R survfit function not producing Kaplan-Meier

Question

I'm using the survfit function from the survival package, and get results I don't understand. I have simple data - an object with survival times, and another with an indicator variable (0 = alive, 1 = dead).

Survival_Time_Months[1:50]
# [1] 165   3 119  92  88   3  25   3  56  18 100 114  17  97 141 145 103 156  37  91 101  43  41 143 108  93 136   4 116
# [30]  85 166   0  92  26   9   8  55 136  10  99   1  20   6  95  85  79 119 109  41  23

Vital_Status_RECODE[1:50]
# [1] 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0
# Levels: 0 1

I can then run this through survfit as:

my.surv.fit <- survfit(Surv(Survival_Time_Months,Vital_Status_RECODE) ~ 1, data=mydata)

Then run the plot command:

plot(my.surv.fit)

I get something that is not a Kaplan-Meier curve, but something that starts at zero and goes upward - it looks like 1.0 - KM. The KM data is in the fit object as $pstate, but I have to mess around extensively to extract it and generate the KM-plot I want.

In trying to resolve this, I've looked through every forum regarding the survfit package, and multiple tutorials, and every single one seems to indicate that sequence of commands should produce a KM curve.

Just discovered the problem. My indicator variable had been set as a factor by R (which R loves to do by default). When the survival package functions see a factorized indicator variable, it defaults to a multi-state model. When I removed the factorization from Vital_Status_RECODE, I got the K-M plots that I was expecting. A newbie mistake in a subtle area. — TPL

thc thc · Accepted Answer · 2017-07-18T20:37:07

Status should be a numeric vector, not a factor. Try this and you will see the difference:

time <- c(165,3,119,92,88,3,25,3,56,18,100,114,17,97,141,145,103,156,37,91,101,43,41,143,108,93,136,4,116,85,166,0,92,26,9,8,55,136,10,99,1,20,6,95,85,79,119,109,41,23)
status <- c(0,0,0,0,0,1,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,1,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0)

plot(survfit(Surv(time, status)~1))

R survfit function not producing Kaplan-Meier

1 Answers