the statistical analysis

Validation - ensure if the results produced are accurate,consistent and reliable

Present scenario:

Source programmer: performs statistical analysis to producethe required results in SASValidator (QC Programmer): independent code in SAS

However, statistical PROC used by both remains same

Sameer Bamnote PhUSE 2013 Cytel 4/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 10: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction

Statistics plays a crucial role in clinical trial reports

SAS is widely used to perform the statistical analysis

Validation - ensure if the results produced are accurate,consistent and reliable

Present scenario:

Source programmer: performs statistical analysis to producethe required results in SASValidator (QC Programmer): independent code in SASHowever, statistical PROC used by both remains same

Sameer Bamnote PhUSE 2013 Cytel 4/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 11: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction Contd...

Source Validator

Data Processing SAS Macro/Code Independent SAS Code

Statistical Analysis SAS PROC SAS PROC

Sameer Bamnote PhUSE 2013 Cytel 5/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 12: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction Contd...

Source ValidatorData Processing SAS Macro/Code Independent SAS Code

Statistical Analysis SAS PROC SAS PROC

Sameer Bamnote PhUSE 2013 Cytel 5/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 13: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction Contd...

Source ValidatorData Processing SAS Macro/Code Independent SAS Code

Statistical Analysis SAS PROC SAS PROC

Sameer Bamnote PhUSE 2013 Cytel 5/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 14: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction Contd...

Source programmer and validator uses same PROC

Results will be same, as they are using the same PROC(especially while performing advanced statistical analysis)

Do we call this independent validation?

Sameer Bamnote PhUSE 2013 Cytel 6/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 15: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction Contd...

Source programmer and validator uses same PROC

Results will be same, as they are using the same PROC(especially while performing advanced statistical analysis)

Do we call this independent validation?

Sameer Bamnote PhUSE 2013 Cytel 6/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 16: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Introduction Contd...

Source programmer and validator uses same PROC

Results will be same, as they are using the same PROC(especially while performing advanced statistical analysis)

Do we call this independent validation?

Sameer Bamnote PhUSE 2013 Cytel 6/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 17: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Alternative Approach For Validation

Statistical analysis plays a vital role in decision making

Can we strengthen our validation process?

Use of a different software for validation

More reliable as both softwares will have different processing

Use R (an open source software) for validation of advancedstatistical analysis

Sameer Bamnote PhUSE 2013 Cytel 7/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 18: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Alternative Approach For Validation

Statistical analysis plays a vital role in decision making

Can we strengthen our validation process?

Use of a different software for validation

More reliable as both softwares will have different processing

Use R (an open source software) for validation of advancedstatistical analysis

Sameer Bamnote PhUSE 2013 Cytel 7/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 19: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Alternative Approach For Validation

Statistical analysis plays a vital role in decision making

Can we strengthen our validation process?

Use of a different software for validation

More reliable as both softwares will have different processing

Use R (an open source software) for validation of advancedstatistical analysis

Sameer Bamnote PhUSE 2013 Cytel 7/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 20: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Alternative Approach For Validation

Statistical analysis plays a vital role in decision making

Can we strengthen our validation process?

Use of a different software for validation

More reliable as both softwares will have different processing

Use R (an open source software) for validation of advancedstatistical analysis

Sameer Bamnote PhUSE 2013 Cytel 7/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 21: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Alternative Approach For Validation

Statistical analysis plays a vital role in decision making

Can we strengthen our validation process?

Use of a different software for validation

More reliable as both softwares will have different processing

Use R (an open source software) for validation of advancedstatistical analysis

Sameer Bamnote PhUSE 2013 Cytel 7/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 22: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Why R for validation?

An open source software for statistical computing and graphics

Statistical programming language with a different approach toprocess large and complex datasets

Highly extensible

Latest statistical techniques are available earlier

Major pharmaceutical companies and regulators started usingR

Sameer Bamnote PhUSE 2013 Cytel 8/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 23: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Why R for validation?

An open source software for statistical computing and graphics

Statistical programming language with a different approach toprocess large and complex datasets

Highly extensible

Latest statistical techniques are available earlier

Major pharmaceutical companies and regulators started usingR

Sameer Bamnote PhUSE 2013 Cytel 8/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 24: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Why R for validation?

An open source software for statistical computing and graphics

Statistical programming language with a different approach toprocess large and complex datasets

Highly extensible

Latest statistical techniques are available earlier

Major pharmaceutical companies and regulators started usingR

Sameer Bamnote PhUSE 2013 Cytel 8/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 25: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Why R for validation?

An open source software for statistical computing and graphics

Statistical programming language with a different approach toprocess large and complex datasets

Highly extensible

Latest statistical techniques are available earlier

Major pharmaceutical companies and regulators started usingR

Sameer Bamnote PhUSE 2013 Cytel 8/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 26: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Why R for validation?

An open source software for statistical computing and graphics

Statistical programming language with a different approach toprocess large and complex datasets

Highly extensible

Latest statistical techniques are available earlier

Major pharmaceutical companies and regulators started usingR

Sameer Bamnote PhUSE 2013 Cytel 8/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 27: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

Background: Oncology Trial: A double blind, randomized,phase 3, international, multi-center, examining the efficacy ofTreatment A in combination with Treatment B.

Objective: To demonstrate superiority of Treatment B plusTreatment A compared to Treatment A plus placebo asmeasured by progression-free survival (PFS), including eventsof radiographic disease progression per RECIST 1.1, clinicalprogression or death from any cause in subjects with recurrentovarian or fallopian tube cancers.

Sameer Bamnote PhUSE 2013 Cytel 9/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 28: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

Background: Oncology Trial: A double blind, randomized,phase 3, international, multi-center, examining the efficacy ofTreatment A in combination with Treatment B.

