topic Calculate 1-RR 95% CI when all participants experience event in Statistical Procedures

Calculate 1-RR 95% CI when all participants experience event

mariko5797 — Fri, 12 Nov 2021 19:00:06 GMT

This is more of a statistics question than a SAS question but I was wondering if it were possible to calculate 1-RR and its 95% CI in SAS when all participants experienced the event of interest.

data have;
 input trt $ cnt tot flag;
 cards;
  1	0	4	0
  1 4	4	1
  2	0	8	0
  2	8	8	1
  3	0	6	0
  3	6	6	1
  ;
run;

proc freq data= have;
 weight cnt;
 tables flag*trt / out= _ve_ relrisk(method= fm);
run;

The above code doesn't work since there are no subjects with FLAG=0 (i.e. did not develop disease). I am trying to calculate vaccine efficacy defined by 1-RR, but if I can't calculate that, then is there a reasonable alternative I can use in SAS?

Re: Calculate 1-RR 95% CI when all participants experience event

Ksharp — Sat, 13 Nov 2021 12:12:59 GMT

Firstly, RELRISK need 2x2 contingency table ,but yours is 2x3 .

Secondly, the count of flag=0 is zero . SAS also can't calculated it .

Check your data firstly.

data have;
 input trt $ cnt tot flag;
 cards;
  1	0	4	0
  1 4	4	1
  2	0	8	0
  2	8	8	1
  3	0	6	0
  3	6	6	1
  ;
run;

proc freq data= have;
where trt in ('1' '2');
 weight cnt/zero ;
 tables flag*trt/nocol norow nopercent  relrisk(NONINFERIORITY method= fm);
 exact relrisk;
run;

Re: Calculate 1-RR 95% CI when all participants experience event

Rick_SAS — Sat, 13 Nov 2021 12:34:00 GMT

A similar topic (CI for a proportion) was discussed in a previous thread.

The suggestion there was to use the EXACT BINOMIAL statement in PROC FREQ.

Re: Calculate 1-RR 95% CI when all participants experience event

mariko5797 — Mon, 15 Nov 2021 18:04:10 GMT

Wouldn't using exact binomial be different from relative risk? I already used binomial(cl= wilson); exact binomial; to obtain the infection rate, but I am trying to find RR to obtain the vaccine efficacy.

Would the hazard ratio make an okay approximation since 100% of subjects experienced the event?

Re: Calculate 1-RR 95% CI when all participants experience event

Rick_SAS — Mon, 15 Nov 2021 21:58:59 GMT

> Wouldn't using exact binomial be different from relative risk?

Correct. I wanted to direct you to the other discussion and its references in case anything there was relevant to your question.