Bioequivalence and Bioavailability Forum

Moloko
☆

Russia,
2020-12-22 12:41
(899 d 06:16 ago)

Posting: # 22153
Views: 1,822

Significant effects [General Statistics]

Hi all,

Could somebody help me with my doubts about significant sequence/treatment/period
effects in BE study? I got the following anova results in 2x2x4 design:

                 DF Sum Sq Mean Sq      F   P-val

sequence          1  0.558   0.558 10.269   0.002

period            3  0.566   0.189  3.472   0.021

treatment         1  0.075   0.075  1.384   0.244

sequence:subject 23  7.632   0.332  6.102  <0.001

residual         69  3.752   0.054

My СI-s are within the range of 80.000% - 125.000%, so formally BE acceptance criterion is met. But these p-values scare me... :confused:

I don't have much experience in BE and I think it is not good then period or sequence factor influences PK-parameter.

thanks!

Edit: Category changed; see also this post #1. Tabulators changed to spaces and BBcoded; see also this post #6. [Helmut]

Helmut
★★★

Vienna, Austria,
2020-12-22 17:27
(899 d 01:31 ago)

@ Moloko
Posting: # 22154
Views: 1,546

Significant effects: Don’t worry, be happy!

Post reply

Hi Moloko,

❝ Could somebody help me with my doubts about significant sequence/treatment/period

❝ effects in BE study?

Search the forum for sequence effect, treatment effect, and period effect. Lots of posts, references, etc. See also the articles about carryover and a significant treatment effect.

❝ My СI-s are within the range of 80.000% - 125.000%, so formally BE acceptance criterion is met.

Excellent. Open a bottle of champagne. :party:

❝ But these p-values scare me... :confused:

❝

❝ I don't have much experience in BE and I think it is not good then period or sequence factor influences PK-parameter.

Unless the the study was extremely imbalanced (say, almost all subjects in one sequence and the rest in the other), period effects mean out, since both T and R are affected to the same extent. Nothing to worry about.
Sequence effects (actually unequal carry-over) cannot be “corrected” by a statistical method – only avoided by design (sufficiently long washout). It was shown in a large meta-analysis that in properly designed studies a significant effect is seen in ~ the level of the test and hence, likely is a statistical artifact (false positive). Forget it.
Even a significant treatment effect is not important. Only the CI-inclusion counts.
The fact that subjects differ is trivial. Only if you don’t see a significant effect you should start to worry (monozygotic twins or triplets in the study?) because it would violate the assumption of independence.

See also here for the sequence effect and there for the treatment effect. The period effect cancels out in a crossover study anyway.

—
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

Moloko ☆ Russia, 2020-12-23 14:24 (898 d 04:34 ago) @ Helmut Posting: # 22158 Views: 1,461	Significant effects: Don’t worry, be happy! Post reply
	Helmut, thank you very much for response! It seems like I couldn't find the correct keywords when searching for posts earlier.. So now I can immerse myself in reading

Significant effects [General Sta­tis­tics]

Significant effects: Don’t worry, be happy!

Significant effects: Don’t worry, be happy!

Significant effects [General Statistics]