I put a macro from one PharmaSUG paper here to seek you help to resovle the issue that : ERROR: Invalid reference value for _group.
/*
Data requirements:
1. Input dataset assumed to be one record per subject.
2. Stratification (macro var _stratavar) needs a numeric variable.
Data need another variable <_stratavar>_c has corresponding label for
proc format. e.g. TRTPN/TRTPN_C: 1/Drug A, 2/Placebo.
*/
%macro mKMplot ( _indata= /*Required.Input dataset assumed to be one record per
subject*/
,_timevar= /*Required. Time variable*/
,_censorvar= /*Required. Censoring variable, should include value
to identify a censored observation e.g: HYPER(0),
0 is censor*/
,_stratavar= /*Permissible. Analysis stratification for KM,
usually treatment group.*/
,_stratalis= /*Permissible. Study stratification groups*/
,_datafl= /*Permissible. Data filter for LIFETEST and PHREG
(where=(&datafl)), please use SAS logic. e.g.
GROUP=1*/
,_tinterval= /*Permissible. Time interval for displaying.
Default =3*/
,_header= /*Permissible. The header of the figure*/
,_Xlabel= /*Permissible. The label of X-axis (Time interval)*/
,_Ylabel= /*Permissible. The label of Y-axis (Survival)*/
,_ARtitle= /*Permissible. The label of Number of subjects at
risk*/
,_Refline= /*Permissible. Present reference line for median
survival rate. Default=N.*/
,_ref= /*Permissible. Study reference group*/
,_fnmed= /*Permissible. The footnote shows survival median or
not. Default=N.*/
,_fnhr= /*Permissible. The footnote shows hazard ratio or
not. Default=N.*/
,_fnp= /*Permissible. The footnote shows pvalue for hazard
ratio or not. Default=N.*/
,_outtype= rtf /*Permissible. Output document type. Default=rtf.*/ );
/*-----------------------*/
/* ticks for X-axis */
/*-----------------------*/
%if &_TINTERVAL = %then %let _TINTERVAL = 3; /*default=3*/
proc sql noprint;
select &_TINTERVAL.*ceil((max(&_TIMEVAR.)/&_TINTERVAL.)) into: _maxi
from &_INDATA.
;
quit;
%put &_maxi.;
/*-----------------------*/
/* strata */
/*-----------------------*/
%if &_STRATAVAR.= %then %do;
data &_INDATA.;
set &_INDATA.;
DMY = 1;
DMY_C = "All Subjects";
run;
%let _STRATAVAR = DMY;
%end;
/*-----------------------------*/
/* label for strata: STRATA */
/*-----------------------------*/
proc sort data= &_INDATA out = fmt nodupkey;
by &_STRATAVAR.;
run;
%let _STRATAVAR_C = %sysfunc(cats(%scan(&_STRATAVAR.,1),_C));
%put &_STRATAVAR_C;
data fmt (keep=fmtname type start label &_STRATAVAR.);
set fmt end=eos;
by &_STRATAVAR.;
retain fmtname "STRATA" type "N";
length start end $8. label $40.;
start = strip(put(_N_,best.));
end = strip(put(_N_,best.));
label = "&_STRATAVAR_C.";
if eos then call symputx ('_Nstrata', _N_);
run;
proc format library=work cntlin=fmt;
run;
/*-------------------------------------------------------------------------------*/
/*PROC LIFETEST for KM-plot, subject at risk, survival rate median, event numbers*/
/*-------------------------------------------------------------------------------*/
ods graphics / reset;
ods exclude all;
