Oncology Efficacy Tables

7
PharmaSUG2010 - Paper AD03 Tips for Creating Oncologic Efficacy Summary Tables using PROC LIFETEST and PROC PHREG Scott Michael Ward, i3 Statprobe, Cary, North Carolina ABSTRACT Survival statistics are used frequently within Oncologic Efficacy Summary tables. They are used to measure several types of outputs including Progression Free Survival, Time to Treatment Failure, Time to Disease Progression, Duration of Survival, and Duration of Response. This paper will show how to achieve the following types of outputs with PROC LIFETEST and PROC PHREG by offering some tips and techniques for fast and efficient results and outputs. KEYWORDS: SAS, PROC LIFETEST, PROC PHREG, DURATION, SURVIVAL, HAZARD RATIOS, DISEASE PROGRESSION, TREATMENT FAILURE, PROGRESSION FREE SURVIVAL, RESPONSE INTRODUCTION To create these Oncologic Efficacy Summary Tables use the SAS procedures PROC LIFETEST and PROC PHREG. Some commonly created efficacy outputs used for these analyses are: Progression Free Survival is the length of time during and after treatment in which a patient is living with a disease that does not get worse. Progression-free survival may be used in a clinical study or trial to help find out how well a new treatment works. Time to Treatment Failure is the time from the start of a new drug regimen to the end of time where the drug regimen is no longer taken because of disease progression, study endpoint, or death. Time to Disease Progression is defined as time from randomization to first event of disease progression. Disease progression is a composite end point. Duration of Survival is the time from diagnosis date to the date of death. Duration of Response is the time from first Stable Disease (SD) or Partial Regression (PR) or Complete Regression (CR) until death or disease progression. EXAMPLE – Basic Syntax to create any of the analyses listed above. This particular example use Progression Free Survival data points. The common statistics that you output from PROC LIFETEST are Median, 95% Confidence Intervals, 25th-75th percentiles, Minimum and Maximum, and p-values for Log-Rank and Wilcoxon. Additionally, you can use PROC PHREG to create Hazard Ratios and 95% Confidence Intervals.

Transcript of Oncology Efficacy Tables

Page 1: Oncology Efficacy Tables

PharmaSUG2010 - Paper AD03

Tips for Creating Oncologic Efficacy Summary Tables using PROC LIFETEST and PROC PHREG

Scott Michael Ward, i3 Statprobe, Cary, North Carolina

ABSTRACT Survival statistics are used frequently within Oncologic Efficacy Summary tables. They are used to measure several types of outputs including Progression Free Survival, Time to Treatment Failure, Time to Disease Progression, Duration of Survival, and Duration of Response. This paper will show how to achieve the following types of outputs with PROC LIFETEST and PROC PHREG by offering some tips and techniques for fast and efficient results and outputs. KEYWORDS: SAS, PROC LIFETEST, PROC PHREG, DURATION, SURVIVAL, HAZARD RATIOS, DISEASE PROGRESSION, TREATMENT FAILURE, PROGRESSION FREE SURVIVAL, RESPONSE INTRODUCTION To create these Oncologic Efficacy Summary Tables use the SAS procedures PROC LIFETEST and PROC PHREG. Some commonly created efficacy outputs used for these analyses are:

• Progression Free Survival is the length of time during and after treatment in which a patient is living with a disease that does not get worse. Progression-free survival may be used in a clinical study or trial to help find out how well a new treatment works.

• Time to Treatment Failure is the time from the start of a new drug regimen to the end of

time where the drug regimen is no longer taken because of disease progression, study endpoint, or death.

• Time to Disease Progression is defined as time from randomization to first event of

disease progression. Disease progression is a composite end point.

• Duration of Survival is the time from diagnosis date to the date of death.

• Duration of Response is the time from first Stable Disease (SD) or Partial Regression (PR) or Complete Regression (CR) until death or disease progression.

EXAMPLE – Basic Syntax to create any of the analyses listed above. This particular example use Progression Free Survival data points. The common statistics that you output from PROC LIFETEST are Median, 95% Confidence Intervals, 25th-75th percentiles, Minimum and Maximum, and p-values for Log-Rank and Wilcoxon. Additionally, you can use PROC PHREG to create Hazard Ratios and 95% Confidence Intervals.

Page 2: Oncology Efficacy Tables

(Note – to see the name of the output datasets, you can turn on ODS Trace=ON and to see what the names of the statistical datasets created in the SAS Log.) *** Create mean, median, min, max, 25th-75th percentile, 95% C.I. ***; ODS TRACE ON; ODS OUTPUT CensoredSummary =_censorsum Means =_mean Quartiles =_quartiles; PROC LIFETEST data=test1dts outsurv=_surv alphaqt=0.05 ties=EFRON; time pfstm*pfscen(1); strata treatment; Run; ODS OUTPUT CLOSE; ODS TRACE OFF; (Note – again to see the name of the output datasets, you can turn on ODS Trace=ON and to see what the names of the statistical datasets created in the SAS Log.) *** Create Kaplan-Meier Curve ***; PROC LIFETEST data=test1dts outsurv=_surv alphaqt=0.05 ties=EFRON plot=(s); time pfstm*pfscen(1); strata treatment; Run; See Figure 1.Sample Kaplan-Meier Curve (Note – Adding the option plot=(s) with create a simple Kaplan-Meier Curve.) *** Create Hazard Ratio for Stratified Analysis ***; ODS TRACE ON; ODS OUTPUT ParameterEstimates=_parmests; PROC PHREG data=test1dts; where trtnum = 0; model pfstm*pfscen(1)= trtnum / rl alpha = 0.05 ties = EFRON; strata treatment; Run; ODS OUTPUT CLOSE; ODS TRACE OFF; *** Create Log-Rank and Wilcoxon p-values ***; ODS TRACE ON; ODS OUTPUT WilUniChiSq =_wilcox1 LogUniChiSq =_logrank1; PROC LIFETEST data=test1dts; where trtnum in (0,1); time pfstm*pfscen(1);

