a modular reprinted from the official journal of ispe ... · pharmaceutical production facility...

Case Study of ANSI ISA S88.01

JANUARY/FEBRUARY 2003 PHARMACEUTICAL ENGINEERING 1©Copyright ISPE 2003

The ANSI ISA S88.01 Standard:A Case Study of its Application in thePharmaceutical Industry

by M. Mangiarotti and M. Rizzi

A modularmanagementproductionsystem basedon the ANSIISA S88.01standardprovides acomparison ofEquipmentModule andControl ModuleOptions.

Introduction

In the light of an innovative productionsystems development project, and in linewith the guidelines issued by its US parentcompany, an Italian company has based its

new model for a modular management produc-tion system on the ANSI ISA S88.01 standard.

The system is intended to guarantee thatthe control of process-related activities are safelyperformed and to ensure that a highly flexiblemanagement of recipes and equipment isachieved, as well as a substantial improvementin the management of the production units.

S88.01 StandardsIn February 1995, following an extended pe-riod of preparation, the final version of theANSI ISA S88.01 standard was published.

The standard is intended to “define referencemodels for batch control as used in the processindustries and terminology that helps explainthe relationship between these models andterms.” (ANSI/ISA-S88.01, Batch Control Part1: Models and Terminology, October 23, 1995.)

More specifically, the ANSI ISA S88.01 stan-dard outlines standard models and terminolo-gies aimed at defining the requirements of

Figure 1. The fluidized-bed granulation process.

Reprinted from The Official Journal of ISPE

PHARMACEUTICAL ENGINEERING® January/February 2003, Vol. 23 No. 1


2 PHARMACEUTICAL ENGINEERING JANUARY/FEBRUARY 2003 ©Copyright ISPE 2003

Figure 3. Procedural and physical model.

batch control for a manufacturing plant. These models andthe terminology help emphasize the Good ManufacturingPractices (GMPs) for the implementation and running of abatch-type manufacturing plant.

The standard may be used to improve control of existingbatch-type manufacturing plants, and in particular, may beapplied regardless of the automation level of the plant.

The ANSI ISA S88.01 standard does not:

• suggest a single way to implement batch control and doesnot guarantee the outcome of a project

• require a change in the habitual way of developing batchprocesses

• limit the scope of new developments relating to batchcontrols

Figure 2. Process model.

Set Up and Management of BatchProcesses

Batch processes manage the production of predeterminedquantities of material (batches) using preset quantities ofraw materials within established operation cycles, but em-ploying different equipment.

The set up and management of batch processes in apharmaceutical production facility raises several questionsin relation to process automation. A common example iswhether to have prescriptions managed by a ProgrammableLogic Controller (PLC) or a Personal Computer (PC).

Although they are certainly safer than personal comput-ers during the process, PLCs are less flexible, i.e., the higherflexibility linked to recipes within a PLC-based system re-sults in a more complex code and the associated issues thatthis can bring to the development and testing processes.

Conversely, a PLC/PC interface-based system may yieldbetter results in terms of flexibility and control although twoother issues need to be considered:

1. computer reliability and performance, which become criti-cal parts of the process

2. the potential need to implement bespoke batch manage-ment software, resulting in an additional workload duringdevelopment and validation of the system

To provide batch-type manufacturing plants with a practicalsolution, a few software companies have studied and devel-oped the implementation model outlined by the ANSI ISAS88.01 standard, creating applications that are able to meetthe requirements of ANSI ISA S88.01 standard. The opportu-nity of using such applications in automated pharmaceutical



batch-type manufacturing plants is becoming increasinglyimportant.

The structure of a batch-type production plant must betightly linked to its control system and consists of two maincomponents:

• the physical components of the plant (Physical Model)

• the operating processes (Procedures)

The physical components (Physical Model) are defined as:

• Process Cell - a functionally complete plant area

• Unit - a plant unit with independent functions (e.g., areactor or a mixer)

• Equipment Module - set of devices with specific functions(e.g., temperature control, pressure control)

• Control Module - device (e.g., a valve, an agitator, or apump)

Case Study DescriptionThis case study uses as an example a fluidized-bed granula-tion process - Figure 1. Without analyzing the pharmaceuti-cal process in detail, it will be sufficient to say that fluidized-bed granulation is used in the manufacture of tablets, andprimarily requires a solution preparation tank, a hopper or apowder loading system, a granulator, one or more desiccators(fluid bed driers), and a system unloading granulate into steelcontainers.

