bobmarlo
☆    

2017-01-15 18:22
(2629 d 03:53 ago)

Posting: # 16953
Views: 12,173
 

 WinNonlin: NCA of multiple dosing plasma PK profile [Software]

I want to do non-compartmental analysis on plasma PK data from two cycles of short infusions using WinNonlin. Unfortunately, I do not have the plasma concentrations from first cycle and the only available concentrations are from second cycle. Doses used in first and second cycle are same, however, the predose conc. for second cycle is 10. Concentration for first time point of second cycle being 10 (not 0 at 0 hr for second cycle) or >LLOQ although close to LLOQ (with Cmax of 200, last measured conc. of second cycle 10, achieving steady state), makes the baseline look like 10. I want to know if NCA can be done directly on this data set from second cycle? Or it may lead to error because of drug still being present in the system before second infusion? I am primarily concerned about parameters including terminal t1/2, AUClast & inf, CL, MRT, Vss? I believe I cannot subtract baseline value from all the samples as commonly done in case of endogenously produced molecules and therefore all the listed parameters except t1/2 may not accurately represent drug profile. What would be the best approach to analyze such data to get parameters from second cycle?
Thanks


Edit: Category and subject line changedchanged; see also this post #1 and #2. [Helmut]
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2017-01-16 14:15
(2628 d 08:00 ago)

@ bobmarlo
Posting: # 16955
Views: 11,554
 

 WinNonlin: steady state setup

Hi Bob,

❝ I want to do non-compartmental analysis on plasma PK data […] Doses used in first and second cycle are same, however, the predose conc. for second cycle is 10. Concentration for first time point of second cycle being 10 …


OK. By design WinNonlin can calculate NCA for single dose data and for data in steady state. According to the superposition principle (H. Dost, 1953) nothing else makes sense. :-D

❝ … (not 0 at 0 hr for second cycle) or >LLOQ although close to LLOQ (with Cmax of 200, last measured conc. of second cycle 10, achieving steady state), makes the baseline look like 10.


The definition when steady state is reached in practice is based on convention. Most people are happy with 5×t½ (96.88% of steady state) whilst others are more strict (7×t½ 98.44%, 10×t½ 99.38%). In your case we have 95.00%. Fine with me. Since the dosing interval was less than 5×t½ I leave it to you to judge whether that is sufficient. The concentration at the end of the second interval – which was identical to the first one – gives some confidence that no more relevant accumulation occurred.

❝ I want to know if NCA can be done directly on this data set from second cycle?


Yes.

❝ Or it may lead to error because of drug still being present in the system before second infusion?


That’s perfectly fine.

❝ I am primarily concerned about parameters including terminal t1/2, AUClast & inf, CL, MRT, Vss?


AUC0–∞ in (pseudo) steady state does not make sense. Either AUC0–∞ (single dose) or AUC0–τ (steady state). The relationship AUC0–∞ = AUC0–τ holds for linear PK (superposition principle).

❝ I believe I cannot subtract baseline value from all the samples as commonly done in case of endogenously produced molecules and therefore all the listed parameters except t1/2 may not accurately represent drug profile.


Correct.

Example: Five minutes infusion, dose 2× 100 mg, τ four hours. I simulated data in such a way that Cmax in the second profile is 200 µg/mL and Cτ is 10 µg/mL.

 t        C  
0.0000   0.00
0.0833 191.04
[image]0.2500 168.17
0.5000 138.90
0.7500 114.73
1.0000  94.76
1.2500  78.27
1.5000  64.64
2.0000  44.10
3.0000  20.52
4.0000   9.55
4.0833 200.00
4.2500 176.06
4.5000 145.42
4.7500 120.11
5.0000  99.21
5.2500  81.94
5.5000  67.68
6.0000  46.17
7.0000  21.49
8.0000  10.00


Note that the first Cmax is <200 µg/mL. That’s as expected since there is some accumulation. Also the second pre-dose concentration is only 9.55 µg/mL.
In WinNonlin you should define Dose Options | Type IV Infusion and Calculation Method Linear Up Log Down. For the second profile Setup | Dosing ☑ Use Internal Worksheet.

