Part I: Process Characterization

New Perspectives on Cleaning. Coordinated by Rizwan Sharnez

For Client Review Only. All Rights Reserved. Advanstar Communications Inc. 2007

]

Parametric Release for Cleaning

Part I: Process Characterization Rizwan Sharnez and Laura Klewer

“New Perspectives on Cleaning” is an ongoing series of articles dedicated to cleaning validation and monitoring. This column addresses situations faced in cleaning and cleaning validation. Reader questions, comments, and suggestions are requested for future discussion topics and can be forwarded to either the author at [email protected] or to journal coordinating editor Susan Haigney at [email protected]

INTRODUCTION A previous series in this column on cycle development and validation for cleaning shows how design space and control space are established for cleaning processes (1). The series also shows how small-scale models can be used to determine worst-case operating conditions within the control space of the cleaning process (2). These models were used to demonstrate that the cleaning cycle would meet the imposed performance requirements if the cycle is operated within the established control space (3). This article demonstrates how this quality-by-design (QbD) approach to cycle development can be used to implement parametric release of equipment in real time. An important regulatory expectation for cleaning is that the validated state of the cleaning process be maintained throughout the lifecycle of the product (4, 5). This expectation may be met by performing scheduled reviews of data and documentation such as clean-and-use logs, process deviations, investigations, impromptu inspections, management reviews, and combinations thereof. These activities, along with change control, are the key elements of a robust monitoring program. Some organizations verify the effectiveness of their cleaning procedures periodically as part of their monitoring program. This is typically accomplished by perform-

ing a validation run at a predetermined frequency, such as semi-annually or annually. A significant limitation of this periodic testing strategy is that it is based on limited data. Typically, the cleaning procedure may be performed several times during any given period. Periodic verification of the effectiveness of a cleaning procedure provides limited assurance that the procedure was effective throughout that period. Another issue with periodic testing is that it presents significant risk to product disposition. If any of the tests were to fail, it would call into question all the lots manufactured during that period, which in turn could lead to arduous investigations, quarantining of unreleased lots, regulatory notifications, disruption of supply, and even product recalls.

PARAMETRIC RELEASE— AN ALTERNATIVE TO PERIODIC TESTING The above consequences can be avoided through parametric release based on real-time monitoring of critical process parameters (CPPs). With real-time monitoring of CPPs one would know of a failure prior to manufacturing the next lot. Thus, procedures could be put in place to ensure that the equipment is re-cleaned before being released, thereby eliminating the risk associated with periodic testing. Real-time monitoring of CPPs could provide the following additional benefits: • Reduced time and cost of cycle development • More efficient, effective, and consistent cleaning cycles • Higher cleaning assurance levels.


For more Author information, go to gxpandjvt.com/bios

Depending on the type of monitoring and the relationship between the CPPs and the critical quality attributes of the cleaning process, real time monitoring of CPPs

[

gxpandjv t.com

ABOUT THE AUTHORS Rizwan Sharnez, Ph.D., is principal engineer at Amgen Colorado. Rizwan has 15 years experience in the pharmaceutical and biotech industries. He may be contacted at [email protected]. Laura Klewer is senior engineer at Amgen Colorado. Laura has eight years experience in the pharmaceutical industry. She may be contacted at [email protected].

Journal

of

Validation Technology [Summer 2009]

29

New Perspectives on Cleaning.


could also eventually eliminate the need for traditional three-run validation.

QUALITY BY DESIGN QbD is a systematic approach to product and process design and development (6, 7). The goal of QbD is to enhance robustness by building quality into the manufacturing process rather than testing it in. It also provides a structured framework for innovation and continuous improvement. “Robustness is the ability of the process to tolerate the expected variability in raw materials, operating conditions, process equipment, environmental conditions, and human factors (8).” A process is considered robust if it can consistently meet performance requirements even when it is run under worst-case operational conditions. QbD is based on a better understanding of the relationship between operating conditions (inputs) and performance requirements (outputs). Operating conditions are defined in terms of CPPs and their respective operating ranges; performance requirements are defined in terms of critical quality attributes (CQAs) and their respective acceptable limits. Most processes have multiple CPPs and CQAs. Each CPP or CQA can vary over a specified range defined by its lower and upper acceptable limits (LAL and UAL) as shown on Figure 1. The acceptable ranges for the CPPs and the CQAs are defined as the control space and the performance requirements of the process, respectively. The dotted line represents the set point (SP) of the CPP, and the discontinuous line represents its worst case operating point. Experimental strategies for identifying worst-case operating conditions and for leveraging them to establish the control space for cleaning have been described previously (1-3). An important element of QbD is process characterization at small scale. The small-scale experiments are used to enhance process understanding and predict performance at full scale. The data are also used to determine the worst-case operating conditions and to verify that the process can consistently meet performance requirements under these conditions. Thus, the QbD approach provides a high degree of assurance that the product will meet the imposed performance specifications if the process is operated within its established control space. The following sections describe the role of process characterization in establishing acceptance criteria for parametric release. These sections also demonstrate how these criteria are expressed in terms of operating conditions and performance limits that can be

