Bias and Its Remedy in Interval-Censored Time-to-Event Applications - Interval-Censored-Time-to-Event Data: Methods and Applications - page 323

center prospective study. Observed intervals (Li, i ;R i ] are in quarterly increments.
The data are listed as Table A.2 and also analyzed in Sun (2006). These 368 subjects
were at risk for HIV-1 infection because they received donated blood from a blood
bank to which thousands of donors with factor VIII had donated. In the study,
the subjects were placed into two different groups of 236 and 132 each depending
on the average annual dose of blood they received, denoted as \No" or \Low"-dose
factor VIII concentrate. The data analysis was performed using R software. We
start the analysis to estimate the survival functions using both (Turnbull, 1974)
nonparametric estimator and the \IntCox" for interval-censored data. The R program
codes to fit the Turnbull estimator are listed in (Chen and Peace, 2010) and the R
code for \IntCox" can be simply called as follows:
#loadthelibrary
library(intcox)
#callfunctionintcoxtofittheIntCoxapproach
intcox.fit=intcox(Surv(left,right,type="interval2")~group,
data=dat)
#printthesummaryfitsummary(intcox.fit)
where left and right are intuitively the left/right ends of the observed intervals.
group is the group factor, and dat is the HIV data in R dataframe. The estimated
survival functions for both groups are shown in Figure 11.4 with dashed lines for
Turnbull and solid lines for \IntCox." In this gure, the bold lines reect the \No"
group and the thin lines reect the \Low-dose" dose factor VIII concentrate group. It
can be seen in this gure for both methods that patients receiving \Low-dose" factor
VIII have significantly lower survival probability and are significantly exposed to
higher risk of being infected by HIV-1 than those receiving \No" factor VIII. The
test for statistical signicance can be performed using the \logrank" test for interval-
censored data from Sun (1996) as implemented in the R function \ictest" from R
library \interval" (more information about this R library can be found in Chapter
13) as follows:
#loadthelibrary
library(interval)
#callfunctionictesttodologranktest