Dose │ Time of Dose │ Length of Infusion │ Tau
(mg) │              │                    │    
─────┼──────────────┼────────────────────┼────
 100 │            4 │             0.0833 │   4

It is important to give Tau 4. If you leave the cell empty (the default) WinNonlin will “assume” that the profile came from a single dose and steady state PK metrics will not be calculated.
You should get:

Cmax        200      µg/mL
Cmin          9.55   µg/mL
Clast        10      µg/mL
Cavg         64.286  µg/mL
Fluctuation 296.25   %
AUC_TAU     257.15   h*µg/mL
Vz          508.44   mL
Vss         508.98   mL
Acc_Index     1.0492

Note also that AUC_TAU is calculated for the specified τ (by inter-/extrapolation if the last sampling time point deviates from τ). That’s important since in such a case AUC_TAU  AUClast.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
BE-proff
●    

2017-01-17 08:54
(2627 d 13:20 ago)

@ Helmut
Posting: # 16960
Views: 10,967
 

 WinNonlin: steady state setup

Hi Helmut,

❝ Most people are happy with 5×t½ (96.88% of steady state) whilst others are more strict (7×t½ 98.44%, 10×t½ 99.38%


Did you get this information from literature or these figures are your own observations?

One more question - how would you check if steady state has been reached? :confused:


Edit: Standard quotes restored; see also this post #8. [Helmut]
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2017-01-18 03:22
(2626 d 18:52 ago)

@ BE-proff
Posting: # 16970
Views: 10,842
 

 Accumulation

Hi BE-proff,

❝ ❝ Most people are happy with 5×t½ (96.88% of steady state) whilst others are more strict (7×t½ 98.44%, 10×t½ 99.38%

❝ Did you get this information from literature or these figures are your own observations?


We discussed that already, right?

❝ One more question - how would you check if steady state has been reached? :confused:


There are some fancy methods (MANOVA, regression of the pre-dose concentrations and testing whether the slope ≠ 0, Hotelling’s T2; see this thread). None of them are really good. Luckily regulators realized that recently and are generally satisfied with:
  1. A justification of the dosing regimen in the study (e.g., EMA: saturation ≥5t½) based on the (average?) half-life reported in the literature.
    • AFAIK, only in Taiwan and China estimation of t½ in the study is required. Since – especially for MR products – this might be difficult within the dosing interval (flat profile) these poor guys have to sample beyond τ.
  2. A table and plots of ≥3 pre-dose concentrations in the saturation phase.
You can also use the quick & dirty method from below: % of steady state = 100×Cpre-dose/Cτ. Please, don’t ask me for a reference!

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
bobmarlo
☆    

2017-01-17 21:36
(2627 d 00:39 ago)

@ Helmut
Posting: # 16968
Views: 10,953
 

 WinNonlin: steady state setup

Dear Helmut,
Thank you for your response. Your simulation and explanation makes complete sense to me. However, I am still curious about calculation of CL and MRT. As CL=Dose/AUC and MRT=(AUMC/AUC) - (Duration of Infusion/2). These parameters are dependent on AUC & AUMC. And the problem with AUC & AUMC I see is that baseline being 10, the calculated AUC is sum of exposure due to second dose + exposure due to the baseline concentration of drug already present, which can contribute to as high as 10% of total AUC (200 at Cmax, 10 at 0h). In this situation do you think these parameters (CL and MRT) will still be good representative of Plasma PK for this drug? or there is way to analyze it better?
Thanks again to you and Mittyri for links.
Bob


Edit: Full quote removed. Please delete everything from the text of the original poster which is not necessary in understanding your answer; see also this post #5! [Helmut]
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2017-01-18 02:41
(2626 d 19:34 ago)

@ bobmarlo
Posting: # 16969
Views: 11,095
 

 NCA agrees with PK-modeling

Hi Bob,

❝ I am still curious about calculation of CL and MRT. As CL=Dose/AUC and MRT=(AUMC/AUC) - (Duration of Infusion/2). These parameters are dependent on AUC & AUMC.


Correct so far.

❝ And the problem with AUC & AUMC I see is that baseline being 10, the calculated AUC is sum of exposure due to second dose + exposure due to the baseline concentration of drug already present, which can contribute to as high as 10% of total AUC (200 at Cmax, 10 at 0h).


