Bioequivalence and Bioavailability Forum • FDA’s RSABE code (lenghty!)

Helmut
★★★

Vienna, Austria,
2012-01-03 04:27
(5274 d 05:55 ago)

Posting: # 7854
Views: 13,804

FDA’s RSABE code (lenghty!) [RSABE / ABEL]

Hi to all copypasters!

I have mislaid the Apr 2010 draft, but in the current Feb 2011 version five lines of SAS code are missing. Well, not really missing, but printed in white on a white background. Here is the entire stuff in all its beauty (missing lines in red):

Example SAS Codes: partial reference-replicated 3-way design

Dataset containing TEST observations:

data test;

  set pk;

  if trt='T';

  latt=lauct;

run;

Dataset containing REFERENCE 1 observations:

data ref1;

  set ref;

  if (seq=1 and per=2) or (seq=2 and per=1) or (seq=3 and per=1);

  lat1r=lauct;

run;

Dataset containing REFERENCE 2 observations:

data ref2;

  set ref;

  if (seq=1 and per=3) or (seq=2 and per=3) or (seq=3 and per=2);

  lat2r=lauct;

run;

Determine I_ij and D_ij

data scavbe;

  merge test ref1 ref2;

  by seq subj;

  ilat=latt-(0.5*(lat1r+lat2r));

  dlat=lat1r-lat2r;

run;

Intermediate analysis - ilat

proc glm data=scavbe;

  class seq;

  model ilat=seq/clparm alpha=0.1;

  estimate 'average' intercept 1 seq 0.3333333333 0.3333333333 0.3333333333;

  ods output overallanova=iglm1;

  ods output Estimates=iglm2;

  ods output NObs=iglm3;

  title1 'scaled average BE';

run;

IGLM2:

pointest=exp(estimate);

x=estimate**2–stderr**2;

boundx=(max((abs(LowerCL)),(abs(UpperCL))))**2;

Intermediate analysis - dlat

proc glm data=scavbe;

  class seq;

  model dlat=seq;

  ods output overallanova=dglm1;

  ods output NObs=dglm3;

  title1 'scaled average BE';

run;

DGLM1:

dfd=df;

s2wr=ms/2;

From the above parameters, calculate the final 95% upper confidence bound:

theta=((log(1.25))/0.25)**2;

y=-theta*s2wr;

boundy=y*dfd/cinv(0.95,dfd);

sWR=sqrt(s2wr);

critbound=(x+y)+sqrt(((boundx-x)**2)+((boundy-y)**2));

Example SAS Codes: fully replicated 4-way design

Dataset containing TEST 1 observations:

data test1;

  set test;

  if (seq=1 and per=1) or (seq=2 and per=2);

  lat1t=lauct;

run;

Dataset containing TEST 2 observations:

data test2;

  set test;

  if (seq=1 and per=3) or (seq=2 and per=4);

  lat2t=lauct;

run;

Dataset containing REFERENCE 1 observations:

data ref1;

  set ref;

  if (seq=1 and per=2) or (seq=2 and per=1);

  lat1r=lauct;

run;

Dataset containing REFERENCE 2 observations:

data ref2;

  set ref;

  if (seq=1 and per=4) or (seq=2 and per=3);

  lat2r=lauct;

run;

Determine I_ij and D_ij

data scavbe;

  merge test1 test2 ref1 ref2;

  by seq subj;

  ilat=0.5*(lat1t+lat2t-lat1r-lat2r);

  dlat=lat1r-lat2r;

run;

Intermediate analysis - ilat

proc mixed data=scavbe;

  class seq;

  model ilat =seq/ddfm=satterth;

  estimate 'average' intercept 1 seq 0.5 0.5/e cl alpha=0.1;

  ods output CovParms=iout1;

  ods output Estimates=iout2;

  ods output NObs=iout3;

  title1 'scaled average BE';

  title2 'intermediate analysis - ilat, mixed';

run;

IOUT2:

pointest=exp(estimate);

x=estimate**2–stderr**2;

boundx=(max((abs(lower)),(abs(upper))))**2;

Intermediate analysis - dlat

proc mixed data=scavbe;

  class seq;

  model dlat=seq/ddfm=satterth;

  estimate 'average' intercept 1 seq 0.5 0.5/e cl alpha=0.1;

  ods output CovParms=dout1;

  ods output Estimates=dout2;

  ods output NObs=dout3;

  title1 'scaled average BE';

  title2 'intermediate analysis - dlat, mixed';

run;

DOUT1:
s2wr=estimate/2;
DOUT2:
dfd=df;

From the above parameters, calculate the final 95% upper confidence bound:

theta=((log(1.25))/0.25)**2;

y=-theta*s2wr;

boundy=y*dfd/cinv(0.95,dfd);

sWR=sqrt(s2wr);

critbound=(x+y)+sqrt(((boundx-x)**2)+((boundy-y)**2));

Calculation of unscaled 90% bioequivalence confidence intervals:

PROC MIXED

  data=pk;

  CLASSES SEQ SUBJ PER TRT;

  MODEL LAUCT = SEQ PER TRT/ DDFM=SATTERTH;

  RANDOM TRT/TYPE=FA0(2) SUB=SUBJ G;

  REPEATED/GRP=TRT SUB=SUBJ;

  ESTIMATE 'T vs. R' TRT 1 -1/CL ALPHA=0.1;

  ods output Estimates=unsc1;

  title1 'unscaled BE 90% CI - guidance version';

  title2 'AUCt';

run;

data unsc1;

  set unsc1;

  unscabe_lower=exp(lower);

  unscabe_upper=exp(upper);

run;

Enjoy!

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

jag009 ★★★ NJ, 2013-01-14 20:47 (4896 d 13:36 ago) @ Helmut Posting: # 9837 Views: 8,689	FDA’s RSABE code (lenghty!) Post reply
	Hi Helmut, I was just checking the 4-way full rep code. You wrote: ❝ From the above parameters, calculate the final 95% upper confidence bound: ❝ `theta=((log(1.25))/0.25)2;` From the progesterone FDA guidance Dec 2012. I see different equation for theta: `theta=((log(1.11111))/0.1)2;` Yours is wrong? Thanks John

Helmut
★★★

Vienna, Austria,
2013-01-14 22:40
(4896 d 11:43 ago)

@ jag009
Posting: # 9839
Views: 8,893

FDA’s RSABE code (lenghty!)

Post reply

Hi John,

❝ ❝ From the above parameters, calculate the final 95% upper confidence bound:

❝ ❝ theta=((log(1.25))/0.25)**2;

❝

❝ From the progesterone FDA guidance Dec 2012. I see different equation for theta:

❝

❝ theta=((log(1.11111))/0.1)**2;

❝

❝ Yours is wrong?

Nope. Maybe you had the warfarin-guidance also open in your reader. :-D

Yours would mean downscaling. In the Dec 2012 revision FDA corrected only the ‘visibility’ (missing lines which were formatted in white on a white background). The formula I gave is still the same (upscaling) like in previous versions. See p.6 for the partial replicate and p.8 for the full replicate.

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes