Nonparametric survival curves with delayed entry

[Madeline Bauer <mbauer@usc.edu>]

Greetings,
I have survival data where a subset of patients start a new (nonstudy) treatment at random times during the study. I would like to estimate two survival curves, one curve for the patients who are never started on the new treatment and the second for those that did.
This seems just like the Stanford Heart transplant data set, where the second group should be included in the denominator for the first group until they start the new treatment.
So I've set up a data frame with Start, Stop, Event, and a stratum variable to indicate the start of the new treatment for the second records, following the counting process setup.
The Surv function happily returns a Surv object, but survfit returns an error saying that it only handles right-censored data.
This seems so straightforward, so I am obviously missing something that is right in front of my face.
TIA!
Madeline
S-Plus 2000 Professional on Win98SE (if it works don't fix it)

===
Madeline Bauer, Ph.D.        University of Southern California
Keck School of Medicine (Infectious Diseases)
IRD Room 620 (MC9520), 2020 Zonal Ave, Los Angeles 90033
(323) 226-2775 [Voice & FAX]
mbauer@usc.edu  [Fastest communication method]