Estimating and modelling relative survival using SAS
I have written SAS code (version 7 or higher) for estimating and modelling relative survival. A sample data set containing information on colon carcinoma diagnosed in Finland is provided. All required data and SAS files can be downloaded in a ZIP archive. Full documentation is provided in the file readme.pdf in the ZIP archive.
If the contents of the ZIP archive are extracted to the directory
then the code should run without requiring alteration. The code provided will
reproduce the estimates reported in Table I of the paper Dickman
et al (2004). Two input data files are provided;
contains the cancer patient data and
data on expected probabilities of death for the Finnish general population.
The SAS code in
survival.sas produces life table estimates of
relative survival stratified by sex, age, and calendar period of diagnosis.
In addition, two output data sets are created (one containing grouped data and
one containing individual patient data) which are used as input data sets for
modelling. The SAS code in
models.sas estimates a relative survival
regression model using several different approaches (described in Dickman
et al (2004)).
20040619 Version 1.0
20101025 Version 1.1 (only file updated in survival_period.sas)
Summary: Standard errors for period analysis were not correct and have now been corrected.
Details: I have traditionally used the actuarial approach to estimate survival and used Greenwood's method for estimating standard errors.However, for period analysis (implemented in survival_period.sas) I estimated survival by transforming the estimated cumulative hazard. Unfortunately standard errors were still based on Greenwood's method using an incorrect value of the effective number at risk.Note that there is no problem with Greenwood's method per se, just that I didn't calculate the effective number at risk correctly since I wasn't using it to estimate survival. I've now used the appropriate method for calculating the standard error. Technical details can be found in this document, which described the Stata implementation. Output from SAS, using both the v1.0 and v1.1 code can be found here.