The solution production tank is fitted with a solutiontemperature control system, a mixer, and a loading systemfor liquids. Once they have been added and properly heatedby means of a recirculation system, the liquids are pouredinto the mixer.

Here, the solution is mixed with the powder. After apredetermined time, the resulting granulate is transferred tothe desiccators where it is air-dried under controlled dewpoint conditions.

Figure 4. Relation between physical/procedural model and process.



Figure 5. Equipment module.

After drying, the product is placed in containers andconveyed onto the next phase of the process.

Studying and Identifying the Phasesof the Process

Studying and identifying the phases of the process is essen-tial for successful plant automation. Both the process engi-neer and the automation engineer should be involved in thisactivity (it’s a milestone project phase based on S88.01modeling) and it is considered beneficial for them to collabo-rate from the start of the project.

The first step involves identifying the individual phases,which compose the process, and analyzing each phase indi-vidually.

In the example given, for the solution preparation tank,the process can be subdivided into a number of related actions(The action is the part of the process related to the phase, asshown in Figure 4) (Process Stage):

1. addition of liquids2. recirculation3. mixing4. unloading into the mixer

To prepare a solution, each of the four actions is required.Each of these actions can be further divided into a number ofoperations and each operation can, itself, comprise a numberof actions. Figure 2 shows an example of the liquid (water)addition phase.

Production Prescription AnalysisAfter studying the process, the production recipes and theequipment interactions are analyzed, taking into account thephysical (site/plant/equipment) model and the procedural(process) model, as shown in Figure 3.

This analysis will result in the definition of the mostcritical part of the whole system: the Area Model or basicsystem configuration. The Area Model describes the equip-ment necessary to perform the process by means of recipes.

Within the area model, every process management andimplementation phase is identified along with the equipmentnecessary to perform the process on site.

This defines the next steps, which are:

• to define the process to be carried out (down to the level ofdefining individual ‘actions’)

• to define the equipment necessary to execute the process(Area Model)

• to define the “procedures,” i.e., the production prescrip-tions which will be issued by means of the area model

Relation Between Physical/ProceduralModel and Process

Figure 4 shows the connections between the various partswhich comprise the correct definition of a batch process.

This analysis of the identified process phases provides thecomplete definition of the physical model of the process andits associated phases.

Addition of LiquidsThe liquid addition phase [PROCESS] (three types of liquids)requires a physical system.

The physical system is made up of three valves (one pertype of liquid) and, e.g., three dosing pumps.

• This physical system (composed of the three valves andthree pumps) is considered a single phase [PHYSICALMODEL]

The recirculation phase [PROCESS] requires one pump,several valves, and one heat exchanger.

• The system (composed of one pump, several valves, andone heat exchanger) is considered a single phase [PHYSI-CAL MODEL].

The mixing phase [PROCESS] requires a mixer.

• The mixer is considered a single phase [PHYSICALMODEL].

The mixer unloading phase [PROCESS] is made up of anunloading valve and a Mixer loading valve.

• The unloading valve and mixer loading valve are consid-ered a single phase [PHYSICAL MODEL].

The next step in the procedure is to translate the PhysicalModel into an appropriate PLC-based automation system.

There are two types of approaches, based on the sametheory:

1. the “Equipment Module” approach



2. the “Control Module” approach

Equipment Module ApproachThe “Equipment Module” approach requires the creation of acoded structure within the PLC that enables the operation ofthe relevant part of the physical model. In the example given,the three loading valves and the three dosing pumps are thecomponents of a single “Equipment Module” within the PLC,i.e., a single coding structure that has parameters and oper-ating controls which are summarized in the schematic dia-gram shown in Figure 5.

The equipment module starts from an idle state, and ifevery condition enabling it to start has been implemented,the module is brought into a running state after receiving astart command. The module will stop following if it receivesan alarm signal or a stop command.

