topic Re: How can I draw spline curve using adjusterd Cox proportional hazard model? in New SAS User

How can I draw spline curve using adjusterd Cox proportional hazard model?

Paako — Thu, 13 Dec 2018 01:55:33 GMT

Hi, I'm new to SAS and having a trouble with drawing a cubic spline curve. I'm using SAS 9.4.

I analyzed my data with the multivariable-adjusted Cox proportional hazard regression model. The program is like this;

proc phreg data=data1;

model personyears*case(0)=mets sex age DM/rl;

run;

"mets" means METs which is used to estimate physical activity level. Sex, age, DM are covariates.

I would like to draw a cubic spline curve. The y-axis is hazard ratio and the x-axis is METs. The hazard ratio is 1 when METs is 0.

I'm wondering if "proc icphreg" may be an applicable program, but I cannot find the program which fits my request. Could you tell me the program to solve my question? Thank you.

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Reeza — Thu, 13 Dec 2018 04:36:53 GMT

Did you see this example?

https://documentation.sas.com/?docsetId=statug&docsetVersion=15.1&docsetTarget=statug_icphreg_examples01.htm&locale=en

@Paako wrote:

Hi, I'm new to SAS and having a trouble with drawing a cubic spline curve. I'm using SAS 9.4.

I analyzed my data with the multivariable-adjusted Cox proportional hazard regression model. The program is like this;

proc phreg data=data1;

model personyears*case(0)=mets sex age DM/rl;

run;

"mets" means METs which is used to estimate physical activity level. Sex, age, DM are covariates.

I would like to draw a cubic spline curve. The y-axis is hazard ratio and the x-axis is METs. The hazard ratio is 1 when METs is 0.

I'm wondering if "proc icphreg" may be an applicable program, but I cannot find the program which fits my request. Could you tell me the program to solve my question? Thank you.

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Paako — Thu, 13 Dec 2018 07:18:00 GMT

Hi, Reeza, thank you for the reply.

I had already checked the page you mentioned and tried first this program;

proc icphreg data=data1;

model personyears*case(0)=mets sex age DM/basehaz=splines;
hazardratio activemets;
run;

but the log says there is a syntax error and it doesn't work. Do you have any idea?

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Reeza — Thu, 13 Dec 2018 16:07:55 GMT

Post your log. But given that you haven't included activemets as a variable but do have it in your hazardratio statement I suspect that alone would cause an error.

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

ballardw — Thu, 13 Dec 2018 16:37:13 GMT

@Paako wrote:

Hi, Reeza, thank you for the reply.

I had already checked the page you mentioned and tried first this program;

proc icphreg data=data1;

model personyears*case(0)=mets sex age DM/basehaz=splines;
hazardratio activemets;
run;

but the log says there is a syntax error and it doesn't work. Do you have any idea?

From the documentation on the HAZARDRATIO statement in ICPHREG:

The HAZARDRATIO statement enables you to request hazard ratios for any variable in the model at customized settings.

which agrees with @Reeza's diagnostic statement. If the variable isn't on the model statement it should not be on the Hazardratio.

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Paako — Fri, 14 Dec 2018 11:45:23 GMT

Thank you for the reply. I made a mistake when I posted my program. I wrote the program as an example, and this is my actual program;

proc icphreg data=jalskaiseki10;
model gyear*FIRST_HASSHO(0)=activemets sexa age DMC/basehaz=splines;
hazardratio activemets;
run;

but the log says there is a syntax error again.

here is the log;

159 proc icphreg data=jalskaiseki10;
160 model gyear*FIRST_HASSHO(0)=activemets sexa age DMC/basehaz=splines;
           -----
           79
           76
ERROR 79-322: (designate
ERROR 76-322: there is a syntax error
161 run;

My sas is not in English, so " designate" and "there is a syntax error" is my translation.

In the program, "hazardratio" is red-letter allthough "proc icphreg", "model" and "run" is blue-letter. I suspect there is a syntax error in

"hazardratio activemets;", but I cannot find the problem.

