Bioequivalence and Bioavailability Forum

jag009 ★★★ NJ, 2015-09-22 01:19 (3465 d 23:24 ago) Posting: # 15434 Views: 19,663	New D. Guidance – Dabigatran [Regulatives / Guidelines] Post reply
	Hi folks, Any thoughts? Guidance John

nobody nothing 2015-09-22 13:04 (3465 d 11:39 ago) @ jag009 Posting: # 15438 Views: 17,324	New D. Guidance – Dabigatran Post reply
	Hi John! Any info on what has changed from 2012 draft to 2015 version? Regards nobody — Kindest regards, nobody

Helmut
★★★

Vienna, Austria,
2015-09-22 13:43
(3465 d 11:00 ago)

@ nobody
Posting: # 15439
Views: 18,236

The FDA’s fourth BE approach!

Post reply

Hi nobody,

❝ Any info on what has changed from 2012 draft to 2015 version?

Jun 2012

Design: Single-dose, two-way crossover in vivo
Additional Comments: Applicants may consider using a reference-scaled average bioequivalence approach for dabigatran etexilate. Please refer to Progesterone Capsules Guidance for information regarding statistical analysis method using the reference scaled average bioequivalence approach.
Bioequivalence based on (90% CI): free (non-conjugated) dabigatran

Sep 2015

Design: Single-dose, 2-treatment, 2-sequence, 4-period, fully replicated crossover in vivo
Additional comments: Dabigatran demonstrated a steep exposure-response relationship for both efficacy and safety. Therefore applicants should not use the reference-scaled average bioequivalence (BE) approach to widen the BE limits for dabigatran BE evaluation. Applicants should use the average BE approach, with BE limits of 80–125%. The within-subject variability of test and reference products should be compared and the upper limit of the 90% confidence interval for the test-to-reference ratio of the within-subject variability should be ≤ 2.5. For details about the Method for Statistical Analysis comparing within-subject variability of test and reference products, refer to Guidance on Warfarin Sodium.

All subjects should have prothrombin time (PT), activated partial thromboplastin time (aPTT), and creatinine clearance (CrCl) measured. The PT and aPTT should be within the normal range and the CrCl value should be more than 50 mL/min for each subject before dosing in order to prevent or avoid the possibility of bleeding.
Bioequivalence based on (90% CI): free (non-conjugated) dabigatran and total dabigatran (non-conjugated plus conjugated dabigatran)

Apart from the additional safety parameters for inclusion and the BE requirement of total dabigatran the statistics are interesting – going from RSABE (widening of implied limits) to sumfink similar to Warfarin. No narrowing of BE limits but still the comparison of variances.*

This gives yet a fourth BE approach stated by the FDA:

Generally average BE; acceptance range 80.00–125.00%. Designs: Two-period crossover, higher-order crossovers, parallel, full (4-period) or partial replicate (3-period).
RSABE for HVDs/HVDPs if \(\small{s_\text{wR}\geq 0.294}\). GMR-restriction 0.8000–1.2500. Designs full (4-period) or partial replicate (3-period).
RSABE for NTIDs. Narrowing BE-limits based on \(\small{s_\text{wR}}\). Must pass conventional ABE as well. Upper CL of \(\small{s_\text{wT}/s_\text{wR}\leq 2.5}\). Design full (4-period) replicate.
As above but conventional BE limits (no downscaling).

How could one estimate the sample size? Yet another method for the wish-list of PowerTOST?
You could work by trial-and-error; assessing p(BE-ABE) and p(BE-sratio):
library(PowerTOST) theta0 <- 0.95 # GMR n <- 16 # minimum mandated by the FDA CV <- 0.2 # average (pooled) CV ratio <- 1 # ratio of SEs CVs <- CVp2CV(CV = CV, ratio = ratio^2) print(CVs) power.NTIDFDA(theta0 = theta0, CV = CVs, n = n, design = "2x2x4", details = TRUE)
For a ratio of 1 I get p(BE-ABE)=0.95041 and p(BE-sratio)=0.95128. Fine.
The maximum ratio for 80% power seems to be 1.25 (CV_T 0.221, CV_R 0.176): p(BE-ABE)=0.95041 and p(BE-sratio)=0.80843.

