Truncated power splines

Romain69100 — Mon, 31 Oct 2022 19:25:13 GMT

Dear SAS users,

I have a question truncated power basis spline. I am studying effect of Lymphocyte/Monocyte ratio (LMR) on progression free survival. To do so, i first centered the LMR on its median value which is 2.5. Then i used the 2 following code which conducted to same results:

/* 1 creation of truncated power basis spline */

data RELEVANCE_LMR;
set RELEVANCE_LMR;
d_ratio_l_m_c= d_ratio_l_m-2.5; /*centering the variable on its median value*/
d_ratio_l_m_c_sq = d_ratio_l_m_c**2;/*quadratic term*/
d_ratio_l_m_c_sq_plus = (d_ratio_l_m_c>=0)*d_ratio_l_m_c_sq;/*positive quadratic term*/

run;	


/* QUADRATIC SPLINE;*/

proc phreg data = RELEVANCE_LMR;
model D_TIME_PFS_IRC_EMA*D_CENS_PFS_IRC_EMA(1) = d_ratio_l_m_c d_ratio_l_m_c_sq d_ratio_l_m_c_sq_plus ;
output out=mydatsq Xbeta=mylinpredsq;
run;

/*second code*/
proc phreg data = RELEVANCE_LMR;
effect slmr = spline(d_ratio_l_m_c / basis=tpf(noint) degree=2 knotmethod=LIST(0));
model D_TIME_PFS_IRC_EMA*D_CENS_PFS_IRC_EMA(1) = slmr;
output out=mydatsquare Xbeta=mylinpredsquare;
run;

My first question is the following: When i extract linear predictors of these models and exponiate them, could you confirm me that i obtain Hazards ratio values relative to the value of 2.5?

I would like to add a second knot for a value of LMR=20. How can i modify the previous code to do it? ( Can i keep the d_ratio_l_m_c variable which is centered on 2.5 and add 17.5 in the list option? and how hazard ratios will be interpreted (still elative to 2.5)?

Thanks a lot for your help!

topic Truncated power splines in Statistical Procedures

Truncated power splines