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Reeza — Fri, 14 Dec 2018 16:07:33 GMT

What version of SAS are you using? I'm guessing ICPHREG is a newer proc and you don't have a version that supports it.

Check this via, and see what the SAS/STAT version is.

proc product_status;run;

@Paako wrote:

Thank you for the reply. I made a mistake when I posted my program. I wrote the program as an example, and this is my actual program;

proc icphreg data=jalskaiseki10;
model gyear*FIRST_HASSHO(0)=activemets sexa age DMC/basehaz=splines;
hazardratio activemets;
run;

but the log says there is a syntax error again.

here is the log;

159 proc icphreg data=jalskaiseki10;
160 model gyear*FIRST_HASSHO(0)=activemets sexa age DMC/basehaz=splines;
           -----
           79
           76
ERROR 79-322: (designate
ERROR 76-322: there is a syntax error
161 run;

My sas is not in English, so " designate" and "there is a syntax error" is my translation.

In the program, "hazardratio" is red-letter allthough "proc icphreg", "model" and "run" is blue-letter. I suspect there is a syntax error in

"hazardratio activemets;", but I cannot find the problem.

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Paako — Sun, 16 Dec 2018 12:27:44 GMT

The log says;

For SAS/STAT ...
   Custom version information: 13.2
For SAS/GRAPH ...
   Custom version information: 9.4_M2
For High Performance Suite ...
   Custom version information: 2.2_M3
   Image version information: 9.04.01M0P042215

I think my version is 13.2.

Doesn't it have a way to draw a spline curve? Or another syntax?

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Reeza — Mon, 17 Dec 2018 01:28:51 GMT

ICPHREG is available in 13.2 so that's not the issue. You are definitely on an older version though, 9.4M2 is from 2014. Anyways, please post your full code and log, since we can't really help with what you've posted so far. The error could be in your MODEL statement or somewhere else.

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Paako — Fri, 21 Dec 2018 05:36:10 GMT

Here is my sas code. In the code, EGFRC=1 means eGFR<50. If eGFR>=50. EGFRC=0;

data jalshassho;
set base4;
if (area=88 and 2005<=eyear) or (area=23 and 2004<=eyear) ;
IF EGFRC=1 THEN ; ELSE DELETE;
run;
proc sort data = jalshassho;
by REGI eyear;
run ;

DATA jalshassho1;
SET jalshassho;
BY REGI eyear ;
if FIRST.REGI=1 THEN FIRST_HASSHO=1;
IF FIRST_HASSHO=1 THEN; ELSE DELETE;

KEEP REGI KENKAID eyear area EMONTH EDAY CRT EGFR_E EGFRC FIRST_HASSHO;

RENAME
eyear = EYEAR_CASE
EMONTH = EMONTH_CASE
EDAY = EDATE_CASE
CRT = CRT_CASE
EGFR_E = EGFR_CASE
EGFRC = EGFRC_CASE;
run;
proc sort data=jalshassho1;
by EYEAR_CASE area;
run;

DATA jalslast;
SET base4;
BY REGI eyear ;
if LAST.REGI=1 THEN LAST_NENDO=1;
IF LAST_NENDO=1 THEN; ELSE DELETE;

KEEP REGI KENKAID eyear area EMONTH EDAY CRT EGFR_E EGFRC LAST_NENDO;

RENAME
eyear = EYEAR_LAST
EMONTH = EMONTH_LAST
EDAY = EDATE_LAST
CRT = CRT_LAST
EGFR_E = EGFR_LAST
EGFRC = EGFRC_LAST ;
run;

data jalsbase0;
set base4;
activemets=sum(mets_shigoto_d,mets_hoko_d,mets_kaji_d,mets_undo_d,mets_yoka_d);
run;
data jalsbase;
set jalsbase0;
if activemets="." then delete; run;
proc sort data=jalsbase;
by regi;
run;
proc sort data=jalshassho1;
by regi;
run;
proc sort data=jalslast;
by regi;
run;