monitored in real time. An operating strategy for implementing parametric release will be described in Part II of this series.

EXPERIMENTAL MODELS Experimental models for cleaning characterization have been described in the literature (2, 10-14). In these experiments, the kinetics of soil removal (mass transfer) from a surface is measured at small scale. A critical element of these models is to identify the hardest-to-clean (worst case) location in the equipment from a cleaning standpoint and then scale it down in the lab. It is important to note that we do not need to simulate the entire cleaning process in the lab; instead, we only need to simulate the location within the equipment that is the hardest to clean. If we can clean the worst-case location, we should also be able to clean the other locations in the equipment. Further, with this approach, the cleaning times obtained at small scale would be indicative of those at full scale provided that the system is qualified to provide adequate spray coverage to the surfaces that need to be cleaned. For tanks that are cleaned-in-place, we can assume that the worst-case location within the equipment is cleaned by the action of a gravity-induced falling film (i.e., fluid flowing only under the influence of gravity, such as along the underside of an impellor blade). Thus, from a fluid dynamics perspective, the experimental set up should simulate this condition at the soil-solvent interface. The fluid velocity for the small-scale experiments should be set to the worstcase fluid velocity in the equipment (VMIN) (3). If VMIN is not known, it could be set to ≤ 10 cm/sec. This estimate of VMIN is based on the average velocity of a gravity-induced falling film of water along a 30° incline at 60°C (15). Under these conditions, the Reynolds Number is less than 20, and the flow is laminar with negligible rippling (16). Other important operating parameters such as material of construction, surface finish, soil-to-solution ratio, spotting volume (droplet size), and drying rate are determined in accordance with full-scale operating conditions. It should also be emphasized that in the design of small-scale experiments it is important that the application of the soil to the coupons be representative of the full-scale process. This can be quite challenging when the process dynamics at full scale are characterized by high temperature or high impact interactions between the soil and the equipment surfaces. Examples of these types of soil-surface interactions are


30

Journal

of


iv thome.com

Rizwan Sharnez, Coordinator.

For Client Review Only. All Rights Reserved. Advanstar Communications Inc. 2007 Figure 1: Schematic representation of a typical process.

Each Critical Process Parameter (CPP) or Critical Quality Attribute (CQA) can vary over a specified range defined by its lower and upper acceptable limits (LAL and UAL). The dotted and discontinuous lines represent the set point and worst case operating point of the CPP, respectively.

found in heat exchangers, centrifuges, homogenizers, tablet presses, and fluidized bed granulators.

CRITERION FOR PARAMETRIC RELEASE From a QbD perspective, the criterion for parametric release may be stated as follows: If the CPPs are controlled within their respective acceptable ranges (i.e., within the established control space of the process), then the performance of the process is acceptable, and as a result, the equipment can be released in real time. This section presents a case study to demonstrate how this criterion for parametric release can be expressed in terms of operating conditions and performance limits. We will see how this QbD approach, when combined with real time monitoring of CPPs, obviates the need for periodic monitoring. The case study is based on an automated cleaning process; however, because the underlying principle of parametric release is the same for all processes, the approach described here can be extended to semi-automated and manual cleaning processes provided that the CPPs are appropriately identified and monitored. Consider a dedicated formulation tank that is used to hold an aqueous solution of an active ingredient. The tank is cleaned in place (CIP) with a validated cleaning cycle (see Table I). The tank is sterilized in place (SIP) with a validated steaming cycle before it is released for manufacturing. In-process testing for the batch includes performancebased specifications for bioburden and endotoxins. Additionally, the facility has appropriate methods for controlling entry and proliferation of microbes. The

batch is also sterile filtered before being processed further. The performance specifications for the system are given in Table II. Does the system in Table II meet the requirements for parametric release? To answer this question, we need to verify that the performance of the process would be acceptable if the CPPs are controlled within their respective operational ranges (Table I). From a QbD perspective, this can be ascertained through the following steps.

