I am trying to use staggered entry in Cox regression in SPSS but do not know how to solve this. I have talked to the head statistician in our research unit, he runs SAS, and SAS can do it. Can SPSS (because I have always used SPSS).
I have two groups:
One group with type 1 diabetes; the other one are matched controls. The two groups were identified according to date of diabetes diagnosis. This means that :
John, 22 years, in 1992 when he had diabetes
will be matched with
5 other males, aged 22 years, in 1992.[these controls will be assigned the same data of study entry as John's date of diagnosis]
We now want to study if survival among patients with lymphoma is affected by diabetes. In the two groups (diabetes +controls) we have 800 cases of lymphoma. Hence, our data-set is restricted to patients with lymphoma (everyone has it).
The follow-up begins with diagnosis of lymphoma and ends with end of follow-up (could be death, emigration, or just last date of follow-up: Dec 31, 2009). Follow-up ends with death (censored=1; ends without death: censored=0).
The problem is that for a patient with lymphoma diagnosed before diabetes diagnosis or before date of matching, this patient "cannot die" until he/she has his diabetes diagnosis or is matched into the study. This creates immortal time. Let's say John had lymfoproliferative cancer in 1990, but was identified for my study through a diabetes diagnosis in 1992, this means he could not die within the first 2 years, because then he would not have been in my study...
SAS solves this through using staggered entry, and will run a Cox regression to estimate survival according to diabetes-study. Below is the critical syntax.
Entry= time after lymphoma when the patient is diagnosed with diabetes, or matched into the study [i.e. the staggered entry].
Exit=follow-up time (from lymphoma to end of follow-up)
RL will just give us the confidence intervals for the HRs.
Q1. Can SPSS handle staggered entry?
Q2. If not, is there any other way I could solve this question? use other syntax? restructure the data-set etc?
Best wishes, Jonas Ludvigsson (Sweden)