Confidence interval for a relative risk in a 2x2 table - using PROC PHREG

SteffenF — Mon, 18 Sep 2017 15:42:37 GMT

I want to calculate the confidence interval for a relative risk in a 2x2 table (X * Y).

For a reason - not relevant for this topic - I would like to use PROC PHREG to accomplish this.

Strangely, with PROC PHREG I only get the same confidence interval as with procedure PROC FREQ or PROC GENMOD,
when I use the EVENTCODE=1 option in PHREG. Since there are only two possible levels for Y (0/1), I do not understand
how this option can have an impact on the result.

When I do not use the EVENTCODE=1 option, the confidence interval for the relative risk is much wider and seems inappropriate given the discrepancy with the results from PROC FREQ and PROC GENMOD

Any suggestions are more than welcome!

/*simulate data such the the relative risk equals 1.5 (60% if x=1 versus 40% if x=0)*/
data test;
do x=0 to 1;
 do id=1 to 100;
  if x=0 then  y=(id>40);
  if x=1 then  y=(id>60);
  output;
 end;
end;
run;

title 'Using PROC FREQ';
ods select none;
proc freq data=test;
 tables x*y/nocol  nocum nopercent relrisk  ;
 ods output relativerisks=relrisk_FREQ;
run;
ods select all;
proc print data=relrisk_FREQ noobs label; var value LowerCL UpperCL;where statistic="Relative Risk (Column 2)";format _numeric_ 8.5;run;

title 'Using PROC GENMOD';
ods select none;
proc genmod data=test ;
 model y=x/dist=bin link=log;
 estimate 'RR' x 1/exp;
 ods output estimates=relrisk_GENMOD;
run;
ods select all;
proc print data=relrisk_GENMOD noobs label; where label="Exp(RR)";var LBetaEstimate LBetaLowerCL LBetaUpperCL;format _numeric_ 8.5;run;

title 'Using PROC PHREG - with eventcode specification';
ods select none;
data test;set test;FUtime=1;run;
proc phreg data=test;
class x;
model FUtime*y(0)=x/eventcode=1 rl;
ods output parameterestimates=RR_phreg;
run;
ods select all;
proc print data=RR_phreg noobs;var hazardratio HRLowerCL HRUpperCL;format _numeric_ 8.5;run;

title 'Using PROC PHREG - without eventcode specification';
ods select none;
data test;set test;FUtime=1;run;
proc phreg data=test;
class x;
model FUtime*y(0)=x/ rl;
ods output parameterestimates=RR_phreg;
run;
ods select all;
proc print data=RR_phreg noobs;var hazardratio HRLowerCL HRUpperCL;format _numeric_ 8.5;run;

Re: Confidence interval for a relative risk in a 2x2 table - using PROC PHREG

Norman21 — Thu, 21 Sep 2017 17:01:43 GMT

I get errors in the log when I run your code, but output in results viewer.

Have you checked your log to make sure the program is running as it should?

Re: Confidence interval for a relative risk in a 2x2 table - using PROC PHREG

SteffenF — Fri, 22 Sep 2017 07:04:15 GMT

Dear Norman,

Thanks for your response!

I do not get any error in the log.

The version I'm working with is SAS 9.4 (TS1M3) X64_8PRO WIN 6.2.9200 (hence, slightly different)

Kind regards,

Steffen

topic Re: Confidence interval for a relative risk in a 2x2 table - using PROC PHREG in Statistical Procedures

Confidence interval for a relative risk in a 2x2 table - using PROC PHREG

Re: Confidence interval for a relative risk in a 2x2 table - using PROC PHREG

Re: Confidence interval for a relative risk in a 2x2 table - using PROC PHREG