—
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

nobody nothing 2015-09-22 15:06 (3465 d 09:37 ago) @ Helmut Posting: # 15442 Views: 17,273	The FDA’s fourth BE approach! Post reply
	...nice question for statistics exam: Please elucidate the scientific background for the "upper limit of the 90% confidence interval for the test-to-reference ratio of the within-subject variability should be ≤ 2.5" of FDA dabigatran BE guidance Just kidding! — Kindest regards, nobody

Helmut
★★★

Vienna, Austria,
2015-09-22 15:25
(3465 d 09:18 ago)

@ nobody
Posting: # 15443
Views: 17,289

The FDA’s fourth BE approach!

Post reply

Hi nobody,

❝ Please elucidate the scientific background for the "upper limit of the 90% confidence interval for the test-to-reference ratio of the within-subject variability should be ≤ 2.5" of FDA dabigatran BE guidance

As if pulling a rabbit out of the hat. László Endrényi regularly goes ballistic on this issue. He showed already twenty years ago (in the context of IBE) that this limit is arbitrary and the “method” flawed.

—
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

jag009
★★★

NJ,
2015-09-23 22:29
(3464 d 02:14 ago)

@ Helmut
Posting: # 15461
Views: 16,997

The FDA’s fourth BE approach! Yada Yada Yada

Post reply

Hi Helmut,

❝ As if pulling a rabbit out of the hat. László Endrényi regularly goes ballistic on this issue. He showed already twenty years ago (in the context of IBE) that this limit is arbitrary and the “method” flawed.

I was in one local BA/BE meeting last week in NJ. One of the ex-FDA presenter (consultant) made a short "blasting" speech on the Sub*formulation interaction test on Concerta... I wonder when my old prof Laszlo will start his own campaign....

John

Helmut
★★★

Vienna, Austria,
2015-09-24 00:06
(3464 d 00:37 ago)

@ jag009
Posting: # 15464
Views: 16,878

Yucky…

Post reply

Hi John,

I like how you modified the subject line. ;-)

❝ I was in one local BA/BE meeting last week in NJ. One of the ex-FDA presenter (consultant) made a short "blasting" speech on the Sub*formulation interaction test on Concerta...

I see! Which kind of eternal wisdom did he segregate?

❝ I wonder when my old prof Laszlo will start his own campaign....

I talked to him a couple of times on this. It’s on his to-do-list…

—
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,
2015-09-22 16:08
(3465 d 08:35 ago)

@ Helmut
Posting: # 15447
Views: 17,193

The FDA’s fourth BE approach!

Post reply

Dear Helmut, dear all,

Interesting. Seems another Mighty Oracle that went into labour and delivered a mouse.

❝ How could one estimate the sample size? Yet another method for the wish-list of PowerTOST?

Your wish is my command!
Should not that hard to modify power.NTIDFDA() to not shrinken the acceptance range. But in the mean time I suffer from names :cool:

for the functions in PowerTOST.
How should we baptize the new baby? Or better having an additional argument in the already mature power.NTIDFDA()? If yes, how to name this?

—
Regards,

Detlew

Helmut
★★★

Vienna, Austria,
2015-09-22 16:27
(3465 d 08:16 ago)

@ d_labes
Posting: # 15448
Views: 17,368

New PowerTOST function(s)

Post reply

Dear Detlew,

welcome back! How was vacation on the lonesome isle far out on the high seas?

❝ Interesting. Seems another Mighty Oracle that went into labour and delivered a mouse.

❝ ❝ How could one estimate the sample size? Yet another method for the wish-list of PowerTOST?

❝

❝ Should not that hard to modify power.NTIDFDA() to not shrinken the acceptance range. But in the mean time I suffer from names :cool: for the functions in PowerTOST.

❝ How should we baptize the new baby? Or better having an additional argument in the already mature power.NTIDFDA()? If yes, how to name this?

I would not add other argument(s) to power.NTIDFDA(). AFAIK dabigatran is actually pretty highly variable (see here: C_max 59.9%, AUC_∞ 52.9% and the Australian PAR). BTW, if you add arguments you would have to change the default theta0 from 0.975 (to 0.95 or even 0.90?).
Will [image]

confuse users. I would suggest new functions like power.RatioSFDA() and sampleN.RatioSFDA() instead. Let’s ask Ben as 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,
2015-09-22 16:59
(3465 d 07:44 ago)

@ Helmut
Posting: # 15449
Views: 17,313

New PowerTOST function(s)

Post reply

Dear Helmut,

❝ welcome back! How was vacation on the lonesome isle far out on the high seas?

Thanks! It was very recreative. And I could practise what I have to do if I will be a rendeer (German Rentier) next time.
Doing nothing the whole blessed day but "Maulaffen feil halten" (to gawk, to sit on one's hands) :cool:

❝ I would not add other argument(s) to power.NTIDFDA(). AFAIK dabigatran is not a NTID.

