Date: Fri, 6 Feb 1998 10:41:33 +0000
Reply-To: Sarah Lewington <sarah.lewington@CTSU.OX.AC.UK>
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: Sarah Lewington <sarah.lewington@CTSU.OX.AC.UK>
Subject: QUERY: Stratifying by age at risk
Content-Type: text/plain; charset=US-ASCII
I have data from several prospective observational studies with
information on age at screening, age at death (or last follow-up),
the exposure of interest and several other covariates. I am
analysing survival using stratified Cox regression analysis.
If STATUS is my censoring variable (1=alive) and DURATION is the
time from screening (when exposure, e.g. SBP, was measured) to death
or censor, then to find the RR of death per 1 mmHg increase in SBP I
use the model:
model DURATION*STATUS(1) = SBP;
In previous analyses I have adjusted for cohort, sex and age at
screening (AGES) by stratification, so I've added a STRATA statement,
strata COHORT SEX AGES(35, 40 to 85 by 5);
However I should really be stratifying by age at death rather than
screening (i.e. comparing people who died at age x with
people who contributed person-years at risk through the
interval in which x lies). Initially I thought I could replace AGES
in the STRATA statement by OUTAGE (= age at death or censor), but of
course this doesn't quite work because the "duration" variable should
be the person-years at risk in that stratum for each individual. For
example, someone who is screened at age 46 and dies at age 58 would
contribute 4 years (and be censored) to the 45-50 stratum, 5 years
(and be censored) to the 50-55 stratum and 3 years (and die) to the
Can this be done (easily) within SAS or do I have to write the
algorithm to calculate person-years at risk in each stratum myself?
If so how do I then incorporate this into PROC PHREG?
If I have not made myself clear please let me know.
Thanks for your time.
Prospective Studies Collaboration
Oxford OX2 6HE