Control Module ApproachUsing the example given, at the point that the water additionphase must start: the control module shown in Figure 6 willbehave as follows:

If the starting conditions are met (module in idle state, i.e.valves in place and pump at rest) the start command (givenby a supervisor or by the prescription management system,depending on whether the system is in semi-automatic orautomatic mode) will bring the module into a starting state.

In this state, the water dosing valves receive the position-ing command (by managing a parameter which could resultfrom the prescription) and the pump receives the runningcommand. If the feedback signals from the field are correct,i.e. if the valves are properly positioned and the pump isrunning, the module will go into running mode. Upon recep-tion of a stop command, the module stops the pump and closesthe valves.

If no fault arises, the module will go into an idle stateagain, ready to restart. An emergency condition will put themodule into a fault mode. In order to set the module runningagain, either the restart control needs to be initiated (e.g., ifthe fault has been identified and is easily removed), or thestop control operated (e.g., if the fault requires maintenance).

The “Control Module” approach is based on a differentstructure, i.e., on the operation of single devices or physicalcomponents.

Individual pumps and valves are operated in a standardmanner, as shown in Figure 6.

The two Manual/Automatic command signals originatingfrom the supervising system are filtered by a code structurewhich is able to control the safety interlocks/alarms, as wellas the devices to be operated first (e.g., a device driver).

The manual command is given by the Supervisory Controland Data Acquisition (SCADA) system (e.g., for the forcing ofsingle devices) whereas the automatic command is generatedby the batch software which operates the various devices inaccordance with the prescription steps. A fault in the batchsystem will cause the devices associated with the faulty stepto stop immediately.

Defining Phase ModulesBoth the equipment module and the control module structureneed that part of the PLC code to interface with the batchsystem. The structure is called a phase module.

The main components of the phase module are the Com-mands, Parameters, and Reports. The phase module sendsCommands and Parameters to the equipment module. Theequipment module sends the Reports to the phase module.

In the example given, the liquid addition equipment mod-ule receives the following parameters from the phase module:

• type of liquid (water)• quantity to be added• start/stop commands

The equipment module returns the following report:

• added quantity (back to the phase module)

The report will then be used by the batch report structures inconjunction with the report from the equipment module togenerate the batch records for the manufacturing process.

Analysis of the discussion to this point indicates that theequipment module parameters are nothing more than therecipe parameters of the pharmaceutical process.

ConclusionComparison of the Equipment Module and theControl Module OptionsThe control module structure guarantees considerably sim-pler PLC software which on the one hand reduces its capabil-ity, but on the other provides a greater storage capacity. Thegreatest disadvantage is the dependence on the batch pro-gram when executing the recipes. There is no alternative.Any faults in the recipe management program will result inan inevitable production standstill which can only be reducedby means of robust back-up structures, e.g., single devicescould be manually operated (forced) by means of the SCADAsystem. However, this would only enable relatively simpleprocesses to be recovered.

On the other hand, although the equipment module struc-ture provides a complex PLC code structure, it ensures that

Figure 6. Control module.



process recipes can be manually completed with subsequentsteps even if the batch system crashes. In such case, theoperator will use each single equipment module following themaster formula to achieve the final product. This type ofstructure is recommended for highly critical processes suchas pharmaceutical manufacturing processes.

In order for the batch control system and its validation tobe successful, it is worth stressing that it is essential that thevarious players involved in the project need to collaborateduring every phase, that the process traceability is achievedby means of a common systematic approach, and that anappropriate batch system development process is used inorder to avoid having to repeatedly develop a new methodol-ogy. Since pharmaceutical manufacturers no longer operatein local markets, the recommendations of the various boardsin charge of guaranteeing observance of those regulationscannot be neglected.

Reference1. ANSI/ISA - S88.01 - 1995.

About the AuthorsMaurizio Mangiarotti is ManufacturingAutomation Manager for the Pavia Engi-neering Department of Merck Sharp & DohmeItaly. With an educational background inTechnical Studies, he received a Degree inelectronic engineering from the Universitàdi Pavia with a thesis on the GSM datatransmissions. He worked as automation

project engineer in various application fields such as mechan-ics, food, petrochemical; and in the pharmaceutical industryworking with engineering companies, production companies,plant maintenance, and validation services including everyaspect of automation engineering. He has shared his experi-ence by publishing articles and presenting papers at interna-tional conferences. He is an active member of the AssociazioneItaliana per l’Automazione (ANIPLA).

