Log in
Register|
sciguy ☆ Canada, 2012-03-23 05:23 (4790 d 02:47 ago) Posting: # 8320 Views: 14,767 |
|
|
Hi all, I just had a quick question for you all - quite basic but I'm forgetting things in my old age  For AUC0-72, the actual time at 72 hrs should be used to calculate AUC0-72 and not the nominal 72 hrs? Same goes for all truncated AUCs calculated? Many thanks, sciguy Edit: Category and subject line changed. [Helmut] |
|
Dr_Dan ★★ Germany, 2012-03-23 11:25 (4789 d 20:45 ago) @ sciguy Posting: # 8321 Views: 13,500 |
|
|
Hi sciguy According to CPMP/QWP/EWP/1401/98 Rev. 1 "4.1.5 Characteristics to be investigated Pharmacokinetic parameters Actual time of sampling should be used in the estimation of the pharmacokinetic parameters." This holds true for all sampling points. Kind regards Dan — Kind regards and have a nice day Dr_Dan |
|
d_labes ★★★ Berlin, Germany, 2012-03-27 11:22 (4785 d 21:48 ago) @ Dr_Dan Posting: # 8337 Views: 13,782 |
|
|
Dear Dan, dear sciguy! The question and the answer are little bit ambiguous to me. Do you refer to how to act in case of some time deviations at 72h? Use AUC up to the actual time (f.i. AUC0-73.5h) as your measure for AUC0-72h? My habit (and I know others here in the forum  ) was up to now to do an inter-/extrapolation to the scheduled time 72h and then calculate AUC0-72h with that inter-/extrapolated concentration.This also is in accordance with Dan's excerpt of the EMA BE guidance, I think. Of course this will usually not make a great difference if the time devs are small and the concentrations at 72h also. The 'missing 72h' is inherently quite another cup of tea (see f.i. here). But for other 'truncated' AUC also 'small' time deviations may be of great influence. Imagine early exposure partial AUC's, f.i. AUC0-30min. — Regards, Detlew |
|
sciguy ☆ Canada, 2012-04-02 22:44 (4779 d 10:26 ago) @ d_labes Posting: # 8373 Views: 13,303 |
|
|
Hello Dr.Dan and d_labes! ❝ Do you refer to how to act in case of some time deviations at 72h? ❝ Use AUC up to the actual time (f.i. AUC0-73.5h) as your measure for AUC0-72h? Yes, this is what I meant. I couldn't remember if AUC0-72 really means AUC0-72 or AUC0-72.xx (whatever your actual time at 72h is) ❝ My habit (and I know others here in the forum I see...thanks for the link. I set up partial areas in WNL to define the actual time rather than inter/extrapolate at 72. Is it clear from a regulatory perspective (ie: FDA/EMA) that inter/extrapolation is required? Obviously, if time deviations are large, it makes sense to do this since there would be a major difference between AUC0-72 vs say, AUC0-76.3. Thanks, sciguy |
|
Helmut ★★★   Vienna, Austria, 2012-04-03 02:33 (4779 d 06:37 ago) @ sciguy Posting: # 8374 Views: 13,399 |
|
|
Dear sciguy! ❝ […] I set up partial areas in WNL to define the actual time rather than inter/extrapolate at 72. That’s a comfortable way of doing exactly the same. Be sure to set the method of calculation to lin-up/log-down.❝ Is it clear from a regulatory perspective (ie: FDA/EMA) that inter/extrapolation is required? If you want to get a statement which you can use as a reference literally – no, there is none. One the other hand I had never any problems with inter-/extrapolations – if specified in great detail in the protocol. ❝ Obviously, if time deviations are large, it makes sense to do this since there would be a major difference between AUC0-72 vs say, AUC0-76.3. Correct. — 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 |
|
sciguy ☆ Canada, 2012-04-03 17:50 (4778 d 15:21 ago) @ Helmut Posting: # 8377 Views: 13,358 |
|
|
Hi Helmut! ❝ ❝ […] I set up partial areas in WNL to define the actual time rather than inter/extrapolate at 72. ❝ That’s a comfortable way of doing exactly the same. I think setting your pAUC to 72 when your actual time is 72.6 would be the same as interpolating. But I set my pAUC to 72.6 which would just be an AUC estimate at actual time rather than interpolation. Sorry, just wanted to clarify.. -sciguy |
|
Helmut ★★★ Vienna, Austria, 2012-04-03 17:56 (4778 d 15:14 ago) @ sciguy Posting: # 8378 Views: 13,423 |
|
|
Dear sciguy! ❝ I think setting your pAUC to 72 when your actual time is 72.6 would be the same as interpolating. Yes. ❝ But I set my pAUC to 72.6 which would just be an AUC estimate at actual time rather than interpolation. Sorry, just wanted to clarify.. You mean that your nominal time was 72 h and in all subjects in all periods the actual time was 72.6 h? — Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
sciguy ☆ Canada, 2012-04-03 18:10 (4778 d 15:00 ago) @ Helmut Posting: # 8379 Views: 13,380 |
|
|
