scaling of the x-axis for the survival curves

[clementinelrx]

Hello, I am having some difficulties with the scaling of my x-axis for a survival curve.

Here is the lifetest procedure used and the results. I would like to scale from 0 to 15 by 5. I have already used xOptions without success and I can't find any other solution on the sas users forum. Thanks

proc lifetest data = &data outsurv = tx method=km plots=survival(atrisk=0 to 15 by 5 outside(0.1)) ;
time &time*event_&time(0);
strata &arm;
ods output Quartiles = med (where=(Percent=50)) CensoredSummary = event HomTests = pv ;
run;

[ger15xxhcker]

You can achieve this by adding the option "axis=(0 to 15 by 5)" to the PLOTS statement in the PROC LIFETEST step. For example:

PROC LIFETEST DATA = &data 
    OUTSURV = tx 
    METHOD = KM 
    PLOTS = SURVIVAL(ATRISK = 0 TO 15 BY 5 OUTSIDE(0.1) AXIS=(0 TO 15 BY 5));
    TIME &TIME*EVENT_&TIME(0);
    STRATA &ARM;
    ODS OUTPUT 
        Quartiles = med (WHERE=(Percent=50)) 
        CensoredSummary = event 
        HomTests = pv
    ;
RUN;

[clementinelrx]

Thank you very much for your answer

[FreelanceReinh]

Hello @clementinelrx and welcome to the SAS Support Communities!

I think using xOptions was a good idea. To test it, I have used a slightly modified version VALUNG1 of the VALUNG dataset from Example 76.1 Product-Limit Estimates and Tests of Association of the PROC LIFETEST documentation

data valung1;
set valung;
survtime=survtime/50;
event_survtime=1-censor;
run;

together with the macro variable definitions

%let data=valung1;
%let time=survtime;
%let arm=cell;

Now, using

%ProvideSurvivalMacros
%let xOptions   = shortlabel=XNAME offsetmin=.05
                  linearopts=(viewmax=15 tickvaluelist=(0 5 10 15)
                              tickvaluefitpolicy=XTICKVALFITPOL);
%CompileSurvivalTemplates

(the highlighted values are the changes to the original value of xOptions as found in section "The Macro Variables")

followed by your PROC LIFETEST code I've got the expected x-axis scaling.

With the original templates the scaling would be 0 to 20 by 5. With the modified tickvaluelist= option alone, i.e., using viewmax=MAXTIME, the tick mark 20 would be dropped. Adding viewmax=15 truncated the longest survival curve at x=15+offsetmax.

Similarly, using

viewmax=25 tickvaluelist=(0 5 10 15 20 25)

the x-axis was extended beyond the largest survival or censoring time in the VALUNG1 data.

[clementinelrx]

Thank you very much for your help