Merck Sharp & Dohme Italy S.p.A., Automation Depart-ment, Pavia, V. Emilia 21, 27100 Italy, tel: +39 0382 433371,e-mail: [email protected].

Marco Rizzi joined Rockwell AutomationItaly five 5 years ago as application managerand then assumed the marketing responsi-bility for the Process Business. Currently, heis Application and Marketing Manger, Pro-cess Control Italy. He has 15 years of experi-ence in the process control industry withpositions such as system engineer, project

leader, and project manager. He has worked in countries suchas the Peoples Republic of China and Saudi Arabia. He hasbeen involved in the MS&D Pavia project, spoken at interna-tional conferences, and published articles in technical Italianmagazines.

Rockwell Automation Italy, Viale De Gasperi, 126-20026Mazzo Di Rho, Italy.

Product Development


Manufacturing and QualityPartnership in Support of ProductDevelopment

by Carmen M. Wagner and Frank S. Kohn

This articlefocuses on oneaspect ofproductdevelopmentteamwork,namely thepartnershipbetween qualityandmanufacturing.

Introduction

Vaccine companies are entrusted withthe responsibility to develop vaccinesthat are safe, efficacious, meet regula-tory compliance, and are not too costly.

This is a tall order, considering that companiesmust risk up to 15 years, and as much as $800million, knowing these investments of timeand dollars may never pay off .1 In fact, many ofthe products now under development will neveractually reach the market. Consequently, meet-ing these time and investment requirements,while remaining profitable, requires creative

and “out of the box” thinking.The above business demands are just a part

of the story. In order to be successful, vaccinecompanies also must invest heavily in regula-tory compliance. They must apply the appro-priate level of current Good ManufacturingPractices (cGMPs) to the development of newvaccines, and must ensure the right level ofquality oversight and the application of appro-priate manufacturing principles.2 As stated inthe FDA guidance document,3 “when drug de-velopment reaches the stage where the drugproducts are produced for clinical trials in hu-

Figure 1. Key factors forsuccessfulmanufacturing and QApartnership.



Product Development


mans or animals, then compliance with cGMPs is required.For example, the drug product must be produced in a quali-fied facility, using laboratory and other equipment that hasbeen qualified, and processes must be validated.” Accordingto FDA expectations, the quality and manufacturing func-tions must not only ensure the availability of sufficient andadequate product, but also must assure their compliancewith the appropriate regulatory requirements.

Business, technological, and regulatory demands con-tinue to pressure companies to search for more efficient andcost-effective ways to develop and market new products. Inour experience, one of the answers lies in the organization ofeffective teamwork and in the constructive relationship be-tween the manufacturing and quality portions of the organi-zation. All teams, including the manufacturing/quality teammust work synergistically to shorten timelines and containcost. We believe it is never too early to start involving thequality and manufacturing specialists in the product devel-opment process. Successful companies should start thinkingof these two functions as soon as a candidate is considered tobe a potential commercial product. The role of the qualityfunction should increase in step-wise fashion, along with theincrease in level of cGMPs. Quality oversight should be in fullmode by Phase III, when full cGMPs should apply.

Given the potential for the Manufacturing and Qualityfunctions to significantly impact the outcome of the develop-ment process, we believe that a partnership between qualityand manufacturing can increase the potential for companiesto achieve their goal of getting to market quickly, whilemarketing vaccines that meet pre-determined regulatoryrequirements and quality standards, and are not too costly.

This article focuses on one aspect of product developmentteamwork, namely the partnership between quality andmanufacturing. It includes a discussion of key business,technical, and regulatory challenges that may impact prod-uct development and describes the attributes of the effectivequality and manufacturing partners. It ends with a presenta-tion of the Total Product Quality concept, and a discussion ofhow Quality and Manufacturing must work together to achieveproduct quality and support product development.

The Vaccine Business - Key ChallengesFor those who are unfamiliar with the vaccine industry, it isworth mentioning that being first to market is often key tosuccess and profitability. To remain competitive, it is oftencritical to pay strict attention to activities that can shortenthe time it takes to achieve product commercialization. Short-ening the time to market may be difficult since this is usuallytied to a company’s willingness and ability to dedicate theproper resources and budget to a given project.

