July / August 2015

Comparison of Two Relevant Statistical Approaches to Assess Content Uniformity

James Bergum, PhD
William E. Brown
Jon Clark, M.S.
Thomas P. Parks, PhD
Thomas P. Garcia, PhD
James Prescott
Charles P. Hoiberg, PhD
Samir Patel
Mr. Ravindra Tejwani, Ph.D.

This article compares the performance of two statistical approaches (tolerance interval and ASTM E2709/E2810) to assess dosage unit uniformity. The potential impact that the approaches can have on the USP <905> monograph is also discussed


The ISPE Blend Uniformity and Content Uniformity (BUCU) Group was formed in August 2013 to address the gap resulting from the withdrawal of the draft stratified sampling guidance document. The Group’s proposed modifications address the US Food and Drug Administration’s (FDA’s) concerns, including insufficient blend testing and the use of USP <905> for release testing. The framework defined by the Group provides flexibility for sampling plans and statistical approaches/acceptance criteria used for the assessment of dosage unit uniformity. The following article compares the performance of two statistical approaches to assess dosage unit uniformity: One is based on a tolerance interval, and the other is the ASTM E2709/E2810 approach. The potential impact that the framework will have on the USP <905> monograph is also discussed.


The FDA withdrew the draft stratified sampling guidance document in August 2013 because it wasn’t consistent with its current thinking.1 2 3  The reasons for its withdrawal included:

  1. The desire to test triplicate blend samples to allow variance component analysis to detect non-uniform locations in the mix
  2. The acceptance criteria were based on USP <905>,4 which is insufficient for batch release
  3. The desire to use of statistically based sampling plans
  4. Linking the assessment of blend and content uniformity to the 2011 validation guidance document5

ISPE sponsored the formation of the Group in July/August. The Group’s recommendations were published,6 7 and consisted of a framework that could be used to assess blend and content uniformity throughout the three stages of process validation. The framework provides greater confidence that future samples of dosage units taken from the batch will comply with USP <905>. It also allows flexibility for the use of different sample sizes and statistical approaches to assess dosage unit uniformity by inserting them into the appropriate boxes. Figure 1 can be used for both Stage 1 Process Design and Stage 2 Process Qualification, and Figure 2 contains an approach that can be used during Stage 3 Continued Process Verification. The diagrams contain statistically valid sampling plans that are but one set of plausible sampling plans that can be used. They are for example purposes only and should not be considered firm numbers or requirements. Justification for the sampling plans and acceptance criteria selected should be based on stage appropriateness, existing product and process knowledge, and the consideration of consumer and producer risks.

  • 1Guidance for Industry, “Powder Blends and Finished Dosage Units – Stratified In-Process Dosage Unit Sampling and Assessment,” US Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER), October 2003, Pharmaceutical CGMPs.
  • 2Federal Register/Vol. 78, No. 152, pp. 48175–48176, 7 August 2013/Notices.
  • 3Questions and Answers on Current Good Manufacturing Practices, Good Guidance Practices, Level 2 Guidance – Production and Process Controls, CDER/OC Office of Manufacturing and Product Quality: CGMP Subject Matter Contacts, 6 August 2013,
  • 4 United States Pharmacopeia Convention, USP 37 NF 32, USP General Chapter <905> Uniformity of Dosage Units, General Notices and Requirements, Section 3.10, Applicability of Standards.
  • 5Guidance for Industry, “Process Validation: General Principles and Practices,” US Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research (CDER), Center for Biologics Evaluation and Research (CBER), Center for Veterinary Medicine (CVM), January 2011, Current Good Manufacturing Practices, Revision 1.
  • 6Garcia, Thomas, James Bergum, James Prescott et al., “Recommendations for the Assessment of Blend and Content Uniformity – Part 1: Modifications to Withdrawn FDA Draft Stratified Sampling Guidance,” Journal of Pharmaceutical Innovation, 2015, 10:76- 83, 10.1007/s12247-014-9207-0.
  • 7Bergum, J., T. Parks et al., “Assessment of Blend and Content Uniformity: Technical Discussion of Sampling Plans and Application of ASTM E2709/E2810,” Journal of Pharmaceutical Innovation, (2015) 10:84-97, DOI 10,1007/s12247-014-9208-z.
Figure 1
Current USP <905> Developments
Process flow diagram for assessment of blend and content uniformity for process design and process qualifications batches

Comparison of ASTM E2709/E2810 and Tolerance Interval Approaches

Using a sampling plan that tests one dosage unit from multiple locations, Figure 3 contains operating characteristic (OC) curves that demonstrate the performance of a tolerance interval approach and the ASTM E2709/E2810 approach for the same level of confidence (90 percent) and probability of passing the USP uniformity of dosage unit (UDU) test (95 percent) for various sample sizes.8  All curves are to the left of that for the USP <905> test. As the sample size increases, estimates of both the true mean and true standard deviation become more precise causing the curves to move to the right (lowering the producer’s risk while maintaining the same level of confidence without increasing the consumer’s risk). For the same sample size, the tolerance interval curves are to the right of those for the corresponding (more conservative) ASTM E2709/E2810 curves. Although not shown, OC curves for other statistical approaches could be generated and compared to the tolerance interval and ASTM E2709/E2810 approaches contained in Figure 3.

Figure 4 contains an approach that can be used for routine release testing during Stage 3 Continued Process Verification that demonstrates the impact that decreasing the confidence from 90 percent to 50 percent (while maintaining a 95 percent probability of passing the USP UDU test) has on the position of the curves. Decreasing the confidence level to 50 percent still results in curves far to the left of the USP curve. The plans are two-tiered using 10 dosage units in the first stage and 20 dosage units in the second stage when needed (referred to as Tier 10:30).