Objective: To demonstrate superiority of Treatment B plusTreatment A compared to Treatment A plus placebo asmeasured by progression-free survival (PFS), including eventsof radiographic disease progression per RECIST 1.1, clinicalprogression or death from any cause in subjects with recurrentovarian or fallopian tube cancers.

Sameer Bamnote PhUSE 2013 Cytel 9/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 29: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

Mock Shell:

Sameer Bamnote PhUSE 2013 Cytel 10/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 30: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

Source ValidatorData Processing SAS Macro/Code Independent SAS Code

Hazard Ratio PROC PHREG PROC PHREG

SAS Code:proc phreg data = phreg1;

model months*pfsc(1) = trt / alpha = .05 ties=exact risklimits;

ods output ParameterEstimates=coxph;

run;

Sameer Bamnote PhUSE 2013 Cytel 11/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 31: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

R Code:#Calling Required Library (Package)

library(survival)

#Cox Proportional Hazard Model

HR<-coxph(Surv(MONTHS, PFSC) ~ TRT_N, data=ph, method=’exact’)

#Printing default results

HR

#Summary of the Object HR

summary(HR)

Sameer Bamnote PhUSE 2013 Cytel 12/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 32: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

SAS Output:

R Output:

Sameer Bamnote PhUSE 2013 Cytel 13/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 33: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

SAS Output:

R Output:

Sameer Bamnote PhUSE 2013 Cytel 14/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 34: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

What we observe

# of events in SAS & R outputs do not match

Hazard Ratio & 95% CI do not match

# events in R & # censored in SAS match

Why MISMATCH?

Reason: Difference in the way SAS & R defines the model

Solution: Modify the model defined in R

Sameer Bamnote PhUSE 2013 Cytel 15/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 35: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

What we observe

# of events in SAS & R outputs do not match

Hazard Ratio & 95% CI do not match

# events in R & # censored in SAS match

Why MISMATCH?

Reason: Difference in the way SAS & R defines the model

Solution: Modify the model defined in R

Sameer Bamnote PhUSE 2013 Cytel 15/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 36: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

What we observe

# of events in SAS & R outputs do not match

Hazard Ratio & 95% CI do not match

# events in R & # censored in SAS match

Why MISMATCH?

Reason: Difference in the way SAS & R defines the model

Solution: Modify the model defined in R

Sameer Bamnote PhUSE 2013 Cytel 15/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 37: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

What we observe

# of events in SAS & R outputs do not match

Hazard Ratio & 95% CI do not match

# events in R & # censored in SAS match

Why MISMATCH?

Reason: Difference in the way SAS & R defines the model

Solution: Modify the model defined in R

Sameer Bamnote PhUSE 2013 Cytel 15/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 38: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

What we observe

# of events in SAS & R outputs do not match

Hazard Ratio & 95% CI do not match

# events in R & # censored in SAS match

Why MISMATCH?

Reason: Difference in the way SAS & R defines the model

Solution: Modify the model defined in R

Sameer Bamnote PhUSE 2013 Cytel 15/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 39: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

What we observe

# of events in SAS & R outputs do not match

Hazard Ratio & 95% CI do not match

# events in R & # censored in SAS match

Why MISMATCH?

Reason: Difference in the way SAS & R defines the model

Solution: Modify the model defined in R

Sameer Bamnote PhUSE 2013 Cytel 15/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 40: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

Surv function in R define the survival model asSurv(time, event, ...)

R Code:#Cox Proportional Hazard Model

HR<-coxph(Surv(MONTHS, PFSC) ~ TRT_N, data=ph, method=’exact’)

R Code (Model Modified):

#Cox Proportional Hazard Model

HR<-coxph(Surv(MONTHS, 1-PFSC) ~ TRT_N, data=ph, method=’exact’)

Sameer Bamnote PhUSE 2013 Cytel 16/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 41: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

Sameer Bamnote PhUSE 2013 Cytel 17/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 42: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Survival Analysis(Hazard Ratio)

Sameer Bamnote PhUSE 2013 Cytel 17/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 43: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Summary

R can be used to validate the results of SAS

Platform independent validation

Strengthen the validation process

Ensures the validity of the results

Similarly other softwares can be explored

Sameer Bamnote PhUSE 2013 Cytel 18/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 44: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Summary

R can be used to validate the results of SAS

Platform independent validation

Strengthen the validation process

Ensures the validity of the results

Similarly other softwares can be explored

Sameer Bamnote PhUSE 2013 Cytel 18/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 45: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Summary

R can be used to validate the results of SAS

Platform independent validation

Strengthen the validation process

Ensures the validity of the results

Similarly other softwares can be explored

Sameer Bamnote PhUSE 2013 Cytel 18/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 46: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Summary

R can be used to validate the results of SAS

Platform independent validation

Strengthen the validation process

Ensures the validity of the results

Similarly other softwares can be explored

Sameer Bamnote PhUSE 2013 Cytel 18/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 47: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Summary

R can be used to validate the results of SAS

Platform independent validation

Strengthen the validation process

Ensures the validity of the results

Similarly other softwares can be explored

Sameer Bamnote PhUSE 2013 Cytel 18/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 48: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Conclusion

Use of another statistical software for replicating results ofstatistical analysis strengthens the validation process.

Sameer Bamnote PhUSE 2013 Cytel 19/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?

Page 49: Are we really doing Independent Validation of Advanced Statistical

Introduction Alternative Approach For Validation Why R for validation? Example Summary Conclusion

Questions?

Contact Detail:Sameer Bamnote

[email protected]

Sameer Bamnote PhUSE 2013 Cytel 20/20

Are we really doing Independent Validation of Advanced Statistical Procedures in SAS?