Bioequivalence and Bioavailability Forum

Dear AB!

❝ we used the code as given in FDA draft guidance

❝

❝ So we used the below statement of code for calculation of point estimate

❝ Geomean_ratio= pointest/ poinREF;

❝ However one of our other statisticain used the below statement

❝ Geomean_ratio= pointest;

Hhm. Why didn’t you use the code exactly as given by the FDA (see here)? Since the analysis is done on log-transformed data you must back-transform the estimate.

❝ […] So please let us know what is correct ASAP.

ASAP sends me into hibernation state. We are posting here in our free time!

I love deadlines.
I love the whooshing sound they make
as they go by. (Douglas Adams)

Dear AB,

Sorry but none is correct . If you have transformed your PK metric (AUC or Cmax) logarithmically, as the Progesterone guidance recommends, the ESTIMATE statement in the step "Intermediate analysis - ilat" gives you the difference estimate=log(µT)-log(µR). Recalling your school math you will obtain the point estimator back-transformed to the original values by

exp(estimate) = exp(log(µT)-log(µR)) = exp(log(µT))/exp(log(µR)) = µT/µR.

To cite the guidance:
"From the dataset IGLM2, calculate the following:
IGLM2: pointest=exp(estimate);
      ..."

The ratio µT/µR is sometimes called geometric mean ratio (GMR). But this is only an approximately valid statement.
See for that

Hauschke, Steinijans, Pigeot
"Bioequivalence Studies in Drug Development"
Wiley, Chichester (2007)
Chapter 3.2.2 "The lognormal distribution"

---

Ouch! Helmut has outpaced my post .
Full ACK with his statement concerning ASAP.