ods trace on;
ods output survivalplot=zz_plotdata CensoredSummary=zz_esum Quartiles=zz_qtp
%if &_Nstrata.>= 2 %then HomTests=zz_htest ;;
proc lifetest data=&_INDATA %if %bquote(&_DATAFL.) ne %then %do; %str((where=(&_DATAFL.))) %end;
method=pl
plots=survival(atrisk(outside)=0 to &_maxi. by &_TINTERVAL.);;
ods select SurvivalPlot CensoredSummary Quartiles %if &_Nstrata.>= 2 %then Homtests;;
time &_TIMEVAR. * &_CENSORVAR.;
strata &_STRATAVAR.;;
run;
data zz_plotdata;
set zz_plotdata;
by STRATUM;
retain CUMEVENT SURVPROB;
if first.STRATUM then do;
CUMEVENT=0;
SURVPROB=1;
end;
else do;
if EVENT ^= . then CUMEVENT=CUMEVENT+EVENT;
if SURVIVAL ^= . then SURVPROB=SURVIVAL;
end;
run;
/*-----------------------*/
/* Data for plot */
/*-----------------------*/
data fSURVIVAL;
set zz_plotdata;
format STRATUMNUM strata.;
run;
/*------------------------*/
/* footnote event/ median */
/*------------------------*/
proc sql;
create table fn1_0 as
select a.&_STRATAVAR., a.STRATUM, a.TOTAL, a.FAILED, b.ESTIMATE, b.LOWERLIMIT, b.UPPERLIMIT
from zz_esum as a
left join zz_qtp (where=(PERCENT=50)) as b
on a.&_STRATAVAR.=b.&_STRATAVAR.
having ^missing(&_STRATAVAR.)
order by STRATUM
;
quit;
%if &_REF= %then %do;
data _null_;
set fmt end=eos;
by START;
if eos then call symputx ('_REF', input(&_STRATAVAR.,best.));
run;
%end;
proc sql;
select LABEL into: _REFGP
from fmt
where &_STRATAVAR. = &_REF
;
quit;
data fn1;
set fn1_0;
by STRATUM;
/*length setting- for output space*/
%if &_fnmed=Y %then %do; length x $100.; %end;
%else %do; length x $120.; %end;
if missing(TOTAL) then TOTAL=0;
if missing(FAILED) then FAILED=0;
array zero1(*) ESTIMATE LOWERLIMIT UPPERLIMIT;
array zero2(*) $8. ESTIMATE_ LOWERLIMIT_ UPPERLIMIT_;
do i=1 to dim(zero1);
if missing(zero1(i)) then zero2(i)="NA";
else zero2(i) = strip(put(zero1(i),5.1));
end;
x1 = cats("Events : ", put(FAILED,best.) , "/", put(TOTAL,best.));
x2 = cats("Median : ", ESTIMATE_, " (", LOWERLIMIT_, ",", UPPERLIMIT_, ")" );
%if &_fnmed=Y %then %do; x = catx(", ", x1, x2); %end;
%else %do; x = x1; %end;
if &_stratavar = &_ref then call symputx ('_FAILED', FAILED); /*Failed - event number of ref
group*/
run;
/* depends on statification numbers */
%do i=1 %to &_Nstrata.;
%global _event_&i;
data _null_;
set fn1;
if _n_=&i. then call symput("_event_&i.",x);
run;
%end;
/*footnote HR and P, at least Straum >1*/
%if &_Nstrata.>= 2 %then %do;
ods output ParameterEstimates=zz_hazard;
proc phreg data=&_INDATA. %if %bquote(&_DATAFL.) ne %then %str((where=(&_DATAFL.)));;
ods select ParameterEstimates;
class &_STRATAVAR. (ref="&_REF.") &_STRATALIS.;
model &_TIMEVAR.* &_CENSORVAR.=&_STRATAVAR./ties=EXACT risklimits alpha=0.05 ;
strata &_STRATALIS.;
run;
ods exclude none;
proc sql;
create table fn2_0 as
select a.*,b.LABEL as _STRATANAME, monotonic() as _N_
from zz_hazard as a
left join FMT as b
on input(a.CLASSVAL0,best.)=b.&_STRATAVAR.
;;
quit;
%do i=1 %to %eval(&_Nstrata.-1);
proc sql;
select _STRATANAME into: _STRATANAME_&i
from fn2_0
where _N_ = &i.