As discussed below, the key challenges can be classifiedinto technical, regulatory, and business. These challengesmust be addressed keeping in mind the overall project goalsand timelines. This may seem simple, but in large companies,with many departments and with the same people involved inseveral projects, it is easy for project team members to beconfronted with conflicting assignments, project target comple-

tion dates and goals that are not clearly defined.The information below outlines the issues classified as

technical, regulatory, and business; and illustrates how qual-ity and manufacturing can work together to maximize thechances for business success.

1. TechnicalThe manufacturing areas in a vaccine plant often resemble alaboratory, including specialized equipment such as the onesused in fermentation, purification and/or conjugation of vac-cines. This kind of operation generally requires:

• high level of technical expertise• use of automated and customized equipment• costly and complex processes• long lead times for manufacturing

These technical requirements tend to add to the overalloperational cost, and impose long lead times that greatlyimpact materials management, production planning, andproduct distribution. Effective communication between Qual-ity and Manufacturing can help overcome some of the techni-cal issues by ensuring that quality systems take into consid-eration the appropriate manufacturing requirements.

Our experience: In order to deal with some of the existingconflicts between manufacturing and quality, and to addressmanufacturing’s technical needs, we instituted weekly meet-ings between the two groups. These meetings lasted from 30-60 minutes (sometimes less) and focused on technical andcompliance issues that were often resolved before they be-came a problem. The meetings helped the Quality Depart-ment design and update systems, taking into considerationthe technical needs of the manufacturing process, but withattention to regulatory requirements. The interaction helpeddevelop better understanding of each group’s function andhelped build mutual respect between the quality and manu-facturing staff. Quality was able to prevent certain concernsfrom escalating into full compliance problems.

2. Regulatory ComplianceAn overriding imperative for a company in the vaccine indus-try is to stay ahead of the competition, while dealing withtechnological changes in a highly regulated environment.The nature of these complex biological processes, and theresources needed to comply with increasing regulatory de-mands are numerous and costly.

The Quality/Manufacturing partnership can help achievecompliance with regulatory expectations by developing ajoint strategy to ensure the following:

• that the Quality oversight is part of every step of productmanufacturing

• that cGMP increases in a step-wise manner, as recom-mended by FDA

Product Development


• that product manufacturing complies with the appropri-ate level of cGMP at different stages of development

• that full cGMP compliance is achieved by Phase III Clini-cal Trials.

• that in the case of global commercialization, the require-ments imposed by the Center for Biologics Evaluation andResearch (CBER) in the US, and by several Boards ofHealth worldwide, are understood and applied

• validation of critical processes is completed as early aspossible, no later than Phase III (examples: sterilization,lyophilization, and media fills)

• cGMP activities are segregated from research and fromdevelopment

• change control is in place for cGMP manufacturing ofclinical supplies and consistency runs

• preliminary validation data for cleaning manufacturingequipment is available as early as possible

Our experience: Quality encouraged Manufacturing to pro-pose solutions to problems associated with manufacturingcontrol. These solutions were then discussed with Quality toensure that they complied with regulatory requirements.Manufacturing’s active participation in the troubleshootingprocess lead to a feeling of ownership, made them feel part ofthe solution, and encouraged greater compliance with theimplemented solution.

3. BusinessThe development of new vaccines is often lengthy and costly.A company can lose as much as $1 million a day for every daythat market entry is delayed. Besides aggressive timelinesand pressures to be first-to-market, additional business is-sues include:

• capital intensive facilities, calling for specialized equip-ment and work area design

• pricing strategy is difficult• complex and varied customer requirements• integrating quality requirements in the business plan

Our experience: Manufacturing and Quality agreed on anapproach for self-audit in the manufacturing department.This did not replace the Quality audits, but encouraged themanufacturing department to discover their own deficien-cies, and propose their own solutions to correct them. Thedeficiencies and proposed solutions were then discussed withQuality and joint solutions were frequently agreed upon.However, it is important to note that Quality did reserve theright to disagree and impose alternative solutions, whenappropriate.

