Log in
Register|
vezz ☆ Erba (CO), Italy, 2009-12-09 22:52 (6032 d 12:52 ago) Posting: # 4450 Views: 13,758 |
|
|
Dear All, I'm Stefano Vezzoli, biostatistician at CROS NT, a CRO based in Verona, Italy. Most of my (little: I'm just 28) experience is with Phase III trials, but I'm currently working on a pilot bioavailability study. In this forum I have already found a lot of useful suggestions and provocative thoughts, and I hope that my questions will lead to equally interesting discussions. My question concerns the inclusion of sequence effect in models used to analyse pharmacokinetic parameters. I know this is a very debated issue. For this reason, I tried to gather as much information as possible from previous topics and available resources. The model described by Helmut Schütz in his excellent lecture (pag. 15) does not include sequence effect. I think that the reasons behind this choice are fully explained in the following section (pag. 44-50) of the same presentation. However, as noted in the slides, EMEA Draft GL on BE explicitly requests sequence effect to be included in the model (line 509). The same model is described by the FDA guideline Statistical Approaches to Establishing Bioequivalence (section VI.B.1.b, pag. 10). But in the section VII.B (pag. 13) of the same document, it is stated that in a 2X2 crossover design "the only statistical test available for the presence of unequal carryover effects is the sequence test", and circumstances are provided under which "the possibility of unequal carryover effects is considered unlikely in a BE study". My question is the following: if the conditions defined in the FDA GL, the inclusion of the sequence effect is mandatory? Thank you very much for your help. — Kind regards, Stefano |
|
Pankaj Mishra ☆ Mumbai, India, 2009-12-10 08:21 (6032 d 03:24 ago) (edited on 2009-12-10 10:19) @ vezz Posting: # 4453 Views: 11,643 |
|
|
Dear Stefano, Answer to your question is NO. As far as bioequivalence studies for US-FDA are concerned, it is not necessary to keep the "sequence" effect in the model. But most of the time we keep this "sequence" effect in the model. Obviously if we remove the sequence effect, our total error variance will increase and consequently the intra-subject CV will rise which will further give us a wider 90% CI as compared to a model with "sequence" effect. So it is in our favor to keep “sequence” effect in a model especially for a BA/BE study. Further, I have observed that in Phase-I trials some statisticians don't prefer to use sequence effect as they are more keen to test the "treatment" or "period" effects. In my opinion, sequence effect should always be considered for a cross-over trial as the basic principle of ANOVA is to segregate the total variability into different components of known variability and "sequence" is not an exception in a cross-over trial. Regards, Pankaj Mishra Edit: Full quote removed. Please delete anything from the text of the original poster which is not necessary in understanding your answer; see also this post! [Ohlbe] — Pankaj Mishra |
|
d_labes ★★★ Berlin, Germany, 2009-12-10 10:24 (6032 d 01:21 ago) @ Pankaj Mishra Posting: # 4455 Views: 11,585 |
|
|
Dear Pankaj Mishra, ❝ ... Obviously if we remove the sequence effect, our total error variance will increase and consequently the intra-subject CV will rise which will further give us a wider 90% CI as compared to a model with "sequence" effect. This is definitely not true  (at least not for a classical 2x2 cross-over). Try an analysis (ANOVA) with effects formulation, period and subject. Try a second analysis with effects formulation, period, sequence and subject within sequence. Look at the result for the error term and the 90% confidence intervals of the formulation difference. If all is ok with your software and model formulation you will get the same MSE and 90% CI from both models. Try it! And note that the sum of squares for sequence and subject within sequence of the second analysis sum up to the sum of squares of subject effects in the first. — Regards, Detlew |
|
d_labes ★★★ Berlin, Germany, 2009-12-10 10:01 (6032 d 01:43 ago) @ vezz Posting: # 4454 Views: 11,647 |
|
|
Dear Stefano, ❝ My question is the following: if the conditions defined in the FDA GL, the inclusion of the sequence effect is mandatory? Interesting question. Following Stephen Senn[1] the answer is NO. He is a strong advocat of not testing carry-over in bioequivalence studies. And the inclusion of a sequence effect is effectively a test of unequal carry-over in a 2x2 crossover (its at least one possibility for a significant sequence effect). But as you have noticed by yourself EMEA and FDA (and other regulatory bodies also) mention the inclusion of sequence in the ANOVA. Also they complain on the other hand that unequal carry over is unlikely (your citation) and should not tested for (EMEA Draft Guideline on BE, line 514 ff). This is scientifically inconsistent I think. Why to include an effect, that is unlikely to occur and that should not be tested for? But regulatory rules are not always scientifically sound  .Thus, to be on the safe side I would recommend you to perform the usual ANOVA with formulation, period, sequence and subject within sequence as effects. This was my habit up to now, also I am an adept of Stephen Senn's thinking. If you are a "hero for truth" try to leave the sequence effect out and fight against deficiencies letters  .BTW: Your results concerning the bioequivalence test via the 90% confidence intervals will not be influenced by a decision of including or not a sequence effect in the ANOVA as opposed to Pankaj Mishra's post. This is because the sequence effect is a decomposition of the between total subject effect into sequence and subject within sequence in the ANOVA. The error MSE is not influenced. [1] S. Senn, "Crossover Trials in Clinical research", Second edition, Wiley 2002 — Regards, Detlew |
|
vezz ☆ Erba (CO), Italy, 2009-12-10 12:02 (6031 d 23:42 ago) @ d_labes Posting: # 4457 Views: 12,051 |
|
|
Dear D. Labes and Pankaj Mishra, thank you for your answers. My personal feeling is that the inclusion of the sequence effect is more strongly requested by the regulator for BE studies than for other kinds of trials. This seems to me to be a quite inconsistent attitude. Do you agree with me? However, this regulatory requirement leads me to another question: which effect should we include for the analysis of a crossover design with number of periods = number of treatments > 2 (i.e., 3x3, 4x4, etc.)?