DATA jalskaiseki ;
MERGE
jalsbase
jalslast
jalshassho1;
by REGI ;
run;
data jalskaiseki1;
set jalskaiseki;
GYEAR = (MDY(EMONTH_LAST,EDATE_LAST,EYEAR_LAST)-MDY(EMONTH, EDAY, eyear))/365.25 ;

IF FIRST_HASSHO=1
then GYEAR = (MDY(EMONTH_CASE, EDATE_CASE, EYEAR_CASE)-MDY(EMONTH, EDAY, eyear))/365.25 ;

IF GYEAR=.    THEN delete;
IF GYEAR=0   THEN delete;
IF GYEAR<0           THEN delete;
IF FIRST_HASSHO=. THEN FIRST_HASSHO=0;

IF GYEAR=0 and FIRST_HASSHO=0 THEN delete;;
run;

data jalskaiseki2;
set jalskaiseki1;

DMC=.;
IF GLU>0 or HBA1CNGSP>0 or HBA1CJDS>0 THEN DMC=0;
IF FSTIME<8 AND GLU >=200 THEN DMC=1;
IF GLU=. and HBA1CNGSP >=6.5 THEN DMC=1;
IF GLU=. and HBA1CJDS >=6.1 THEN DMC=1;
IF 8<=FSTIME<24 AND GLU >=126 THEN DMC=1;
run;

data jalskaiseki3;
set jalskaiseki2;
del = 0;

if CHOL = .   then del=1;
    if TG = .   then del=1;
if CRT = .  then del=1;

IF GYEAR=.   then del=1;
if sex =.   then del=1;
if AGE =.   then del=1;
if HT =.   then del=1;
if WT =.   then del=1;
if CRT =.  then del=1;

if eGFRC=1 then del=2;

run;

data jalskaiseki4;
set jalskaiseki3;
if del=1 then delete;
if del=2 then delete;

RUN;

proc sort data = jalskaiseki4 ; by sex area;run;
data jalskaiseki5;
set jalskaiseki4;
if activemets="." then delete; run;
data jalskaiseki6;
set jalskaiseki5;
areas=.;
if area=88 then areas=0;
if area=23 then areas=1;
sexa=.;
if sex=1 then sexa=0; /*male*/
if sex=2 then sexa=1; /*female*/
run;

/*spline curve*/
proc icphreg data=jalskaiseki6;
model gyear*FIRST_HASSHO(0)=activemets sexa age DMC/basehaz=splines;
hazardratio activemets;
run;

******************************************

And here is the log;

274 data jalshassho;
275 set base4;
276 if (area=88 and 2005<=eyear) or (area=23 and 2004<=eyear) ;
277 IF EGFRC=1 THEN ; ELSE DELETE;
278 run;

NOTE: dataset WORK.BASE4 is 88814 observations

NOTE: dataset WORK.JALSHASSHO is 5915observations, 746 variables

279 proc sort data = jalshassho;
280 by REGI eyear;
281 run ;

NOTE: dataset WORK.JALSHASSHO is 5915 observations

NOTE: dataset WORK.JALSHASSHO is 5915observations, 746 variables

282
283 DATA jalshassho1;
284 SET jalshassho;
285 BY REGI eyear ;
286 if FIRST.REGI=1 THEN FIRST_HASSHO=1;
287 IF FIRST_HASSHO=1 THEN; ELSE DELETE;
288
289 KEEP REGI KENKAID eyear area EMONTH EDAY CRT EGFR_E EGFRC FIRST_HASSHO;
290
291 RENAME
292      eyear = EYEAR_CASE
293      EMONTH = EMONTH_CASE
294      EDAY = EDATE_CASE
295      CRT = CRT_CASE
296      EGFR_E = EGFR_CASE
297      EGFRC = EGFRC_CASE;
298 run;

