Log in
Register|
kevan ☆ 2009-05-25 18:33 (6231 d 00:13 ago) Posting: # 3750 Views: 9,535 |
|
|
Hi, I faced a problem in calculation of CI of transformed and non-transformed data. The CI of transformed data was outside the accepted range (80-125) while the non-transformed data in the accepted range (80-120). Most of the guidelines focus on the transformed data. Any statistical method to show that the transformed data is not valid in this case? How can I make conclusion for the study? Hope someone can help. Thanks. Kevan |
|
Helmut ★★★   Vienna, Austria, 2009-05-25 19:01 (6230 d 23:44 ago) @ kevan Posting: # 3751 Views: 8,574 |
|
|
Dear Kevan! ❝ I faced a problem in calculation of CI of transformed and non-transformed data. The CI of transformed data was outside the accepted range (80-125) while the non-transformed data in the accepted range (80-120). ❝ ❝ Most of the guidelines focus on the transformed data. Right (at least for AUC, Cmax,...). The use of the log-transformation is supported by two assumptions:
❝ Any statistical method to show that the transformed data is not valid in this case? I hope you have stated a method in the protocol? It does not make sense to select between log-transformed and untransformed analyses. Give it a try. Run any normality test (Shapiro-Wilk, Anderson-Darling,…) on the intra-subject residuals of both the log-transformed and the untransformed data. If you don't have discordant outliers in the data set I bet that both tests will come up with a nonsignificant result. In other words, you can not select the distribution based on a statistical test given the small sample size in a BE study. Therefore you have to stick with the log-transformed analysis. The only exception I know comes from Japan (Q&A Document, November 2006, my private translation): Q-32. Is logarithmic transformation always necessary? Is it not acceptable to carry out logarithmic transformation only when required? ❝ How can I make conclusion for the study? Your study was not able to demonstrate BE. Sorry. — 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 |
|
kevan ☆ 2009-05-26 09:41 (6230 d 09:04 ago) @ Helmut Posting: # 3755 Views: 8,329 |
|
|
Hi HS, ❝ Run any normality test (Shapiro-Wilk, Anderson-Darling,…) on the intra-subject residuals of both the log-transformed and the untransformed data. If you don't have discordant outliers in the data set I bet that both tests will come up with a nonsignificant result. You means that i need to find each intra-subject residuals to run the normality test? How can I identify the outlier in the data set which consist the intra-subject residues? ❝ If it is not appropriate to analyse with transformed data, for example, when untransformed data shows normal distribution but transformed data shows non-normal distribution, you may state as such and use untransformed data for the assessment. Should I mention above statements in the protocol before start the study? Anyway, the study already finished, what can I do now is to mention this phenomenon in report and conclude the treatments is not bioequivalence? Thanks very much. Kevan |
|
Helmut ★★★ Vienna, Austria, 2009-05-26 14:29 (6230 d 04:16 ago) @ kevan Posting: # 3762 Views: 8,424 |
|
|
Dear Kevan! ❝ ❝ Run any normality test (Shapiro-Wilk, Anderson-Darling,…) on the intra-subject residuals of both the log-transformed and the untransformed data. If you don't have discordant outliers in the data set I bet that both tests will come up with a nonsignificant result. ❝ You means that i need to find each intra-subject residuals to run the normality test? ❝ How can I identify the outlier in the data set which consist the intra-subject residues? The statistical model consists of a global mean µ ± effects (treatment, period, sequence,…) + ε. ε is a the error term (normal distribution) with µ zero and variance σ. In the general applied model the (unknown) ε is estimated by inter- and intra-subject residuals (in other words, the variability not 'explained' by effects). In a cross-over design inter-subject variability has no influence on the BE assessment – only intra-subject variability (the higher CVintra the wider the CI). If you want to