Dear Helmut! ❝ ❝ But I set my pAUC to 72.6 which would just be an AUC estimate at actual time rather than interpolation. Sorry, just wanted to clarify.. ❝ You mean that your nominal time was 72 h and in all subjects in all periods the actual time was 72.6 h? Right, nominal time = 72 but different subjects had different actual times. I should point out that this is a single period exploratory Phase 1 study, not a BE study. |
|
Helmut ★★★ Vienna, Austria, 2012-04-03 18:16 (4778 d 14:55 ago) @ sciguy Posting: # 8380 Views: 13,473 |
|
|
Dear sciguy! ❝ ❝ You mean that your nominal time was 72 h and in all subjects in all periods the actual time was 72.6 h? ❝ ❝ Right, nominal time = 72 but different subjects had different actual times. Now we are going in circles.  Was 72.6 h the “latest” sampling in any of the subjects? If yes and you set pAUC in PHX/WNL to 72.6 h all AUCs of subjects with tlast<72.6 will be extrapolated. I would rather set the limit for pAUC to 72 h.— Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
sciguy ☆ Canada, 2012-04-05 18:27 (4776 d 14:44 ago) @ Helmut Posting: # 8387 Views: 13,374 |
|
|
Dear Helmut! ❝ Now we are going in circles. Yes, 72.6 was the latest sampling time observed. For a BE study, I understand the rationale of setting the pAUC limit to 72hr. But....for an exploratory PK study (where there is no equivalence assessment), should we still set this limit (ie: is there any reason to just use each subject's actual time at 72hr?) My guess is to just stick with the 72 hr limit.. Thanks! sciguy |
|
Helmut ★★★ Vienna, Austria, 2012-04-05 20:25 (4776 d 12:45 ago) @ sciguy Posting: # 8391 Views: 13,340 |
|
|
Dear sciguy! ❝ Yes, 72.6 was the latest sampling time observed. For a BE study, I understand the rationale of setting the pAUC limit to 72hr. But....for an exploratory PK study (where there is no equivalence assessment), should we still set this limit (ie: is there any reason to just use each subject's actual time at 72hr?) My guess is to just stick with the 72 hr limit.. In a PK study maybe you are interested in modeling as well? Just use the actual time points. Even if you stick with NCA, more emphasis in this type of study lies on AUC∞ anyway (prediction of steady state based on superposition etc.). For the reasoning of the truncation time point of 72 hours see my final remark in this post. In real life we rarely see negative time deviations (i.e., sample taken too early); might only be relevant if the last sample is missing. Generally the sample is taken either at the nominal time point or is delayed. If you have taken the latest sample in the study at 72.6 hours it would mean that you have to extrapolate all the others – a procedure I (!) would not prefer. — Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
SDavis ★★ UK, 2012-04-17 13:13 (4764 d 19:58 ago) @ sciguy Posting: # 8426 Views: 13,331 |
|
|
Hi I think I concur with Helmut. To summarise; Actual sampling times are the preferred input since deviations are then accounted for already. Then you are requesting partial AUC to nominal 72 h to have the EQUIVALENT time period for all subjects (as requested by the protocol). That's how I've always analysed my data over the last 15+years. Slightly more tricky is when protocol requests 0-72 and some samples were not quantifiable at that timepoint, again if you have a value for Lz this will be extrapolated for you but for profiles with no Lz then WNL will do it's best to extrapolate, effectively using AUCall (or a fraction thereof) with all the inherent assumptions and approximations of that metric. On these occasions I would be tempted to consider reporting an AUC to last COMMON timepoint e.g. 48. Again AUCinf(pred) would probably be a preferred comparison parameter, where available. Simon — Simon Senior Scientific Trainer, Certara™ [link=https://www.youtube.com/watch?v=xX-yCO5Rzag[/link] https://www.certarauniversity.com/dashboard https://support.certara.com/forums/ |
|
ElMaestro ★★★ Denmark, 2012-04-17 13:25 (4764 d 19:45 ago) @ SDavis Posting: # 8427 Views: 13,154 |
|
|
Hi Simon, ❝ That's how I've always analysed my data over the last 15+years. Slightly more tricky is when protocol requests 0-72 and some samples were not quantifiable at that timepoint, again if you have a value for Lz this will be extrapolated for you but for profiles with no Lz then WNL will do it's best to extrapolate, effectively using AUCall (or a fraction thereof) with all the inherent assumptions and approximations of that metric. On these occasions I would be tempted to consider reporting an AUC to last COMMON timepoint e.g. 48. Sounds very interesting. Could you explain what this means in practice? Can you give an example? Many thanks. — Pass or fail! ElMaestro |
