Log in
Register|
rana ☆ India, 2012-02-13 08:51 (5235 d 15:36 ago) Posting: # 8114 Views: 7,386 |
|
|
Why only 90% Confidence Interval is being used in generally to prove bioequivalence..why can't 80% or 70% ??? What is rationale behind this? Edit: Category changed. Please follow the Forum’s Policy. [Helmut] |
|
ElMaestro ★★★ Denmark, 2012-02-13 09:07 (5235 d 15:21 ago) @ rana Posting: # 8115 Views: 6,579 |
|
|
Hello Rana, ❝ Why only 90% Confidence Interval is being used in generally to prove bioequivalence..why can't 80% or 70% ??? What is rationale behind this? Generally in statistics an alpha of 5% is considered appropriate. Whether this alpha level is appopriate or not is of course also a good question, but it has stuck for years and seems to be the tradition now. When that alpha is applied in the equivalence world it implies a 90% CI (1-2*alpha, two-sided). — Pass or fail! ElMaestro |
|
Helmut ★★★   Vienna, Austria, 2012-02-13 15:29 (5235 d 08:58 ago) @ ElMaestro Posting: # 8119 Views: 6,341 |
|
|
Hi ElMaestro! ❝ Generally in statistics an alpha of 5% is considered appropriate. Whether this alpha level is appopriate or not is of course also a good question, but it has stuck for years and seems to be the tradition now. <bean counting mode>Not in statistics but in parts of science/engineering where variability is so high that the risk of going for larger sample sizes is considered to give no further benefit. In industry 6σ is the buzzword. </bean counting mode>— 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 |
|
ElMaestro ★★★ Denmark, 2012-02-13 15:35 (5235 d 08:53 ago) @ Helmut Posting: # 8120 Views: 6,365 |
|
|
Hi HS you terrrible person, ❝ In industry 6σ is the buzzword. </bean counting mode>Ewwww, I am allergic to terms such as Lean, 6σ and CPK  — Pass or fail! ElMaestro |
|
Helmut ★★★ Vienna, Austria, 2012-02-13 15:42 (5235 d 08:46 ago) @ ElMaestro Posting: # 8121 Views: 6,375 |
|
|
❝ HS you terrrible person, Welcome!  ❝ Ewwww, I am allergic to terms such as Lean, 6σ and CPK Me too – but what’s wrong with Lean? Thanks for teaching me a new term. I guess you mean this one rather than that one.  — 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 |
|
ElMaestro ★★★ Denmark, 2012-02-13 16:15 (5235 d 08:13 ago) @ Helmut Posting: # 8122 Views: 6,450 |
|
|
Hi again, ❝ Me too – but what’s wrong with Lean? Ooooh... [rant_on] Lean is an interesting concept. Common sense, some would say. In practice however, the combination of Lean and vendor oversight is a nasty cocktail. Sometimes lean is practiced by corpoquacks who have a lot of insight into ways to minimise cost by cutting away the fat but the fat needs to be identified objectively. I am not explicitly saying these guys know spreadsheets better than the real world, but ... An area where fat is, or can be, quite subjective is supplier qualification. No FDA or EU guideline or law says specifically how frequently it has to be done or when and how etc. And this can be toxic, because it often leads to the conclusion that cost can be cut a lot then. I see this in several companies and I have seen the consequences such as studies that are suddenly not of submission quality. Usually discussions around need for supplier audits end up in hopeless argumentation like "But they were inspected by the Finnish authority in 1963", "My uncle knows the head of QC, and he is a really nice guy" or "Show me concrete proof that we need to allocate an audit here". In the wake of mishaps there will be a lot of finger-pointing because of the cost. Fortunately, FDA is now starting to be very interested in vendor oversight principles. I am not exactly sure if EU inspectorates go in that direction too for the time being. [/rant off] — Pass or fail! ElMaestro |
|
Helmut ★★★ Vienna, Austria, 2012-02-13 13:41 (5235 d 10:47 ago) @ rana Posting: # 8117 Views: 6,487 |
|
|
Dear Rana! ❝ Why only 90% Confidence Interval is being used in generally to prove bioequivalence. Let me reformulate ElMaestro’s answer (see also this presentation, slide 53). α (aka error type I or error of the first kind) is the patient’s risk of falsely accepting bioequivalence (hardcore-statistics: rejecting the null hypothesis of bioinequivalence). Therefore the risk for a particular patient to be treated with a non-BE formulation is 5% at the lower end of the acceptance range (<80%) and 5% at the upper end (>125%). In other words we accept that there are two groups of patients: 5% with expected BA of <80% and 5% with expected BA of >125%. If we look at the entire population of patients we get 5% (below) + 5% (above) = 10%. Since in a particular patient BA cannot be <80% and >125% at the same time, his/her risk it still 5% although we deal with a 90% (1–2×α) confidence interval. ❝ why can't 80% or 70% ??? 90% CI ⇒ 5% risk80% CI ⇒ 10% risk70% CI ⇒ 15% risk❝ What is rationale behind this? Convention; seems to work. Note that in Brazil for narrow therapeutic index drugs α has to be kept at 2.5% (i.e., 95% CI). This is unique because in many other regulations for NTIDs α is kept at 5% but the acceptance range is narrowed to 90–111% instead. — 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