Log in
Register|
Sriraj ☆ India, 2010-02-26 15:32 (5557 d 06:07 ago) Posting: # 4826 Views: 14,939 |
|
|
If we don't have information regarding the variability of the treatments, how do we estimate the sample size? While estimating the sample size, for BE studies we use 90% CI limits as per guidance, but for BA studies what should be the 90% CI limits and how it will be decided? Thanks in advance. SriKanth *SriRaj* |
|
Helmut ★★★   Vienna, Austria, 2010-03-01 15:00 (5554 d 06:39 ago) @ Sriraj Posting: # 4840 Views: 13,387 |
|
|
Dear SriKanth! ❝ If we don't have information regarding the variability of the treatments, how do we estimate the sample size? Get the CVintra from a pilot study. ❝ While estimating the sample size, for BE studies we use 90% CI limits as per guidance, but for BA studies what should be the 90% CI limits […] No; in BA report 95 % CI. For sample sizes in BA, see this post. ❝ […] and how it will be decided?  — 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 |
|
yjlee168 ★★★  Kaohsiung, Taiwan, 2010-03-01 20:25 (5554 d 01:14 ago) @ Helmut Posting: # 4843 Views: 13,513 |
|
|
Dear Helmut, ❝ ❝ how do we estimate the sample size? ❝ ❝ Get the CVintra from a pilot study. a very interesting topic! Can this CVintra from a pilot study really represent a CVintra from a main study? What sample size will be required for a pilot study? Usually we use 4-6 subjects as the sample size in a BE pilot study here. I am confused by this question for a while. Thanks in advanced. — All the best, -- Yung-jin Lee bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee Kaohsiung, Taiwan https://www.pkpd168.com/bear Download link (updated) -> here |
|
Helmut ★★★ Vienna, Austria, 2010-03-01 22:01 (5553 d 23:38 ago) @ yjlee168 Posting: # 4844 Views: 13,546 |
|
|
Dear Yung-jin! ❝ Can this CVintra from a pilot study really represent a CVintra from a main study? Well, you may see CVintra as an estimate from the population of possible main studies.  This estimate is uncertain. Therefore, based on the sample size of the pilot study, you may calculate a confidence interval of the CV (see this post). Generally one is only interested in the one-sided (upper) limit. If the CV in the main study is lower than expected, you will loose some money, but still demonstrate BE. If you base your sample size calculation one the ‘carved-from-stone’ CV in the pilot study, and the actual CV is higher - you loose money and fail to show BE. Bad luck.In the thread mentioned above I used a 95% CI, which is quite over-weary… Patterson & Jones (2006) recommend an 80% one-sided-interval (or the common producer's risk of 20%). If sponsors have some problems in understanding this concept, just tell them, that if they use the carved-from-stone CV the chance of a higher CV in the main study is 50%. ❝ What sample size will be required for a pilot study? Hey, that’s a really good question! We should leave this to some Bayesian statisticians, because we would need some priors in this case. Actually the questions is: how small can a pilot study be that with a certain confidence the sample size estimation for the pivotal study is reliable? Main hint: The larger, the better. OK, that's trivial, but let's play around with the formula (balanced 2×2 studies), upper 80% CL of CVintra:
n 10% 15% 20% 30% 40% 50%For an expected CV of 20% I would say 12 subjects are OK, for HVDs the minimum in a pilot is 16-24. Well, of course you can perform a pilot study in six subjects, obtain a CV of 50% and run the main study as a 4×2 replicate study (PE 95%, 80% power, conventional AR) in 50 subjects. But: you have a 50% chance, that the CV will be >50% and the study will fail. Or you use the upper CL (84.8%), which will call for 118 subjects. Now you have only a 20% chance of failure. Since it's a HVD, perform the pilot in 24 subjects, base the sample size on 59.8% and have an expected chance of success in 68 subjects. ❝ Usually we use 4-6 subjects as the sample size in a BE pilot study here. Wow, four! That’s brave. Six is Vinod Shah’s magic number – I call that a dice game with the devil. See one of my old lectures (1.7MB, slides 60-63). I cheated in the example: I took a real dataset (24 subjects) and sliced it into either four subsets of six subjects each, or in two subsets of twelve subjects. Again: the bigger, the better. — 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 |
|
yjlee168 ★★★ Kaohsiung, Taiwan, 2010-03-02 00:47 (5553 d 20:52 ago) @ Helmut Posting: # 4846 Views: 13,190 |
|
|