That astonishes me. I had interpreted the sentence "Dabigatran demonstrated a steep exposure response relationship for both efficacy and safety." as some sort of NTID.

❝ Let’s ask Ben as well.

Good idea. I will wait for his reaction.

BTW: Do you think this sort of BE evaluation is not an "Eintagsfliege" (nine day wonder, ephemera), specific for dabigatran?

—
Regards,

Detlew

Helmut
★★★

Vienna, Austria,
2015-09-22 17:16
(3465 d 07:27 ago)

@ d_labes
Posting: # 15450
Views: 17,872

Pradaxa = HVDP (dabigatran = HVD?)

Post reply

Dear Detlew,

❝ It was very recreative. And I could practise what I have to do if I will be a rendeer (German Rentier) next time.

❝ Doing nothing the whole blessed day but "Maulaffen feil halten" (to gawk, to sit on one's hands) :cool: .

That’s not the worst thing one can do!

❝ ❝ I would not add other argument(s) to power.NTIDFDA(). AFAIK dabigatran is not a NTID.

❝

❝ That astonishes me.

See the linked references I added in the meantime to my previous post.

❝ I had interpreted the sentence "Dabigatran demonstrated a steep exposure response relationship for both efficacy and safety." as some sort of NTID.

Me too. Seems that the innovator (Boehringer Ingelheim) assessed the variabilities in its studies.
This is the first example I’m aware of where the observation “HVDs/HVDPs are safe drugs since they have a flat dose/response-curve” seemingly does not hold.
Another question comes into my mind: How the heck did Boehringer’s formulation “survive” phase III and/or bridging from the formulation used in phase III to the marketed formulation?
:confused:

A dirty story? Wikipedia:

On July 26, 2014, the British Medical Journal (BMJ) published a series of investigations that accused Boehringer of withholding critical information about the need for monitoring to protect patients from severe bleeding, particularly in the elderly. Review of internal communications between Boehringer researchers and employees, the FDA and the EMA revealed that Boehringer researchers found evidence that serum levels of dabigatran vary widely. The BMJ investigation suggested that Boehringer had a financial motive to withhold this concern from regulatory health agencies because the data conflicted with their extensive marketing of dabigatran as an anticoagulant that does not require monitoring.^1,2

❝ BTW: Do you think this sort of BE evaluation is not an "Eintagsfliege" (nine day wonder, ephemera), specific for dabigatran?

No idea. Let’s ask generic companies to follow the path the innovator walked?

Cohen D. Dabigatran: how the drug company withheld important analyses. BMJ. 2014;349:g4670. doi:10.1136/bmj.g4670. free resource.
Moore TJ, Cohen MR, Mattison DR. Dabigatran, bleeding, and the regulators. BMJ. 2014;349:g4517. doi:10.1136/bmj.g4517.

—
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

Shuanghe
★★

Spain,
2015-09-22 18:53
(3465 d 05:50 ago)

@ Helmut
Posting: # 15451
Views: 17,394

Pradaxa = HVDP (dabigatran = HVD?)

Post reply

Hi all,

❝ ❝ BTW: Do you think this sort of BE evaluation is not an "Eintagsfliege" (nine day wonder, ephemera), specific for dabigatran?

❝

❝ No idea. Let’s ask generic companies to follow the path the innovator walked?

There's another guidance of rivaroxaban which takes the same approach as dabigatran.

Unfortunately, I've done BE studies for both dabigatran and rivaroxaban and one of the dossiers have been submitted. Now I have to do some additional analysis according to the new requirements and waiting for the deficient letter to come. :-(

By the way, dabigatran is highly variable under fasting condition but not under fed condition. At least based on my 6 studies (EU/US, pilot/pivotal), all fed study showed ISCV < 30% while all fasting study showed ISCV > 40%.

As for rivaroxaban, it's high variable only for 20mg under fasting. under fed, not so much.

