Competing Risk Survival AnalysisUsing PHREG in SAS 9.4Lovedeep GondaraCancer Surveillance & Outcomes (CSO)Population OncologyBC Cancer Agency
Competing riskDefinitionCompeting risk are said to be present when a patient is at risk ofmore than one mutually exclusive event, such as death fromdifferent cause which will prevent any other from happening.
Competing risk
Competing risk
Competing risk
When & Why? Should be considered when the observation of eventof interest is made impossible by a precedingcompeting event. Competing risk models provide real world probabilitiesof death when competing events are present asopposed to standard survival models by allowing us toseparate the probability of event into different causes.
When & Why? Frequently pointed out that in presence of competingevents, standard product limit method of estimatingsurvivor function for event of interest yields biasedresults as the probability of occurrence is modified byan antecedent competing event.
Data Structure
Data Structure
Competing risk Key concepts– Cumulative incidence function (CIF)– Sub distribution hazard– Cause specific hazard
Competing risk Key concepts– Cumulative incidence function (CIF)– Sub distribution hazard– Cause specific hazard
Cumulative incidence function (CIF)
Cumulative incidence function (CIF)
Key concepts Competingrisk– Cumulative incidence function (CIF)– Sub distribution hazard– Cause specific hazard
Fine and Grays model Introduces covariates in context of competing risks Focuses on cumulative incidence function Descriptive approach, focusing on probability of eachevent type
Fine and Grays model
Fine and Grays model
Fine and Grays model
Fine and Grays modelGraphically
Dataset usedData presented by Klein and Moeschberger whichcontains data for bone marrow transplant for 137patients, grouped into three risk categories based ontheir status at the time of transplantation: acutelymphoblastic leukemia (ALL), acute myelocytic leukemia(AML) low-risk, and AML high-risk.
Dataset used During the follow-up period, some patients mightrelapse or some patients might die while in remission. Relapse is the event of interest, death from any othercause is a competing risk because death impedes theoccurrence of leukemia relapse.
How?With the release of version 9.4(SAS/STAT 13.1) ofSAS software, Fine and Gray’s sub-distributionhazard model can be fitted by specifying eventcodeoption in PROC PHREG.proc phreg data Bmt plots(overlay stratum) cif;class Disease (order internal ref first);model T*Status(0) Disease / eventcode 1;run;
How?Plot CIFproc phreg data Bmt plots(overlay stratum) cif;class Disease (order internal ref first);model T*Status(0) Disease / eventcode 1;run;
How?proc phreg data Bmt plots(overlay stratum) cif;class Disease (order internal ref first);Code for event of interestmodel T*Status(0) Disease / eventcode 1;run;
How?Analysis of Maximum Likelihood imateErrorHazard95% Hazard RatioRatioConfidence LimitsPr 10.2931.233Disease 573Disease 2
Key conceptsCompeting risk– Cumulative incidence function (CIF)– Sub distribution hazard– Cause specific hazard
Cause specific hazard
Cause specific hazardCause specific hazard in PHREG– Can be used to assess the effect of competing events onoutcome which otherwise would have been censoredproc phreg data Bmt;class Disease (order internal ref '3');model T*Status(0,2) Disease;run;
Cause specific hazardCause specific hazard in PHREGproc phreg data Bmt;Competingevents censoredclass Disease(order internalref '3');model T*Status(0,2) Disease;run;
Cause specific hazard Output
Cause specific hazard Effect of competing eventsproc phreg data Bmt;class Disease (order internal ref '3');model T*Status(0,1) Disease;run;
Cause specific hazard Output
Checking PH assumptions(for CSH model)– For cause specific hazards Use “assess ph” optionproc phreg data Bmt;class Disease (order internal ref '3');model T*Status(0,2) Disease;assess ph/resample seed 47337;run;
Checking PH assumptions(for CSH model)- For cause specific hazards Use “assess ph” optionproc phreg data Bmt;class Disease (order internal ref '3');Assess PHassumptionmodelT*Status(0,2) Disease;assess ph/resample seed 47337;run;
