2 replies to this topic

[Helmut Schütz]

Dear all,

I set up workflows to calculate partial AUCs by different methods.

• Cut at the individual t_max.
  Might be awful!
• Cut at a pre-specified time.
  Recommended by the FDA (‘old’ methlylphenidate guidance, zolpidem) and the EMA (biphasic products).
• Cut at the reference treatment’s median t_max.
  The FDA’s “early exposure” for IR products. The 2^nd pAUC is not required.
• Cut at the individual reference treatment’s t_max.
  Health Canada’s “early exposure”. The 2^nd pAUC is not required.
• Cut at half of the intended dosing interval tau.
  EMA MR guideline (if both pAUCs pass BE, the otherwise mandatory MD study can be waived).
  Note that the end-time of the 2^nd partial AUC is tau – and not t_last as in the other methods. If you don’t have samples at exactly tau in all cases (e.g., missing samples, BQLs, deviations from the sampling scheduled at tau) you need a reliable estimate of λ_z in order to perform the required inter- or extrapolation.

Example with three subjects: t_max of T (2, 3, 2) and t_max of R (3, 2, 3).

#1: Cut-off times for T (2, 3, 2) and for R (3, 2, 3).

#3: Cut-off-time 3 for all subjects irrespective of treatment.

#4: Cut-off times (3, 2, 3) for all subjects irrespective of treatment.

The attached project is a test case:
Bimodal formulations, unbalanced sequences, and a lot of deviations from scheduled sampling times. Since PHX/WNL interpolates lin/log between decreasing concentrations (regardless which trapezoidal method was selected) I suggest to use the lin-up/log-down trapezoidal method for consistency. BTW, in the study we used #2 (4 hours).

Limitations:

• 2×2×2 crossover.
• Treatments have to be coded T and R.
• Only one cut-off time (hence, not suitable for the current FDA guidance for methylphendidate).

Enjoy!

Attached Files

•  pAUC example.phxproj   6.29MB   635 downloads

Edited by Helmut Schütz, 12 November 2016 - 02:30 PM.

[mittyright]

Hi Helmut!

 

Thx for posting this!

Could you please clarify what do you mean by this:

Since PHX/WNL interpolates lin/log between decreasing concentrations (regardless which trapezoidal method was selected)

 

 

If Linear Trapezoidal Linear Interpolation method is selected, then PHX/WNL is using linear interpolation between 2 measurable decreasing concentrations:

 

Computing partial area from 0.000000 to 1.000000:

piece from (0.000000, 200.000000) to (1.000000, 150.000000) = 175.000000

Partial area from 0.000000 to 1.000000 = 175.000000

 

BR,

Mittyright

[Helmut Schütz]

Hi Mittyright!

 

Could you please clarify what do you mean by this:

If Linear Trapezoidal Linear Interpolation method is selected, then PHX/WNL is using linear interpolation between 2 measurable decreasing concentrations

 

THX, I stand corrected! I should have RTFM.

 

I mixed sumfink up. Only (!) if you ask for an end time which is beyond the last time point, regardless the method selected and if you have an estimate of λ_z, the extrapolation is always done lin/log. Makes sense.