—
All the best,
Shuanghe

d_labes ★★★ Berlin, Germany, 2015-09-23 10:34 (3464 d 14:09 ago) @ Shuanghe Posting: # 15455 Views: 16,963	HVNTID's Post reply
	Dear Shuanghe, ❝ There's another guidance of rivaroxaban which takes the same approach as dabigatran. THX for this information. So it seems it is worth the efforts to implement this BE approach in `PowerTOST`. — Regards, Detlew

Shuanghe ★★ Spain, 2015-09-25 18:14 (3462 d 06:29 ago) @ d_labes Posting: # 15469 Views: 16,341	HVNTID's Post reply
	Dear Detlew, ❝ ... So it seems it is worth the efforts to implement this BE approach in `PowerTOST`. Wow! That's something worth waiting for!! — All the best, Shuanghe

d_labes
★★★

Berlin, Germany,
2015-09-23 10:23
(3464 d 14:20 ago)

@ Helmut
Posting: # 15454
Views: 17,310

Highly variable NTID

Post reply

Dear Helmut,

❝ ❝ I had interpreted the sentence "Dabigatran demonstrated a steep exposure response relationship for both efficacy and safety." as some sort of NTID.

❝

❝ Me too. Seems that the innovator (Boehringer Ingelheim) assessed the variabilities in its studies.

❝ This is the first example I’m aware of were the observation “HVDs/HVDPs are safe drugs since they have a flat dose/response-curve” seemingly does not hold.

Oh I see. That's the reason for the sentence "Therefore applicants should not use the reference-scaled average bioequivalence (BE) approach to widen the BE limits for dabigatran BE evaluation" in the dabigatran guidance.

Seems it's time for a new term: HVNTID :cool:

❝ A dirty story? Wikipedia ...

A really dirty story :vomit:

. So much to "To cheat is Indian" (My exculpation to the Indian members of this forum).

—
Regards,

Detlew

Helmut
★★★

Vienna, Austria,
2015-09-23 15:20
(3464 d 09:23 ago)

@ d_labes
Posting: # 15460
Views: 17,095

What a mess!

Post reply

Dear Detlew,

❝ ❝ This is the first example I’m aware of were the observation “HVDs/HVDPs are safe drugs since they have a flat dose/response-curve” seemingly does not hold.

❝ ❝ A dirty story? Wikipedia ...

❝

❝ A really dirty story :vomit: .

Indeed! On another note I don’t understand why the FDA came up with their “BE approach #4”. It would have been sufficient to state “apply the NTID-approach given in the Warfarin guidance”. With the reported CVs (@Shuanghe: Did you ever observe one <21.42%?) downscaling would be unlikely. The sample size would be driven by part of the requirement “must pass 80.00–125.00%”. Sample sizes (80% power, 4-period full replicate, CV_T=CV_R):

expected GMR CV 0.90 0.95 0.25 32 18 0.30 40 22 0.35 54 28 0.40 68 34 0.45 84 42 0.50 102 50 0.55 120 60 0.60 140 68

However; I guess the FDA wanted to play it ”safe”. But: Given that for dabigatran no dose-titration and TDM – as for Warfarin – is (falsely?) suggested by Boehringer (despite the observed bleeding-rates!) I would classify it as a NTID and ask even for the EMA’s more strict limits of 90.00–111.11% regardless the variability. Sample sizes (80% power, GMR 0.95, CV_T=CV_R):

design CV 2×2×2 2×2×4 0.25 258 130 0.30 366 184 0.35 492 246 0.40 630 316 0.45 782 392 0.50 946 474 0.55 1120 560 0.60 1302 652

In this post I linked to Boehringer’s ‘world record’ study. Incidentally it was dabigatran… Unfortunatelly the pranial state was not given (fasting?). Did not pass FDA’s RSABE for total dabigatran (primary) and conventional ABE for free dabigatran (secondary).
Total dabigatran

metric  upper 95% lin. crit.   PE

AUCt     -0.082 (pass; ≤0)   1.2633 (fail; outside 0.8000–1.2500)

Cmax     -0.085 (pass)       1.2549 (fail)

No RSABE-analysis was presented for AUC_∞; failed ABE (90% 118.32–131.51).
Free dabigatran

metric    PE       90% CI

AUCt    126.22  119.36–133.47 (fail)

AUC∞    124.49  118.11–131.21 (fail)

Cmax    125.30  118.42–132.59 (fail)

The FDA’s 2012 guidance recommended RSABE for free dabigatran only – which was not presented by Boehringer on ClinTrials.gov. Nevertheless, given the PEs of AUC_t and C_max the study would have failed to show RSABE. The highest ISCV was observed for C_max of free dabigatran with ~48%. Were the primary and secondary targets switched by Boehringer in May 2012 (from free to total dabigatran)?

