Log in
Register|
Frieda ☆ 2008-08-01 16:36 (6126 d 00:56 ago) Posting: # 2132 Views: 10,346 |
|
|
Dear All, Thanks in advance for any help in this problem which I do not know how to deal with! We have done a cross-over study in 24 subjects to evaluate the effect of co-administration of drug B on the bioavailability of A. The purpose of the study was to hopefully demonstrate that B intake does not affect exposure to A, and thus no efficacy studies will be required for development of a combination product. The subjects received A + placebo and A + B in randomized order for several weeks, with a wash-out of several weeks between the treatment periods. The combination A+B is hereafter called the test T and A + placebo the reference R. Samples were drawn on Day 1 of each period; troughs to evaluate attainment of steady state and complete profiles after the last dose of each period. We have gone through a lot of effort to ensure medication intake and believe that the study was properly designed. Initially one group of 10 subjects and another group of 14 were to be enrolled, as 24 subjects in one go was not feasible. Treatment order was randomized and blocked for these two groups. But eventually due to recruitment problems one group of 8 subjects and one group of 16 subjects was enrolled, three weeks apart. In the group of 8 subjects 5 got the order R-T; 2 got the order T-R and one (T-R) dropped out. In the group of 16 subjects 7 got the order R-T; 7 got the order T-R and two (T-R) dropped out. So in all, 12 got the order R-T and 9 the order T-R (three T-R's dropped out, unrelated to the treatment). The ratio T to R for AUCtau on the last dosing day (n=21) was 0.89, 90% CI 0.85 to 0.94; similar values were obtained for Cmax. Although the 90%CI falls within 0.80-1.25, 1.0 is not included so I looked at the data in more detail, mostly plots and geometric means for the two treatment sequences, and this seemed to point at a sequence effect. The group receiving sequence R-T (n=12) had a perfect geometric mean AUCtau ratio (T to R) of 1.01, minimum individual ratio 0.80, maximum was 1.26 (I am not making this up (  ).The group receiving sequence T-R (n=9) had a geometric mean AUCtau ratio (T to R) of 0.78, minimum individual ratio 0.68, maximum was 0.98. The low ratios for the group receiving T-R (n=9) appeared to be due to a raised AUCtau for the reference treatment in Period 2. Geometric mean AUCs for the test in Period 1 (n=9) or test in Period 2 (n=12) and for the reference in Period 1 (n=12) were virtually the same; the only AUC that was substantially higher was R in Period 2 (n=9). Then I looked at the Day 1 data and the troughs and all these profiles indicated a higher exposure in the nine subjects receiving the Reference in Period 2. The wash out was such that a carry-over effect was not to be expected, so we seem to be left with an unexplained sequence effect. As I have not encountered this before, and am not much of a statistician, any help is appreciated. Do we need to go into all this detail in the study report or just stick to the fact that technically speaking the treatments are bioequivalent? Is there any way to find out by whatever statistical test what has been going on? How will regulatory agencies look upon such a study result? Should we go talk with them? Helmut Schuetz has already kindly offered suggestions and thought it would be interesting for all and suggested to put my problem on the forum. One other explanation could be the following, far-fetched maybe as too many assumptions all need to be true at the same time: 1) there is a Period effect (which may for instance have to do with different food intake in Autumn and in Winter) leading to a higher exposure in Period 2 for both sequences; 2) co-administration of B does reduce the exposure to A, irrespective of sequence 3) exposure to A is intrinsically higher in the n=9 subjects than in the n=12 subjects due to inter-subject variability (CV's were appr. 30-50%). Then the group R-T (n=12) has a raised exposure to A in Period 2 due to 'winter' and a reduced exposure due to B intake which levels out 'winter' and R and T come out the same (as they did). The group T-R (n=9) has a raised exposure to A in Period 2 due to 'winter' which is not counteracted by B, so a high R, and a reduced exposure in Period 1 due to B (Test). However, since they had a higher intrinsic exposure to A to start with, their exposure to test in Period 1 comes out the same as for the n=12 in Period 2 and also the same as R for n=12 in Period 1. Does this make any sense? Thanks in advance for any thoughts or suggestions, Frieda |
|
ElMaestro ★★★ Denmark, 2008-08-01 18:36 (6125 d 22:57 ago) @ Frieda Posting: # 2133 Views: 9,013 |
|
|
Hi, there's a lot to think of here. Perhaps I got somethin wrong, but I think you have another factor possible in your anova: Time of Recruitment with levels "early" (n=7) or "late" (n=14). Did you try and run an anova with this extra factor? Does it make sense? What were the results like? Far-fetched and utterly crackheaded hypothesis: B has a long-term effect that causes a decrease in intestinal motility (or a decrease in metabolising enzyme activity or whatever); nothing is seen in Seq RT. But in TR the transit time (or elim. rate or whatever) for the tablet is longer (decreased) in P2 becuase of the effect from B in period 1. Therefore AUC gets higher for R in this seq. Madness? Is B known to induce any enzymes? EM. |
