MD design based on SD data [Design Issues]
Hi BE-proff,
I guess you are talking BE and not PK (modeling)?
Nitpicking: PK parameters are estimated by a PK model. PK metrics are obtained by NCA/SHAM.
If linear PK is applicable, 5–7 times t1/2 (AFAIK, only ANVISA requires 10×t1/2 ~99.9% of steady state). Exact for the one-compartment model and approximate for >1 compartments. In theory steady state (output = input) is reached after infinite time. We don’t want to wait that long, right? Formula to estimate the percentage of steady state reached: 100(1–½i), where i is the i-th dose administered.
As said above all this is valid for linear PK only (check the literature). For nonlinear PK it might get extremely complicated (see this thread) and designing an MD study without knowing the kind of nonlinearity and establishing a PK model first might be almost impossible.
You need to have single dose data (preferably of your own study). Some people design the MD dose study already based on literature data of the IR formulation and black magic assumptions about the new MR formulation. Not a good idea!
Based on SD data you could use a method called Nonparametric Superposition (see this thread for software and this post for an example). Essentially at every dosing you stack the single dose profile on top of the expected concentrations of previous doses estimated by λz. If the software allows that you may explore different sampling schedules (i.e., interpolating time points not existing in the SD profiles). Remember that a precise estimate of Css,max is crucial. Since tss,max for IR and MR will be different, try to find a sampling schedule which will give you sufficient information about both.
Some answers, I do hope.
❝ If I need to compare kinetics of immediate-release tablet and modified-release medication how to plan a study?
I guess you are talking BE and not PK (modeling)?
❝ Which parameters are to be used?
Nitpicking: PK parameters are estimated by a PK model. PK metrics are obtained by NCA/SHAM.

- AUC0–τ (if there is circadian variability in PK and the drug is administered more than once additionally AUC0–24).
- Css,max (descriptively tss,max).
- For originators Css,min (the minimum within the dosage interval τ) and for generics Css,τ (the concentration at the end of τ).
- %PTF (the Peak-to-Trough Fluctuation in percent): 100(Cmax–Cmin)/Cav, where Cav is the average concentration within the dosage interval given by Cav = AUC0–τ/τ.
- In Russia T>75%Cmax: The time interval where the concentration is ≥75% of Cmax. In the literature also called “Plateau Time” (t75%) or “Peak Occupancy Time” (POT-25: the time period during which concentrations are within 25% of Cmax). It’s variability may be high.
You can try to convince regulators to assess the more stable “Half Value Duration” (HVD) aka POT-50 (time period during which concentrations are within 50% of Cmax) instead. Good luck.
❝ How to determine how many days and how often tablets are to be taken?
If linear PK is applicable, 5–7 times t1/2 (AFAIK, only ANVISA requires 10×t1/2 ~99.9% of steady state). Exact for the one-compartment model and approximate for >1 compartments. In theory steady state (output = input) is reached after infinite time. We don’t want to wait that long, right? Formula to estimate the percentage of steady state reached: 100(1–½i), where i is the i-th dose administered.
As said above all this is valid for linear PK only (check the literature). For nonlinear PK it might get extremely complicated (see this thread) and designing an MD study without knowing the kind of nonlinearity and establishing a PK model first might be almost impossible.
❝ How to determine what time point must be in PK-session?
You need to have single dose data (preferably of your own study). Some people design the MD dose study already based on literature data of the IR formulation and black magic assumptions about the new MR formulation. Not a good idea!
Based on SD data you could use a method called Nonparametric Superposition (see this thread for software and this post for an example). Essentially at every dosing you stack the single dose profile on top of the expected concentrations of previous doses estimated by λz. If the software allows that you may explore different sampling schedules (i.e., interpolating time points not existing in the SD profiles). Remember that a precise estimate of Css,max is crucial. Since tss,max for IR and MR will be different, try to find a sampling schedule which will give you sufficient information about both.
❝ A lot of questions...
Some answers, I do hope.
—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!![[image]](https://static.bebac.at/pics/Blue_and_yellow_ribbon_UA.png)
Helmut Schütz
![[image]](https://static.bebac.at/img/CC by.png)
The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
![[image]](https://static.bebac.at/pics/Blue_and_yellow_ribbon_UA.png)
Helmut Schütz
![[image]](https://static.bebac.at/img/CC by.png)
The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
Complete thread:
- How to plan steady state study? BE-proff 2016-03-13 18:28 [Design Issues]
- MD design based on SD dataHelmut 2016-03-15 16:39
- Something new from Russian regulators mittyri 2016-03-19 14:25