Step 1: Are there any assumptions that can be leveraged to simplify the problem? For this case study, the following assumptions are

For Client Review Only. All Rights Reserved. Communications Inc. 2007 known toAdvanstar be valid:

gxpandjv t.com

• The cleanability of the process soil decreases with hold time (i.e., the process soil becomes harder to clean as it dries) and increases with the concentration and temperature of the cleaning solution. Thus, the worst-case conditions for product removal are as follows: hold time of 7 days; 0.75% concentration of NaOH; and 60 °C. Note that if the cleaning agent is appropriately selected, the lower limits of the concentration and temperature ranges would represent a worst-case from the standpoint of product removal. • The rinsibility of the cleaning solution is independent of hold time, decreases with increasing concentration, and increases with increasing temperature of the cleaning solution. (This assumption is valid for most cleaning agents.) Thus, the worst-case conditions for removal of cleaning agent are as follows: 1.25% concentration of NaOH, and 60 °C. If these assumptions cannot be substantiated, then Journal

of


31



Table I: Operating ranges for critical process parameters (control space) LAL, SP, and UAL are the lower operating limit, set point, and upper operating limit, respectively. Critical Process Parameters Temperature (°C)

Concentration (% NaOH[v/v])

LAL

LAL

Critical Steps

Duration

Velocity*(cm/sec)

1. Hold time

1-7 days

NA

SP

UAL

2. Caustic wash

≥ 10 min

≥ 10

60

70

80

3. Water rinses

≥ 2 min (recirculated) ≥ 1 min (burst)

≥ 10

60

70

80

NA**

SP

UAL

NA 0.75

1.00

1.25

NA

*Since velocity is a function of flow rate, it can be controlled by regulating the flow. ** Temperature is not a critical parameter for the dirty hold step.

Table II: Performance specifications. Acceptable concentration of cleaning agent (NaOH) in rinse water

≤ 2ppm*

Conductivity of NaOH in the rinse water at a concentration of 2 ppm (17)**

≤ 4.2 μS/cm (in-line) at 60 oC ≤ 4.7 μS/cm (off-line) at 25 oC

Critical quality attributes (CQA)

Visual inspection: no visible process residue In-line conductivity of burst rinse: ≤ 4.2 μS/cm

*This specifications should be based on calculated safety limits for acceptable carryover of NaOH into subsequent batch of product. ** The conductivity of the rinse water depends on the concentration of the cleaning agent and the purity of the water.

the worst-case operating conditions for removal of product and cleaning agent should be determined from first principles. A seven-run experimental strategy for estimating the worst-case operating conditions has been described previously (3).

Step 3: Do the CPPs and their operating ranges meet the criterion for parametric release?

the CPPs. For this type of convective means of cleaning, the CPPs are the duration of the wash and rinse steps, and the velocity, temperature, and concentration of the cleaning solution (2). CPPs for other mechanisms of cleaning such as sonication, scrubbing (manual cleaning), or fluid impingement can be identified using a similar approach. The worst-case operating conditions within the control space of the process were identified in Step 1 and are summarized in Table IV. The ability of the process to meet the performance specifications under these conditions needs to be verified at small scale. Thus, for the small-scale characterization runs, the CPPs should be set to their respective worst-case operating values at full scale (Table IV). The acceptance criteria for these runs are summarized as follows: • Absence of any visible process residue at the end of the run • The in-line conductivity should be ≤ 4.2 µS/cm at ≥ 60°C, or the off-line conductivity should be ≤ 4.7 µS/cm at 25°C, for the burst rinse.

The criterion for parametric release is that the process must be able to meet the imposed performance specifications (Table III), when it is operated within the control space. The control space is defined in Table I in terms of

Note that since the worst-case concentration of cleaning agent for product removal (0.75%) is different than that for cleaning agent removal (1.25%), the above condi-

For Client Review Only. All Rights Reserved. Advanstar Communications Inc. 2007 Step 2: Are the CQAs and their specifications appropriate for demonstrating adequate clearance of process residues? The CQAs and their specifications should be based on the performance requirements for the system. The operating strategy for meeting the performance requirements is summarized in Table III. Based on the assessment shown in Table III, the CQAs and their specifications are appropriate for demonstrating adequate clearance of process residues.

