coxph "an id statement is required for multi-state models" error

Question

[jnmcluo]

hi everyone, it keeps show me a error when I use "survival" package coxph() for multivariate survival analysis : Error in coxph(Surv(days_to_last_followup, fustat) ~ NRG3 + FSTL5 + ASTN2 + : an id statement is required for multi-state models .

my input data is:

id    days_to_last_followup    fustat     NRG3 FSTL5 ASTN2

TCGA-3L-AA1B                   154  Alive   WT    WT    WT

TCGA-4N-A93T                     8  Alive   WT    WT    WT

TCGA-4T-AA8H                   160  Alive   WT    WT    WT

TCGA-5M-AAT4                     0   Dead   WT    WT    WT

and when i provide id for the coxph method:cox <- coxph(Surv(days_to_last_followup, fustat)~NRG3+FSTL5+ASTN2+MROH2B+ZNF43+COL4A4+AXIN2+SMARCA4+ADAM12+SYNE1+THSD7B, data =multiCox.input ,id =id) , there are no errors show up but I can't see gene names in cox result:

 coef exp(coef) se(coef) robust se      z Pr(>|z|)  
1   0.29929   1.34890  0.80904   0.67471  0.444    0.657  
2   0.03079   1.03127  0.57015   0.40335  0.076    0.939  
3  -0.21452   0.80693  0.49745   0.39913 -0.537    0.591  
4  -0.16062   0.85161  0.67287   0.64364 -0.250    0.803  
5   0.31424   1.36922  0.81711   0.65210  0.482    0.630

what's wrong in my steps? thanks

Answer 1

[MatthewP]

Hey, no need to specific id , try to convert Alive/Dead to int 1/2. I don't think days_to_last_followup is survival time. See this post

[jnmcluo]

Thanks a lot! convert the state to 1,2 does fix the problem! and yes It's always not clear for me how to set the proper survival time for TCGA data. I used to do it like: clinicalData %>% mutate(futime=ifelse(fustat=="Alive",days_to_last_followup,days_to_death)) and then use futime as censored time, but I found that days_to_last_followup could give out a better result. I checked out the links you provided, but still not completely understand the concept. So could you show me some simple lines about how to do this? many thanks.

[1021jiyoung]

Hey I am stuck with the same error and I can't seem to understand why it's happening. I've posted the issue here and if anyone could help me out, it would be much appreciated!!!!

[Hamid Ghaedi]

check this. You should change the type of vector that contains your event data. If your vector is a factor, recoding from Dead/Alive into 1/2 is not going to work. You need to specify the type of vector as.numeric.