;
quit;
%put &&_STRATANAME_&i ;
%end;
data fn2;
set fn2_0;
array zero1(*) HAZARDRATIO HRLOWERCL HRUPPERCL PROBCHISQ;
array zero2(*) $8. HAZARDRATIO_ HRLOWERCL_ HRUPPERCL_ PROBCHISQ_;
do k=1 to dim(zero1);
%if &_FAILED. ^= 0 %then %do;
if missing(zero1(k)) then zero2(k)="NA";
else zero2(k) = strip(put(zero1(k),5.2));
%end;
end;
if . < PROBCHISQ < 0.001 then PROBCHISQ_ = "<0.001";
else if PROBCHISQ > 0.999 then PROBCHISQ_ = ">0.999";
%if &_fnhr=Y %then %do;
%do i=1 %to %eval(&_Nstrata.-1);
%global _hr1_&i _hr2_&i _hr3_&i _hr4_&i;
if _n_ = &i then do;
%if &_Nstrata. = 2 %then
%let _hr1_&i = %str(Hazard ratio (reference=%trim(&_REFGP.)) : );
%if &_Nstrata. > 2 %then
%let _hr1_&i = %str(Hazard ratio for %trim(&&_STRATANAME_&i)
(reference=%trim(&_REFGP.)) : );
%if &_FAILED. = 0 %then %do;
%let _hr2_&i = {unicode '221e'x};
call symput ("_hr3_&i" , " (NA,NA)");
%end;
%else %do;
call symput ("_hr2_&i" , cats(HAZARDRATIO_));
call symput ("_hr3_&i" , " (" || cats(HRLOWERCL_, ",",
HRUPPERCL_, ")"));
%end;
%if &_fnp=Y %then call symput ("_hr4_&i" , cats(", p-value : ", PROBCHISQ_));;
end;
%put &&_hr1_&i &&_hr2_&i &&_hr3_&i &&_hr4_&i;
%end;
%end;
run;
%end;
/*------------------------*/
/* for output */
/*------------------------*/
%let _lev2=%sysevalf((%sysevalf(&_Nstrata)*.04)+.01);
%let _lev3=%sysevalf(&_lev2);
%let _lev4=%sysevalf((%sysevalf(&_Nstrata-1)*.04)+.01);
%let _lev1=%sysevalf(1-&_lev2-&_lev3-&_lev4);
%put &_lev1 &_lev2 &_lev3 &_lev4;
%macro temp;
proc template;
define statgraph f_SURVIVAL;
dynamic TOTAL FAILED ;
mvar _Nstrata _ARtitle _Ylabel _Xlabel _header _lev1 _lev2 _lev3 _lev4;
begingraph / border=FALSE designwidth=20cm designheight=14cm;
layout lattice / rows=4 rowweights=(&_lev1 &_lev2 &_lev3 &_lev4) rowgutter=0
columndatarange=union ;
/*1st cell- KM plot*/
cell;
layout overlay / border=false
yaxisopts=(offsetmin=0 linearopts=(thresholdmin=0 viewmin=0
tickvaluesequence=(start=0 end=1 increment=.1))
%if &_Ylabel^= %then %do; label=&_Ylabel. labelattrs=(size=9pt) %end;)
xaxisopts=(offsetmin=0 display=(line label ticks tickvalues)
%if &_Xlabel^= %then %do; label=&_Xlabel. labelattrs=(size=9pt) %end;
linearopts=(thresholdmin=0 tickvaluesequence=(start=0 end=%eval(&_maxi)
increment=%eval(&_tinterval) )));
scatterplot x=TIME y=CENSORED / group=STRATUMNUM name="scat";
stepplot x=TIME y=SURVIVAL / group=STRATUMNUM name="step";
%if &_Refline=Y %then %do; referenceline y=0.5 / lineattrs=(color=lightgray); %end;
endlayout;
endcell;
/*2nd cell- AT Risk*/
cell;
%if &_ARtitle^= %then %do;
cellheader;
entry halign=left textattrs=(size=8pt weight=bold) &_ARtitle. / border=FALSE;
endcellheader;
%end;