Checking PH assumptions(for CSH model)– For cause specific hazards Use assess ph optionproc phreg data Bmt;class Disease (order internal ref '3');Perform amodel T*Status(0,2) Disease;Kolmogorov-type supremum testassess ph/resample seed 47337;run;
Checking PH assumptions(for CSH model)– For cause specific hazards Use “assess ph” option
Checking PH assumptions(for CSH model)– Using Schoenfeld residuals Check for non-zero slope ZPH option in PHREG(v 9.4)can be used for causespecific hazard
Checking PH assumptions(for CSH model)– Using Schoenfeld residualsRequest ZPH test for non proportional hazardsproc phreg data Bmt zph;class Disease (order internal ref '3');model T*Status(0,2) Disease;run; ZPH: diagnostics based on weighted residuals, residualsplotted against transformed rank(default)
Checking PH assumptions(for CSH model)– Using Schoenfeld residuals
Checking PH assumptions(for CSH model)– Using Schoenfeld residualsRequest ZPH test for non proportional hazardsUsing Log transformationproc phreg data Bmt zph(transform log);class Disease (order internal ref '3');model T*Status(0,2) Disease;run;
Checking PH assumptions(for CSH model)– Using Schoenfeld residuals
Checking PH assumptions (for FG model)– Checking PH assumption: Bit more complicated Use a new dataset with more covariates and events.filename rawfoll a follic;infile rawfoll firstobs 2 delimiter "," DSD;input age path1 hgb ldh clinstg blktxcat relsite ch rt survtime stat dftimedfcens resp stnum;run;data follic;set follic;if resp 'NR' or relsite '' then evcens 1; else evcens 0;if resp 'CR' and relsite '' and stat 1 then crcens 1; else crcens 0;cens evcens 2*crcens;agedecade age/10;if ch 'Y' then chemo 1; else chemo 0;run;
Checking PH assumptions (for FG model)– Checking PH assumption: Export Schoenfeld residuals from PHREGproc phreg data follic plots(overlay stratum) cifcovs(aggregate) out estimates;model dftime*cens(0) agedecade hgb clinstg chemo /eventcode 1;output out test ressch WSR agedecade WSR hgb WSR clinstgWSR chemo;run;
Checking PH assumptions (for FG model)– Checking PH assumption: Export Schoenfeld residuals from PHREGOutput model estimatesproc phreg data follic plots(overlay stratum) cifcovs(aggregate) out estimates;model dftime*cens(0) agedecade hgb clinstg chemo /eventcode 1;output out test ressch WSR agedecade WSR hgb WSR clinstgWSR chemo;run;
Checking PH assumptions (for FG model)– Checking PH assumption: Export Schoenfeld residuals from PHREGproc phreg data follic plots(overlay stratum) cifcovs(aggregate) out estimates;model dftime*cens(0) agedecadehgb clinstg chemo /Output Schoenfeld residualseventcode 1;output out test ressch WSR agedecade WSR hgb WSR clinstgWSR chemo;run;
Checking PH assumptions (for FG model)– Checking PH assumption: Merge estimates with residuals and create an adjustedestimate(beta(t))data schoenfeld data;merge test(keep dftime by agedecade2 hgb2clinstg2 chemo2) estimates;by by;rescaled WSR agedecade agedecade2 agedecade;rescaled WSR hgb hgb2 hgb;rescaled WSR clinstg clinstg2 clinstg;rescaled WSR chemo chemo2 chemo;ldftime log(dftime 1);label rescaled WSR agedecade "beta(t) of age per decade"rescaled WSR hgb "beta(t) of haemoglobin"rescaled WSR clinstg "beta(t) of stage"rescaled WSR chemo "beta(t) of chemotherapy"ldftime "log of time";run;
Checking PH assumptions (for FG model)– Checking PH assumption: Plot using Proc Loessods select fitplot;proc loess data schoenfeld data plots residuals(smooth);model rescaled WSR agedecade ldftime /CLM smooth 0.5;run;
Checking PH assumptions (for FG model)– Checking PH assumption: Plot using Proc Loess
Difference made by using FG When competing events are rare and distributed towards endof follow-up.EventFrequency1 (Event of interest)772 (Competing event)60 (Censored)54
Difference made by using FG Fit two models to this data.– Cox proportional hazard model censoring all competingevents– Fine and Grays sub distribution hazard modelCovariateCoxFGCox ll0.760.760.0099Disease-HR1.131.13 0.0001FG P-valueCox HazardFG Hazardratioratio0.00982.132.13 0.00013.083.08
Difference made by using FGCIF from both models
Difference made by using FG– Frequent competing eventsEventFrequency1 (Event of interest)422 (Competing event)410 (Censored)54Both models fitted again.
