Log in
Register|
Anu ★ India, 2013-07-29 15:04 (4284 d 23:11 ago) (edited on 2013-07-30 07:49) Posting: # 11109 Views: 10,516 |
|
|
Hi all, Greetings! I've had done a Partially Replicate BE Study. After running the WinNonlin software for the Partially Replicate study I couldn't get Intrasubject variability directly (Unlike Fully Replicate and Cross over studies) but had done manual calculations on the available Var(Residual) by using the formula: Intrasubject CV = sqrt(exp(Residual_Variance) - 1). The same problem had cropped up for Three way crossover study as well. Kindly guide me how can I get the Intrasubject CV value directly from the software itself or can I put whole formula somewhere to get the CV. Thanks & Regards Anu |
|
Helmut ★★★   Vienna, Austria, 2013-07-30 15:51 (4283 d 22:24 ago) @ Anu Posting: # 11124 Views: 9,517 |
|
|
Hi Anu, ❝ After running the WinNonlin software… Which version? ❝ … for the Partially Replicate study I couldn't get Intrasubject variability directly (Unlike Fully Replicate and Cross over studies) … Was the setup for conventional ABE (= unscaled) or RSABE (= my implementation of FDA’s code)? In the latter case only CVWR is meaningful and part of the output. I’m surprised that you got some output for a fully replicated design (I don’t). AFAIK Intrasubject CV is part of the standard output only for 2×2 cross-overs. ❝ … had done manual calculations on the available Var(Residual) by using the formula: ❝ Intrasubject CV = sqrt(exp(Residual_Variance) - 1). Correct. ❝ The same problem had cropped up for Three way crossover study as well. Also correct (see above). ❝ Kindly guide me how can I get the Intrasubject CV value directly from the software itself or can I put whole formula somewhere to get the CV. All examples for ABE:
— 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 |
|
Anu ★ India, 2013-08-02 12:49 (4281 d 01:26 ago) @ Helmut Posting: # 11166 Views: 9,361 |
|
|
Hi Helmut, Thank you very much for the reply. ❝ Which version? WinNonlin version 5.3 ❝ Was the setup for conventional ABE (= unscaled) or RSABE (= my implementation of FDA’s code)? In the latter case only CVWR is meaningful and part of the output. It was RSABE. Used the codes given in these paper to run the design in Phoenix  ❝ I’m surprised that you got some output for a fully replicated design (I don’t). My mistake sorry (Actually I had run the fully replicate taking 2 blocks at a time in a study, so that was there in my mind) ❝ • classical WinNonlin (≤5.3) ❝ — Run the Bioequivalence Wizard ❝ — Detach the result worksheet ❝ — Enter the formula in an empty cell, referring to the cell containing Var(Residual) ❝ — Refresh Thanks . Though treated the WinNonlin sheet as excel sheet and even without detaching it worked. Is it right?Have one more doubt, Can be a stupid one but still a qn please: Point estimate: T/R (Do we consider it as geometric mean ratio or Geometric least square mean ratio?) Thanks & Regards Anu Jaswal |
|
Helmut ★★★ Vienna, Austria, 2013-08-02 16:24 (4280 d 21:51 ago) @ Anu Posting: # 11167 Views: 9,428 |
|
|
Hi Anu, ❝ WinNonlin version 5.3 I strongly suggest to upgrade to Phoenix. WinNonlin 5.3 is not supported any more and buggy (to name a few: λz based on max. R²adj may include Cmax, nonparametric comparison of tmax always in lexical order of treatments, CI obtained from the BE wizard is wrong, whereas from LME is correct – requiring a workaround, …) Inspectors might ask why you still use an outdated software – although the upgrade was free of charge for 3+ years. ❝ Though treated the WinNonlin sheet as excel sheet and even without detaching it worked. Is it right? Maybe. Since I installed my latest PHX-license WNL5.3 refuses to start, so I can’t check. ❝ Point estimate: ❝ T/R (Do we consider it as geometric mean ratio or Geometric least square mean ratio?) I would prefer the latter (aka adjusted mean). Only in a balanced study geometric mean = adjusted mean. — 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 |
|
BRN ☆ Turkey, 2014-02-04 19:00 (4094 d 18:15 ago) @ Helmut Posting: # 12336 Views: 8,870 |
|
|