32

Journal

of


iv thome.com



Table III: Operating strategy for meeting performance requirements. Performance requirements: for demonstrating adequate clearance of the following process residues

Operating strategy: For meeting the performance requirement during regular operation

Performance specification (CQA): For validation runs

Product (active ingredient)

Verify CPPs are within acceptable ranges.

Visual inspection: no visible process residue

Equipment is dedicated (9)

Conductivity of rinse water: ≤ 4.2 μS/cm (in-line) at 60 °C ≤ 4.7 μS/cm (off-line) at 25 °C

At this conductivity, carryover of cleaning agent into subsequent batch is ≤ 2 ppm, which is acceptable for this system based on calculated safety limits for NaOH

NA*

NA

Cleaning agent (NaOH)

Bioburden and endotoxins

• Controls hold time within validated limit (≤ 7 days) • SIP equipment with a validatied steaming cycle before it is released for manufacturing • Test in-process samples of batch for bioburden and endotoxins • Sterile filter batch before it is processed any further • Implement appropriate methods for controlling entry and proliferation of microbes.

Rationale

*Based on the operating strategy, there are no CQAs associated with cleaning for the removal of bioburden and endotoxins.

tions must be satisfied at both the lower and the upper limits of cleaning agent concentration. Thus, for this example, a minimum of two process characterization runs is necessary to satisfy the criterion for parametric release (Table V). For the equipment to be released in real time, the smallscale experiments should meet the above performance specifications under the worst-case operating conditions, and the CPPs should be controlled within their respective acceptable ranges. If the small-scale experiments fail to meet the performance requirements, the duration of the caustic wash and the rinse steps could be extended until the performance criteria are met. Additionally, an overage of about 20% may be added to the durations at full scale to account for errors associated with instrumentation and controls, and variability of raw materials and equipment over the lifecycle of the process.

approach, we design and develop a process to ensure that it can meet predefined performance requirements. This requires that we understand the impact of critical process parameters on product quality, and that the process be continually monitored and upgraded as needed to assure consistent quality over the lifecycle of the process. An important element of QbD is process characterization at small scale. The small-scale experiments are used to enhance process understanding and predict performance at full scale. The small-scale data can also be used to determine the worst-case operating conditions within the control space of the process. The small-scale experiments are designed to simulate the velocity of the cleaning solution at the worstcase location in the equipment. All other variables such as hold time, concentration, and temperature of the cleaning solution are set to their corresponding worst-case values at full scale. Setting the average fluid velocity at small scale to the worst-case operating value for the equipment provides assurance that if the soil can be cleaned by the simulated wash at small


CONCLUSION Quality by design provides a framework for implementing a systematic approach to process design, development, and monitoring. With the QbD gxpandjv t.com

Journal

of


33



Table IV: Worst-case operating conditions within the control space of the process. Critical Process Parameters Critical Steps

Duration (min)

Velocity (cm/sec)

Temperature (°C)

Concentration (% NaOH[v/v])

1. Hold time

7 days

NA

NA*

NA

2. Caustic wash

≥ 10 min

10

60

0.75 for removal of product; 1.25 for removal of NaOH

3. Recirculed rinse

≤ 2 min

10

60

NA

4. Burst rinse

≤ 1 min

10

60

NA

*Temperature is not a critical parameter for the hold time step.

Table V: Summary of worst-case operating conditions for the small-scale experimental runs. Experimental Run

Hold Time (days)

Velocity (cm/sec)

Temperature (°C)

Concentration (%)