|
Frieda ☆ 2008-08-06 11:57 (6121 d 05:35 ago) @ ElMaestro Posting: # 2152 Views: 8,843 |
|
|
Hi EM Thanks for you response! Meanwhile I have run the ANOVA again with group in it, and this did not make a whole lot of difference, but thanks for the suggestion anyway. One other helper has looked at the data, and there is both a sequence and a period effect (p<0.05), but although statistically significant, not clinically relevant because the study still passes. You hypothesis is not crackheaded, it is not impossible that something has happened, although by having this long washout, we thought to have avoided that. I guess I wil never know. But there has been sufficient reassurance that the study will pass as is, and suggestions how to word it the report have helped as well! Frieda |
|
Helmut ★★★   Vienna, Austria, 2008-08-01 20:58 (6125 d 20:34 ago) @ Frieda Posting: # 2134 Views: 8,927 |
|
|
Dear Frieda! ❝ […] The combination A + B is hereafter called the test T and A + placebo the reference R. I kept this line as a reference.  PK interaction studies are similar to BE studies, but not equal. Let's start with the fundamental pharmacokinetic relationship fa x D = AUC x CLor, rewritten fa × D AUC = ———— CLIn BE testing we are relying on two assumptions (unfortunatelly many people are not aware of #1): (1) CL remains constant during the study (2) D is the same (no potency correction)For two formulations (T, R) we have AUCT fT × DT × CLR ——— = ———————— AUCR fR × DR ×CLTapplying assumptions (1) and (2) we get AUCT frel = ——— AUCRIn a PK interaction study assumption #1 (constant clearances) may not hold! Contrary to BE studies, where the AUC-ratio is only a measure of the absorption characteristics of formulations, the concomitant treatment of drug B may influence the clearance of drug A (AUC therefore is a composite parameter consiting of extent of absorption and clearance). See the pioneering paper published in 1994 by Schall et al. [1] and the chapter ‘Analysis of pharmacokinetic interactions’ in Hauschke et al. (2007).[2]If the CI falls with the commonly applied acceptance range (AR) ‘lack of interaction’ is likely (because a ratio close to unity would imply either no influence of T on fR and CLR or an influence on both fR and CLR in opposite directions to the same degree – which is considered improbable). The authors suggest in any case no further analyses, but to include such an possible influence on the clearance in the wording of the conclusion, like ‘the amount of drug A available in the systemic circulation is not affected by concomitant administration of drug B’rather than ‘drug B does not affect the absorption of drug A’.If an influence is observed (CI not within AR), the authors suggest a second step in the analysis, namely (1) the elimination half life of drug A, (2) the ratio AUC/t½ (or AUC × kel).The second metric was also proposed by Abdallah (1998) [3] in another context (to ‘reduce variability’ of HVDs).This is only an entry point from my side; though you didn’t see residual concentrations after the washout, some residual effects of T on the PK cannot be excluded (higher concentration of the reference in the TR group - this was also ElMaestro’s idea in his post). IMHO I would go with the second part of the analysis notwithstanding you have already shown a CI within the AR. Anyhow, from both clinical and regulatory points of view for a combination product any demonstrated PK interaction will go into the labeling (US) / SmPC (EU). Must sleep over it… 
— 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 |
|
Frieda ☆ 2008-08-06 20:35 (6120 d 20:58 ago) @ Helmut Posting: # 2156 Views: 8,696 |
|
|
Dear Helmut, Thanks for your thoughts and the references! Sorry to be responding this late, was a bit sick for a couple of days. When we designed the study we did realise that there was more than just the absorbtion part, but since we compared AUCtau at SS all possible kinetic interactions (absorption, distribution and elimination) are in there, I think anyway. We did sample extensively (4-5 terminal half-lives) at the end of each period (just in case, and to demonstrate that thalf was not affected..) and I do have thalf data. So I will do as you advise, if only to satisfy my own curiosity. In the meantime I know through another 'helper' that there is a period and a sequence effect (p<0.05) despite the randomisation. Will check the subject of period effects on the forum to see if I can work that one out all by myself  !Best wishes, Frieda |
|
d_labes ★★★ Berlin, Germany, 2008-08-04 11:23 (6123 d 06:09 ago) @ Frieda Posting: # 2135 Views: 8,777 |
|
|
Dear Frieda, ❝ The ratio T to R for AUCtau on the last dosing day (n=21) was 0.89, 90% CI 0.85 to 0.94; similar values were obtained for Cmax. Although the 90%CI falls within 0.80-1.25, 1.0 is not included so I looked at the data in more detail, mostly plots and geometric means for the two treatment sequences, and this seemed to point at a sequence effect ... IMHO: What you do is to look for differences (and explanations for it) which you consider in your first test (the bioequivalence test using the 90% CIs) as not important enough and not contradictory to the equivalence hypothesis. This is illogical to me. I have seen in my carrier a lot of such analysis but in the context of failed superiority trials seeking for reasons of that and seeking for subgroups with effect present. All of them useless, at least in a regulatory sense. — Regards, Detlew |