Dear Helmut, I have results from a pilot study with 2 test products (T1 and T2) against R, administered in 3 periods, two analytes. 17 subjects out of 18 completed the study. I'd like to have some guidance for the BE analysis for WinNonlin 6.3. What would be the best approach? Could you please kindly guide? I included T1 and T2 in one NCA and run the bioquivalence as one dataset for each analyte. To run the NCA I included columns to distinguish formulations (my columns are: analyte, subject, period, formulation, sequence, time, concetration). Do I need to calculate intrasubject variability as described below or it's just written in the final variance parameters as var(residual) (which seems to be little bit low)? I can separate T1 and T2 and run them against the R separetely also but would it make a difference? If I need to calculate CVintra manually, I couldn't figure out the step below. Do I hit OK after Column [Parameter] and then another step to exclude? Thank you very much in advance for your answer. Regards, BRN ❝ – Options > Filter > Add > ● Built In > Add… > Action [Include] > Find [Var(Residual)] > Column [Parameter] > Exclude ● Units, Parameter > OK ❝ – Options > Transformation > Add > Transformation Type [Custom] > New Column Name [CVintra] > Formula [100*sqrt(exp(Estimate)-1)] I figured out how to calculate the CVintra, so nevermind the related question. However, still can't figure out what's the best approach for presenting results. I used A, B and C letters for T1, T2 and Reference products respectively not to enter T1T2R in the sequence column. Then I did 2 types of analysis. 1-Run the analysis on the complete data set. Getting these results (PE, 90%CI U - L): A, Cmax, 103.92 (83.42-129.46), CVintra: 39.1% B, Cmax, 96.01 (77.07-119.61), CVintra: (Getting one CVintra for Cmax, shown above) 2-Run the analysis on two different datasets. One including A and C, the other including B and C. A, Cmax, 104.24 (85.84-126.57), CVintra: 32.9% B, Cmax, 95.55 (76.54-119.29), CVintra: 38.0% Which way to go? In both cases it is a highly variable drug, widened limits and bioequivalent, Right? Thank you for your help, Kind regards, BRN Edit: Merged from a later post; you can edit posts for 24 hours. [Helmut] — Regards, BRN |
|
Helmut ★★★ Vienna, Austria, 2014-02-05 15:41 (4093 d 21:35 ago) @ BRN Posting: # 12345 Views: 9,014 |
|
|
Merhaba BRN, ❝ I figured out how to calculate the CVintra, so nevermind the related question. Fine. ❝ I used A, B and C letters for T1, T2 and Reference products respectively not to enter T1T2R in the sequence column. ![[image]](img/uploaded/image220.png) Why not? But if you want to make the life of inspectors easier, I would suggest to use the treatment’s coding. You can still enter A/B/C for convenience and then replace A → T1, B → T2, and C → R. This procedure is transparent (documented in the “History”-tab of the worksheet).A convenient way is to have two worksheets. One containing the blinded concentrations with columns subject / period and the other one the randomization (subject / period / sequence / treatment). In the workflow you can merge or join them sorting by subject / period and include the other columns. Generally I perform the NCA still blinded for treatments and link to the randomization afterwards.❝ Then I did 2 types of analysis. ❝ 1-Run the analysis on the complete data set. Getting these results (PE, 90%CI U - L): ❝ A, Cmax, 103.92 (83.42-129.46), CVintra: 39.1% ❝ B, Cmax, 96.01 (77.07-119.61), CVintra: (Getting one CVintra for Cmax, shown above) ❝ ❝ 2-Run the analysis on two different datasets. One including A and C, the other including B and C. ❝ A, Cmax, 104.24 (85.84-126.57), CVintra: 32.9% ❝ B, Cmax, 95.55 (76.54-119.29), CVintra: 38.0% ❝ ❝ Which way to go? The first analysis (full model) assumes a common variance of all treatments. This is the one you would use in pivotal studies. Examples
❝ In both cases it is a highly variable drug, widened limits and bioequivalent, Right? A high CV from a non-replicated cross-over is just a hint (CVintra is pooled from CVWR and CVWT). The limits can be only widened based on the CV of the reference in a replicated design. — 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 |
|
BRN ☆ Turkey, 2014-02-06 09:38 (4093 d 03:37 ago) @ Helmut Posting: # 12353 Views: 8,740 |
|
|
Thank you so much! Another question came into my mind. Does this replicate design has to be with minimum two reference administration? If we want to design a 3 period crossover, what should we take into consideration, would you suggest a lecture, thread or any other source? Thanks for your help, Regards. BRN Edit: Merged from a later post; you can edit posts for 24 hours. [Helmut] — Regards, BRN |
|
Helmut ★★★ Vienna, Austria, 2014-02-06 15:42 (4092 d 21:34 ago) @ BRN Posting: # 12357 Views: 8,814 |
|
|
Merhaba BRN! In the future please don’t reply to your own posts; edit the original one instead. THX! I don’t know what the current thinking of the Turkish MoH is – or are you aiming at another country? A historical overview of approaches used in the EU:
❝ Does this replicate design has to be with minimum two reference administration? Yes. ❝ If we want to design a 3 period crossover, what should we take into consideration, would you suggest a lecture, thread or any other source? I would suggest my lectures. There are hundreds (!) of ![[image]](img/search.png) posts covering this topic and many references (literature, guidelines) within.— 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