## Cherry-picking? [Study Assessment]

»

3. The within-subject standard deviation of test and reference products will be compared, and the upper limit of the 90% confidence interval for the test-to-reference ratio of the within-subject variability should be ≤ 2.5.

» […] 90% confidence interval for the test-to-reference ratio of the within-subject variability ≤ 2.5 were not meet the criteria for all PK variables (Cmax, AUCt and inf).

Failed to demonstrate BE due to the higher within-subject variability of the test product. Full stop.

» Exercises, Observations and Analysis:

What do you mean by „Exercises”? Since the study failed are you asking for a recipe to cherry-pick?

»

1. We have taken subjects who have completed at least 2R or 2T in Reference Scaled Average Bio equivalence calculation (existing study).

That’s my interpretation as well. Although only the calculation of

*s*is given in Step 1 of the guidance by analogy the same procedure should be applicable for

_{WR}*s*.

_{WT}»

2. We have done the exercise who have completed all four treatments and did the statistical calculation- still failing on the same criteria marginally.

Leaving cherry-picking aside: By doing so, you drop available information. One should always use all. The more data you have, the more accurate/precise an estimate will be. Have a look at the formula to calculate the 100(1–α) CI of σ

_{WT}/σ

_{WR}:$$\left(\frac{s_{WT} / s_{WR}}{\sqrt[]{F_{\alpha /2,\nu_1,\nu_2}}},\frac{s_{WT} / s_{WR}}{\sqrt[]{F_{1-\alpha /2,\nu_1,\nu_2}}} \right)$$We have two different degrees of freedom (

*ν*

_{1},

*ν*

_{1}), the first associated with

*s*and the second with

_{WT}*s*.

_{WR}»

3. It was observed that if the “SWT” value should be closed to SWR value or lower, then 90% CI for the test-to-reference ratio of the within-subject variability ≤ 2.5 will meet the criteria.

Of course.

»

1. Which Reference Scaled Average Bioequivalence approach is acceptable in regulatory?

» Approach 1: Subject completed at least two test product will consider for SWT calculation and subject who completed at least two reference will consider for SWR calculation.

Yes.

»

Approach 2: Subjects who completed four period will be consider for SWR & SWT calculation.

No.

»

or both.

Which one will you pick at the end if one

*passes*and the other one

*fails*? The passing one, right? The FDA will love that. Be aware that the FDA recalculates every study.

BTW, how would you describe that in the SAP?

»

2. which are the factors adding variability to SWT?

That’s product-related. The idea behind the FDA’s reference-scaling for NTIDs is not only the narrow the limits but also to prevent products with higher variability than the reference’s entering the market.

»

3. Whether same formulation can be taken for the repeat bio-study with some clinical restrictions? If yes then what are the clinical factor to be considered?

As I wrote above, the failure to show BE was product-related. If you introduce clinical restrictions* in order to reduce within-subject variability – due to randomization – both products will be affected in the same way and

*s*/

_{WT}*s*be essentially the same like in the failed study.

_{WR}Reformulate.

PS: I changed the category of your post to yesterday and you to today. Wrong. Don’t test my patience – your problems are definitely study-specific (see the Policy for a description of categories).

- Which ones are you thinking about? I don’t see how it could be possible to reduce
*within*-subject variability by any means. Given, chaining volunteers to their beds might help.

*Dif-tor heh smusma*🖖

Helmut Schütz

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

Science Quotes

### Complete thread:

- NTI drug Bioequivalence study Statistical approach Sukalpa Biswas 2019-08-06 09:15 [Study Assessment]
- Cherry-picking?Helmut 2019-08-07 11:09
- digging out the possible reasons of the failure Sukalpa Biswas 2019-08-09 06:11
- Fed state: T higher variable than R Helmut 2019-08-09 12:36

- digging out the possible reasons of the failure Sukalpa Biswas 2019-08-09 06:11

- Cherry-picking?Helmut 2019-08-07 11:09