Dear Helmut, Thank you for your message. ❝ statisticians, because we would need some priors in this case. Actually the questions is: how small can a pilot study be that with a certain confidence the sample size estimation for the pivotal study is reliable? Exactly. ❝ Main hint: The larger, the better. OK, that's trivial, but let's play around with the formula (balanced 2×2 studies), upper 80% CL of CVintra: ❝ ❝ ❝ ❝ ❝ ❝ ❝ Uhh... Sorry about that I don't know how to interpret/use this table. ❝ For an expected CV of 20% I would say 12 subjects are OK, for HVDs the minimum in a pilot is 16-24. O.k.. could be a good guess. ❝ Wow, four! That's brave. Six is Vinod Shah's magic number - I call that a dice game with the devil. That's commonly used with 4-6 subjects in a BE pilot study here... I just cannot believe that the subj# like that could give a good estimation of CVintra at all. It almost becomes SOP with local CROs here to use 4-6 subjects to perform a BE pilot study. Then they replicate the values they get from these 4-6 subjects up to 24 or so, and finally do an anova to obtain the MSE. And they get their estimated CVintra! I am thinking that it should be very to prove that CVintra is poorly estimated in that way. ❝ subjects each, or in two subsets of twelve subjects. Again: the bigger, the better. Yes, but the question is: how bigger is the better? It's a pilot study only, but it seems very critical. — All the best, -- Yung-jin Lee bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee Kaohsiung, Taiwan https://www.pkpd168.com/bear Download link (updated) -> here |
|
Helmut ★★★ Vienna, Austria, 2010-03-03 20:04 (5552 d 01:35 ago) @ yjlee168 Posting: # 4856 Views: 13,158 |
|
|
Dear Yung-jin, ❝ ❝ Actually the questions is: how small can a pilot study be that with a certain confidence the sample size estimation for the pivotal study is reliable? ❝ ❝ Exactly. Yes, but I think the only tools we have is either the confidence limit approach or bootstrapping. Both methods may give us a nasty (i.e., large CV) if the sample size of the pilot is small - or on the other hand will suggest a small size, because the CV in the pilot was low due to pure chance. But that's trivial. The larger any sample size is, the lower the variability (aka the 1/sqrt(N) game). I'm afraid, there is not a simple answer to the question above. ❝ ❝ ❝ ❝ ❝ ❝ ❝ ❝ Uhh... Sorry about that I don't know how to interpret/use this table. That's easy. You have a pilot study (n=4) which gave you a CV of 20%. The upper 80% confidence interval of the CV is 43.8%. In other words you have a 50% chance that in the main study the CV will be larger than 20% (trivial, because it's an estimate) and a 20% chance that the CV will be larger than 43.8%. Now it's up to you: Believe in the CV of 20% (well, that't like flip of a coin, or just a little bit better than red/black in roulette) or go with the CV of 43.8%. Oops, is it a HVD? Now look at the second row: The pilot study was larger (n=16) and again we got a CV of 20%. But now the upper CL is only 24.4%. Due to the larger pilot, we are more certain about the value of the CV. ❝ ❝ Wow, four! That's brave. Six is Vinod Shah's magic number - I call that a dice game with the devil. [cut] ❝ That's commonly used with 4-6 subjects in a BE pilot study here... I just cannot believe that the subj# like that could give a good estimation of CVintra at all. It almost becomes SOP with local CROs here to use 4-6 subjects to perform a BE pilot study. Haha. I guess they apply a nice safety margin on the sample size. ❝ Then they replicate the values they get from these 4-6 subjects up to 24 or so, and finally do an anova to obtain the MSE. And they get their estimated CVintra! I'm not sure whether I understand this procedure. Sounds like some kind of resampling (bootstraping), but I don't get the point. Why 24? Bootstraping relies on a sufficiently large and representative sample. If the sample is small and the variability is particularly small or high due to chance, any resampling will reproduce this behaviour in the simulated data set. Sorry, but n=4 is rubbish, IMHO. ❝ Yes, but the question is: how bigger is the better? It's a pilot study only, but it seems very critical. Well, easy question - difficult answer. I'm really in favour of sequential designs; you guessed right: D Potvin et al. (2008) is my favourite method. It's nice to have a method to deal with the uncertainty of CV, but another point is the variability of the point estimate. To my knowledge, there is no statistical method available to deal with that. So again, what to use? The ratio in the pilot (the ‘carved-from-stone-method’)? A safety margin? How large; ±5% - or what? — 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 |
|