layout overlay/ walldisplay=none xaxisopts=(display=none);
blockplot x=TATRISK block=ATRISK / class=STRATUMNUM blockindex=STRATUMNUM
repeatedvalues=true display=(label values)
valuehalign=start valuefitpolicy=truncate
valueattrs=graphdatatext(size=8pt)
labelposition=left labelattrs=graphvaluetext
labelattrs=(size=8pt) includemissingclass=false;
endlayout;
endcell;
/*3rd cell- Event Number*/
cell;
layout overlay/ walldisplay=none xaxisopts=(display=none) pad=(top=0 left=5 bottom=0);
discretelegend "step" / valueattrs=(size=8pt) border=false displayclipped=true halign=left;
layout gridded / rows=&_Nstrata. border=false;
%do i=1 %to &_Nstrata.;
entry halign=left textattrs=(size=8pt) "%nrbquote(&&&_event_&i)" ;
%end;
endlayout;
endlayout;
endcell;
/*4th cell- Hazard Ratio*/
%if &_Nstrata. >=2 %then %do;
cell;
layout overlay/ walldisplay=none xaxisopts=(display=none);
layout gridded / rows=%eval(&_Nstrata.-1) border=false halign=left;
%do i=1 %to %eval(&_Nstrata.-1);
%if &_FAILED. ^= 0 %then %do; entry halign=left textattrs=(size=8pt) " " "&&_hr1_&i"
"&&_hr2_&i" "&&_hr3_&i" "&&_hr4_&i"; %end;
%else %if &_FAILED. = 0 %then %do; entry halign=left textattrs=(size=8pt) " "
"&&_hr1_&i" &&_hr2_&i "&&_hr3_&i" "&&_hr4_&i"; %end;
%end;
endlayout;
endlayout;
endcell;
%end;
endlayout;
endgraph;
end;
run;
quit;
%mend temp;
%temp;
options orientation=landscape papersize=letter nodate nonumber;
ods noproctitle escapechar="`";
ods listing style=listing;
ods graphics / reset noborder imagename="fSURVIVAL" ;
ods listing gpath="&_outputPath\";
ods &_OutType file ="&_outputPath\survival.&_outtype"
style=listing nogfootnote nogtitle bodytitle toc_data headery=720 footery=720;
proc sgrender data=fSURVIVAL template=f_SURVIVAL
%if &_HEADER^= %then description=&_HEADER.;;
run;
ods &_OutType close;
ods listing ;
%mend mKMplot;
data bmt;
set sashelp.bmt;
T_M=T/30.4375;/*Time from Unit of Days to Month*/
if group="ALL" then _group=1;
else if group = "AML-Low Risk" then _group=2;
else if group="AML-High Risk" then _group=3;
run;
PROC FORMAT;
VALUE STRATA 1="ALL"
2="AML-Low Risk"
3="AML-High Risk";
RUN;
proc lifetest data=bmt method=pl alpha=.05 outsurv=ci95 plots=survival;
time T_M*STATUS(1);
strata GROUP;
run;
%let outputpath=D:;
/*===By use of the above macro to generate enhanced K=M Plot===*/
%mKMplot ( _indata= BMT
,_timevar= T_M
,_censorvar= STATUS(1)
,_stratavar= GROUP
,_tinterval= 10
,_ref= 0
,_stratalis=
,_datafl=
,_header= "Figure 1: Kaplan-Meier Plot of Median Follow-up"
,_Xlabel= "Time (Month)"
,_Ylabel= "Probability of follow-up"
,_ARtitle="Number of Subjects of at risk"
,_Refline= Y
,_fnmed= Y
,_fnhr= N
,_fnp= N
,_outtype=rtf);
