Bioequivalence and Bioavailability Forum • ANVISA and replicate designs

d_labes
★★★

Berlin, Germany,
2016-05-11 11:47
(3238 d 16:46 ago)

Posting: # 16291
Views: 13,456

ANVISA and replicate designs [Regulatives / Guidelines]

Dear All (especially our Brazilian colleges),

is there any more specific information what type of statistical evaluation ANVISA demands for replicate crossover designs?

I have only found the Resolution - RE n. 898, of May 29, 2003 which states in Section 3.2 Data analysis: "A variance analysis (ANOVA) must be used in the pharmacokinetic parameters ASC and Cmax using generalized linear models. Appropriate statistical models according to the design chosen in the study must be employed."

Only the case of classical 2x2x2 crossover is then exemplified in the next sentences.

Is the same ANOVA model demanded for replicate designs (like the EMA recommendation)?
Or is a mixed model approach the better choice (like Proc Mixed SAS code of the FDA)?
Is an analysis via intra-subject contrasts acceptable or recommended (like FDA's progesterone guidance)?

—
Regards,

Detlew

Mauricio Sampaio
★

Brazil,
2016-05-18 08:14
(3231 d 20:19 ago)

@ d_labes
Posting: # 16319
Views: 11,539

ANVISA and replicate designs

Post reply

Dear d_labes,

Unfortunatly, until the moment, the ANVISA did not show us a specific guideline published and official to explain how we should to proceed with replicate crossover designs. :-(

The RE 898 from 2003 is old and do not follow the advance of bioequivalence studies.

On the other hand, in some non official meetings, in regulatory terms, ANVISA clarified that accept the use of the statistical method for scaling proposed by the European Agency with the exception that for scaling can only be applied in the cases that the intra-subject coefficient of variation (CV%) exceeds 40% for the originator product.

As we do not have a new regulation, each protocol must be previously submitted to the technical evaluation of Therapeutic Equivalence Coordination (Anvisa) before conducting the bioequivalence study that you want to consider the scaling.

There are many view points as you can check in "Anvisa's home brewed scaling" posted by Helmut in this forum. Check it out!

During a plan to bioequivalence study for submission in Brazil, frequently I avoid replicate designs. Why? :confused:

Because two reasons:

First, in Brazil we do not have criteria or regulations established and everything can change according a new understanding from ANVISA. Black on white is much better! :yes:

Second, according Helmut's post (ANVISA’s home brewed scaling?), if ANVISA wants 40% as the switching condition, a very little gain in terms of the acceptance range could be observed! :surprised:

Best regards!

earlybird
☆

2016-05-19 18:10
(3230 d 10:23 ago)

@ d_labes
Posting: # 16334
Views: 11,551

ANVISA and replicate designs

Post reply

Dear D. Labes,

recently we got a letter from ANVISA with the following ...

"The protocol proposes the possibility to scale up Cmax values based on a within-subject coefficient of variation (CV%) higher than 40% for the reference drug.

More recently, Anvisa decided to follow the WHO instructions specified in “MULTISOURCE (GENERIC) PHARMACEUTICAL PRODUCTS: GUIDELINES ON REGISTRATION REQUIREMENTS TO ESTABLISH
INTERCHANGEABILITY. REVISION (JULY 2014)”, where the parameters to be followed in this kind of study are defined. We will accept parameters starting at CVWR > 30%. Therefore, please follow the guidelines above or EMA guidelines that have the same study assessment criteria, and the correct regulatory constant. Only Cmax intervals can be extended and the regulatory constant is K = 0.760. GMR should be within the 80%-125% range.
The correct study design for this case is the full replicate design, i.e. a 4-period design. Therefore, the study design is adequate for its purpose."

telling us to do the same as EMA.

Hope this helps,
Greatings from earlybird, Oberbayern, Bavaria

Helmut
★★★

Vienna, Austria,
2016-05-19 18:23
(3230 d 10:10 ago)

@ earlybird
Posting: # 16335
Views: 11,553

ANVISA = EMA’s ABEL!

Post reply

Hi earlybird,

that’s a fine piece of cake! We have to change the code of the next release of PowerTOST.
@Detlew: Only ANOVA also in Section 1.j) of the current RESOLUÇÃO – RE Nº 1.170, DE 19 DE ABRIL DE 2006

❝ The correct study design for this case is the full replicate design, i.e. a 4-period design.

The WHO’s Working Document does not specify which replicate designs are acceptable (contrary to the EMA mentioning 3- and 4-period replicates). IMHO, the 4-period replicate is the best design anyway…

—
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

Lucas
★

Brazil,
2016-05-19 20:32
(3230 d 08:01 ago)

@ Helmut
Posting: # 16336
Views: 11,467

ANVISA = EMA’s ABEL!

Post reply

Hi guys.

Earlybird is right. Anvisa is now suggesting the use of EMA's ABEL based on the 2014 WHO guideline. I've discussed with one of ANVISA's statistician and he said that this will probably be included in their own guidelines/resolutions soon. I showed him the thread we had here about TIE inflation but he said they were going to follow the 30% ISCV criteria anyway, because of WHO's guidance, abandoning the >40% "home brewed" criteria. Since it is not yet published as guidelines or resolutions we always submit the protocol for evaluation before initiating the study, even though the answer is always the same. The use of semi or partial replicated designs is also allowed, since blood volume might be a problem for 4-period studies.

Detlew:
As Mauricio said, there isn't a specific guideline determining the ANOVA structure for replicated designs, but I've always used FDA's suggestion (appendix E of 2001 guidance) without problems.

Hope I've been of help.

Lucas Teixeira

Helmut
★★★

Vienna, Austria,
2016-05-19 21:04
(3230 d 07:29 ago)

@ Lucas
Posting: # 16337
Views: 11,461

Method A, B (or C?)

Post reply

Hi Lucas,

❝ Hi guys.

Don’t forget the girlzzz!

❝ Anvisa is now suggesting the use of EMA's ABEL based on the 2014 WHO guideline. […] The use of semi or partial replicated designs is also allowed, since blood volume might be a problem for 4-period studies.

Good to know. Would mean that we could drop the argument regulator="ANVISA" in some functions of PowerTOST. There are two functions which adjust α in order to preserve the consumer risk at 0.05: scABEL.ad() and sampleN.scABEL.ad(). I had to introduce a lot of special conditions for ANVISA since large adjustments were needed for CV_wR close to 40%.

❝ I've always used FDA's suggestion (appendix E of 2001 guidance) without problems.

Yep. The early versions of ANVISA’s guidelines were practically single-handed developed by Salomon Stavchansky. Hence, a lot of US-background. If ANVISA is following the WHO’s document (which in this section is almost 1:1 to the EMA’s GL) could we assume that we should also use the crippled models given in the EMA’s Q&A-document (Method A and B)? These are not the mixed-effects model given by the FDA (termed in the Q&A document Method C).

—
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

Mauricio Sampaio
★

Brazil,
2016-05-19 21:57
(3230 d 06:36 ago)

(edited on 2016-05-20 10:27)
@ Lucas
Posting: # 16338
Views: 11,399

ANVISA = EMA’s ABEL!

Post reply

As I told you,,....everything can change according a new understanding from ANVISA. But, it is a good news!

About the use of semi or full or replicated, I am encouraged to peform a semi-replicated study with 3-periods because to calculate ISCV, I just need replicate the reference.

Adding: There is no reason to restrict the study only with use of total replicated.

Regards.

Mauricio Sampaio

Edit: two successive messages merged. You can edit your posts within 24 h if needed [Ohlbe]

d_labes
★★★

Berlin, Germany,
2016-05-24 11:46
(3225 d 16:47 ago)

@ Lucas
Posting: # 16357
Views: 11,123

ANVISA = EMA’s ABEL?

Post reply

Dear discutants!

Thanks to all (especially from Oberbavaria :-D

) for sharing your knowledge.
Seems a tiny step towards international harmonization :cool:

.

One question is left, if I see this correct: The estimation method. EMA's crippled ANOVA(s), or FDA's mixed model approach, or intra-subject contrasts (ISC)?
Unfortunately I have no clue how to simulate FDA's mixed model approach avoiding time consuming subject data sims. ISC may be a substitute, at least w.r.t. the degrees of freedom.

For the aim of sample size estimation the differences (if any) are only marginal, but for evaluation of the TIE or other explorations of power it may count.
Upcoming PowerTOST version:

# Defaults: partial replicate design, regulator EMA

# sample size, GMR=0.9 (theta0)

# EMA's crippled ANOVA(s)

sampleN.scABEL(CV=0.3, print=FALSE, details=FALSE)

  Design alpha CVwT CVwR theta0 theta1 theta2 Sample size Achieved power Target power nlast

1  2x3x3  0.05  0.3  0.3    0.9    0.8   1.25          54        0.81593          0.8    54

# Evaluation via robust ISC

sampleN.scABEL2(CV=0.3, print=FALSE, details=FALSE)

  Design alpha CVwT CVwR theta0 theta1 theta2 Sample size Achieved power Target power nlast

1  2x3x3  0.05  0.3  0.3    0.9    0.8   1.25          54        0.80765          0.8    54

# Now power at border of widened acceptance limits (TIE), GMR=0.95

# EMA's crippled ANOVA(s)

power.scABEL(CV=0.3, n=24, theta0=scABEL(CV=0.3)[2], nsims=1e6)

[1] 0.069039

# Evaluation via 'robust' ISC

power.scABEL2(CV=0.3, n=24, theta0=scABEL(CV=0.3)[2], nsims=1e6)

[1] 0.07631

—
Regards,

Detlew