Log in
Register|
Cytokrom C ☆ Sweden, 2022-04-29 16:24 (27 d 04:02 ago) Posting: # 22942 Views: 368 |
|
|
Hi, I am planning a study where I want to include the time until a pre-defined plasma concentration threshold is reached as a PK parameter. Two options/methods for defining this parameter has been discussed: A) The simplest approach would be to use the first sampling time-point where the concentration exceeds the threshold. B) The other approach would be to use linear interpolation to find a more precise estimate of the time the concentration actually crosses the threshold. From a methodological point of view, I find option B somewhat more appealing, as this would give a more "correct" result. However, there are a few arguments making me lean more towards option A anyhow, including: 1. Option A requires no imputation/assumptions of plasma levels between samples. 2. Data based on actual sampling times will be analogous to Tmax and perhaps more recognizable/acceptable by the authorities. 3. I also plan to present the percentage of subjects exceeding the threshold value at each sampling timepoint. A definition according to option A would better correspond with this analysis. -Is there a "standard" approach? -Does anyone have any advice on the best approach or experience regarding regulatory acceptability of data? Thanks, /Martin |
|
ElMaestro ★★★ Denmark, 2022-04-29 17:49 (27 d 02:37 ago) @ Cytokrom C Posting: # 22943 Views: 293 |
|
|
Hi Martin, I have been using both approaches with success in deficiency letters. This is not the type if thing you are expected to do by any standard. Your choice may reflect your own preference, the challenge at hand and the particulars of your drug but you will most likely not find a guideline telling you what & how. — Pass or fail! ElMaestro |
|
Helmut ★★★   Vienna, Austria, 2022-04-29 19:03 (27 d 01:23 ago) @ Cytokrom C Posting: # 22944 Views: 297 |
|
|
Hi Martin, see this (lengthy) thread as a starter. Since I’m using the linear-up / log-down trapezoidal method (see this article why), I use the corresponding interpolation. In the meantime implemented both in commercial software as well as in numerous ![[image]](img/uploaded/Rlogo_15_12.svg) packages.— Dif-tor heh smusma 🖖 ![[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 |
|
Cytokrom C ☆ Sweden, 2022-05-01 22:44 (24 d 21:42 ago) @ Helmut Posting: # 22948 Views: 247 |
|
|
Thanks Helmut and ElMaestro. I will go for the interpolation then. Would you also use mean (SD), rather than median (min, max), for the descriptive stats then, given that this then becomes a continuous variable? |
|
Helmut ★★★ Vienna, Austria, 2022-05-02 00:05 (24 d 20:21 ago) @ Cytokrom C Posting: # 22949 Views: 240 |
|
|
Hi Martin, » Would you also use mean (SD), rather than median (min, max), for the descriptive stats then, given that this then becomes a continuous variable? Yes. The jury is out for years trying to figure out whether the distribution is normal (use \(\small{\bar{x},\,s_x}\)) or lognormal (use \(\small{\bar{x}_\textrm{geo},\,CV_\textrm{geo}}\)). See this thread for the Half-Value-Duration HVD (aka Peak-Occupancy Time 50) and the Plateau-time t75% (aka Peak-Occupancy Time 25). Like in you case both are also continuous in the time-domain. Answering an unasked question: For discrete variables (e.g., tmax) I would give the median / IQR – not the range. The latter is not informative. — Dif-tor heh smusma 🖖 Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
Ing. Helmut Schütz