timeROC: Time-dependent ROC curve estimation

Description Usage Arguments Details Value Author(s) References See Also Examples

View source: R/timeROC_3.R

Description

Inverse Probability of Censoring Weighting (IPCW) estimation of Cumulative/Dynamic time-dependent ROC curve. The function works in the usual survival setting as well as in the competing risks setting. Computation of the iid-representation of areas under time-dependent ROC curves is implemented. This enables computation of inference procedures: Confidence intervals and tests for comparing two AUCs of two different markers measured on the same subjects.

Usage

1
2
timeROC(T, delta, marker, other_markers = NULL, cause,
	    weighting = "marginal", times, ROC = TRUE, iid = FALSE)

Arguments

T

The vector of (censored) event-times.

delta

The vector of event indicators at the corresponding value of the vector T. Censored observations must be denoted by the value 0.

marker

The vector of the marker values for which we want to compute the time-dependent ROC curves. Without loss of generality, the function assumes that larger values of the marker are associated with higher risks of events. If lower values of the marker are associated with higher risks of events, then reverse the association adding a minus to the marker values.

other_markers

A matrix that contains values of other markers that we want to take into account for computing the inverse probability of censoring weights. The different columns represent the different markers. This argument is optional, and ignored if method="marginal". Default value is other_markers=NULL.

cause

The value of the event indicator that represents the event of interest for which we aim to compute the time-dependent ROC curve. Without competing risks, it must be the value that indicates a non-censored obsevation (usually 1). With competing risks, subjects can undergo different type of events; then, it must be the value corresponding to the event of interest, for which we aim to compute the ROC curve (usually 1 or 2).

weighting

The method used to compute the weights. weighting="marginal" uses the Kaplan-Meier estimator of the censoring distribution. weighting="cox" and weighting="aalen" model the censoring by the Cox model and the additive Aalen model respectively. Default value is weighting="marginal".

times

The vector of times points "t" at which we want to compute the time-dependent ROC curve. If vector times contains only a single value, then value zero is added.

ROC

A logical value that indicates if we want to save the estimates of sensitivities and specificties. Default value is ROC = TRUE.

iid

A logical value that indicates if we want to compute the iid-representation of the area under time-dependent ROC curve estimator. iid = TRUE is required for computation of all inference procedures (Confidence intervals or test for comparing AUCs). For large sample size (greater than 2000, say) and/or large length of vector times, the computation of the iid representations might be time-consuming. Default value is iid = FALSE.

Details

This function computes Inverse Probability of Censoring Weighting (IPCW) estimates of Cumulative/Dynamic time-dependent ROC curve. By definition, time-dependent ROC curve intrinsically depends on the definitions of time-dependent cases and controls.
Let T_i denote the event time of the subject i.

Without competing risks : A case is defined as a subject i with T_i <=t. A control is defined as a subject i with T_i > t.

With competing risks : In this setting, subjects may undergo different type of events, denoted by δ_i in the following. Let suppose that we are interested in the event δ_i=1. Then, a case is defined as a subject i with T_i <=t and δ_i = 1. With competing risks, two definitions of controls were suggested: (i) a control is defined as a subject i that is free of any event, i.e with T_i > t, and (ii) a control is defined as a subject i that is not a case, i.e with T_i > t or with T_i <=t and δ_i != 1 . For all outputs of this package, objects named with _1 refer to definition (i). For instance AUC_1 or se_1 refer to time-dependent area under the ROC curve and its estimated standard error according to the definition (i). Objects named with _2 refer to definition (ii) .

Value

Object of class "ipcwsurvivalROC" or "ipcwcompetingrisksROC", depending on if there is competing risk or not, that is a list. For these classes, there are print, plot and confint methods. Most objects that they contain are similar, but some are specific to each class.

Specific objects of class "ipcwsurvivalROC" :

Specific objects of class "ipcwcompetingrisksROC" :

Objects common to both classes :

Author(s)

Paul Blanche [email protected]

References

Hung, H. and Chiang, C. (2010). Estimation methods for time-dependent AUC with survival data. Canadian Journal of Statistics, 38(1):8-26

Uno, H., Cai, T., Tian, L. and Wei, L. (2007). Evaluating prediction rules for t-years survivors with censored regression models. Journal of the American Statistical Association, 102(478):527-537.

Blanche, P., Dartigues, J. F., & Jacqmin-Gadda, H. (2013). Estimating and comparing time-dependent areas under receiver operating characteristic curves for censored event times with competing risks. Statistics in medicine, 32(30), 5381-5397.

P. Blanche, A. Latouche, V. Viallon (2013). Time-dependent AUC with right-censored data: A Survey. Risk Assessment and Evaluation of Predictions, 239-251, Springer, http://arxiv.org/abs/1210.6805.

