Date: Sat, 14 Dec 2002 12:03:25 -0500
Reply-To: ananthcv <ananthcv@epi.umdnj.edu>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: ananthcv <ananthcv@EPI.UMDNJ.EDU>
Subject: Cox model with repeated measures and time-dependent exposure
Content-Type: TEXT/plain; charset=us-ascii
Dear SAS-L users:
Please pardon if this question has already been posted.
I'm trying to evaluate if anemia during pregnancy (ANEMIA; yes/no) results in
greater risk of stillbirths (SB; yes/no). I have serial anemia assessments
(unequally spaced measurements throughout gestation) for about 13,000 women.
Women may have had anywhere from 1 to 5 anemia assessments during her pregnancy.
I have the data set up as as longitudinal data (attached below).
For the 2 sample patients shown below, the first patient (SUBJ=18) did not have
any anemia during her pregnancy (and delivered a stillborn at 31 weeks). The
second patient (SUBJ=20) did not have anemia at 28 weeks (GA_ANEMIA), but had
anemia at 32 weeks, and delivered a live born infant at 39 weeks. The variable
GA_DEL is when the pregnancy ended, and SB is the outcome (or censoring
variable).
What will be the SAS code for PROC PHREG? I realize that I have a time-dependent
exposure (anemia status that varies across pregnancy), as well as repeated
measures. I looked up SAS online "Sample programs" help pages, but were not too
helpful. Any help with figuring out the PROC PHREG code will be much
appreciated.
regards,
--
Cande V. Ananth, PhD, MPH tel: 732.235.7940
Associate Professor of OB/GYN fax: 732.235.6627
Director, Section of Epid and Biostats
Dept of OB/GYN and Reproductive Sciences ananthcv@EPI.UMDNJ.EDU
UMDNJ-Robert Wood Johnson Medical School rwjms.umdnj.edu/obgyn/
---------------------------------------------------------------------
SAMPLE DATA (2 patients)
SUBJ ANEMIA GA_ANEMIA GA_SB SB TRIMESTER
18 0 6 31 1 1
18 0 9 31 1 1
18 0 30 31 1 3
20 0 28 39 0 3
20 1 32 39 0 3
