95% Confidence interval for overall survival

I try to get overall survival and its 95% confidence interval.
With upper 95% CI, I get 1.000 that means the upper boundary is at 100%.
I also get 1 for another group. Does "1" mean not valid or that cannot be calculated?
Please help.

I'm not sure what you're asking. To help us help you, could you please prepare a reproducible example (reprex) illustrating your issue? Please have a look at this guide, to see how to create one:

It'd be best if you used a dataset included in the survival package, or at least a small and anonymized version of your dataset that leads to the same problem.

Call: survfit(formula = Surv(time = futime, event = rec) ~ grade, data = mydata_1015)

213 observations deleted due to missingness 
                grade=Moderate 
 time n.risk n.event survival std.err lower 95% CI upper 95% CI
   20    348       1    0.997 0.00271        0.992        1.000
   40    141       4    0.978 0.01019        0.958        0.998
   60     24       1    0.961 0.01925        0.924        1.000

                grade=Poor 
 time n.risk n.event survival std.err lower 95% CI upper 95% CI
   20    121       1    0.993 0.00692        0.980            1
   40     59       1    0.984 0.01112        0.963            1
   60     11       0    0.984 0.01112        0.963            1

Please see the output above. at the upper 95% CI, there are 1.00 and 1. What's the difference between them? Thanks for your help.

The short answer is that there's no difference in the actual values (they're both 1, or 100% survival probability). They're just displayed differently.

It looks like this is a result of R displaying all values in a column with the same number of decimal places. In the first table (grade=Moderate) the upper CI = 0.998 for time=40, meaning that this value needs at least three decimal places to be displayed with full precision. The other two confidence limits are therefore also displayed with three decimal places, or 1.000, which is the same as 1 (100% survival probability). For grade=Poor, the upper confidence limit is 1 for all three values of time, so zero decimal places are needed to show full precision, and R therefore rounds the displayed result to 1, which is the same at 1.000 (100% survival probability).

2 Likes

Thank Jeols for your prompt reply.

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.