Parametric Tolerance Interval (PTI) Test for Delivered Dose Uniformity (DDU) for Orally Inhaled and Nasal Drug Products (OINDP) Michael Golden On behalf of IPAC-RS Advisory Committee of Pharmaceutical Science October 25, 2005

2 Overview IPAC-RS appreciates the opportunity to discuss the DDU quality standard with FDA Progress made to date Comments on FDA position IPAC-RS current position, if choice must be made today Strong preference for further accelerated dialogue –Agreement of quality standard

3 Agreements – IPAC-RS Perspective PTI Test provides a more thorough characterization of the batch and facilitates improved decision making. PTI Test is a good scientific approach to setting DDU specifications in accord with emerging QbD and risk management approaches. –Zero tolerance criteria are eliminated. Quality is best defined by coverage of the target interval Sample size is determined and set by the applicant –Flexibility in sample size selection carries no penalty –Reduces manufacturer’s risk without compromising product quality. Technical agreements: –1/3 of the sample in 1 st tier and 2/3 of the sample in the 2 nd tier. –Beginning and End testing from the same OINDP unit. –The Pocock approach to split the Type I error between the two tiers.

4 Agreements Conditional on an Acceptable Quality Standard The FDA-proposed methodology for control of upper and lower “tails” The FDA-proposed methodology for calculation of PTI Test coefficients (k’s) A target interval (‘goalposts’) of % LC

5 FDA’s 4 October 2005 proposal A standard PTI Test applied to beginning and end doses separately Target interval (’goalposts’) % LC Sample sizes n=20/60, 40/120, and 60/180 in the 1 st /2 nd tier (half Beginning & half End, for each tier) Coverage of 87.5% was proposed as the standard Coverages ranging from 82.5 to 90% were also presented by FDA 10/10/30/3020/20/60/6030/30/90/90 CoverageTier ITier IITier ITier IITier ITier II Test coefficients “Smallest” test

6 Comments on FDA Proposal Application of PTI test to each life-stage separately has not been discussed previously by Joint WG. –This makes the test significantly tighter than all previously discussed approaches. Implementation of PTI test in this manner results in a coverage requirement that is greater than the design point (95.8% for “10/10/30/ %” test) Even higher coverage is required when the mean is off target This proposal results in very significant increase in sample size compared to current requirements due to frequent use of tier 2. In addition, 40/120, 60/180 can not be considered for routine release and stability testing The FDA October 4 proposal is tighter than the 1998 MDI/DPI draft guidance test

7 Batch mean at 97% LC FDA October 4 proposal is tighter than the 1998 MDI/DPI draft guidance test

8 IPAC-RS Case Studies Case studies are presented for 3 US-marketed HFA MDIs (comprising 77 batches) –IPAC-RS database > 2000 batches of OINDP 1117 batches have been released to the US market (the others were released to non-US markets or were late development) Of these, 1045 batches are MDIs or DPIs –96 batches of these have enough data to allow evaluation (3 HFA MDIs, 4 CFC MDIs) Judicious pooling of data was performed to create samples of sufficient size to evaluate the 10/10/30/30 test Approach identical to that presented by FDA at the October 4, 2005 meeting

9 IPAC-RS case studies (1) US commercial solution HFA MDI (preventer) 23 batches evaluated Sample means % LC Sample SDs % LC Coverage# pass B & E of of of of 23 Results In this case study, 19 of 23 batches (83%) comply with the FDA proposal of 87.5% coverage Eleven batches require 2 nd tier testing (48%) Average sample size needed is 32

10 IPAC-RS case studies (2) US commercial suspension HFA MDI (preventer) 28 batches evaluated Sample means % LC Sample SDs % LC Coverage# pass B & E of of of of 28 Results In this case study, 19 of 28 batches (68%) comply with the FDA proposal of 87.5% coverage Twenty-two batches require 2 nd tier testing (79%) Average sample size needed is 43

11 IPAC-RS case studies (3) US commercial suspension HFA MDI (reliever) 26 batches evaluated Sample means % LC Sample SDs typically 3-9% LC but values up to 14% Coverage# pass B & E of of of of 26 Results In this case study, 13 of 26 batches (50%) comply with the FDA proposal of 87.5% coverage Twenty-two batches require 2 nd tier testing (85%) Average sample size needed is 48

12 Conclusions from Case Studies on US Commercial Products 26 of the 77 batches would have failed the FDA proposed ”87.5%” test and thus would not have been released. However, all 77 batches passed their approved specifications for uniformity and were suitable for their intended use The lowest coverage presented by FDA on October 4 th (82.5%) resulted in a pass rate of 58-91%, which is still unacceptable from both compliance and business perspectives These case studies illustrate why the FDA October 4, 2005 proposal is not acceptable to IPAC-RS

13 IPAC-RS Current Proposal If a choice must be made today, IPAC-RS requests that the Advisory Committee approve a test and quality standard with which the majority of OINDP can comply, such as the PTI test based on Lieberman-Resnikoff* (LR) methodology proposed to FDA by IPAC-RS in 2004 –82.5% coverage, –80-120% LC target interval, –Beginning and End evaluated together, –Sample size determined by the applicant Exceptions to this could be proposed by the applicant with adequate scientific justification * Lieberman, G. J. and Resnikoff, G..J. Sampling plans for inspection by variables. American Statistical Association Journal, June 1955, p

14 Proposed Two-Tier LR PTI Test Measure n 1 doses, calculate sample mean and SD –Estimate proportion outside goalposts (P 1 ) –Accept in 1 st tier if P 1  p 1 and life-stage means within % LC Otherwise measure n 2 -n 1 more doses, calculate sample mean and SD based on total sample of n 2 doses –Estimate proportion outside goalposts (P 2 ) –Accept in 2nd tier if P 2  p 2 and life-stage means within % LC P depends on goalposts, sample size, mean and SD p depends on sample size, coverage and  

15 Benefits of the IPAC-RS Proposal Provides a quality standard with which the majority of modern OINDP can comply Correctly controls the design-point coverage for batches on and off target -Tighter SDs are required for off-target batches Scientifically rigorous Has precedent in the literature Uses Pocock’s alphas as suggested by FDA

16 Comparison of different DDU tests Batch mean at 97% LC The figure indicates (1) that current FDA standard (green) is much more restrictive than other international requirements (brown and purple), (2) that IPAC-RS current positions (blue and yellow) represent a significant tightening compared to the initial proposals (black and pink) and provide a reasonable compromise between different international standards, and (3) that the FDA October 4 proposal is significantly tighter than current US requirements.

17 OC Curves for US Approved OINDP ( Frequent Exceptions From Guidance) Subset of DDU Specifications Approved by FDA (24 products from ) Red: FDA draft guidance Green: FDA OCT4 20/ % coverage Batch mean at 100% LC

18 IPAC-RS Conclusions Further accelerated dialogue is needed to reach a consensus quality standard for DDU IPAC-RS is flexible about the methodology of the test as long as the required quality standard is attainable –The majority of modern OINDPs should be able to comply with the agreed quality standard using the smallest sample size (20/60)

19 Question to ACPS Would you support the quality standard and test as proposed by IPAC-RS in slide 13?