NOTE: dataset WORK.JALSHASSHO is 5915 observations

NOTE: dataset WORK.JALSHASSHO1 is 1595observations, 10 variables

299 proc sort data=jalshassho1;
300 by EYEAR_CASE area;
301 run;

NOTE: dataset WORK.JALSHASSHO1 is 1595 observations

NOTE: dataset WORK.JALSHASSHO1 is 1595observations, 10 variables

302
303
304 DATA jalslast;
305 SET base4;
306 BY REGI eyear ;
307 if LAST.REGI=1 THEN LAST_NENDO=1;
308 IF LAST_NENDO=1 THEN; ELSE DELETE;
309
310 KEEP REGI KENKAID eyear area EMONTH EDAY CRT EGFR_E EGFRC LAST_NENDO;
311
312 RENAME
313      eyear = EYEAR_LAST
314      EMONTH = EMONTH_LAST
315      EDAY = EDATE_LAST
316      CRT = CRT_LAST
317      EGFR_E = EGFR_LAST
318      EGFRC = EGFRC_LAST ;
319 run;

NOTE: dataset WORK.BASE4 is 88814 observations

NOTE: dataset WORK.JALSLAST is 15893observations, 10 variables

320
321 data jalsbase0;
322 set base4;
323 activemets=sum(mets_shigoto_d,mets_hoko_d,mets_kaji_d,mets_undo_d,mets_yoka_d);
324 run;

NOTE: dataset WORK.BASE4 is 88814 observations

NOTE: dataset JALSBASE0 is 88814observations, 746 variables

325 data jalsbase;
326 set jalsbase0;
327 if activemets="." then delete; run;

NOTE: dataset WORK.JALSBASE0 is 88814 observations

NOTE: dataset WORK.JALSBASE is 3216observations, 746 variables

328 proc sort data=jalsbase;
329 by regi;
330 run;

NOTE: dataset WORK.JALSBASE is 3216 observations

NOTE: dataset WORK.JALSBASE is 3216observations, 746 variables

331 proc sort data=jalshassho1;
332 by regi;
333 run;

NOTE: dataset JALSHASSHO1 is 1595 observations

NOTE: dataset WORK.JALSHASSHO1 is 1595observations, 10 variables

334 proc sort data=jalslast;
335 by regi;
336 run;

NOTE: dataset WORK.JALSLAST is 15893 observations

NOTE: dataset WORK.JALSLAST is 15893observations, 10 variables
337
338
339 DATA jalskaiseki ;
340 MERGE
341      jalsbase
342      jalslast
343      jalshassho1;
344 by REGI ;
345 run;

NOTE: dataset WORK.JALSBASE is 3216 observations

NOTE: dataset WORK.JALSLAST is 15893observations

NOTE: dataset WORK.JALSHASSHO1 is 1595 observations

NOTE: dataset WORK.JALSKAISEKI is 15894 observations, 761 variables

346 data jalskaiseki1;
347 set jalskaiseki;
348 GYEAR = (MDY(EMONTH_LAST,EDATE_LAST,EYEAR_LAST)-MDY(EMONTH, EDAY, eyear))/365.25 ;
349
350 IF FIRST_HASSHO=1
351      then GYEAR = (MDY(EMONTH_CASE, EDATE_CASE, EYEAR_CASE)-MDY(EMONTH, EDAY, eyear))/365.25
351! ;
352
353 IF GYEAR=.          THEN delete;
354 IF GYEAR=0         THEN delete;
355 IF GYEAR<0           THEN delete;
356 IF FIRST_HASSHO=. THEN FIRST_HASSHO=0;
357
358 IF GYEAR=0 and FIRST_HASSHO=0 THEN delete;;
359 run;