Maybe you were confused by the fact that you measured 10 both at the start and the end of the second infusion (sorry for drawing the red line in my first post). Due to the fact that you are close to the LLOQ that might be just a coincidence (random noise). Since this was not an endogenous compound, there is no „baseline”. You have residual concentrations of the first infusion which gradually are eliminated. In my simulation the residual concentration is 9.55 at four hours and just 0.45 at eight hours (note that 9.55+0.45=10). Below the model:

[image]


❝ In this situation do you think these parameters (CL and MRT) will still be good representative of Plasma PK for this drug?


NCA is fine as long as
  • you assume linear PK (superposition principle holds),
  • you assume that you are sufficiently close to steady state, and
  • you use my setup. ;-)

❝ or there is way to analyze it better?


There is no free lunch. Of course modeling might be an alternative. Depends on your experience and the purpose of the study. Let’s compare NCA (lin-up log-down trapezoidal) with PK models (one compartment, parameterized in CL & V, duration of infusions five minutes):

                 NCA      1       2       3       4  
VSS  mL        508.98  507.12  507.13  507.13  484.40
CL   ml/h      388.88  387.89  387.87  387.88  370.49
AUC  h×µg/mL   257.15  257.81  257.82  257.81  269.91
AUMC h²×µg/mL  294.97  337.05  337.08  337.07  352.90
MRT  h         1.3088  1.3074  1.3075  1.3074  1.3075

  1. The data of the first infusion (which you don’t have).
  2. The data of the second infusion; both doses given in the setup.
  3. The data of both infusions.
  4. The data of the second infusion but only the second dose given in the setup.
Given that NCA works with the log-trapezoidal method and there was no noise in the data I’m fine with the results (agree with models 1–3; max. relative error 0.36%).
t½ was 0.90624 h. That means we reached 100(1–½4/0.90624)=95.31% of steady state at the start of the second infusion. Should be sufficient. Quick & dirty: 100×C4/C8=95.5%.
Since we didn’t give the first dose (classical user error) in model 4, Phoenix does its best to fit the data but the results are wrong. Rubbish in, rubbish out.
BTW, it is one of the mysteries of Phoenix/WinNonlin why the NCA’s MRT is correct though the AUMC isn’t. My pocket calculator gives 294.97/257.15=1.1471.
:PCchaos:

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
bobmarlo
☆    

2017-01-18 06:06
(2626 d 16:09 ago)

@ Helmut
Posting: # 16971
Views: 10,737
 

 NCA agrees with PK-modeling

Dear Helmut,
Thank you very much for your explanation :-).
Warm Regards,
Bob
mittyri
★★  

Russia,
2017-01-18 19:00
(2626 d 03:15 ago)

(edited by mittyri on 2017-01-18 20:20)
@ Helmut
Posting: # 16975
Views: 10,753
 

 Winnonlin AUMC_TAU and others

Hi Helmut and all,

❝ BTW, it is one of the mysteries of Phoenix/WinNonlin why the NCA’s MRT is correct though the AUMC isn’t. My pocket calculator gives 294.97/257.15=1.1471.


the formula for MRTINF (using Cpred version) is the following:

MRTINF_pred = [image]

AUMC_TAU is a'cut-off' version like AUC_TAU.
BTW AUMC is calculated inside modeling like that:

AUMC = A1InfDose/(tvV*Ke*Ke) + (Tinf*A1InfDose/(2*tvV*Ke)

So the modeling version is AUMC extrapolated to infinity. If you remove Tau from the Dosing sheet in NCA object, you'll see AUMCINF_pred which is close enough to the AUMC value in PK model

Kind regards,
Mittyri
mittyri
★★  

Russia,
2017-01-16 18:29
(2628 d 03:46 ago)

@ bobmarlo
Posting: # 16957
Views: 11,191
 

 WinNonlin: NCA of multiple dosing plasma PK profile

Dear Bob,

❝ I am primarily concerned about parameters including terminal t1/2, AUClast & inf, CL, MRT, Vss? I believe I cannot subtract baseline value from all the samples as commonly done in case of endogenously produced molecules and therefore all the listed parameters except t1/2 may not accurately represent drug profile.


Some pages from my bookmarks marked as must read each time before MD study performed/analyzed:-D:
Statistics in multiple dose study
half-life in multiple-dosed steady-state BE/BA study?

Kind regards,
Mittyri
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
84 visitors (0 registered, 84 guests [including 11 identified bots]).
Forum time: 22:15 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5