test for normality, intra-subject residuals are the term you should consider (neither the individual PK-responses nor the individual T/R-ratios). ❝ ❝ If it is not appropriate to analyse with transformed data, for example, when untransformed data shows normal distribution but transformed data shows non-normal distribution, you may state as such and use untransformed data for the assessment. ❝ Should I mention above statements in the protocol before start the study? I guess that no country except Japan will accept such a statement! ❝ Anyway, the study already finished, what can I do now is to mention this phenomenon in report and conclude the treatments is not bioequivalence? I don't know which country you are aiming at, but you should have stated the planned statistical model/methods in the protocol. You can run any method you like and show results in the report, but anything but log-transformed analyses most likely will not be accepted. I have seen a couple of reports 15+ years ago, where people tried to come up with an argument like this: the sample size is too low to make a clear decision between log-transformed and untransformed data analyses. Therefore the analysis with the narrower CI should be considered the “valid” one. The acceptance by regulatory authorities even at that time was close to zero. Nowadays: no chance. — 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, 2009-05-26 16:07 (6230 d 02:38 ago) @ Helmut Posting: # 3763 Views: 8,336 |
|
|
Dear HS, ❝ ε is a the error term (normal distribution) with µ zero and variance σ. ...so the sd is the square root of σ?  EM. |
|
Helmut ★★★ Vienna, Austria, 2009-05-26 16:23 (6230 d 02:23 ago) @ ElMaestro Posting: # 3764 Views: 8,294 |
|
|
Dear ElMaestro! ❝ ❝ ε is a the error term (normal distribution) with µ zero and variance σ. ❝ ...so the sd is the square root of σ? Oh wow, that’s about the brain the size of a walnut… Do you know the phrase “Lights are on, nobody at home.”? Just for the archive: … µ zero and variance σ2. ![[image]](img/uploaded/image249.png) — 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 |
|
Ohlbe ★★★ France, 2009-05-26 16:24 (6230 d 02:21 ago) @ Helmut Posting: # 3765 Views: 8,258 |
|
|
❝ Therefore the analysis with the narrower CI should be considered the “valid” one. The acceptance by regulatory authorities even at that time was close to zero. Nowadays: no chance. Nowadays: the analysis with the wider CI would be considered as the 'valid' one. 'Conservative' approach, they would call it. Regards Ohlbe |
|
Helmut ★★★ Vienna, Austria, 2009-05-26 17:41 (6230 d 01:04 ago) @ Ohlbe Posting: # 3766 Views: 8,363 |
|
|
Dear Ohlbe! ❝ ❝ Therefore the analysis with the narrower CI should be considered the “valid” one. The acceptance by regulatory authorities even at that time was close to zero. Nowadays: no chance. ❝ ❝ Nowadays: the analysis with the wider CI would be considered as the 'valid' one. 'Conservative' approach, they would call it. Not quite so. I was quoting an argument which was used in the past. Considering the wider CI to be valid just because it’s wider, is no justification as well. We simply cannot decide which distribution is the ‘correct’ one based on the small sample size. Therefore a consensus was reached based on assumptions: Lognormal for AUC, Cmax. Remark: many people would go for an additive model (untransformed) for urine data… — 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 |
|
Ohlbe ★★★ France, 2009-05-26 20:42 (6229 d 22:03 ago) @ Helmut Posting: # 3768 Views: 8,265 |
|
|
Dear Helmut, ❝ Considering the wider CI to be valid just because it’s wider, is no justification as well. True. But Agencies will not try to find out which model is the correct one, or whether you demonstrated which is correct. They will just keep their ass in a safe place (excuse my French  ) and reject your study.Regards Ohlbe |
|
Helmut ★★★ Vienna, Austria, 2009-05-26 20:46 (6229 d 21:59 ago) @ Ohlbe Posting: # 3769 Views: 8,304 |
|
|
Dear Ohlbe, agree 100% – especially with your French.  My main point is that one shouldn’t try to “justify“ an assumption. That’s simply futile. — 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