Difference made by using FG– Frequent competing mateDisease-All0.890.800.04Disease-High1.501.31 0.0001riskCox P-valueFG P-value Cox HazardFG Hazardratioratio0.062.452.230.00074.53.71
Difference made by using FG– CIF
Difference made by using FG– Results show that in presence of competingevents, using Cox proportional hazard model canyield biased results affecting inference.– CIF plot makes it clear that CPH model is overestimating hazard.– Degree of over estimation depends on frequencyand distribution of competing events.
Explained variation in Cox model– Explained variation and predictive accuracy “EV” option in PHREG can be used to get estimates ofexplained variation and predictive accuracy of Coxmodel (Schemper and Henderson (2000) ).
Explained variation in Cox model– Explained variation and predictive accuracy Use it in conjunction with cause specific hazard toassess the importance of competing events
Explained variation in Cox model– Explained variation and predictive accuracy Use it in conjunction with cause specific hazard toRequest explained variation and accuracy estimatesassess the importance of competing eventsproc phreg data Bmt ev;class Disease (order internal ref '3');model T*Status(0,2) Disease;run;
Explained variation in Cox model– Explained variation and predictive accuracy
When, Why & How?– If new release of SAS is not available: %CIF (To estimate and plot CIF)(http://support.sas.com/kb/45/997.html) %PSHREG (Fine and Grays sub distribution dara-SASvR.pdf)
Thanks!Questions?
Difference made by using FG Fit two models to this data. -Cox proportional hazard model censoring all competing events -Fine and Grays sub distribution hazard model Covariate Cox Parameter Estimate FG Parameter Estimate Cox P-value FG P-value Cox Hazard ratio FG Hazard ratio Disease-All 0.76 0.76 0.0099 0.0098 2.13 2.13
Practice basic survival skills during all training programs and exercises. Survival training reduces fear of the unknown and gives you self-confidence. It teaches you to live by your wits. Page 7 of 277. FM 21-76 US ARMY SURVIVAL MANUAL PATTERN FOR SURVIVAL Develop a survival pattern that lets you beat the enemies of survival. .
survival guide book, zombie apocalypse survival guide government, zombie apocalypse survival guide essay, zombie apocalypse survival guide movie, zombie apocalypse survival guide apk, zombie apocalypse survival guide video meetspaceVR the home to the UK's greatest free-roam virtual reality experiences in London, Nottingham and Birmingham. Oct .
Estimating survival non-parametrically, using the Kaplan-Meier and the life table methods. Non-parametric methods for testing di erences in survival between groups (log-rank and Wilcoxon tests). 1. Analysis of Time-to-Event Data (survival analysis) Survival analysis is us
analysis, many conditions are considered simultaneously. time and where we can only analyze a single disease. We now describe the survival filter, a model that addresses these goals to perform large-scale joint survival analysis of EHR. Survival Filter. In the discrete time setting, let the observa-
Introduction to Survival Analysis 4 2. The Nature of Survival Data: Censoring I Survival-time data have two important special characteristics: (a) Survival times are non-negative, and consequently are usually positively skewed. – This makes the na
Test Bank for Competing Visions A History of California 2nd Edition by Cherny Author: Cherny" Subject: Test Bank for Competing Visions A History of California 2nd Edition by ChernyInstant Download Keywords: 2nd Edition; Castillo; Cherny; Competing Visions A History of California; Lemke-Santangelo;
Outline Introduction to competing risks Kaplan-Meier for cause-specific survival Cumulative Incidence Function Cause-specific hazards; conditional probability
Hegemony or Survival: America's Quest for Global Dominance by Noam Chomsky Research Notes NOTES TO CHAPTER 1 1 . November 2002, sec. A, p. 1. Andrew Card cited in Doug Saunders, with reports from Associated Press New York Times,,. Hegemony or Survival., American Empire, Hegemony or Survival Hegemony or Survival. , New York , . 2 . Deterring .