Process Capability Eng
Process Capability Eng
Process Capability Eng
PHARMACEUTICAL ENGINEERING
www.PharmaceuticalEngineering.org
This article presents summary report of the main points from the 28th IFPAC
Annual Meeting Process Capability Symposium.
Opening Remarks
On behalf of symposium chair Dr. Lawrence Yu (Acting
Director, Office of Pharmaceutical Science, Center for Drug
Evaluation and Research (CDER)/FDA), Dr. Daniel Peng
delivered opening remarks. Dr. Peng began with a quotation
from H. Thomas Johnson: What you measure is what you
get. More likely, what you measure is all youll get. What
you dont (or cant) measure is lost.2 This highlights the importance of measuring the right things and measuring them
correctly. Dr. Peng emphasized that, in the past, it was usual
and customary to set acceptance criteria based on process
capability (the variability observed in the data). However,
this practice unintentionally allows manufacturers with
poor manufacturing and process controls to have products
with relatively wider acceptance criteria compared to good
manufacturing and controls with tight specifications. This
also could be one of the fundamental reasons why the pharmaceutical industry only gets 2-3 sigma since, the specification is set based on process capability.3 To break the vicious
cycle, our view may need to be fundamentally changed so
that the role of specifications is to confirm (control) product
quality rather than process robustness. Therefore, prior to
PHARMACEUTICAL ENGINEERING
SEPTEMBER/OCTOBER 2014
SEPTEMBER/OCTOBER 2014
PHARMACEUTICAL ENGINEERING
Dr. Dafni Bika (Vice President, Global Manufacturing Science and Technology, Bristol-Myers Squibb) provided an
innovator pharmaceutical company perspective on how to
use process capability indices as a tool by which to drive
operational excellence at Bristol-Myers Squibb (BMS).
Dr. Bika started the presentation by describing that
process robustness is the ability of a
process to demonstrate acceptable
quality and performance while tolerThe symposium, Use of Process Capability to Ensure Pharmaceutical
ating input variability. Then, Dr.
Product Quality, was held on 23 January 2014, in Arlington, Virginia
Bika discussed that product robustness
(USA) during the 28th International Forum on Process Analytical Chemis a holistic and comprehensive strategy
istry (IFPAC) annual meeting. This symposium provided an opportunity
at BMS that is used to ensure product
for pharmaceutical scientists to discuss the potential applications of
quality, compliance, and uninterrupted
supply. The process capability concept is
process capability indices to ensure product quality. Various presentatherefore applied across the value chain
tions from both the innovator and generic pharmaceutical industries
from raw materials to manufacturing
and the Food and Drug Administration (FDA) were included. The panel
processes and distribution to reduce varidiscussion provided an excellent source of information regarding the
ability from different sources (materials,
challenges and opportunities when using process capability indices.
equipment, methods, processes, etc).
This summary report documents the main points from this symposium:
Process robustness monitoring plans
are applied to many new, growth, and
1. clinically relevant specifications (CRSs) are needed in the modern
mature products. This is supported by
world, and are worth investing in because of their benefits, 2. process
automated data gathering and analysis
capability indices can be a powerful tool by which to ensure drug
at the growth and launch manufacturing
product quality and process robustness, 3. case studies from both the
sites with future technology implemeninnovator and generic pharmaceutical industries demonstrated that
tation planned for most sites. Product
process capability indices can be an useful tool by which to drive oprobustness and process capability analyses also have been extended to analytical
erational excellence and ensure delivery of superior product quality, 4.
methods, equipment, distribution, and
use of process capability indices should always go hand-in-hand with
selected external manufacturers. BMS
enhanced scientific understanding, and 5. some technical and culture
has updated its business processes and
challenges in implementing these tools still exist. Hence, further discusquality systems and engaged employees
sion and broadly engaging industry and other stakeholders on details
in developing the required skills and setof the implementation are still needed.
ting the expectations.
PHARMACEUTICAL ENGINEERING
SEPTEMBER/OCTOBER 2014
Dr. Bika then discussed the application of process monitoring plans and data analysis (including process capability) across all stages of validation lifecycle, with emphasis
on early and routine commercial manufacturing (Stage
3).7 Product performance across the network is tracked by
leading indicators (e.g., process capability) and resulting
business outcomes such as reduction of product recalls,
batch rejections, and investigations. Dr. Bika also presented
specific process monitoring examples, robustness improvement cases, and future applications.
Dr. Bika concluded that even with some challenges that
exist, the return on investment is evident and manifests
itself every day across BMSs manufacturing network in
savings from yield and cycle time improvements, avoidance
of rejects, write offs, and time lost on investigations. It also
continues to building an engaging and proactive culture
where monitoring data and related metrics are used to anticipate and prevent manufacturing issues.
