Bioequivalence and Bioavailability Forum

Dear Ioanam!

❝ In almost all cases, the PK software auto-detects the best interval to calculate terminal half – life.

Whatever the algorithm defines as ‘the best interval’. Have a look at your software’s manual.

❝ But there are situations when the user may define this interval manually.

Yes.

❝ Can you explain me when the user can intercede in the software and which points must be selected for the optimum calculation of t half?

Tricky. First see a discussion we had in this rather lenghty thread. Two quotes from Hauschke et al. (2007)^[1] and Scheerans et al. (2008)^[2]:

1. The selection of the most suitable time interval cannot be left to a programmed algorithm based on mathematical criteria, but necessitates scientific judgment by both the clinical pharmacokineticist and the person who determined the concentrations and knows about their reliability.
   
2. It should be emphasised that the TTT method has been introduced in this paper to provide a reasonable tool to support visual curve inspection for reliably identifying the mono-exponential terminal phase. Moreover, the TTT method should not be utilised without visual inspection of the respective concentration-time course. Thus, before using this new approach the monophasic shape post the peak of the curve has to be checked visually by means of a semilogarithmic diagram.

The automatic method in Kinetica (up to the current release v5) selects data points based on the adjusted R². The same method is used by WinNonlin up to version 5.2.1. Phoenix/WinNonlin 6+ uses the same method, with the exception that the data point at C_max/t_max is not included by default. See also this thread about the implementation of different methods in bear for . The current version 2.4.0 supports following automatic methods:

1. Last 3 data points
   
2. Adjusted R² (ARS) excluding C_max/t_max
   
3. Akaike Information Criterion (AIC)
   
4. Two-Times-Tmax (TTT)
   
5. TTT and ARS
   
6. TTT and AIC

In any (!) case visual inspection of the fit is a must. For a multi-segment profile the intersection of the slowest part with the preceeding faster one is most likely a good starting point. Knowing the limitations of any algorithm you should include a statement about visual inspection of fits at least in an SOP, or – better – in the protocol.
I’m working with two members of the forum on a large simulation study comparing automatic methods – but we are not finished yet.

IMHO, automatic methods need ‘close supervision’ if:

• Data is noisy,
  
• Absorption and elimination rates are less than ~5 apart (i.e., \(\small{k_\text{a}\leq5\times k_\text{e}}\)); especially nasty are ‘flat’ profiles in flip-flop PK (\(\small{k_\text{a}=k_\text{e}}\)),
  
• Multiphasic release formulations.