Technical, business, and regulatory considerations must beintegrated into the overall development plan to preventdelays in timelines and prevent profit loss. Moreover, as apractical consideration, when the common conflicts betweenmanufacturing and quality occur, if dealt with constructivelyand quickly, these potential roadblocks can be turned intosuccesses.

As mentioned previously, addressing business, techno-logical, and regulatory demands requires more effective part-nership between quality and manufacturing. These two func-tions need to develop a joint strategy and work to preventproblems, develop proactive solutions, and practice cost avoid-ance. The remainder of this article will focus on two Modelsthat illustrate the effective quality/manufacturing partner-ship attributes.

Quality as an Effective PartnerThe scope and list of tasks associated with product manage-ment and control can seem overwhelming, but a well de-signed, cross-functionally derived quality program, based onproactive quality thinking, can help break the tasks intomanageable pieces. It also can help promote strong partner-ships with Manufacturing.

Quality principles integrated into routine product manu-facturing can be great problem prevention and cost avoidancetools. However, for this to happen, quality needs to be inte-grated into routine manufacturing planning. This integra-tion should start with clinical manufacturing and should befocused on product and package development, from clinicalevaluation through scale-up, to full-scale manufacturingstart-up and commercialization/distribution. Thus, qualitymust be an integral part of the development and the commer-cial supply chain.

Achieving total product quality requires consideration ofcustomer requirements and several other critical factors asillustrated in the proposed TPQ model - Figure 2. The ulti-mate goal in process development and routine manufactur-ing is to build efficiency and cost effectiveness, while main-taining product quality.

To emphasize the point, bringing quality and manufactur-ing into the development process as early as possible, helpsensure that the processes are carried out in a controlled anddocumented way, thus facilitating technology transfer andexpediting launching of new products. This in turn helpsestablish effective quality systems and practices, applicableto routine product manufacturing and quality control duringthe commercial phase.

Manufacturing as an Effective PartnerThe primary mission of the Manufacturing Department is toproduce high quality products, help control cost, and meetproduction schedules. Quality’s responsibility is to developquality systems, test, audit, release or reject products, andassure that released lots meet all quality attributes andregulatory requirements. The Quality Department cannotbuild quality into the process or product, but it can help setthe policies, quality culture, and practices that encourage the

Product Development


Figure 2. This figure shows the different components of the Total Product Quality (TPQ) concept. These components must work insynergy, and must be embraced by both quality and manufacturing so that they can contribute to the goal of shortening productcommercialization timeline, and to control cost.

manufacturing staff to meet acceptable quality standards. Inthe ideal situation, product quality must start within themanufacturing function in partnership with the QualityDepartment. Manufacturing management can ensure a suc-cessful partnership with Quality by helping foster an envi-ronment of cGMP compliance as part of the department’sdaily routine.

Our experience has demonstrated that the factors identi-fied in Figure 1 are essential for a successful partnership.Without clear partnership goals, communication, strong lead-ership, clearly defined and implemented policies and proce-dures, staff training, of cross-functional team membership,and Standard Operating Procedures (SOPs), there can be nosuccessful partnership between Manufacturing and Quality.

Figure 1 identifies the necessary factors to ensure asuccessful Manufacturing – Quality partnership.

Manufacturing, Quality, and the TotalProduct Quality (TPQ) Concept

Figure 2 shows the different components of the Total ProductQuality (TPQ) concept. Both Quality and Manufacturingmust integrate the TPQ concept into their routine practicesin order to ensure the effective application of cGMPs toproduct development. The first and central component of the

model identifies the factors that influence the overall productquality. These factors or disciplines should be initiated inPhase I with focus on safety related aspects of the cGMPs. Asthe product moves through the development stages, theemphasis on cGMPs also should increase. DocumentationManagement and Control is a key discipline since documen-tation is critical even in the early stages of product develop-ment.

The second component illustrates the timeline from dis-covery to commercialization. It also depicts the GMP con-tinuum and the Technology Transfer (TT) steps during theproduct development life cycle. Tracking and trending, thethird component is used to indicate that all the factors shownshould be monitored and their performance documented inorder to measure the effectiveness of the concept, and itsapplication to the development process. The more automatedthe tracking and trending process, the easier it is to gatherand access the information.

