Log in
Register|
Mike_270 ☆ Australia, 2011-02-01 12:39 (5618 d 00:42 ago) Posting: # 6525 Views: 9,543 |
|
|
Hello everyone, Firstly I would like to say thankyou to the contributors to this forum – it has been an invaluable resource for me in learning more about BE and other statistical methods. I am writing about a clinical bioavailability study our research group plan to undertake using a drug that is known to have saturable absorption; we are investigating the effect of co-administering a range of doses of a nutritional supplement (until reaching steady state) with the drug using a crossover design as follows: A - Drug alone for 5 days B - Drug + 'x' mL supplement for 5 days C - Drug + 'x' mL × 2 supplement for 5 days D - Drug + 'x' mL × 4 supplement for 5 days Sampling would be carried out on the morning of Day 6 for each treatment (to calculate AUC over the dosing interval). Since the study will be carried out in patients and due to budget limits, we only expect to be able to include around 12 subjects. We expect that the supplement will increase bioavailability. I have read through some useful threads on the forum (#1368, #1423) and some very helpful lecture slides posted by HS (here and here). From these it seems that a Williams design for a four-treatment, four-period crossover is suitable (as drawn by HS in this post). This design results in 4 treatment sequences, with 3 patients assigned to each sequence. My question is how would I best calculate power based on this design and sample size? As we expect AUC will differ between treatments, and are not interested in testing Bioequivalence limits between treatments (not a regulatory study - only interested in comparing AUC's between treatments), I am unsure whether I can use sample size calculations intended for BE studies in this context. For reference the CVintra% I estimated using the FARTSSIE spreadsheet for AUC at steady state is 32.19%. Another question (possibly silly) is would we be better off using only a single treatment sequence (eg. ABCD) for all patients due to the small sample size and accept that a period effect may be present? Sorry for the long post and thanks in advance for any advice  Kind regards, Michael |
|
ElMaestro ★★★ Denmark, 2011-02-01 13:08 (5618 d 00:14 ago) @ Mike_270 Posting: # 6526 Views: 8,237 |
|
|
Hi Mike, one thing is not overly clear to me in this case: What are you actually trying to investigate (what are you hoping to show)? — Pass or fail! ElMaestro |
|
Mike_270 ☆ Australia, 2011-02-01 14:18 (5617 d 23:04 ago) @ ElMaestro Posting: # 6527 Views: 8,202 |
|
|
Hi ElMaestro, We are hoping to show that the nutritional supplement increases the absorption and thus bioavailability of the drug. Clinically this medication is known to have low bioavailability, often leading to low systemic exposure below the therapeutic target in patients and potentially reducing its efficacy. It has been shown that meals (particularly high fat meals) substantially increase the bioavailability, so we are hoping to see this effect with a nutritional supplement. We are also interested in examining if the different doses of the supplement affect the bioavailability of the drug to a different extent, so that if increased bioavailability is observed, we will be able to make a recommendation of how much supplement is needed to optimise the systemic availability of the drug. Thus we are interested in comparing the AUC's between the reference (drug alone) and the drug + differing doses of the supplement (treatments B - D). Michael |
|
ElMaestro ★★★ Denmark, 2011-02-01 15:24 (5617 d 21:58 ago) @ Mike_270 Posting: # 6528 Views: 8,224 |
|
|
Hi again, ❝ We are hoping to show that the nutritional supplement increases the ❝ absorption and thus bioavailability of the drug. ❝ (...) ❝ We are also interested in examining if the different doses of the ❝ supplement affect the bioavailability of the drug to a different extent If you have 'enough' subjects in your pivotal trial you may end up with a significant value for the ingredient addition. However, in order to dimension such a thing you will need to have an idea about the effect size the dietary supplement. In the absence of a model for it I guess you need to either conduct a pilot trial to get an idea about the effect size in your pivotal trial or press the 'I feel lucky'-button. — Pass or fail! ElMaestro |
|
Mike_270 ☆ Australia, 2011-02-02 01:47 (5617 d 11:35 ago) @ ElMaestro Posting: # 6533 Views: 8,215 |
|
|
Hi ElMaestro, ❝ In the absence of a model ❝ for it I guess you need to either conduct a pilot trial to get an idea ❝ about the effect size in your pivotal trial or press the 'I feel ❝ lucky'-button. Thankyou for the advice. In terms of randomisation, did you have any comment about using the Williams design 4 treatment sequence design vs. a single treatment sequence (eg. ABCD)? Is it always better to randomise the treatment sequences even with small patient numbers? Thanks again, Michael |
|
ElMaestro ★★★ Denmark, 2011-02-02 09:48 (5617 d 03:34 ago) @ Mike_270 Posting: # 6536 Views: 8,241 |
|
|
Hi Mike, ❝ In terms of randomisation, did you have any ❝ comment about using the Williams design 4 treatment sequence design vs. a ❝ single treatment sequence (eg. ABCD)? Is it always better to randomise the ❝ treatment sequences even with small patient numbers? There are pros and cons. Of course you get rid of a period effect by randomising and that in itself can be an advantage. On the other hand, if your drug has en evil safety profile then your dietary supplement may precipitate AEs or SAEs due to increased absorption. In that case it might be a good idea to consistently start low and go high higher. Specific ADME properties of the drug and effects of the diet supplement on metabolism will also play a role. Tell some more, please. — Pass or fail! ElMaestro |