yjlee168 ★★★ Kaohsiung, Taiwan, 2010-03-04 23:35 (5550 d 22:04 ago) @ Helmut Posting: # 4863 Views: 13,081 |
|
|
Dear Helmut, Thanks again for your response. ❝ variability. I'm afraid, there is not a simple answer to the question above. Agree. ❝ ❝ Uhh... Sorry about that I don't know how to interpret/use this table. ❝ That's easy. You have a pilot study (n=4) which gave you a CV of 20% I see. Great. ❝ I'm not sure whether I understand this procedure. Sounds like some kind of resampling (bootstraping), but I don't get the point. Why 24? Bootstraping Kind of, but it's absolutely not. ❝ Well, easy question - difficult answer. I'm really in favour of sequential designs; you guessed right: D Potvin et al. (2008) is my favourite method. Well, thanks to ElMaestro too. After I got the replied message from EM, I did some search and found some nice articles regarding adaptive two-stage desing with pilot-pivotal trials for average BE. I think the optimal design of a average BE pilot study is an important issue, although we do not see many investigators published the results of their pilot study. Probably it's kind of trivial as you said. — All the best, -- Yung-jin Lee bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee Kaohsiung, Taiwan https://www.pkpd168.com/bear Download link (updated) -> here |
|
d_labes ★★★ Berlin, Germany, 2010-03-19 14:02 (5536 d 07:37 ago) @ Helmut Posting: # 4938 Views: 12,992 |
|
|
Dear Helmut! ❝ In the thread mentioned above I used a 95% CI, which is quite over-weary… Patterson & Jones (2006) recommend an 80% one-sided-interval (or the common producer's risk of 20%). .In Patterson, Jones "BIOEQUIVALENCE and STATISTICS in CLINICAL PHARMACOLOGY", Chapter 5.7, page 164 (of my copy) in the text it is mentioned "... with an upper 50% confidence bound ..." which is clearly a typo. If I try to verify their example (Table 5.6) I get the upper confidence bound mentioned only if I use alpha=5%. BTW: I nevertheless are the opinion that a 95% CI is too stringent. But have no reference for it. Patterson and Jones cite Gould in this context. Gould, A.L. (1995) Group sequential extensions of a standard bioequivalence testing procedure. J. Pharmacokinetics and Biopharmaceutics, 23, 57–86. Since I don't have this on file yet, I can't verify if this is also only a typo? — Regards, Detlew |
|
Helmut ★★★ Vienna, Austria, 2010-03-19 15:12 (5536 d 06:27 ago) @ d_labes Posting: # 4939 Views: 12,941 |
|
|
Dear D Labes, you are are right; this is a typo - 5 % is correct. ❝ BTW: I nevertheless are the opinion that a 95% CI is too stringent. Yes; I re-read the entire chapter - where did I get the 80 % from?! ❝ But have no reference for it. Patterson and Jones cite Gould in this context. Gould (p.71) uses the upper 75 % confidence bound (Chi²0.25,df). — 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, 2010-03-01 23:23 (5553 d 22:16 ago) @ yjlee168 Posting: # 4845 Views: 13,267 |
|
|
Hello bears, ❝ Can this CVintra from a pilot study really represent a CVintra from a main study? ❝ What sample size will be required for a pilot study? Usually we use 4-6 subjects as the sample size in a BE pilot study here. I am confused by this question for a while. Yes to the first Q. In a pivotal trial the important uncertainty is that on the T/R ratio. We use this uncertainty it to construct a confidence interval around T/R and to satisfy regulators (hopefully). In a pilot study like the one we discuss here the important uncertainty is that on the estimated CVintra. If we underestimate it, we may underpower the pivotal study (saved only by luck, or a better T/R ratio then expected). I have seen pilots ranging from 12 to perhaps somewhere in the 30'ies, I think. Not exactly sure. 4 to 6 in a pilot trial sounds like it may result in a CVintra that is too poorly estimated. There are no rules, and 'poorly' is in this regard purely my personal subjective unqualified view. As a good alternative, it is getting more and more common to do a two-stage study in which the first, say, 20 patients are analysed whereafter an interim stats evaluation is carried out; the interim evaluation tries to answers a quesiton like: How many additional patients should be recruited in order to satisfy a goal of 80% (or 90% whatever one likes) power given the CVintra that can be estimated from the first 20? The latter generally also comes with an expectation that the variability does not increase in the last patients, i.e. the first patients should be representative of the last patients. Best regards EM. |
|
yjlee168 ★★★ Kaohsiung, Taiwan, 2010-03-02 01:02 (5553 d 20:37 ago) @ ElMaestro Posting: # 4847 Views: 13,226 |
|
|