|
SDavis ★★ UK, 2012-04-17 15:35 (4764 d 17:35 ago) @ ElMaestro Posting: # 8430 Views: 14,379 |
|
|
Hi - I just warning people of the simple assumptions that AUCall makes as illustrated in this image http://www.pharmpk.com/pk/AUCinf.jpg ![[image]](img/uploaded/image25.jpg) Basically draws a straight line to the next SAMPLED point from Tlast. SO if we look at the 3 profiles below, Subject 1 and 2 since they are either sampled to, or beyond 24 h, but subject 3 is not quantifiable after 12h. The redline I drew on here (12-24h) shoes how AUCall would be described and you can see that it could be quite different to an AUC24 that have been calculated by extrapolating. ![[image]](img/uploaded/image26.jpg) Note in the NCA results AUClast=AUCall for Subjects 1 & 2 where last sampled point was quantifiable. ![[image]](img/uploaded/image27.jpg) Simon. — Simon Senior Scientific Trainer, Certara™ [link=https://www.youtube.com/watch?v=xX-yCO5Rzag[/link] https://www.certarauniversity.com/dashboard https://support.certara.com/forums/ |
|
ElMaestro ★★★ Denmark, 2012-04-17 15:42 (4764 d 17:29 ago) @ SDavis Posting: # 8431 Views: 13,135 |
|
|
Hi Simon, I am not entirely sure I understand. This in effect means that for such 'problematic' subjects a value is actually imputed on basis of knowledge from the other subjects in the same study or period or sequence or getting the same treatment? — Pass or fail! ElMaestro |
|
SDavis ★★ UK, 2012-04-17 16:09 (4764 d 17:01 ago) @ ElMaestro Posting: # 8432 Views: 13,163 |
|
|
nope - I'm sorry to confuse you WNL NCA is treating each profile independently, I just overlaid them in a chart to show visibly how they may look different. The 'problematic' bit is just understanding that IF the last scheduled timepoint is not quantifiable, WNL has an 'extra' parameter that crudely extrapolates from actual Clast to the NEXT sampled timepoint. The conc at that sample point had been set to ZERO in the raw data, if I copy that same profile and instead left that BLQ as missing then you will see for that profile Subject 3miss, AUCall=AUClast ![[image]](img/uploaded/image29.jpg) The main point is use AUCall with caution, and as a consequence what I was trying to show in this case is that; in the case of Subject 3zero, partial areas can be calculated after 12 h when no Lz is available, BUT it relies on a very simple and quite possibly incorrect assumption that the conc will be zero by the next sample timepoint (as in fact the source data told it). Simon. PS I also put back in the original profile that was quantifiable at 24h just so you can see how this data looked before I started hacking about with it to so badly illustrate my point ;0( — Simon Senior Scientific Trainer, Certara™ [link=https://www.youtube.com/watch?v=xX-yCO5Rzag[/link] https://www.certarauniversity.com/dashboard https://support.certara.com/forums/ |
|
ElMaestro ★★★ Denmark, 2012-04-17 19:20 (4764 d 13:51 ago) @ SDavis Posting: # 8435 Views: 13,082 |
|
|
Ah OK Simon, ❝ The 'problematic' bit is just understanding that IF the last scheduled timepoint is not quantifiable, WNL has an 'extra' parameter that crudely extrapolates from actual Clast to the NEXT sampled timepoint. The conc at that sample point had been set to ZERO in the raw data (...) This approach probably leads to a systematic negatively biased AUCinf. I would say this could be considered a kind if imputation although a somewhat naïve one. — Pass or fail! ElMaestro |
|
d_labes ★★★ Berlin, Germany, 2012-04-17 17:45 (4764 d 15:25 ago) @ ElMaestro Posting: # 8433 Views: 13,055 |
|
|
Hi ElMaestro, hi Simon, have a look at this thread. Eventually it helps to have an explanation from a third point of view .— Regards, Detlew |
|
SDavis ★★ UK, 2012-04-17 18:29 (4764 d 14:41 ago) @ d_labes Posting: # 8434 Views: 13,200 |
|
|
Agreed, My original suggestion was essentially for the 'method 2' described in posting Posting: # 1337, when you have some profiles it was not possible/reasonable to extrapolate to AUCinf AND your Tlast is varying across subjects. AUC to last common timepoint gives as useful secondary parameter when you suffer this unfortunate combination of (relatively) high LoQ and sub-optimal sampling schedules. Simon (OFF-topic if you decide to model this with NLME then the Phoenix model uses the probability that the censored observation is below the quantification limit in the likelihood function when BLQ box is checked ![[image]](img/uploaded/image30.jpg) )— Simon Senior Scientific Trainer, Certara™ [link=https://www.youtube.com/watch?v=xX-yCO5Rzag[/link] https://www.certarauniversity.com/dashboard https://support.certara.com/forums/ |
Ing. Helmut Schütz