|
d_labes ★★★ Berlin, Germany, 2011-02-02 11:36 (5617 d 01:46 ago) @ Mike_270 Posting: # 6540 Views: 8,220 |
|
|
Dear Mike, ❝ ... Since the study will be carried out in patients and due to budget ❝ limits, we only expect to be able to include around 12 subjects ... Whatever your final chosen statistical test method for your problem is (superiority t-test of treatment groups Bonferroni corrected, Dunnett test of treatment groups with supplement versus treatment with drug alone, regression analysis of AUC vs. 'dose' of supplement ...), I bet one of my monthly salary*)  that the power you calculate based on your chosen statistical test is low if sample size is 12, given your mentioned intra-subject CV of greater then 30%.What does this imply for you  ? Would you abandon your study because power is low? Or would / can you wait for a sponsor coming around next time with some million bucks in his long pocket to allow you to recruit enough patients to attain a higher power? Or would you go for it to investigate your question nevertheless? Because its so of interest or potential benefit for patients that it is worth to investigate even with low power, low sample size. One may argue that it is unethical to do such a study with low number of subjects. For that I recommend Stephen Senn's book "Statistical issues in drug development" Wiley 2007 Chapter 13.2.7 Let me cite: "The argument here is that one should not ask patients to enter a clinical trial unless one has a reasonable chance of finding something useful. Hence small or ‘inadequately powered’ trials are unethical. There is something in this argument. I do not agree, however, that small trials are uninterpretable and, as was explained in Section 13.2.2, sometimes only a small trial can be run. It can be argued that if a treatment will be lost anyway if the trial is not run, then it should be run, even if it is only capable of ‘proving’ efficacy where the treatment effect is considerable ... ... As Edwards et al. have argued eloquently, some evidence is better than none.". (Emphasis by me) *)BTW: You can't become rich because of that  .— Regards, Detlew |
|
Mike_270 ☆ Australia, 2011-02-02 12:41 (5617 d 00:41 ago) @ d_labes Posting: # 6541 Views: 8,175 |
|
|
Dear ElMaestro & d_labes, Thanks to you both for your posts. ❝ ElMaestro writes... ❝ Specific ADME properties of the drug and effects of the diet supplement on metabolism will also play a role. The drug of interest is primarily excreted unchanged, so we do not expect any change in metabolism to have a significant effect. ❝ if your drug has en evil safety profile then your dietary supplement may precipitate AEs or SAEs due to increased absorption This was also a concern of mine, although the drug has a relatively benign safety profile and most AEs and SAEs are not known to be dose related. ❝ d_labes writes... ❝ I bet one of my ❝ monthly salary*) ❝ your chosen statistical test is low if sample size is 12, given your ❝ mentioned intra-subject CV of greater then 30%. I have done some further research around that figure and I believe I may not have calculated it properly (it was my first time using the FARTSSIE spreadsheet) - one study I have now found has quoted CVintra of <10% for AUC. Admittedly this was in healthy volunteers so I think it would be higher in patients, but maybe still a fair bit better than 32%. ❝ Or would / can you wait for a sponsor coming around next time with some ❝ million bucks in his long pocket to allow you to recruit enough patients ❝ to attain a higher power? The current study budget is <2% of that  I am not the only investigator on the study, and my feeling is that it will probably go ahead with the current budget, it will just be a matter of how far we can stretch it.❝ One may argue that it is unethical to do such a study with low number of ❝ subjects. ❝ sometimes only a small trial can be run. It can be argued that if a ❝ treatment will be lost anyway if the trial is not run, then it should be ❝ run, even if it is only capable of ‘proving’ efficacy where the ❝ treatment effect is considerable ... I liked the quote - in terms of treatment effect, while we do not have a pilot study to characterise it, we expect it be fairly substantial. The drug is known to have a large food effect of around 2 - 4x increase in AUC (depending on meal composition), so we expect the supplement will considerably increase absorption. In addition, although not statistically sound, the supplement has been found to substantially improve the exposure of the drug in a small number of patients at our institution that have initially had poor absorption (not part of any study, supplement was given as part of normal patient care). Kind regards, Michael |
|
d_labes ★★★ Berlin, Germany, 2011-02-02 13:13 (5617 d 00:08 ago) @ Mike_270 Posting: # 6543 Views: 8,186 |
|
|
Dear Michael, ❝ The current study budget is <2% of that ❝ investigator on the study, and my feeling is that it will probably go ❝ ahead with the current budget, it will just be a matter of how far we can ❝ stretch it. Seems you need not a BOSS button but a accountancy button .I recommend you a sample size planning using a superiority t-test with Bonferroni corrected alpha. This will give you the highest sample size among alternative methods. Adjust the CV and the clinically relevant difference until you are comfortable with the number of subjects and you assume that your account staff will bite the bullet. Remember:
— Regards, Detlew |
Ing. Helmut Schütz