Dear ElMaestro, Thank you for your messages. ❝ As a good alternative, it is getting more and more common to do a two-stage study in which the first, say, 20 patients are analysed whereafter an interim stats evaluation is carried out; the interim [cut here] HS had a post previously (Two Stage Design). Was it same as your suggestion? Is it the so-called adaptive BE study design? Sorry, I could be wrong about this. — All the best, -- Yung-jin Lee bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee Kaohsiung, Taiwan https://www.pkpd168.com/bear Download link (updated) -> here |
|
Helmut ★★★ Vienna, Austria, 2010-03-02 03:19 (5553 d 18:20 ago) @ ElMaestro Posting: # 4848 Views: 13,200 |
|
|
Hi ElMaestro! ❝ 4 to 6 in a pilot trial sounds like it may result in a CVintra that is too poorly estimated. There are no rules, and 'poorly' is in this regard purely my personal subjective unqualified view. No, 'poorly' is not purely your personal view, but qualified by the table from my previous post. Part:
n 10% 15% 20% 30% 40% 50%If we calculate a CVintra of 20% in the pilot (n=6), the upper 80% CL of the CVintra is 31.6%. In other words there is a 20% chance that the CV might even be higher than 31.6%. If the pilot study study is larger (n=16), the estimated CV gets more reliable (=less uncertain) with 24.4%. You are right with the sequential designs. If I don't have a lot (!) of previous data, I made them already my new standard. — 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 |
|
yjlee168 ★★★ Kaohsiung, Taiwan, 2010-03-29 15:27 (5526 d 07:12 ago) @ Helmut Posting: # 4979 Views: 13,005 |
|
|
Dear Helmut and Elmaestro, ❝ You are right with the sequential designs. If I don't have a lot (!) of previous data, I made them already my new standard. So, it sounds like that the sequential designs (Method C or D suggested by Potvin D, et al., 2008) have been widely acceptable for regulatory agents (FDA/EMA/Japan/etc.). However, I don't quite remember that there is any regulation mentioned the sequential designs for ABE study. Am I wrong about this? Thanks. — All the best, -- Yung-jin Lee bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee Kaohsiung, Taiwan https://www.pkpd168.com/bear Download link (updated) -> here |
|
ElMaestro ★★★ Denmark, 2010-03-29 15:39 (5526 d 07:00 ago) @ yjlee168 Posting: # 4981 Views: 12,884 |
|
|
Hi Bears, look at page 15/16 of the new EU guidance - "two stage design". Best regards EM. |
|
yjlee168 ★★★ Kaohsiung, Taiwan, 2010-03-29 15:47 (5526 d 06:52 ago) @ ElMaestro Posting: # 4983 Views: 12,894 |
|
|
Dear Elmaestro, YES! I see it. Unfortunately, we still have no such guidance yet in Taiwan. Many thanks. ❝ look at page 15/16 of the new EU guidance - "two stage design". — All the best, -- Yung-jin Lee bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee Kaohsiung, Taiwan https://www.pkpd168.com/bear Download link (updated) -> here |
|
Helmut ★★★ Vienna, Austria, 2010-03-29 16:19 (5526 d 06:20 ago) @ yjlee168 Posting: # 4985 Views: 12,915 |
|
|
Dear bears! ❝ So, it sounds like that the sequential designs (Method C or D suggested by Potvin D, et al., 2008) have been widely acceptable for regulatory agents (FDA/EMA/Japan/etc.). However, I don't quite remember that there is any regulation mentioned the sequential designs for ABE study. Am I wrong about this? Well, in Canada and Japan naïve pooling was acceptable for many years (in Canada for almost the last 20 years!). When the sample size turned out to be too small to demonstrate BE, additional subjects could be included – no correction of the alpha-level… Obviously the patient’s risk may be >0.05, but these countries’ inhabitants are proverbial for their endurance (Lumberjacks. Samurais!). Sequential designs were never part of the official guidelines in the US and the EU. However, the introduction of Potvin’s paper states that sequential design studies were accepted by the FDA in the past. The recent EU-GL allows for a sequential design; the actual method is not stated (but the description matches Potvin’s). Canada in their 2009 draft prefer Gould’s (1995) method. WHO (2006) allows add-ons; for other countries please search the Guidelines. — 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 |
|
yjlee168 ★★★ Kaohsiung, Taiwan, 2010-03-29 21:19 (5526 d 01:20 ago) @ Helmut Posting: # 4989 Views: 13,404 |
|
|