[1]S. Senn, "Crossover Trials in Clinical Research", Second edition, Wiley 2002 (section 10.5.3, pag. 317). — Kind regards, Stefano |
|
d_labes ★★★ Berlin, Germany, 2009-12-10 17:43 (6031 d 18:01 ago) @ vezz Posting: # 4461 Views: 12,366 |
|
|
Dear Stefano, ❝ ... My personal feeling is that the inclusion of the sequence effect is more strongly requested by the regulator for BE studies than for other kinds of trials. Not necessarily so. Since the introduction of the cross-over design researchers using it have to struggle with the question of sequence effect, carry-over ... and so on. This historical burden lasts for BE studies until nowadays, although its meanwhile clear that at least carryover is avoided by proper planning of BE cross-over studies (appropriate wash-out). Carry-over is only one explanation of the more universal term of a treatment by period interaction which can have many sources. This interaction is prominent in cross-over studies, also Senn[2] has argued that this interaction is also of importance for parallel group studies. ❝ ... this regulatory requirement leads me to another question: which effect should we include for the analysis of a crossover design with number of periods = number of treatments > 2 (i.e., 3x3, 4x4, etc.)? I use for such studies exactly the same ANOVA model as for the classical 2x2 crossover. ❝ ... But how the sequence effect should be interpreted in such a design? .Don't waste your time in considering simple models of carry-over. They all do not make sense as you know from Senn [1]. I haven't seen or used any such models of simple carry over in my career up to now (more than 30 years) and was not asked by regulators to do so in the evaluation of BE studies. [2]S. Senn, "Crossover Trials, degrees of freedom, the carryover problem and its dual", Stat. Med., Vol.10, p 1361-1374 — Regards, Detlew |
|
Helmut ★★★   Vienna, Austria, 2009-12-10 17:55 (6031 d 17:49 ago) @ d_labes Posting: # 4462 Views: 11,466 |
|
|
Dear D. Labes and Stefano! ❝ I haven't seen or used any such models of simple carry over in my career up to now (more than 30 years) and was not asked by regulators to do so in the evaluation of BE studies. Agree. Though I can't catch up with your experience (celebrating my 30th anniversary in three weeks). — 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, 2009-12-10 19:10 (6031 d 16:34 ago) @ Helmut Posting: # 4463 Views: 11,429 |
|
|
Dear D. Labes and Helmut, thank you for sharing your experienced viewpoint. As you, I have never used models including simple carryover effect and I'm not pursuing this approach. I was thinking again about the interpretation of sequence effect in designs with more than two treatments and an equal number of periods. Probably, if we assume that carryover of higher order are extremely unlikely, its interpretation is analogous to that given in 2x2 design. As regards the inclusion of sequence effect in BE studies, I think that the book by Jones and Kenward [*] shows an interesting feature. This variable is not included in any model until the last chapter (7. Bioequivalence trials), where it is stated: "Following common practice we will also fit a sequence or group effect" (section 7.1). What I am really interested in is the rationale behind this "common practice". Thank you again for your help. [*] B. Jones, M.G. Kenward, "Design and Analysis of Cross-Over Trials", Second edition, Chapman & Hall/CRC 2003. — Kind regards, Stefano |
|
Helmut ★★★ Vienna, Austria, 2009-12-10 20:19 (6031 d 15:26 ago) @ vezz Posting: # 4464 Views: 11,479 |
|
|
Dear Stefano! ❝ I was thinking again about the interpretation of sequence effect in designs with more than two treatments and an equal number of periods. Probably, if we assume that carryover of higher order are extremely unlikely, its interpretation is analogous to that given in 2x2 design. Yes, and unless a replicate design is used we have no means to sort things out in a simple higher-order Xover. ❝ […] the book by Jones and Kenward shows an interesting feature. Maybe this a historical leftover? According to my - vanishing - memory Byron Jones talked much more about carry-over in the first edition. But Stephen layed siege to this concept (his first edition on Xover studies can be read as a tractatus contra vectum transitus*).  Step by step Byron retreated. I would also recommend to invest in Patterson'/Jones' textbook from 2006 (reference). If you want to dig even deeper into the matter, I can give you Byron's e-mail (he is a very nice guy). * Essay against carry-over P.S.: A nice paper for the bedside table by Stephen. — 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 |