|
Frieda ☆ 2008-08-06 15:49 (6121 d 01:43 ago) @ d_labes Posting: # 2155 Views: 8,730 |
|
|
Hi d_labes, Thanks for your thoughts. Indeed the study has passed and I would not have posted it if the ratio's had been slightly reduced for both sequences; would have accepted that B affects A to a clinically irrelevant degree. However, as the ratios are rather different for the two sequences, one of the two being a reduction of 21% and the other being no effect, I felt that this needed some more thought and hoped that some explanation might be found in the data. For one sequence there is a pretty consistent lower ratio, which might have clinical implications, even though the good ratio for the other sequence - and the sample size maybe - makes that the end result passes. I was wondering how to deal with that or if any explanation might/needs to be found to reassure me. The low ratio for one of the two sequences makes me feel a bit uncomfortable. Regards, Frieda |
|
Helmut ★★★ Vienna, Austria, 2008-08-10 16:25 (6117 d 01:07 ago) @ d_labes Posting: # 2165 Views: 8,736 |
|
|
Dear DLabes! ❝ IMHO: ❝ What you do is to look for differences (and explanations for it) which you consider in your first test (the bioequivalence test using the 90% CIs) as not important enough and not contradictory to the equivalence hypothesis. Isn’t the null-hypothesis inequivalence?  ❝ This is illogical to me. As already said, it’s not that simple in an interaction study, since clearance may be affected. See EMEA’s guideline on drug interactions (Section 5.1.1): In most in vivo pharmacokinetic studies it seems reasonable to focus on the exposure of the drug, AUC and the two variables determining this, i.e. extent of absorption, F, and clearance (CL). Other parameters may also be of importance such as Cmax and t½, especially if the safety issue is dependent on the pharmacological action of the product. — 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, 2008-08-11 16:50 (6116 d 00:42 ago) @ Helmut Posting: # 2166 Views: 8,593 |
|
|
Dear HS, eventually I missed your point. My practical experience regarding interaction studies is zero or below. But if I had understood your suggestion above in the thread, and the chapter in Hauschke et. al., it is enough in a first step of analysis to show that the AUC ratio is with the commonly applied acceptance range. Whatever the correct wording for this result may be. This was the case in Frieda's study. Just to cite her initial post: "The ratio T to R for AUCtau on the last dosing day (n=21) was 0.89, 90% CI 0.85 to 0.94; similar values were obtained for Cmax." Why to seek for an explanation that this result is wrong? This seems illogical to me futher on. (regardless of fiddling with the German purity law for )— Regards, Detlew |
|
Helmut ★★★ Vienna, Austria, 2008-08-11 18:33 (6115 d 22:59 ago) @ d_labes Posting: # 2167 Views: 8,602 |
|
|
Dear DLabes! ❝ My practical experience regarding interaction studies is zero or below. ^^^^^ ❝ […] and the chapter in Hauschke et. al., it is enough in a first step of analysis to show that the AUC ratio is with the commonly applied acceptance range. […] ❝ This was the case in Frieda's study. ❝ Just to cite her initial post: ❝ "The ratio T to R for AUCtau on the last dosing day (n=21) was 0.89, 90% CI 0.85 to 0.94; similar values were obtained for Cmax." Why to seek for an explanation that this result is wrong? ❝ This seems illogical to me further on. I still think, that her reactions were quite right. Though the result of the overall analysis were OK, she also heard all warning bells ringing: differences in groups, period and treatment effects. I also think it's not a bad idea to act beforehand (i.e., run some additonal analyses and discuss the outcome in the report), rather than be forced to react on a deficiency letter - which always upsets everybody and calls for stressed reactions. We all know the classical "Haven't you noticed that?"... ❝ (regardless of fiddling with the German purity law for Oh yeah, as an Austrian I know both laws quite well…  — 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, 2008-08-12 11:55 (6115 d 05:37 ago) @ Helmut Posting: # 2168 Views: 8,654 |
|
|
Dear HS! ❝ [...] I also think it's not a bad idea to act beforehand (i.e., run some additional analyses and discuss the outcome in the report), [...] Full ACK from the scientific view. Although the study was in most cases not planned for that additional analysis. ❝ [...] rather than be forced to react on a deficiency letter [...] This depends. Very heavily. On the assessors and their wisdom. To guess beforehand deficiencies stated by them is like getting the jackpot in lottery . Sometimes.— Regards, Detlew |
|
Helmut ★★★ Vienna, Austria, 2008-08-12 14:36 (6115 d 02:56 ago) @ d_labes Posting: # 2171 Views: 8,629 |
|
|
Dear Dlabes! A cordially ‘Yes’ to all your points! — 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