how do I compare observed annual event rates such as death with a rate from another cohort of similar patients?

[john jose]

In a study, I found an annualized rate of an event, x=3.2% on treatment A (93.4 patient years of follow up) . Expected rate of the event is 6.7% per year on treatment B (gold standard). How can I compare them using a binomial test in STATA? or do I need a Poisson model?

[Carlo Lazzaro]

John:
welcome to the list.
Did you take a look at -help epitab-?

[Svend Juul]

You need to know the actual number of events. First, calculate the expected number of events, next enter the expected and the actual number of events in cii with the poisson option. This gives the rate ratio with confidence interval:

. display 93.4*0.067
6.2578

. cii 6.2578 11 , poisson
                                                         -- Poisson  Exact --
    Variable |   Exposure        Mean    Std. Err.       [95% Conf. Interval]
-------------+---------------------------------------------------------------
             |     6.2578    1.757806    .5299985        .8774906    3.145201

If you have person-level data (t = time at risk (years), d(0/1) = alive/death, er = expected rate (deaths/year), you can use poisson:

gen e = t*er
poisson d , exposure(e)

[john jose]

Thanks Svend for the reply. I have 3 events for a total patient follow up of 93.4 patient years. Treatment B rate is the event rate data obtained from another study. With the method suggested above by you, do I get a p value for the rate ratio?