SEPTEMBER/OCTOBER 2014
PHARMACEUTICAL ENGINEERING
The control plan for mature products starts with the lagging variable and moves upstream to define design Critical
Quality Attributes (CQA) for the product, and progresses to
Critical Control Parameters (CPP) for the manufacture and
incorporates the component CQA. Component CPP was not
included in this study.
Phase 1: identify primary areas of field performance and
monitor for variance over time with Pareto analysis and
control charting.
Phase 2: online process inspection is trended by lot to indicate overall capability of the process to produce defect-free
containers. Defects at the form fill seal pre-sterile inspection
were identified as a leading indicator for field performance.
Phase 3: identify if average performance or product consistency is the primary driver for pre-sterile defects. Average performance of the container is defined by the design of
the device, while the degree of variability in the container is
dependent upon the overall compounded variability of the
process as measured by temperature, equipment alignment,
pressure, and component performance variability. Compounded variability was identified as a leading indicator for
pre-sterile defects.
Phase 4: identify areas of contribution to compounded
variability. Unscheduled process interruption (equipment
failure, intermittent operator intervention, process readjustment, etc.) and component dimensional variability were
identified as primary contributors to process variability.
The interrelationship between component and unscheduled
process interruption was observed when the process steady
state was disturbed by a disruption in component supply or
placement.
A continual improvement project was initiated to standardize component placement and add upstream component
dimensional monitoring at the component supplier. Implementation of the project resulted in a reduction of unscheduled process interruption. Effectiveness of the improvement
was then monitored and was substantiated by a significant
improvement in field performance. By using an integrated
control plan of leading performance variables to provide
indicators for field performance productivity, overall quality
was improved.
PHARMACEUTICAL ENGINEERING
SEPTEMBER/OCTOBER 2014
Panel Discussion
Audience: How do we establish clinical relevant specification, for example, dissolution specification, for a generic
product that generally does not have as much clinical data
as the new drug during Phase 1-3 clinical studies?
SEPTEMBER/OCTOBER 2014
PHARMACEUTICAL ENGINEERING
the method which do not reflect relevant in vivo performance. For example, it is well documented in the literature
that the dissolution vessels hydrodynamics are poorly reproducible and very heterogeneous.14 Minor mechanical and
other changes in dissolution apparatus may have dramatic
effects on dissolution results. Its a persistent limitation of
current approaches that has hindered both in vitro quality
control testing and the development of clinically relevant
dissolution methods and specifications. It would be useful to develop new dissolution apparatus that are simple,
reproducible, robust, and linkable to clinical performance to
improve the antiquated vessel methodology.
Panel: In general, the answer is yes, since data normality is one of the three prerequisites to use process capability indices (Cp/Cpk) to estimate future batch failure rate.
However, there are some remedies when the data are not
normally distributedfor example, data transformation, distribution fit, or reference interval calculation (also known as
the percentile method). The ISO 21747 guidance document15
is cited for readers to gain further detail, as it is beyond the
scope of this paper.
The panel also highlighted that it is important to meet
another two prerequisites to use process capability indices
(Cp/Cpk) to forecast future batch failure rate: 1. sufficient
number of the subgroups is included, and 2. the process is
in a state of statistical control, which means that all special
causes have been eliminated from the systemand therefore
there are no detectable patterns or trends exhibited and the
variation observed in the data is only due to common cause
(process noise) of the system.5,6
The audience also commented that content uniformity
(acceptance value) in general is not normally distributed due
to standard deviation calculation. However, assay data are
generally normally distributed. It is good practice to graph
the raw data to see the shape of the distribution curve and
use science and product understanding to figure out if it is
outliers or any scientific reason that caused the non-normal
distribution.
The panel and audience also further discussed the drawbacks and limitations of the current USP content uniformity
test.16 Other alternative approaches may be considered, for
example the ASTM E2709 where tolerance for variability
(% RSD) is based on the sample size, confidence level, and
sample mean.17 Of course, the sampling plan is also critical
to ensure the samples taken are representative of the batch.
The panel and audience further discussed the distribution
of dissolution data. There is no general agreement about the
shape of the distribution of the dissolution data. Saccone,
et al., assumed the normal or lognormal distribution and
Panel:
1. Many different data sources are collected manually or
through automated systems. It is quite challenging to
refine the data to be usable for process capability analysis and ensure data integrity during format change and
transformation.
2. Both technical knowledge (statistical analysis and SPC)
and cultural gaps (how to manage a change) need to be
addressed.
3. A sizable investment is needed to ensure data gathering,
analysis, and integrity (i.e., appropriate IT platforms).
4. Apart from the technical challenges and accountability,
the firms quality management systems need to be updated accordingly to drive the expectations (SOPs,
validation directives, statistic analysis manuals, procedures
for locking and updating statistic control limits, etc.)
