jag009
★★★

NJ,
2023-12-06 05:19
(88 d 22:16 ago)

Posting: # 23780
Views: 639
 

 Half life longer than literature (or well known) half-life [Study As­sess­ment]

Hi all,

I have a question. What if my NCA calculated half-life of a drug is significantly longer than the literature reported half-life which is well established (and I have old study data to back up the literature half-life as well). The newly calculated half-life is longer because I sampled longer than usual and the bioanalytical method has a much lower BLQ. Note that the elimination phase also has a biphasic appearance. Since the data is the data, I assume there is nothing that I would need to do other than writing something to discuss why the my calculated half-life is longer?

FYI, i just let Phoenix to pick the timepoints itself to calculate the half-life. Of course I can override and select the timepoint as well but then the R2 might not be optimal.

Thx

Jag
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2023-12-06 11:41
(88 d 15:54 ago)

@ jag009
Posting: # 23782
Views: 589
 

 Explain it in the report

Hi John,

❝ What if my NCA calculated half-life of a drug is significantly longer than the literature reported half-life which is well established (and I have old study data to back up the literature half-life as well). The newly calculated half-life is longer because I sampled longer than usual and the bioanalytical method has a much lower BLQ. Note that the elimination phase also has a biphasic appearance.


See this case (slides 16–21). A very old and ‘popular’ drug.

❝ Since the data is the data, I assume there is nothing that I would need to do other than writing something to discuss why the my calculated half-life is longer?


Correct. No worries.

❝ FYI, i just let Phoenix to pick the timepoints itself to calculate the half-life. Of course I can override and select the timepoint as well but then the R2 might not be optimal.


Define “optimal”. :-D
\(\small{R_{\text{adj}}^2}\) is just a tool and a “greedy” algorithm, since it might include too early time points. Further, it regularly fails with controlled release products (flat profiles) and multiphasic release products. Visual inspection of the fits is mandatory. In Phoenix I start with the ‘Best Fit’ algo and subsequently correct questionable start/end times. Yes, it is subjective but better than trusting in Artificial Unintelligence.
See also there and scroll down a bit. It gives also some references.

Heck, writing an article about estimating \(\small{\widehat{\mit{\lambda}}_\text{z}}\) is on my todo-list for years…

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
jag009
★★★

NJ,
2023-12-06 16:46
(88 d 10:49 ago)

@ Helmut
Posting: # 23783
Views: 502
 

 Explain it in the report

Hi Helmut,

❝ See this case (slides 16–21). A very old and ‘popular’ drug.


Thanks! The scenario on page 20 is exactly what I am facing but the difference in t1/2 is much larger than yours (i.e. 2.5 hrs vs 4.5 hrs). It's an IR product and the drug is very old. Previous studies (from archive or literature) had a profile with a monophasic elimination phase due to sampling time and BLQ. Since I sampled longer and has a much much lower BLQ, the biphasic appearance popped up. I double checked the bioanalytical method and I didn't see any issues (calibration curves looked great, intraday and interday CVs looked great, no other issues).

❝ ❝ FYI, i just let Phoenix to pick the timepoints itself to calculate the half-life. Of course I can override and select the timepoint as well but then the R2 might not be optimal.


❝ Define “optimal”. :-D

❝ \(\small{R_{\text{adj}}^2}\) is just a tool and a “greedy” algorithm, since it might include too early time points. Further, it regularly fails with controlled release products (flat profiles) and multiphasic release products. Visual inspection of the fits is mandatory. In Phoenix I start with the ‘Best Fit’ algo and subsequently correct questionable start/end times. Yes, it is subjective but better than trusting in Artificial Unintelligence.


Maybe R2 from PHX was like .99xxx and my R2 was like .7? ;-) PHX only picked timepoints along the "2nd phase" of the curve while I picked data from more timepoints from both the 1nd phase and the 2nd phase of the elimination phase.

thx
John
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2023-12-06 17:30
(88 d 10:05 ago)

@ jag009
Posting: # 23784
Views: 526
 

 Don’t worry, be happy!

Hi John,

❝ ❝ See this case (slides 16–21). A very old and ‘popular’ drug.

❝ Thanks! The scenario on page 20 is exactly what I am facing but the difference in t1/2 is much larger than yours (i.e. 2.5 hrs vs 4.5 hrs).


So? My drug had a half life of 2–3 hours acc. to the literature.
With the proper estimation I got 4.63 ± 1.07 h (test) and 5.59 ± 1.19 h (reference).
Which half lives were given in your case and which ones did you estimate?

❝ It's an IR product and the drug is very old.


Paracetamol?

❝ ❝ ❝ FYI, i just let Phoenix to pick the timepoints itself to calculate the half-life. Of course I can override and select the timepoint as well but then the R2 might not be optimal.

❝ ❝ Define “optimal”. :-D

❝ Maybe R2 from PHX was like .99xxx and my R2 was like .7? ;-) PHX only picked timepoints along the "2nd phase" of the curve …


Makes sense.

❝ … while I picked data from more timepoints from both the 1nd phase and the 2nd phase of the elimination phase.


Don’t. Then you are doing sumfink similar to the example in my slide 17. Even if not that extreme, you must not “mix” phases. It might well be that the second phase is not relevant. That’s why Harold Boxenbaum developed the concept of effective half life.
In a nutshell: Fit a multi-compartment model with hybrid constants (not volumes of distribution, and rate constants or clearances). $$C(t)=A_1\exp(\alpha_1)+A_2\exp(\alpha_2)+\ldots+A_n\exp(\alpha_n).$$ Then $$AUC_{0-\infty}=\frac{A_1}{\alpha_1}+\frac{A_2}{\alpha_2}+\ldots+\frac{A_n}{\alpha_n}.$$ Assess the summands as fractions of \(\small{AUC_{0-\infty}}\) and rank them in descending order. This gives you an idea, which one will be relevant in clinical practice, i.e., when you go to steady state. In many cases the slow phase(s) are not relevant at all.
In my example, the first phase accounted for 98.3% of the \(\small{AUC_{0-\infty}}\) and the second only 1.7%.

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
jag009
★★★

NJ,
2023-12-07 17:36
(87 d 09:59 ago)

@ Helmut
Posting: # 23788
Views: 484
 

 Don’t worry, be happy!

Hi Helmut,

❝ So? My drug had a half life of 2–3 hours acc. to the literature.

❝ With the proper estimation I got 4.63 ± 1.07 h (test) and 5.59 ± 1.19 h (reference).

❝ Which half lives were given in your case and which ones did you estimate?


Half-life from literature is around 2 hours. t1/2 from my previous studies were around 2-3 hrs (w BLQ much much higher than my current studies).

❝ ❝ It's an IR product and the drug is very old.


❝ Paracetamol?


Shhhh... :-)

❝ ❝ … while I picked data from more timepoints from both the 1nd phase and the 2nd phase of the elimination phase.


❝ Don’t. Then you are doing sumfink similar to the example in my slide 17. Even if not that extreme, you must not “mix” phases. It might well be that the second phase is not relevant. That’s why Harold Boxenbaum developed the concept of effective half life.


Understood. I was just goofing around to try and see what would happen to R2 if I just selected more timepoints.

Thx
J
UA Flag
Activity
 Admin contact
22,919 posts in 4,807 threads, 1,643 registered users;
28 visitors (0 registered, 28 guests [including 8 identified bots]).
Forum time: 03:35 CET (Europe/Vienna)

The analysis of variance is not a mathematical theorem,
but rather a convenient method of arranging the arithmetic.    R.A. Fisher

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