I am trying to run a multivariate Cox regression analysis but I have 1 value for 1 variable that has no events reached. This variable was treatment (values = treatment 1, treatment 2, treatment 3, and treatment 4). Treatment 4 was the baseline, and treatment 2 had no events.

On univariate analysis, I am fine with not having a p-value for treatment 2 and saying there were no patients who reached the event.

However, now I need to run a multivariate analysis, and I was wondering if treatment 2 is making the rest of my results invalid. Again, I am fine reporting no statistical results for treatment 2.

Basically, I am wondering if I can use the results below as they are, and just ignore the row for treatment 2 (bolded).

If what is below is not valid, I may consider grouping treatment 2 with the baseline treatment or just removing patients in that group from the analysis entirely (changing the inclusion criteria), but neither of these are ideal based on practical context.

This is what I see in r:

Warning message: In fitter(X, Y, strats, offset, init, control, weights = weights, : Loglik converged before variable 9 ; coefficient may be infinite.

enter image description here


1 Answer 1


Merging group 2 with baseline seems the worst option to me. Dropping group 2 from the analysis would seem better as long as you report what you did and why. I would suggest that the best option is to report the table as you give it but for group 2 not to report the numbers you show but to replace them with dashes or blanks with a footnote stating that they were not estimable as there were no events. Any readers who understand time to event methods will understand what you did.

  • $\begingroup$ Great, thank you. So the inclusion of ALK is not causing the multivariate analysis to be incorrect overall in R, and I'll just dash out that row. Cheers. $\endgroup$
    – petermel12
    Commented Sep 27, 2019 at 19:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.