PROC PHREG interaction

[Cruise]

I’d like to estimate the effect of different types of chemotherapies on survival of cancer patients in considering the effect of other therapies combined with chemo, such as surgery, radiation therapies controlling to cancer stage, race/eth and age. First off, I’d like to construct table shown. How to modify the code below to come up with a table shown? I tried interaction effects with no success.

Table. Multivariate analysis, chemotherapy status on survival adjusting to demographics
Variables	Chemotherapy	Dead (N,%)	Alive (N,%)	HR (95% CI)
Surgery
Type 1	Type1	N	N	1.00 Reference
	Type2	N	N	HR (LCL, UCL)
	Type3	N	N	HR (LCL, UCL)
	Type4	N	N	HR (LCL, UCL)
	Type5	N	N	HR (LCL, UCL)
Type 2	Type1	N	N	1.00 Reference
	Type2	N	N	HR (LCL, UCL)
	Type3	N	N	HR (LCL, UCL)
	Type4	N	N	HR (LCL, UCL)
	Type5	N	N	HR (LCL, UCL)
Type 3	Type1	N	N	1.00 Reference
	Type2	N	N	HR (LCL, UCL)
	Type3	N	N	HR (LCL, UCL)
	Type4	N	N	HR (LCL, UCL)
	Type5	N	N	HR (LCL, UCL)
Type4	Type1	N	N	1.00 Reference
	Type2	N	N	HR (LCL, UCL)
	Type3	N	N	HR (LCL, UCL)
	Type4	N	N	HR (LCL, UCL)
	Type5	N	N	HR (LCL, UCL)
Radiation
Type 1	Type1	N	N	1.00 Reference
	Type2	N	N	HR (LCL, UCL)
	Type3	N	N	HR (LCL, UCL)
	Type4	N	N	HR (LCL, UCL)
	Type5	N	N	HR (LCL, UCL)
Type 2	Type1	N	N	1.00 Reference
	Type2	N	N	HR (LCL, UCL)
	Type3	N	N	HR (LCL, UCL)
	Type4	N	N	HR (LCL, UCL)
	Type5	N	N	HR (LCL, UCL)

/*read in sample data attached*/
proc import datafile="…\have.txt"
out=have
dbms=tab replace;
getnames=yes;
guessingrows=max;
run;
proc phreg data=have;
class
	raceth (ref='2')
	surgery (ref='5') 
	chemo (ref='5')  
	radiation (ref='2')
        stage (ref= '2');
model duration*vital_statusX(1)=chemo
	surgery radiation stage
	age_at_diagnosis raceth 
      /rl selection=stepwise slentry=0.25 slstay=0.35 details;
run;     

[Reeza]

You can’t get it by default, you can capture the ODS tables and customize the output via proc tabulate or report. 

There are examples on lexjansen.com and some on my github though I think it’s not quite what you want. 

[Cruise]

Reeza please, I appreciate for some more hints. Do you have links got your github and lexjansen.com?

[Reeza]

Sorry, my GitHub one is PROC LIFETEST not PHREG, but it's here: 

https://gist.github.com/statgeek/d3bce2a9e2ef0523db9d

The concept is similar. 

1. Capture ODS tables blog post:

https://blogs.sas.com/content/iml/2017/01/09/ods-output-any-statistic.html

2. Creating Complex Reports via Cynthia Zender

https://www.lexjansen.com/pharmasug/2008/sas/SA08.pdf

3. Specific PROC PHREG example here, with code:

https://www.lexjansen.com/nesug/nesug07/po/po22.pdf

I would highly suggest reading the first two but using the code from the last paper which seems pretty close to what you're looking for. It will need modifications though, which is why you need the first two papers. 

[Cruise]

oh ouch, way beyond my SAS literacy. Hats off. But i need to dedicate good amount of time to digest all these and apply.

[Reeza]

Break it into smaller steps and each one on it's own isn't too bad. 

1. Capture ODS tables

2. Format using a data step

3. Proc print or report to display. 

The code shown has basically generalized this so it works for any process but you don't have to do that and probably shouldn't if you've never done so. Just some examples and you can run them to see how they work and if they're close enough to what you want.