compare incidence rate among three groups

[Alejandro Di Sibio]

Dear colleagues. I already have definided the failure event (DEAD) and time variable (TIME), and I have three groups with different treatment (groups 0, 1 and 2; variable GROUP). I calculated the incidence rate in each treatment group using the command "stptime, by(grupo)" and also I have the CI for each rate. I want to compare with a test (and I want to get a p value) the three incidence rates but just can compare 0 vs 1, 1 vs 2 and 1 vs 3 using comand "stir grupo if grupo==0 | grupo==1" and "stir grupo if grupo==1 | grupo==2" and "stir grupo if grupo==0 | grupo==2". Using "stir", I get a p value but just for two groups.

Do you know which command I should use to compare the IR among the three groups?? I know I would not get a incidence rate ratio between 3 groups but I need to compare all the three IR instead 0 vs 1, or 1 vs 2 or 0 vs 2.

Thank you so much, Alejandro Di Sibio

[Paul Dickman]

There may be other ways to do it, but the folllwing should give you what you want:

streg i.grupo, distribution(exponential)

The test statistic will be labelled "LR chi2(2)" and there'll be an associated p-value.

You are fitting a parametric survival model. Assuming survival times are exponential means the incidence rate (hazard) is constant over time. That is, the model is assuming that there is one incidence rate for each group. The predicted rates from the model will be equivalent to those you got from stptime.

You can easily extend this to adjust for, for example, age. Then you would use the testparm or lrtest commands to compare models with and without group.