Dear Helmut and ElMaestro, Thank you for your comments. ❝ Well, in Canada and Japan naïve pooling was acceptable for many years (in Canada for almost the last 20 years!). When the sample size turned out to be too small to demonstrate BE, additional subjects could be included - no So in the case of sequential designs, when we get non BE results, the first thing is to check if the power is o.k. (defined as equal to or greater than 80%). If not, then we can go to Stage 2 by recruiting more subjects. According to Potvin D, et al., 2008, there seems only one chance (Method A-D). The authors considered Method C as the method of choice, and stated that Method B & C has been accepted by FDA (p. 259). If we like to add the sequential designs into bear, we have to implement the methods: to evaluate BE at Stage 1 (alpha = 0.0294) and to calculate sample size based on Stage 1 and alpha = 0.0294, and finally to evaluate BE at Stage 2 using data from both stages (alpha = 0.0294). The alpha level is not commonly used in statistics. However it seems feasible, I guess. ❝ correction of the alpha-level… Obviously the patient's risk may be >0.05, […] ❝ Sequential designs were never part of the official guidelines in the US and the EU. However, the introduction of Potvin's paper states that sequential design studies were accepted by the FDA in the past. The recent EU-GL allows for a sequential design; the actual method is not stated (but the description matches Potvin's). Now I know why you said the sequential designs was your practice standard in the previous thread. Many thanks. I do learn a lots. — All the best, -- Yung-jin Lee bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee Kaohsiung, Taiwan https://www.pkpd168.com/bear Download link (updated) -> here |
|
Helmut ★★★ Vienna, Austria, 2010-03-29 22:43 (5525 d 23:56 ago) @ yjlee168 Posting: # 4990 Views: 13,174 |
|
|
Dear bears! ❝ So in the case of sequential designs, when we get non BE results,... Almost. It is important in any sequential design not to ‘consume’ the entire alpha-risk in the interim looks. In other words - you must not evaluate the study for BE (i.e., calculate the 90% CI and check for inclusion in the AR). If you do that, your entire alpha-risk has gone - nothing left for further looks. This was actually the problem with the Canadian and Japanese method. Sequential designs have a long tradition in clinical phases II/III; the alpha level at each look is selected in such a way, that the overall alpha-risk is maintained at ≤0.05. ❝ … the first thing is to check if the power is o.k. (defined as equal to or greater than 80%). Yes, according to Potvin et al. It’s important to notice that the value calculated here is not some kind of a posteriori power, but only an instrument in deciding whether the study will be stopped after Stage 1 (evaluate at alpha 0.05: pass/fail) or enter the right branch of the flow chart. ❝ If not, then we can go to Stage 2 by recruiting more subjects. You are too fast. If power <80% we evaluate Stage 1 at alpha 0.0294 (instead of 0.05). In my workshops many people are scared by this alpha-value (i.e., a 94.12% CI instead of 90.00%). In the ol’ days of BE testing a 95% was applied all the time. In reality there is not a big difference in sample sizes. For CV 20%, ±5%, 80-125%, 80% power the sample sizes are 19 (alpha 0.05) and 23 (alpha 0.0294) - or the other way 'round: If you just miss the 80% (let's say you had one drop-out; power 79.1%). Power for alpha 0.0294 is 69.3% and that’s still a pretty good chance that we will show BE at Stage 1 and stop. Only if we fail here, we will advance to Stage 2. Another argument I’ve heard a couple of times is: ‘Costs! We will need more subjects in a two-stage design.’ Maybe. Simulations show a penalty of ~10% compared to fixed-sample designs. But only if one assumes that (s)he knows the ‘correct’ variance beforehand. If one has a lot of experiences with a formulation (own studies, same analytical method) and the CV is ‘stable’, fine. But with an uncertain estimate of the CV (literature data only or small pilot study), would you really want to ‘save’ 10% of the budget and end up with an upper CI of 125.94%. My suggestion is to power the first stage of the study as if it is a fixed-sample design. If expectations come true - business as usual: conventional statistical model, 90% CI, everybody is happy. If not, you get a second chance! ❝ According to Potvin D, et al., 2008, there seems only one chance (Method A-D). Yes. The methods were validated only for one interim analysis. There are others, which would allow for more than one look into the data (e.g., Gould 1995). However, in the EU only a two-stage design is acceptable. ❝ If we like to add the sequential designs into bear, we have to implement the methods: to evaluate BE at Stage 1 (alpha = 0.0294) and to calculate sample size based on Stage 1 and alpha = 0.0294, and finally to evaluate