In mdbrown/survCompetingRisk: Evaluating Prognostic Markers under Competing Risk

survCompetingRisk

This package aims to help to evaluate the prognostic accuracy of a marker with multiple competing risk events. Functions to calculate the AUC, ROC, PPV, and NPV are provided. A discrete covariate Z, if available, can be included.

There are five main functions in this package:

1. comprisk.ROC: Calculate the values needed to plot the ROC curve, along with the AUC.

2. comprisk.ROC.CI: Calculate bootstrap confidence intervals for a given set of TPR or FPR values.

3. comprisk.AUC.CI: Calculate bootstrap confidence intervals for the AUC.

4. comprisk.PPV: Calculate PPV and NPV values.

5. comprisk.PPV.CI: Calculate bootstrap confidence intervals for a given set of PPV, NPV, or v (marker quantile).

Tutorial

library(survCompetingRisk)

#simulated data for illustration
data(crdata)

#take a look 
head(crdata)

##ROC curve 

#Set type = 1 if case is defined by the event of interest, and controls are all the rest. Set type = 2 if case is defined by stratifying on event type, and controls are those who have not experienced any events.

myROC.type1 <- comprisk.ROC( times = crdata$times,
                             status1 = crdata$status1, 
                             status2 = crdata$status2, 
                             x = crdata$x, 
                             Z = crdata$Z, 
                             predict.time = 10, 
                             type = 1)

myROC.type1

#plot
tmp <- myROC.type1$ROC

plot(tmp$FPR[tmp$Z==0], tmp$TPR[tmp$Z==0], type="l",lwd=2, main="Type 1", xlab="FPR", ylab="TPR")
lines(tmp$FPR[tmp$Z==1], tmp$TPR[tmp$Z==1], lty=2, lwd=2)
legend(x="bottomright", c("Z = 0", "Z = 1"), lty=c(1,2))
lines(c(0, 1), c(0,1), col="lightgrey")


## type 2 

myROC.type2 <- comprisk.ROC( times = crdata$times,
                             status1 = crdata$status1, 
                             status2 = crdata$status2, 
                             x = crdata$x, 
                             Z = crdata$Z, 
                             predict.time = 10, 
                             type = 2)

myROC.type2


comprisk.AUC.CI( times = crdata$times,
              status1 = crdata$status1, 
              status2 = crdata$status2, 
              x = crdata$x, 
              Z = crdata$Z, 
              predict.time = 10, 
              type = 1, 
              bootstraps = 25) #set to 500-1000 in practice!

References

Zheng Y, Cai T, Jin Y, Feng Z. Evaluating prognostic accuracy of biomarkers under competing risk. Biometrics. 2012 Jun;68(2):388-96.

Zheng Y, Cai T, Feng Z, and Stanford J. Semiparametric Models of Time-dependent Predictive Values of Prognostic Biomarkers. Biometrics. 2010, 66: 50-60.

mdbrown/survCompetingRisk documentation built on May 22, 2019, 3:23 p.m.