5. It is a cross-functional effort that requires many parts of
the organization to collaborate and align (process, analytical, engineering, quality, regulatory, and procurement).
Panel: We are all here for one reason: to ensure that medicines available to the American public are of the highest
quality. It is very important to ensure both generic and innovator drug products meet the same quality standardsan
expectation even more relevant because generic products
account for more than 80% of U.S. medicines.20 It has been
and will continue to be the FDAs policy to ensure the same
quality standard between innovator and generic products.
This will be further supported in the proposed Office of
Pharmaceutical Quality (OPQ) by integrating review and
inspection across product lifecycle.
PHARMACEUTICAL ENGINEERING
SEPTEMBER/OCTOBER 2014
Summary
Key highlights from the symposium were as follows: 1.
SEPTEMBER/OCTOBER 2014
PHARMACEUTICAL ENGINEERING
References
1. 28th International Forum on Process Analytical Chemistry (IFPAC) Annual Meeting, http://www.ifpac.com/cgibin/IFPACProgram2014.pl, accessed 10 February 2014.
2. http://www.isixsigma.com/community/blogs/what-youmeasure-what-you-get/,accessed 10 February 2014.
3. Vase, P., Achieving Six Sigma Quality in Medical Device
Manufacturing by use of Design of Experiments and Statistical Process Control, Pharmaceutical Engineering,
Vol. 27, No. 2, 2007, pp. 1-9, www.PharmaceuticalEngineering.org.
4. ASTM E2281: Standard Practice for Process and Measurement Capability Indices.
5. ISO 8258: Shewhart Control Charts.
6. ASTM E2587: Standard Practice for Use of Control
Charts in Statistical Process Control, www.astm.org.
7. U.S. FDA Guidance for Industry on Process Validation:
General Principles and Practices, 2011, www.fda.gov.
8. ICH Q9 Quality Risk Management, www.ich.org.
9. ICH Q10 Pharmaceutical Quality System, www.ich.org.
10. World Health Organization (WHO) GMP and Inspection
for Pharmaceutical Products: Main Principles, http://
whqlibdoc.who.int/publications/2007/9789241547086_
eng.pdf, accessed on 10 January 2014).
11. Choi, M., Determining the Manufacturability of Druglayered Tablets, Pharmaceutical Manufacturing, No. 4,
2007, pp. 34-42.
12. Quality by Design (QbD): An Example Pharmaceutical Development Report for an Immediate Release (IR)
Dosage Form, (http://www.fda.gov/downloads/Drugs/
DevelopmentApprovalProcess/HowDrugsareDevelopedandApproved/ApprovalApplications/AbbreviatedNewDrugApplicationANDAGenerics/UCM304305.pdf,
accessed 10 February 2014.
13. Quality by Design (QbD): An Example Pharmaceutical Development Report for a Modified Release (MR)
Dosage Form (http://www.fda.gov/downloads/Drugs/
DevelopmentApprovalProcess/HowDrugsareDevelopedandApproved/ApprovalApplications/AbbreviatedNewDrugApplicationANDAGenerics/UCM286595.pdf)
(accessed 10 February 2014).
14. Baxter, J.L., et al., Hydrodynamics-induced Variability in the USP Apparatus II Dissolution Test, Int. J.
Pharm., Vol. 292, No. 1-2, 2005, pp. 17-28.
15. ISO 21747: Statistical Methods -- Process Performance and
Capability Statistics for Measured Quality Characteristics.
16. USP<905> Uniformity of Dosage Units.
17. ASTM E2709: Standard Practice for Demonstrating Capability to Comply with an Acceptance Procedure, www.
astm.org.
18. Saccone, C. D., et al., Statistical Properties of the Dissolution Test of USP, Dissolution Technologies, Vol. 11,
No. 4, 2004, pp. 25-28.
19. The Use of Stratified Sampling of Blend and Dosage Units
to Demonstrate Adequacy of Mix for Powder Blends
(http://www.pqri.org/pdfs/DTC/Articles/Final_BUWG_
Stratfied_Sampling.pdf) (accessed 10 February 2014)
20. Declining Medicine Use and Cost: For Better or Worse?
(www.imshealth.com) (accessed 10 February 2014).
21. Yu, L.X., et al., Understanding Pharmaceutical Quality
by Design, AAPS J., Vol. 16, 2014, pp. 771-783.
Acknowledgments
Andre Raw, Sau (Larry) Lee, Chengjiu Hu, Khalid Khan,
Leila Wieser, Alex Viehmann and Karthik Iyer.
PHARMACEUTICAL ENGINEERING
SEPTEMBER/OCTOBER 2014
10
SEPTEMBER/OCTOBER 2014
PHARMACEUTICAL ENGINEERING
PHARMACEUTICAL ENGINEERING
SEPTEMBER/OCTOBER 2014
11