  • 8Bergum, J., (in press). “Tolerance Interval Alternative to ASTM E2709/E2810 Methodology to Provide Assurance of Passing the USP Uniformity of Dosage Unit (UDU) Test <905>,” accepted by Pharmaceutical Engineering, March 2015.
Figure 2
Current USP <905> Developments
Process flow diagram for assessment of blend and content uniformity for continued process verification (Stage 3B) batches

Using a sampling plan that tests more than one dosage unit per sampling location, Figure 5 shows the OC curves for ASTM E2709/E2810 when the lot mean is 100 percent. OC curves are displayed for both the example validation sampling plans using 20 or 40 locations with three dosage units tested per location (denoted by 20 x 3 and 40 x 3, respectively) at 90 percent confidence and the Tier 10:30 routine release sampling plan. Both the validation and routine OC curves are at 50 percent confidence with a 95 percent probability of passing the USP UDU test. Since the probability of passing the USP test depends on the percentage of total variation due to between locations, two OC curves are presented for the 20 x 3 and 40 x 3 plans, one for no variation due to between locations and the other where 90 percent of the variation is due to between locations. Note that the curves move to the right (reducing the producer’s risk) as the number of locations increases and/or when the percentage of total variation due to locations decreases. The tiered routine sampling plan is generally to the right of the validation OC curves but still far to the left of the USP UDU test curve.

Impact on USP <905> and Future Contributions by USP

The regulations require in-process controls on the adequacy of mixing and variability in drug product characteristics.9  Validating the correlation between blend uniformity and uniformity of the product is a costly process. The tools described in our papers justify this expenditure by promoting an efficient use of data from the finished product that confirms the state of control over the process as well as providing a measure of likelihood that samples taken from the batch will comply with USP <905>. The approach is consistent with principles associated with Quality by Design (QbD), which can have its own set of advantages within individual companies as well as the FDA.10  One outcome of our work may be for USP to provide a way for companies to use manufacturing data to demonstrate compliance with <905> without intrusion into the FDA GMP-compliance role. Would the pharmaceutical industry be willing to accept this shift by USP? In order for this approach to demonstrate compliance with USP <905>, specific information describing how this may be done will need to be added to USP. A current USP Expert Panel is charged with work toward potential revisions that should give alternative approaches for <905>. This will be in the context of harmonization efforts to avoid creating additional testing requirements for distribution in multiple regulatory markets. USP experts on several committees and an Expert Panel are involved.

  • 921 CFR 211.110, Code of Federal Regulations, Title 21, Volume 4, Revised as of 1 April 2014.
  • 10Yu, L.X., G. Amidon, M.A. Khan, S.W. Hoag, J. Polli, G.K. Raju, and J. Woodcock, “Understanding Pharmaceutical Quality by Design,” AAPS Journal, 16(4), 2014, pp. 771–783.
Figures 3 and 4
Current USP <905> Developments
OC curves for various sample sizes at a constant level of confidence (90 pecent) and probability of passing the USP UDU test (95 percent) for the ASTM E2709/E2810 and tolerance for interval approaches (Top), and OC curves for various levels of confidence and coverage for the ASTM E2709/E2810 and tolerance interval approaches (two-tiered 10 + 20 testing approach)

The Group endorses USP efforts to provide guidance on alternative ways a firm decides that a product can meet USP <905> if it is taken from the market and sampled and tested. This decision must be taken at batch release and is subject to risk analysis by the manufacturer. The risk that a sample taken from the market will not meet <905> is the responsibility of the manufacturer. Any USP chapter would presuppose the process validation discussed by the Group’s papers but not specifically discuss it because it is more properly a topic for FDA guidance. Another difficult piece is how to make the correlation between final product and the blend that produced it. That correlation could be investigated and discussed by USP experts with the understanding that no proposed revision will be effective without significant FDA participation.

Two levels of volunteer engagement in the USP standard-setting process are used in this thought exercise: Expert Committee and Expert Panel. Expert Committees are impanelled for a five-year term and participate as individuals with a firm conflict of interest agreement. Expert Panel members participate without the conflict of interest agreement and serve Expert Committees in an advisory role. Expert Committees are responsible for the content of USP. Revisions to USP are proposed in Pharmacopeial Forum (PF), an online publication. Comments received in response to PF proposals are considered and resolved by the Expert Committee before the revision becomes part of the official USP text. It seems like a long road, but we feel that our papers have mapped the journey starting with the limited revisions to USP.

Figure 5
Current USP <905> Developments
OC curves for sampling plans 20 x 3 and 40 x 3, varying confidence/probability (pass USP), true lot mean = 100 percent, and USP UDU vs. true lot standard deviation.

Moving the effort for ensuring quality from relying entirely on final product testing to a model of controlling manufacturing in a way that produces only quality material has been a long evolution. The FDA has been clear that it is ready to accept these practices as part of an overall system of control and has published arguments supporting this. There is the natural wish to avoid testing multiple times to ensure what is effectively the same attribute, even distribution of active ingredient from dose to dose. With the procedures described in this series, potential root cause analysis is strengthened, product manufacturing efficiencies are achieved, and product variability is controlled. It should be possible to add enhanced compendial compliance to this list of benefits.

Additional Information

The Group has a website11 with public access. It contains ASTM E2709/E2810 tables, a list of frequently asked questions (FAQs), references to applicable publications and presentations, and information about current and future activities. The site also includes slides from some of the presentations given during the Group-sponsored BUCU session at the 2015 IFPAC Conference.12  The Group is also willing to have discussions with any regulatory agencies and/or professional organizations that have interest in the topic.  |

  • 12“Blend Uniformity/Content Uniformity,” IFPAC 2015, 29th International Forum and Exhibition, 28 January 2015, Arlington, Virginia.