See Also

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
##-------------Without competing risks-------------------
library(survival)
data(pbc)
head(pbc)
pbc<-pbc[!is.na(pbc$trt),] # select only randomised subjects
pbc$status<-as.numeric(pbc$status==2) # create event indicator: 1 for death, 0 for censored

# we evaluate bilirubin as a prognostic biomarker for death.

# 1) with the Kaplan-Meier estimator for computing the weights (default).
ROC.bili.marginal<-timeROC(T=pbc$time,
                  delta=pbc$status,marker=pbc$bili,
                  cause=1,weighting="marginal",
                  times=quantile(pbc$time,probs=seq(0.2,0.8,0.1)),
                  iid=TRUE)
ROC.bili.marginal

# 2) with a Cox model (with covariates bili, chol and albumin) for computing the weights.
ROC.bili.cox<-timeROC(T=pbc$time,
                      delta=pbc$status,marker=pbc$bili,
                      other_markers=as.matrix(pbc[,c("chol","albumin")]),
                      cause=1,weighting="cox",
                      times=quantile(pbc$time,probs=seq(0.2,0.8,0.1)))
ROC.bili.cox

##-------------With competing risks-------------------


#---------Example with Melano data-------
data(Melano)

# Evaluate tumor thickness as a prognostic biomarker for
# death from malignant melanoma.
ROC.thick<-timeROC(T=Melano$time,delta=Melano$status,
                   weighting="aalen",
                   marker=Melano$thick,cause=1,
                   times=c(1800,2000,2200))
ROC.thick

#---------Example with Paquid data--------
data(Paquid)

# evaluate DDST cognitive score as a prognostic tool for
# dementia onset, accounting for death without dementia competing risk.
ROC.DSST<-timeROC(T=Paquid$time,delta=Paquid$status,
                  marker=-Paquid$DSST,cause=1,
                  weighting="cox",
                  other_markers=as.matrix(Paquid$MMSE),
                  times=c(3,5,10),ROC=TRUE)
ROC.DSST 
plot(ROC.DSST,time=5)        

Example output

  id time status trt      age sex ascites hepato spiders edema bili chol
1  1  400      2   1 58.76523   f       1      1       1   1.0 14.5  261
2  2 4500      0   1 56.44627   f       0      1       1   0.0  1.1  302
3  3 1012      2   1 70.07255   m       0      0       0   0.5  1.4  176
4  4 1925      2   1 54.74059   f       0      1       1   0.5  1.8  244
5  5 1504      1   2 38.10541   f       0      1       1   0.0  3.4  279
6  6 2503      2   2 66.25873   f       0      1       0   0.0  0.8  248
  albumin copper alk.phos    ast trig platelet protime stage
1    2.60    156   1718.0 137.95  172      190    12.2     4
2    4.14     54   7394.8 113.52   88      221    10.6     3
3    3.48    210    516.0  96.10   55      151    12.0     4
4    2.54     64   6121.8  60.63   92      183    10.3     4
5    3.53    143    671.0 113.15   72      136    10.9     3
6    3.98     50    944.0  93.00   63       NA    11.0     3
Time-dependent-Roc curve estimated using IPCW  (n=312, without competing risks). 
         Cases Survivors Censored AUC (%)   se
t=999.2     53       249       10   83.96 2.91
t=1307.4    68       218       26   85.66 2.56
t=1839.5    86       156       70   88.03 2.25
t=2555.7   102        94      116   83.41 3.17
t=3039     108        63      141   80.79 3.48

Method used for estimating IPCW:marginal 

Total computation time : 0.55  secs.
Time-dependent-Roc curve estimated using IPCW  (n=284, without competing risks). 
         Cases Survivors Censored AUC (%)
t=999.2     48       226       10   83.90
t=1307.4    60       198       26   86.07
t=1839.5    76       139       69   87.57
t=2555.7    92        83      109   82.81
t=3039      98        57      129   80.11

Method used for estimating IPCW:cox 

Total computation time : 1.27  secs.
Time-dependent-Roc curve estimated using IPCW  (n=205, with competing risks). 
       Cases Survivors Other events Censored AUC_1 (%) AUC_2 (%)
t=1800    45       124            9       27     76.60     75.71
t=2000    46       103           10       46     76.19     74.97
t=2200    50        83           11       61     73.43     72.45

Method used for estimating IPCW:aalen 

Total computation time : 0.06  secs.
Time-dependent-Roc curve estimated using IPCW  (n=2561, with competing risks). 
     Cases Survivors Other events Censored AUC_1 (%) AUC_2 (%)
t=3     70      2117          194      180     80.83     79.85
t=5    122      1834          313      292     79.55     77.65
t=10   318      1107          545      591     76.40     71.93

Method used for estimating IPCW:cox 

Total computation time : 1.16  secs.

timeROC documentation built on May 29, 2017, 7:54 p.m.