1. To demonstrate removal of product

7 days

≤ 10

60

0.75

2. To demonstrate removal of cleaning agent

Duration at small scale (min) Caustic wash ≤10 Recirculated rinse ≤ 2

NA

≤10

60

1.25

scale, then it can also be cleaned by the actual wash at full scale. This condition is valid only if, for the duration of the wash, there is adequate contact between the cleaning solution and the surfaces being cleaned. For CIP circuits, this condition is generally satisfied if the system is qualified to provide adequate spray coverage to the surfaces that need to be cleaned. This part of “Parametric Release for Cleaning” describes the role of process characterization in establishing acceptance criteria for parametric release. It also describes an experimental strategy to verify that the process is able to meet the imposed performance requirements under worst-case operating conditions. This QbD approach provides a high degree of assurance that the process will meet the performance requirements if it is operated within the established control space (i.e., when the CPPs are within their respective acceptable operating ranges). This continuous monitoring strategy obviates the need for periodic testing within the monitoring program for cleaning. Part II of this series will demonstrate how the acceptance criteria for parametric release can be expressed in terms of operating conditions and performance limits that can be readily monitored in real time. An

Burst rinse ≤ 1

operating strategy for implementing parametric release for cleaning processes will also be described.

ACKNOWLEDGMENTS For Client Review Only. All Rights Reserved. Advanstar Communications Inc. 2007 The authors would like to thank Martin VanTrieste,

34

Journal

of


Paul Lewus, and Sushil Abraham for their helpful suggestions and support.

REFERENCES 1. Sharnez, R., “Taking the Guesswork Out of Validation,” Journal of Validation Technology, Vol 14, No. 3, Spring 2008. 2. Sharnez, R., “Leveraging Small-Scale Models to Streamline Validation,” Journal of Validation Technology, Vol 14, No. 4, Summer 2008. 3. Sharnez, R., “Validating for the Long Haul,” Journal of Validation Technology, Vol. 14, No. 5, Autumn 2008. 4. PIC/S, Guide to Good Manufacturing Practice for Medicinal Products Annex 15, Section 45, January 2009, http:// www.picscheme.org/publis/guides/PE_009-8_GMP_ Guide%20_Annexes.pdf. 5. ICH, Q7 Good Manufacturing Practice Guide for Active Pharmaceutical Ingredients, Sections 12.60 and 12.76, Noiv thome.com



vember 2000, http://www.ich.org/LOB/media/MEDIA433. pdf. 6. FDA, Guidance for Industry: Quality Systems Approach to Pharmaceutical CGMP Manufacturing, September 2006. 7. Chi-wan Chen, Deputy Director ONDQA, CDER, FDA, “A FDA Perspective on QbD,” Regulatory Watch 2007 Conference, Robbinsville, NJ, September 2007. http://pharmtech.findpharma.com/pharmtech/content/ printContentPopup.jsp?id=469915. 8. Glodek, M., et al., “Process Robustness—A PQRI White Paper,” Pharm. Eng. 26 (6), 1-11 (Nov.-Dec. 2006). 9. FDA, Guide to Inspections Validation of Cleaning Processes, US FDA, July 1993, http://www.fda.gov/ICECI/InspectionGuides/ucm074922.htm. 10. Canhoto, A., “A Novel Bench Scale Apparatus to Model and Develop Biopharmaceutical Cleaning Procedures,” Journal of Validation Technology, Vol. 11, Nov 2004. 11. Sharnez, R., et al., “In Situ Monitoring of Soil Dissolution Dynamics: A Rapid and Simple Method for Determining Worst-case Soils for Cleaning Validation,” PDA Journal of Pharm. Sc. & Tech.; Vol. 58, No. 4, p. 203-214, July-Aug 2004.

12. Sharnez, R., et al., “Utilizing Bench-Scale Cleanability Studies and Master Soilants to Develop Cleaning Cycles: A Strategy for Successful Cleaning Validation,” Cleaning Validation and Critical Cleaning Processes, Institute of Validation Technology, Chicago, July 24-27, 2007. 13. Pluta, P., “Laboratory Studies in Cleaning Validation,” Journal of Validation Technology, Vol. 13, No. 4, August 2007. 14. Hoist, B., “Developing a Cleaning Process: Cleaning in Development,” Journal of GXP Compliance, Vol. 10, No. 3, 2006. 15. Sharnez, R., Unpublished results. 16. Bird, R., W. Stewart and E. Lightfoot, Transport Phenomena, John Wiley & Sons, p. 46, 2007. 17. Sharnez, R., Unpublished results. JVT

ARTICLE ACRONYM LISTING CPP CQA LAL QbD SP UAL

Critical Process Parameters Critical Quality Attributes Lower Acceptable Limit Quality by Design Set Point Upper Acceptable Limit


gxpandjv t.com

Journal

of


35