Log in
Register|
shri ☆ 2008-12-12 09:06 (6394 d 12:18 ago) (edited on 2008-12-12 10:19) Posting: # 2909 Views: 10,923 |
|
|
Dear all, What is the rationale behind using the "median" value for Tmax comparisons in place of the "geometric least square mean", which is performed on the rest of the PK parameters analyses? thanks in advance. -- Edit: Category changed. [Jaime] — shri |
|
Jaime_R ★★ Barcelona, 2008-12-12 11:28 (6394 d 09:56 ago) @ shri Posting: # 2911 Views: 9,187 |
|
|
Hi shri, using the search feature (see the Policy) of the forum would have given this post. Simple example: if you have Tmax at 1, 2, and 3 hours (three subjects) the median of 2 hours points to an actual sampling time point, whereas the geometric mean (1.817 h) does not. — Regards, Jaime |
|
Helmut ★★★   Vienna, Austria, 2008-12-12 20:31 (6394 d 00:53 ago) @ shri Posting: # 2920 Views: 9,461 |
|
|
Dear shri! ❝ [...] in place of the "geometric least square mean", which is performed on the rest of the PK parameters analyses? Not the ‘rest of the PK parameters’! There are many different descriptive statistics that can be chosen as a measurement of the central tendency of data (e.g., arithmetic, geometric, harmonic mean, median, mode,…) Based on assumptions of the underlying distribution, different averages are used: AUC: geometric meanAe: arithmetic meantmax: mediant½: harmonic mean (although geometric mean used by some colleagues)These distributional assumptions lead also to different statistical models if comparisons are performed: AUC: multiplicative model (or additive on log-transformed data)Ae: additive modeltmax: nonparametric methodst½: no idea (suggestions? randomization test?)— 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 |
|
vezz ☆ Erba (CO), Italy, 2009-12-11 16:59 (6030 d 04:25 ago) @ Helmut Posting: # 4473 Views: 8,870 |
|
|
Dear Helmut, I'm working on my very first study including PK parameters and I need to define the models to be used for the analysis. My concerns regard the urinary parameters: amount of drug excreted (Ae), maximum excretion rate (Rmax) and elimination half life (t½).
Thank you very much for your help. — Kind regards, Stefano |
|
Helmut ★★★ Vienna, Austria, 2009-12-11 18:02 (6030 d 03:22 ago) @ vezz Posting: # 4474 Views: 9,021 |
|
|
Dear Stefano! ❝ My concerns regard the urinary parameters: amount of drug excreted (Ae), maximum excretion rate (Rmax) and elimination half life (t½). ❝ • Ae. In your post you suggest to use an additive model. Yes. Personal views, as mostly. ❝ However, in the FDA guidance […] an analysis based on log-transformed data is requested I know. FDA also wants subjects to consume a diet of 2500-3500 Calories per day for the 16 days of hospitalization. That's a tough slimming cure!  I would expect volunteers to loose a lot of weight. Not suitable for a cross-over. That's about science in (some) guidelines.❝ From your experience, both these approaches are acceptable? Ae, Rmax: OK, even the last guideline-oriented regulator has heard about the holy grail of log-transformation by now. More science-oriented ones would accept a non-transformed analysis. To be honest my last urine study dates more than 15 years back. And, yes I didn't get complaints about any of them. My reasoning is: Plasma metrics are based on concentrations (ratios! The volume of distribution comes into play), but urine metrics are based on amounts (mass). ❝ • t½. The same considerations as for plasma concentration half life apply for this parameter? Which considerations do you mean? The concept of BE-testing relies on the assumption of time-invariant clearances. Unless we go for stable isotope simultaneous iv administration, we have no means to check this assumption. We simply ignore it. If we have an absorption faster (ka > 5kel) than elimination, the apparent elimination should be roughly the same for both formulations. But I've rarely seen anybody testing for equivalence of t½. An excpetion are PK-interaction studies, where a comparison is mandatory (concept of constant clearances may not hold). ❝ As regards renal clearance (CLr, based also on blood AUC) I think that, if we assume log-normality for Ae, the same distribution can be hypothesized for this parameter. Following your assumption, yes. ❝ But if we assume normality for Ae, how should CLr be analysed? Tricky.  Like D. Labes said: Next question, please.— 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 |
|
vezz ☆ Erba (CO), Italy, 2009-12-13 08:31 (6028 d 12:53 ago) @ Helmut Posting: # 4475 Views: 8,533 |
|
|
Dear Helmut, thank you for your help. ❝ ❝ • t½. The same considerations as for plasma concentration half life apply for this parameter? ❝ ❝ Which considerations do you mean? I was referring to your first post in this thread: "t½: harmonic mean (although geometric mean used by some colleagues)". — Kind regards, Stefano |
Ing. Helmut Schütz