My question from above

❝ ❝ How did Boehringer’s formulation “survive” phase III and/or bridging from the formulation used in phase III to the marketed formulation?

still holds. :angry:

I expect that the FDA’s guidance for yet another NOAC (apixaban) will be revised as 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,
2015-09-24 11:18
(3463 d 13:25 ago)

@ Helmut
Posting: # 15465
Views: 16,824

Messy numbers - PowerTOST V1.3-01 preview

Post reply

Dear Helmut,

❝ ... On another note I don’t understand why the FDA came up with their “BE approach #4”. It would have been sufficient to state “apply the NTID-approach given in the Warfarin guidance”. Given the reported CVs [...] downscaling would be unlikely. The sample size would be driven by part of the requirement “must pass 80.00–125.00%”. Sample sizes (80% power, 4-period full replicate, CV_T=CV_R):

IMHO totally correct. Lets compare NTIDFDA to HVNTID via the upcoming PowerTOSTV1.3-01 new function sampleN.HVNTID():

GMR= 0.90 GMR= 0.95 CV NTIDFDA HVNTID NTIDFDA HVNTID ------------------------------------ 0.20 36 20 18 14 0.25 32 30 18 18 ------------------------------------ 0.30 40 40 22 22 0.35 54 54 28 28 0.40 68 68 34 34 0.45 84 84 42 42 0.50 102 102 50 50 0.55 120 120 60 60 0.60 140 140 68 68

As expected: Identical numbers for higher variabilities, nearly totally driven by the ABE requirement :cool:

.
For instance GMR=0.95, n from table above, power.NTIDFDA(..., details=TRUE)

CV p(BE) p(BE-sABEc) p(BE-ABE) p(BE-sratio) 0.3 0.82570 0.95213 0.83752 0.99070 0.4 0.80727 0.99704 0.80747 0.99974 0.5 0.80496 0.99993 0.80496 1.00000 0.6 0.80286 1.00000 0.80286 1.00000 p(BE-sABEc) = prob. of scaled ABE test alone p(BE-ABE) = prob. of ABE test alone p(BE-sratio)= prob. of test of variabilities alone

Of course HVNTID is simpler. No need to deal with "linearized scaled ABE criterion" and it's 95% upper CI.

—
Regards,

Detlew

Shuanghe ★★ Spain, 2015-09-25 18:32 (3462 d 06:11 ago) @ Helmut Posting: # 15470 Views: 16,469	What a mess! Post reply
	Dear Helmut, ❝ ... Given the reported CVs (@Shuanghe: Did you ever observe one <21.42%?) downscaling would be unlikely. For AUC, yes. as low as 15% under fed, but for C_max, my lowest ISCV of R is about 24% under fed. ❝ I expect that the FDA’s guidance for yet another NOAC (apixaban) will be revised as well. I'd expect it the same as those two. — All the best, Shuanghe

Helmut
★★★

Vienna, Austria,
2015-09-23 02:10
(3464 d 22:33 ago)

@ d_labes
Posting: # 15453
Views: 17,065

OT: power.NTIDFDA()

Post reply

Dear Detlew,

❝ […] the already mature power.NTIDFDA()?

Another suggestion for power.NTIDFDA(). Consider returning a list of the three ps if details=TRUE instead of

cat("p(BE-ABE)=", p["BEabe"],"; p(BE-SABEc)=", p["BEul"], "; p(BE-sratio)=", p["BEsratio"],"\n") cat("\n")

inside the function. Would allow to automate stuff at the end of this post for advanced users (say, solve for a maximum ratio given a sample size, etc).

—
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, 2015-09-23 11:37 (3464 d 13:06 ago) @ Helmut Posting: # 15457 Views: 16,884	OT: power.NTIDFDA() Post reply
	Dear Helmut, as already stated: Your wish is my command . — Regards, Detlew

New D. Guid­ance – Dabigatran [Regulatives / Guidelines]

New D. Guid­ance – Dabigatran

The FDA’s fourth BE approach!

The FDA’s fourth BE approach!

The FDA’s fourth BE approach!

The FDA’s fourth BE approach! Yada Yada Yada

Yucky…

The FDA’s fourth BE approach!

New PowerTOST function(s)

New PowerTOST function(s)

Pradaxa = HVDP (dabi­ga­tran = HVD?)

Pradaxa = HVDP (dabi­ga­tran = HVD?)

HVNTID's

HVNTID's

Highly variable NTID

What a mess!

Messy numbers - PowerTOST V1.3-01 preview

What a mess!

OT: power.NTIDFDA()

OT: power.NTIDFDA()

New D. Guidance – Dabigatran [Regulatives / Guidelines]

New D. Guidance – Dabigatran

Pradaxa = HVDP (dabigatran = HVD?)

Pradaxa = HVDP (dabigatran = HVD?)