## CVintra # Mean/SD [Power / Sample Size]

Dear Mathews!

Unfortunatelly there's no way to calculate CV

Also see this post.

With exception of Canada, where delta should be set to zero (ratio=1) and a correction for actual content should be made in the assessment of BE, delta should be set to the

The commonly applied approach is to set the expected ratio to 0.95 (or -5%). Rationale behind is that release specifications of a batch generally are ±5%; since this applies to both test and reference - and we assume an average deviation from declared content of ±2.5% we end up with ±5% for the study.

Sure. Different model --> different error distribution / power function / sample sizes.

It's a generally accepted assumption that many biological variables (including PK metrics like AUC, C

For t

❝ I know only the Cmax value of the drug from the literature. It is around 42.34 ± 13.27.

❝ How i can calculate the intra subject CV from this Cmax value? (is it required for Sample Size calculation?)

Unfortunatelly there's no way to calculate CV

_{intra}from CV_{total}(CV_{total}= CV_{intra}+ CV_{inter}). Actually you don't even have CV_{total}(in log scale), because CV = SD/mean (what you have) comes from untransformed data.Also see this post.

❝ which value of "delta" we commonly used for calculating Sample Size?

With exception of Canada, where delta should be set to zero (ratio=1) and a correction for actual content should be made in the assessment of BE, delta should be set to the

*expected difference*(*e.g.*, from a**reasonably sized**pilot study).The commonly applied approach is to set the expected ratio to 0.95 (or -5%). Rationale behind is that release specifications of a batch generally are ±5%; since this applies to both test and reference - and we assume an average deviation from declared content of ±2.5% we end up with ±5% for the study.

❝ there are different sample size formulas for additive model and multiplicative model?

Sure. Different model --> different error distribution / power function / sample sizes.

❝ which model we commonly used? why?

It's a generally accepted assumption that many biological variables (including PK metrics like AUC, C

_{max}) follow a lognormal distribution. Furthermore serial dilutions in analytics also lead to multiplicative errors. Since additivity of effects and homoscedasticity (homogeneity of variance) is a prerequisite for ANOVA, we apply a multiplicative model (*i.e.*, work with log-transformed data instead with raw data). By this the distribution skewed to the right becomes symetrical and values below zero are avoided.For t

_{max}an additive model is most suitable; although a nonparametric analysis is also a must (ANOVA is incorrect for discrete data).—

Helmut Schütz

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

Science Quotes

*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

### Complete thread:

- CVintra from Mean/SD mathews 2007-09-17 13:04 [Power / Sample Size]
- CVintra # Mean/SDHelmut 2007-09-17 15:59
- CVintra # Mean/SD mathews 2007-09-18 11:17
- CVintra # Mean/SD Helmut 2007-09-18 13:51
- Lansoprazole: CVintra? mathews 2007-09-24 11:11
- Lansoprazole: CVintra... Helmut 2007-09-25 01:12
- Lansoprazole: CVintra... mathews 2007-09-25 08:22

- Lansoprazole: CVintra... Helmut 2007-09-25 01:12

- Lansoprazole: CVintra? mathews 2007-09-24 11:11

- CVintra # Mean/SD Helmut 2007-09-18 13:51

- CVintra # Mean/SD mathews 2007-09-18 11:17

- CVintra # Mean/SDHelmut 2007-09-17 15:59