NOTE: dataset WORK.JALSKAISEKI is 15894 observations

NOTE: dataset WORK.JALSKAISEKI is 3025 observations, 762 variables

360
361
362 data jalskaiseki2;
363 set jalskaiseki1;
364
365
366 DMC=.;
367      IF GLU>0 or HBA1CNGSP>0 or HBA1CJDS>0 THEN DMC=0;
368      IF FSTIME<8     AND     GLU >=200           THEN DMC=1;
369      IF GLU=. and HBA1CNGSP >=6.5    THEN DMC=1;
370      IF GLU=. and HBA1CJDS >=6.1     THEN DMC=1;
371      IF 8<=FSTIME<24 AND GLU >=126               THEN DMC=1;
372 run;

NOTE: dataset WORK.JALSKAISEKI1 is 3025 observations

NOTE: dataset WORK.JALSKAISEKI2 is 3025observations, 763 variables

373
374 data jalskaiseki3;
375 set jalskaiseki2;
376   del = 0;
377
378      if CHOL = .         then del=1;
379      if TG   = .         then del=1;
380      if CRT = .      then del=1;
381
382      IF GYEAR=.          then del=1;
383      if sex =.           then del=1;
384      if AGE =.          then del=1;
385      if HT   =.          then del=1;
386      if WT   =.          then del=1;
387      if CRT =.      then del=1;
388
389 if eGFRC=1 then del=2;
390
391 run;

NOTE: dataset WORK.JALSKAISEKI2 is 3025 observations

NOTE: dataset WORK.JALSKAISEKI3 is 3025 observations, 764 variables

392
393
394 data jalskaiseki4;
395 set jalskaiseki3;
396 if del=1 then delete;
397 if del=2 then delete;
398
399 RUN;

NOTE: dataset WORK.JALSKAISEKI2 is 3025 observations

NOTE: dataset WORK.JALSKAISEKI4 is 2873observations, 764 variables

400
401 proc sort data = jalskaiseki4 ; by sex area;run;

NOTE: dataset WORK.JALSKAISEKI4 is 2873 observations

NOTE: dataset WORK.JALSKAISEKI4 is 2873 observations, 764 variables

402 data jalskaiseki5;
403 set jalskaiseki4;
404 if activemets="." then delete; run;

NOTE: dataset WORK.JALSKAISEKI4 is 2873 observations

NOTE: dataset WORK.JALSKAISEKI5 is 2873 observations, 764 variables

405 data jalskaiseki6;
406 set jalskaiseki5;
407 areas=.;
408 if area=88 then areas=0;
409 if area=23 then areas=1;
410 sexa=.;
411 if sex=1 then sexa=0; /*male*/
412 if sex=2 then sexa=1; /*female*/
413 run;

NOTE: dataset WORK.JALSKAISEKI5 is 2873 observations

NOTE: dataset WORK.JALSKAISEKI6 is 2873 observations, 766 variables

414
415 /*spline curve*/
NOTE: there is an error

416 proc icphreg data=jalskaiseki6;
417 model gyear*FIRST_HASSHO(0)=activemets sexa age DMC/basehaz=splines;
           -----
           79
           76
ERROR 79-322: designate (
ERROR 76-322: there is a syntax error. We ignored the statement.

418 hazardratio activemets;
419 run;

*****************************

The comment in the log is my translation. Sorry it's long. Could you tell me what is the problem?

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Reeza — Fri, 21 Dec 2018 16:03:04 GMT

Try running the demo here. https://documentation.sas.com/api/docsets/statug/15.1/content/statug_code_icphgex1.htm?locale=en

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Paako — Sat, 22 Dec 2018 13:08:17 GMT

I got the result such as "Analysis of Maximum Likelihood Parameter Estimates".
The log didn't tell any error. Now I have 2 questions. The first one is how I can get the spline curve
by running the demo. The second one is why my code doesn't work.

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Reeza — Sat, 22 Dec 2018 17:27:41 GMT

Check the log, is there a comment about ODS Graphics being off? Make sure to turn it on first:

ods graphics on;

Re: How can I draw spline curve using adjusterd Cox proportional hazard model?

Paako — Thu, 03 Jan 2019 04:04:28 GMT

I'm sorry for the late reply. Thank you for your advice. I'll try it.