survival (version 3.3-1)

rotterdam: Breast cancer data set used in Royston and Altman (2013)

Description

The rotterdam data set includes 2982 primary breast cancers patients whose data whose records were included in the Rotterdam tumor bank.

Usage

rotterdam
data(cancer, package="survival")

Format

A data frame with 2982 observations on the following 15 variables.

pid

patient identifier

year

year of surgery

age

age at surgery

meno

menopausal status (0= premenopausal, 1= postmenopausal)

size

tumor size, a factor with levels <=20 20-50 >50

grade

differentiation grade

nodes

number of positive lymph nodes

pgr

progesterone receptors (fmol/l)

er

estrogen receptors (fmol/l)

hormon

hormonal treatment (0=no, 1=yes)

chemo

chemotherapy

rtime

days to relapse or last follow-up

recur

0= no relapse, 1= relapse

dtime

days to death or last follow-up

death

0= alive, 1= dead

Details

These data sets are used in the paper by Royston and Altman that is referenced below. The Rotterdam data is used to create a fitted model, and the GBSG data for validation of the model. The paper gives references for the data source.

There are 43 subjects who have died without recurrence, but whose death time is greater than the censoring time for recurrence. A common way that this happens is that a death date is updated in the health record sometime after the research study ended, and said value is then picked up when a study data set is created. But it raises serious questions about censoring. For instance subject 40 is censored for recurrence at 4.2 years and died at 6.6 years; when creating the endpoint of recurrence free survival (earlier of recurrence or death), treating them as a death at 6.6 years implicitly assumes that they were recurrence free just before death. For this to be true we would have to assume that if they had progressed in the 2.4 year interval before death (while off study), that this information would also have been noted in their general medical record, and would also be captured in the study data set. However, that may be unlikely. Death information is often in a centralized location in electronic health records, easily accessed by a programmer and merged with the study data, while recurrence may require manual review. How best to address this is an open issue.

References

Patrick Royston and Douglas Altman, External validation of a Cox prognostic model: principles and methods. BMC Medical Research Methodology 2013, 13:33

See Also

gbsg

Examples

status  <- pmax(rotterdam$recur, rotterdam$death)
rfstime <- with(rotterdam, ifelse(recur==1, rtime, dtime))
fit1 <- coxph(Surv(rfstime, status) ~ pspline(age) + meno + size + 
        pspline(nodes) + er,
     data=rotterdam, subset = (nodes > 0))
# Royston and Altman used fractional polynomials for the nonlinear terms