|
d_labes ★★★ Berlin, Germany, 2009-12-11 10:42 (6031 d 01:02 ago) @ vezz Posting: # 4466 Views: 11,394 |
|
|
Dear Stefano, ❝ Probably, if we assume that carryover of higher order are extremely unlikely, its interpretation is analogous to that given in 2x2 design. Not at least: Occurred simply by chance! Since and if it is extremely unlikely. See: Stephen Senn, Giuseppina D’Angelo, and Diane Potvin "Carry-over in cross-over trials in bioequivalence: theoretical concerns and empirical evidence" Pharmaceut. Statist. 2004; 3: 133–142 But I guess, you know that?  Astonishing enough this is already mentioned in the 1997 FDA guidance: "IV. Sequence effect ... Even if there were no true sequence effect, no unequal residual effect (aka carryover) and no period-by-treatment statistical interaction, approximately ten out of every hundred standard two-treatment crossover studies would be likely to show an apparent sequence effect, if the testing is carried out at the ten percent level of significance. ..." ❝ ... What I am really interested in is the rationale behind this "common practice". Like Helmut I'm convinced that this is definitely a historical leftover. But regulated by law. Because we/other/people[1] tend to take guidances in the field of medical research as law. [1] Please delete where inapplicable. — Regards, Detlew |
|
Helmut ★★★ Vienna, Austria, 2009-12-11 14:40 (6030 d 21:04 ago) @ d_labes Posting: # 4470 Views: 11,405 |
|
|
Dear D. Labes! ❝ Astonishing enough this is already mentioned in the 1997 FDA guidance: Even further back in ![[image]](img/uploaded/IAsmall.png) July 1992! Well, I guess they have read Westlake (1988)[1]:Note that the carryover effect is, essentially, the sequence effect, which can be tested against the sum of squares within sequence. If this carryover effect exists, then it confounds the test on formulations. [...] My own experience with a large number of comparative bioavailability trials has led me to believe that significant carryover effects (at the 0.05 level) tend to occur in about 5% of the trials; in other words, I believe that carryover effects do not normally exist. There are several reasons why this assumption is a reasonable one. First subjects in bioavailability trials are usually normal human volunteers whose physical condition is unlikely to change radically from one period of a crossover trial to another, unlike patients whose disease state may progress during the course of even a short trial. Second, the trial usually involves only single doses of the drug formulations and is thus unlikely to result in a marked change in a subject's physical condition. Third, a washout period is scheduled between the periods of the crossover trial which allows for elimination of all drug from the system; and a direct check of the presence of drug is, of course, obtained from the zero-hour blood level. Wagner (1974)[2] has, however, raised the possibility that an unmeasured metabolite might still be present in the body and affect the biotransformation at the succeeding administration. (my emphasis)
— 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 |
|
d_labes ★★★ Berlin, Germany, 2009-12-11 10:09 (6031 d 01:36 ago) @ Helmut Posting: # 4465 Views: 11,365 |
|
|
Dear Helmut! Fart may be acceptable to me, but take back Old! ❝ ... (celebrating my 30th anniversary in three weeks). So accept my congratulations already today  .— Regards, Detlew |
|
Helmut ★★★ Vienna, Austria, 2009-12-11 14:51 (6030 d 20:53 ago) @ d_labes Posting: # 4471 Views: 11,632 |
|
|
Dear D. Labes! ❝ Fart may be acceptable to me, but take back Old! Sorry; "Forever young!" (© Robert Zimmermann) Recently I came across a nice comment: Youths may be able to run faster, but we know the shortcuts! ❝ So accept my congratulations already today THX!  — 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