BE at Stage 2 using data from both stages (alpha = 0.0294). In principle, yes. It is important that you don’t fall into the trap of calculating the sample size based on the point estimate of Stage 1. Potvin’s method is not a full adaptive design; it was not validated to adjust for the effect size, but only for the unknown variance of Stage 1. If you planned the first Stage for an expected point estimate of 95% and get only 90% (or even more tempting 100%), you must not use this value in sample size estimation, but the original one! You may simply be caught by random walks. Another important point: There’s no futility rule in the method (the sample size estimation may come up with 3142 subjects for Stage 2). Ethics committees may be surprised by that, because they are familiar with ‘early stopping-rules’ in clinical trials. Again, the method was not validated for such a rule. There’s only the possibility for the sponsor to pull the ripcord based on non-statistical grounds. ❝ The alpha level is not commonly used in statistics. However it seems feasible, I guess. Oh yes, it is quite common in clinical trials. Actually 0.0294 were first proposed by Armitage (1975) and Pocock (1977), I guess. This value is the one with longest tradition, but there are many, many others as well. For some links, see this post. If you plan to implement the method in bear, don’t forget to modify the model if the study is evaluated after pooling (see this thread). See also the last sentence of the Two-stage design Section of EMA’s GL: When analysing the combined data from the two stages, a term for stage should be included in the ANOVA model. — 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 |
|
yjlee168 ★★★ Kaohsiung, Taiwan, 2010-04-01 20:58 (5523 d 01:41 ago) @ Helmut Posting: # 4998 Views: 12,962 |
|
|
Dear Helmut, The following message was supposed to be posted here yesterday. However, I just could not submit it after I finished it. The forum was hanging there without any response. I wished that I did not do anything stupid at that moment. --- Thank you for your explanations. Very nice comments. ❝ entire alpha-risk in the interim looks. In other words - you must not evaluate the study for BE (i.e., calculate the 90% CI and check My guess is this procedure will not be done unless the power is ≥ 80% (Method C, as the left branch of the flowchart). Do you mean that we should not do this when the power is < 80%? If doing so, I guess the answer is still not BE, isn't it? The question can be if the regulatory agents accept BE when the power < 80%, and evaluate BE at stage 1 with alpha = 0.0294 (Method C, as the left sub branch of the right branch) when we analyze the data as the fixed-sample trial. BTW, I don't know what possibility of getting BE will be with alpha = 0.0294, when it has been not BE with alpha = 0.05. It is more stringent (0.0294 vs. 0.05), isn't it? ❝ for inclusion). If you do that, your entire alpha-risk has gone - nothing left for further looks. This was actually the problem with the Canadian and Japanese method. What kind of the problem with the Canadian and Japanese method can be? ❝ Yes, according to Potvin et al. It's important to notice that the value calculated here is not some kind of a posteriori power... Yes, the calculation equation for the power is different and complicated. ❝ 'round: If you just miss the 80% (let's say you had one drop-out; power 79.1%). Power for alpha 0.0294 is 69.3% and that's still a pretty good chance that we will show BE at Stage 1 and stop. Only if we fail here, we will advance to Stage 2. Smart choices. ❝ (literature data only or small pilot study), would you really want to 'save' 10% of the budget and end up with an upper CI of 125.94%. Very persuasive. only extra 10% of the total costs of the fixed-sample design? or it depends how many subjects should be recruited at stage 2? ❝ In principle, yes. It is important that you don't fall into the trap of calculating the sample size based on the point estimate of Stage 1... O.k.. May I ask why WNL does not implement the data analysis for the two-staged design if FDA/EU have been able to accept the two-staged designs? or it has been implemented with WNL (v6.x)? ❝ this thread). See also the last sentence of the Two-stage design Section of EMA's GL: When analysing the combined data from the two stages, a term for stage should be included in the ANOVA model. This should be no problem with lm() function with R (as PROC GLM with SAS) using stage, sequence, period(stage), trt, and subj(sequence*stage) effects in the model (p. 253 in Potvin D, et al., 2008). — All the best, -- Yung-jin Lee bear v2.9.2:- created by Hsin-ya Lee & Yung-jin Lee Kaohsiung, Taiwan https://www.pkpd168.com/bear Download link (updated) -> here |
