Obtaining P Values from Cox Regression in R
1
0
Entering edit mode
21 months ago
menon_ankita ▴ 30

Hello,

I am a student conducting a survival analysis in R. I am using 6,000 genes from 249 patients each, and am testing each gene separately by putting them in an individual Cox regression model. Here is the code that I have written in order to be able to test them all at the same time instead of creating 6,000 different models (for reference, EFST is the event free survival time and Status is the vital state after the first event):

library(survival)
myFinalData <- dataOne[, -1]
EFST <- myFinalData[ , c("EFST")]
Status <- myFinalData[, c("Status")]
formulas <- sapply(names(myFinalData)[3:ncol(myFinalData)], function(x) as.formula(paste
('Surv(EFST, Status) ~ ',x)))
models <- lapply(formulas, function(x) coxph(x, data=myFinalData))
res <- lapply(models, function(x) return(cbind(HR=exp(coef(x)), exp(confint(x)))))
res


This code works, however this only gives me the hazard ratios and the confidence intervals, whereas I would like the p-values as well. I have tried multiple different things to include the p-value, but to no avail.

Additionally; if I wanted to add other factors (for ex., factor(age)), how would I modify this code in order to do that? I am not able to make it a multivariate regression model whenever I try.

Any help would be extremely appreciated!

Thank you,

Ankita

RNA-Seq survival-analysis p-value cox-regression • 1.6k views
0
Entering edit mode
21 months ago

The p-values can be obtained via the summary() function applied to your model.

The summary function will then have the following values of interest

• coefficients, coefficient p-values, SEs, beta coefficients, etc.
• logtest, likelihood ratio test p-value
• sctest, Score / logRank test p-value
• waldtest, Wald test p-value

Kevin

0
Entering edit mode

Hello Dr. Blighe, Thank you for the response. That is what I initially thought as well, but entering summary(res) gives me an output of something like this:

summary(res)
Length Class  Mode
ACAP3       3      -none- numeric


….

0
Entering edit mode

Hello Professor Menon, that is due to the fact that your res object is not produced directly from coxph(). You will have to apply summary() to your models list object, most likely via another lapply function.

Something like I do in this code chunk from my package RegParallel: