Bioequivalence and Bioavailability Forum

Main page Policy/Terms of Use Abbreviations Latest Posts

 Log-in |  Register |  Search

Back to the forum  Query: 2017-12-13 21:30 CET (UTC+1h)
 
BRB
Junior

Canada,
2016-09-20 19:12

Posting: # 16652
Views: 3,030
 

 EUFEPS/AAPS meeting [Conferences / Workshops]

Hello everyone,

I was able to attend the recent EUFEPS/AAPS Harmonization initiative conference and it was great hearing from many of the highly regarded minds in Bioequivalence from industry, academia, regulatory, etc.

I would be curious to get feedback from those who were there as to what you thought of what was discussed. In particular, I found the discussion regarding the attempts to harmonize the use of scaled BE between the FDA, EMA and Canada quite interesting (i.e. AUC and Cmax for FDA, whereas EMA only allows for Cmax). However, there seemed to be no overall conclusion made, unfortunately.

Regards,
BRB
Helmut
Hero
Homepage
Vienna, Austria,
2016-09-20 20:18

@ BRB
Posting: # 16653
Views: 2,396
 

 A long way to go…

Hi BRB,

» […] the recent EUFEPS/AAPS Harmonization initiative conference…

Was great fun. ;-)

» I would be curious to get feedback from those who were there as to what you thought of what was discussed.

My impression was that the EMA in some parts will have to move towards the other jurisdictions:
  • When it comes to acceptance of Adaptive Two-Stage Designs based on simulations, and
  • the overly “conservative” approach of dealing with outliers, irregular profiles, etc.
It was refreshing to see how open-minded and flexible the American regulators (of the FDA, HC, ANVISA, Mexico’s Secretaría de Salud, and Chile’s ANAMED) were. Different from the EMA’s…

» I found the discussion regarding the attempts to harmonize the use of scaled BE between the FDA, EMA and Canada quite interesting (i.e. AUC and Cmax for FDA, whereas EMA only allows for Cmax).

Actually:

FDA: AUC and Cmax by RSABE – without a clinical justification. The fact that the reference product – despite its high variability – is on the market for years without safety / efficacy issues proofs already the applicability of reference-scaling.
EMA: Cmax (and for MR-products Css,min, Css,τ, partial AUCs) by ABEL (capped at CVwR 50%) – with a clinical justification. Doesn’t make sense to me. There are classes of drugs / formulations (PPIs, bisphosphonates, locally acting MR,…) which show high variability in AUC as well. Remember the example I was talking about at the panel? Details here (slide 17).
HC: Only AUC by ABEL (but capping at CVwR 57.4%)* – with a clinical justification.

For all scaling-methods at higher CVs (say ≥50%) the GMR-restriction leads the decision to a good part. Inflation of the consumer risk might be substantial (even if ABE is used since the CVwR observed in the study was <30%).

» However, there seemed to be no overall conclusion made, unfortunately.

IMHO, László’s presentation gave a good summary of the open issues. It will be a long way to harmonize the requirements. My personal order of importance and difficulties we are facing:
  1. Which PK metrics qualify for reference scaling?
  2. Different statistical models (RSABE for the FDA; ABEL for EMA, HC, ANVISA, and jurisdictions following the WHO-GL).
    1. The FDA’s code for ABE (if RSABE is not applicable since swR <0.294) sometimes fails to converge for the partial replicate design. Study done, computer says no.
    2. The EMA’s implementation of ABEL is crippled. Assumes equal variances of test and reference!
  3. Upper caps (none for the FDA; 50% for EMA, ANVISA, and the WHO; 57.4% for HC).
  4. Inflation of the Type I Error. Most critical in all methods close to the switching CVwR of 30%.
  5. Bias in the estimate.
Statistically the GMR-restriction makes no sense indeed but I guess it will stay for a long time due to political reasons. As Charlie DiLiberti said: “Ask 100% physicians for the definition of bioequivalence and 99 of them will get it wrong.” I would add: 50 of them believe that the common acceptance range means that generics may differ in their effect [sic] by 45%. Oh dear.


  • Seems that HC prefers “nice” numbers. in many countries for NTIDs the narrower acceptance range is based on a clinically not relevant difference of 10%. That leads to an AR of 90.00–111.11%. In Canada the nicer looking upper limit of 112.0% is used. I guess the cap of 57.4% was introduced because the upper scaled limit of 150.0% looks nicer than the EMA’s of 143.19%…

[image]Regards,
Helmut Schütz 
[image]

The quality of responses received is directly proportional to the quality of the question asked. ☼
Science Quotes
mittyri
Senior

Russia,
2016-09-21 00:32
(edited by mittyri on 2016-09-21 13:41)

@ Helmut
Posting: # 16654
Views: 2,292
 

 GMR restrictions?

Hi Helmut!

could you please explain this:

» Statistically the GMR-restriction makes no sense indeed but I guess it will stay for a long time due to political reasons.

My little undestanding:
library(PowerTOST)
library(ggplot2)
rs <- reg_const("EMA")
rs$pe_constr <- FALSE
GMR <- 0.7999
y1 <- data.frame(
  N=integer(), Power=double(), LowerCI=double(), UpperCI=double()
  )

n1 <- 80:150
for(i in seq_along(n1))
{
 
  y1[nrow(y1)+1,] <-c(N=n1[i],
                      Power = power.scABEL(CV=0.5, n=n1[i], theta0=GMR, regulator=rs),
                      LowerCI=CI.BE(pe=GMR, CV=0.5, n=n1[i], design = "2x2x4")[1],
                      UpperCI=CI.BE(pe=GMR, CV=0.5, n=n1[i], design = "2x2x4")[2]   )
}
ggplot(y1)+
  geom_point(aes(N, Power, colour = Power))+
  geom_line(aes(N, LowerCI ), colour = "red")+
  geom_line(aes(N, UpperCI), colour = "red")+
  geom_hline(yintercept = 0.6984, color="red", size=3)+
  ylab("The Power and CIs")+
  ggtitle("CV = 50%, GMR = 79.99%")

[image]

So not necessary that (GMR is inside 0.8-1.25)&(CIs are inside the limits)
May be I understtod something wrong, please feel free to correct me and criticize my code :-D

Kind regards,
Mittyri
Back to the forum Activity
 Thread view
Bioequivalence and Bioavailability Forum | Admin contact
17,558 Posts in 3,759 Threads, 1,090 registered users;
29 users online (0 registered, 29 guests).

The purpose of models is not to fit the data,
but to sharpen the questions.    Samuel Karlin

The BIOEQUIVALENCE / BIOAVAILABILITY FORUM is hosted by
BEBAC Ing. Helmut Schütz
XHTML/CSS RSS Feed