|
Helmut ★★★ Vienna, Austria, 2010-04-02 23:00 (5521 d 23:39 ago) @ yjlee168 Posting: # 5003 Views: 13,074 |
|
|
Dear bears, ❝ The forum was hanging there without any response. I wished that I did not do anything stupid at that moment. Terribly sorry about that. It wasn’t your fault. The forum is hosted on a shared server, and another user installed a faulty script which filled up the server’s error logs until no space was left on the HD. Then the MySQL-database crashed… ❝ ❝ entire alpha-risk in the interim looks. In other words - you must not evaluate the study for BE (i.e., calculate the 90% CI... ❝ ❝ My guess is this procedure will not be done unless the power is ≥ 80% (Method C, as the left branch of the flowchart). Right. According to this post of yours I was afraid that you want to evaluate BE in any case. ❝ Do you mean that we should not do this when the power is < 80%? Yes. Or to be more precise, not with the conventional alpha=0.05. ❝ If doing so, I guess the answer is still not BE, isn't it? Haha, now you are cheating. You should not evaluate BE at 0.05, otherwise the alpha is ‘consumed’.❝ The question can be if the regulatory agents accept BE when the power < 80%, and evaluate BE at stage 1 with alpha = 0.0294 (Method C, as the left sub branch of the right branch) when we analyze the data as the fixed-sample trial. Why not? It was shown in the simulations that it works; at least EMA accepts it. ❝ I don't know what possibility of getting BE will be with alpha = 0.0294, when it has been not BE with alpha = 0.05. It is more stringent (0.0294 vs. 0.05), isn't it? Yes, but that’s the trick. If the decision tree leads you to the right branch (based on power), you don’t evaluate at 0.05, only at 0.0294. Of course it’s more strict, but that’s the penalty in sequential designs. ❝ ❝ left for further looks. This was actually the problem with the Canadian and Japanese method. ❝ ❝ What kind of the problem with the Canadian and Japanese method can be? See the introduction of Potvin’s paper: Statistically, add-on designs cannot preserve the nominal type I error rate if a test is conducted at the nominal level following the completion of the initial trial and then again after the additional subjects are included. (WHO does indicate that the level of consumer risk must be considered; Canada does not.) The argument in favor of add-on designs is that they do make use of the data already collected, and the inflation of the type I error rate is ‘acceptable.’ In other words, the entire patient's risk is already consumed in evaluating the first part for conventional BE (90% CI). If you go for a second part then, what’s the overall risk? Definitely >0.05 (multiplicity!). How large the combined patient’s risk is, is actually unknown (depends on the sizes of the two parts, etc.). Method A in the simulations (which was already more stringent than Canada's/Japan's method) lead to a risk of 6% and was not calculated in all combinations. ❝ only extra 10% of the total costs of the fixed-sample design? or it depends how many subjects should be recruited at stage 2? Both. The ~10% is an estimate for Gould (1995). That’s the penality for the interim look. But on the long run (many studies) sequential designs may be more economical, because the comparison with fixed-sample designs is based on a delusion – namely that the variance is known and the sample size optimal. If the variance is uncertain, and lower than expected, sequential design will stop at the early stage. ❝ May I ask why WNL does not implement the data analysis for the two-staged design if FDA/EU have been able to accept the two-staged designs? or it has been implemented with WNL (v6.x)? Pharsight runs a commercial operation. If they don’t see additional profit, they won’t implement anything. Users have to ask them. Nothing is implemented in v6.1, but it would at least help to publish an application note on the support site (or the new board).Right now there’s no way to get the CV without running the BE-wizard. Another problem starts when it comes to calculating power for the first decision. AFAIK WNL’s result (all versions) is simply wrong (see here). Now for the good part: it is possible to calculate the narrower CI and set up Potvin's model for the combined analysis. ❝ This should be no problem with […] R Sure, R ulez! You even may combine the entire procedure, because power/sample size calculation is possible within bear (no way in WinNonlin). — 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