Log in
Register|
samar sameh ☆ Jordan, 2017-10-04 15:58 (2727 d 00:06 ago) Posting: # 17854 Views: 10,527 |
|
|
Dear Collegues : I have 2 questions regarding calibration range determination, and I will appreciate your response: 1. Is there a limit for Upper/Lower ratio for setting the range?. 2. Should we consider the Linear Model as the simplist regression method? Thanks in advance |
|
Shuanghe ★★ Spain, 2017-10-04 18:23 (2726 d 21:41 ago) @ samar sameh Posting: # 17856 Views: 9,330 |
|
|
Hi Samar, ❝ 1. Is there a limit for Upper/Lower ratio for setting the range?. As far as I know there's no such requirement, at least not in the guidelines that I know of. But just for my curiosity, why do you want to set limit for U/L ratio? Usually, upper limit should be determined based on information of expected Cmax of your product. Consider the variability of Cmax (do not use the reported mean Cmax from literature) when choosing the suitable upper limit. I would choose at least mean + 2sd, preferably mean+3sd, as upper limit. As for lower limit, minimum requirement is <5% of Cmax though personally I usually choose a bit lower than that. In this case, the Cmax I prefer is mean-3sd. Also, I'd check various sources for a better idea of what the "expected maximum concentration" will be. For the same strength and formulation, you might be surprised to find very different Cmax in different sources. ❝ 2. Should we consider the Linear Model as the simplist regression method? Again, I don't think there's guidance requiring only linear model, though the simplest method is preferable. I vaguely recall there's certain guideline mentioned that if the simpler model is ok, then you should not use more complicated model. If you cannot find linear model in the minimum range as mentioned above, I think other model is OK. Let's see if other members have different opinion. — All the best, Shuanghe |
|
nobody nothing 2017-10-04 21:10 (2726 d 18:54 ago) @ Shuanghe Posting: # 17857 Views: 9,390 |
|
|
Range for calibration is given by the maximum concentrations you expect (upper limit) and the reasonably achievable LLOQ, determined by the LLOD as well as the clinical need to determine such low concentrations. So it depends on your study design (dose, population etc.) and the goal you want to achieve (preclinical, pilot, screening, or pivotal study). The larger the range you cover, the more likely you will NOT find your data set to be homoscedastic (i.e. equal variance over the concentration range), which is a necessary prerequisite for simple linear regression to be used. Normally you will find a rather constant CV% over a large part of your calibration range. Here you use weighted linear regression. This will massively improve accuracy of your calculated concentrations at the lower end of your calibration curve. Non-linear methods are not regularly necessary for typical instrumental analytics. These come more into play with stuff like ELISA, radio-receptor assays and stuff like that. Hope that help. — Kindest regards, nobody |
|
Helmut ★★★   Vienna, Austria, 2017-10-04 23:19 (2726 d 16:45 ago) @ nobody Posting: # 17858 Views: 9,236 |
|
|
Hi nobody and everybody, ❝ Non-linear methods are not regularly necessary for typical instrumental analytics. Might be required in MS as well. A classic in fluorescence detection (when quenching sets in). See also this thread and esp. the paper by Almeida er al. — 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 |
|
jag009 ★★★ NJ, 2017-10-05 02:24 (2726 d 13:40 ago) @ samar sameh Posting: # 17859 Views: 9,221 |
|
|
Hi, ❝ 1. Is there a limit for Upper/Lower ratio for setting the range?. ❝ 2. Should we consider the Linear Model as the simplist regression method? John |
|
Helmut ★★★ Vienna, Austria, 2017-10-05 02:37 (2726 d 13:27 ago) @ jag009 Posting: # 17860 Views: 9,271 |
|
|
Hi John, ❝ ❝ 2. Should we consider the Linear Model as the simplist regression method? ❝ Yes.  — 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 |
|
nobody nothing 2017-10-05 09:34 (2726 d 06:30 ago) @ Helmut Posting: # 17861 Views: 9,318 |
|
|
Tu felix Austria... There is always a way, to make things more complicated, BUT...  — Kindest regards, nobody |
|
jag009 ★★★ NJ, 2017-10-06 17:51 (2724 d 22:13 ago) @ Helmut Posting: # 17868 Views: 9,199 |
|
|
Hi Helmut! ❝ ❝ Really? (scroll down) Of course it depends on what kind of data the plot shows and not just applying a straight line and be happy :) J P.S. I had dinner with Laszlo back in Sept in Toronto. He told me he will be attending the Biosimilar conference this week? |
|
Helmut ★★★ Vienna, Austria, 2017-10-06 22:51 (2724 d 17:14 ago) @ jag009 Posting: # 17870 Views: 9,195 |
|
|
Hi John, ❝ ❝ Really? (scroll down) ❝ Of course it depends on what kind of data the plot shows and not just applying a straight line and be happy :) SCNR. ❝ He told me he will be attending the Biosimilar conference this week? Correct. Had dinner with him yesterday. Entertaining as usual. — 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 |
Ing. Helmut Schütz