Page 3: Oncology Efficacy Tables

test trtnum; strata treatment; Run; ODS OUTPUT CLOSE; ODS TRACE OFF; USEFUL TIPS AND OTHER CONSIDERATIONS FOR USING PROC LIFETEST AND PROC PHREG

The trtnum variable is used if there is more than one treatment group, you will have to do a DO Loop to run for each treatment arm.

The timing variable (PFSTM) is the Time to First Event. The censor variable (PFSCEN) is Event Occurred (0/1). The censored value (1) is to show if an Event occurred. The alpha option set the significance level of the confidence limits. The ties option specifies how to handle ties in the failure time. The strata option tells what variable should be stratified (optional). The alphaqt option set the significance level of the confidence limits. The test option specifies a list of numeric (continuous) covariates that you want tested for

association with the failure time. Some additional options in PROC LIFETEST AND PROC PHREG The BY statement is to obtain separate analyses on observations in groups defined by

the BY variables. FREQ statement identifies a variable containing the frequency of occurrence of each

observation. ♦ ID variable values are used to label the observations of the product-limit survival function

estimates.

CONCLUSION All the datasets that are created in PROC LIFETEST and PHREG(_censorsum, _mean, _quartiles, _parmests, _wilcox1, and _logrank1) all need to be manipulated and combined. ID variables should be created for grouping and ordering purposes. A Label field should be created to display a statistic label for each outputted statistics. The statistic columns should be labeled by TREATMENT number in each dataset to ensure proper concatenation of all outputs. A PROC FREQ or a PROC SQL can be used to create the denominator for the denominator for each column: PROC FREQ data= test1dts noprint; tables treatment / out=dentot; Run; Data dentot; set dentot; if treatment = 1 then call symput('total1',compress(count)); if treatment = 2 then call symput('total2',compress(count)); Run;

Page 4: Oncology Efficacy Tables

A PROC FREQ or PROC SUMMARY can be used to the top section of the output for the following categories:

– Number of patients with an event – Earliest contributing event

• Progressive Disease • Death

– Number of patients without an event

A PROC REPORT can be used to output your statistics: PROC Report data=final_ds center headskip headline missing nowindows spacing=2 split='|'; column grp order label (“PHARMA_1|(n=&total1” treatment1) (“PHARMA_2|(n=&total2” treatment2); define grp / order order=internal noprint; define statord / order order=internal noprint; define label / display flow width=50 left ' '; define treatment1 / width=20 center; define treatment2 / width=20 center; break after grp / skip; Run; See Table 1.Progression Free Survival

Page 5: Oncology Efficacy Tables

OUTPUT RESULTS Progression Free Survival Intent to Treat PharmaSUG 2010 PHARMA_1 PHARMA_2 (n=275) (n=225) ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ NO. OF PATIENTS 275 225 NO. OF PATIENTS WITH AN EVENT 120 (43.6%) 70 (31.1%) EARLIEST CONTRIBUTING EVENT: PROGRESSIVE DISEASE 96 54 DEATH 24 16 NO. OF PATIENTS WITHOUT AN EVENT 155 (56.4%) 155 (68.9%) PROGRESSION-FREE SURVIVAL (DAYS) MEDIAN 828.0 947.0 (95% CI) (776.0, 894.0) (905.0, 966.0) 25TH-75TH PERCENTILE 552.0 - 963.0 689.0 - 965.0 RANGE 2.0+ - 963.0 2.0+ - 966.0 UNSTRATIFIED ANALYSIS HAZARD RATIO (RELATIVE TO PHARMA_1) 0.606 (95% CI) (0.450, 0.816) P-VALUE (LOG-RANK) 0.0008 STRATIFIED ANALYSIS HAZARD RATIO (RELATIVE TO PHARMA_1) 0.607 (95% CI) (0.451, 0.818) P-VALUE (LOG-RANK) 0.0009 Table 1.Progression Free Survival

Page 6: Oncology Efficacy Tables

Kaplan-Meier Graph

Sur

viva

l Dis

tribu

tion

Func

tion

0.00

0.25

0.50

0.75

1.00

Days to Event

0 200 400 600 800 1000

STRATA: treatmnt=1 Censored treatmnt=1 treatmnt=2 Censored treatmnt=2

Figure 1.Sample Kaplan-Meier Curve

Page 7: Oncology Efficacy Tables

ACKNOWLEDGEMENTS I would like to thank my colleagues at i3 Statprobe for reviewing and providing feedback on this paper. CONTACT INFORMATION Your comments and questions are valued and encouraged. Contact the author at:

Scott Michael Ward Senior Statistical Programmer I3 Statprobe Weston One 1001 Winstead Drive Suite 200 Cary, North Carolina 27513 Work Phone: 919-678-4725 E-mail: [email protected] SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. ® indicates USA registration.

Other brand and product names are trademarks of their respective companies.