The Manufacturing and Quality Departments must con-sider the following essential factors to achieve TPQ:

Customer Requirements - before a product can be devel-oped, it is important to have an understanding of what themarket needs and wants. What kind of product, for what

Product Development


purpose, in what kind of container presentation, and bywhen. This is a major challenge since market input is notalways clear, and market forecasts are seldom accurate.When your lead manufacturing time is as long as 9 to 12months, it is critical to be able to predict how much productyou can make by when and what technical and businesschallenges you will have to deal with in the future. Qualityand manufacturing must work together to identify the needsof the process and fit regulatory compliance requirementsinto this process.

People - people are the key to success in any organization.The success of this model is dependent on full participation ofall employees, including senior management. The establish-ment of a “win-win” relationship and effective teamwork iscritical to control quality input and ensure continuous evalu-ation and improvement in the manufacturing process. Properteam structure, leadership, and communication strategyshould be in place to support team members.

There is a need to clearly define roles, responsibilities, andaccountabilities. It should be clear to all concerned that thefinal decision relative to product quality acceptance lies withthe senior management in the Quality Group. It should beevident that the manufacturing department must produceproducts that are of quality, properly validated, designed tomeet the regulatory expectations, and properly documentedfor effective and efficient product distribution. The productmust consistently demonstrate quality performance through-out its shelf life. Finally, the Production Department shouldunderstand their role in product development and technologytransfer, pulling the process into manufacturing and check-ing at every step to ensure that all information they need isavailable, including the rationale for critical decision pointsin the manufacturing process.

Product Definition/Design - quality attributes must bepart of the product design from its inception, but it is alsoimportant to take into consideration the ManufacturingDepartment’s needs. Product attributes should be evaluatedand confirmed during pre-clinical and clinical evaluations.Finally, during manufacturing of consistency runs, theseattributes should be controlled through the use of qualitysystems to monitor and measure performance throughout theproduct supply chain.

Facilities, Equipment, and Utilities - the infrastructuremust be designed, calibrated, validated, used, and maintainedaccording to a continuum of quality principles and regulatorycompliance expectations. Control of equipment and facility isthe first line of defense against problems and cost increasesduring development. The Quality and Manufacturing Depart-ments should be involved in every step, from purchasingthrough calibration, to validation, and should have approvalsignature in critical decision points to ensure that the finalfacility and equipment will meet all requirements.

Specifications - specifications must be designed based on

scientific rationale. Quality and Manufacturing should en-sure that documented rationale is available for raw materi-als, intermediates, packaging components, labeling, and thefinal vial. The acceptance program for raw materials receipt,testing, and release should be defined in writing, and vendorsof critical materials and components should be properlyqualified. The program should be designed jointly betweenManufacturing and Quality, together with R&D and theMaterials Management Department.

Control of Material and Processes - all raw materials,components, product intermediates, and final packagingshould be properly controlled through the establishment ofan effective lot numbering system that will enable lot identi-fication and traceability. This system should be designed byboth Manufacturing and Quality and should help establishthe historical documentation for the entire production, test-ing, and release of the final product. Bar coding should beseriously considered.

In addition to lot identification and control, it is necessaryto ensure that the process is capable of consistent, reproduc-ible, and reliable performance. Process manufacturing datashould help minimize reliance on product testing and helpaddress questions that may arise during routine manufactur-ing. A joint effort between Quality and Manufacturing canhelp ensure the development of a program that works totroubleshoot problems during routine production.

Auditing/Monitoring - auditing and monitoring are impor-tant factors in the assurance of quality. More specifically,manufacturing’s involvement in auditing and documenta-tion review should be considered. A written procedure shouldbe available to describe the auditing function. At a minimum,the quality department should audit and monitor the follow-ing:

• control of sterility assurance• process, equipment, utilities, and cleaning of manufactur-

ing equipment• final product release criteria• container vial integrity studies• control monitoring program - the plan and implementa-

tion• development of documentation, including adequate batch

records and associated release documents.• establishment of change control procedures• establishment of investigation documentation• preparation for licensing inspection• documented training

Again, working with manufacturing to practice self-auditingand commitment tracking of findings is an effective way tohelp them accept responsibility for the first line of defenseagainst quality problems and product waste.

Last, it is necessary to comment on documentation. All thefactors discussed require documentation. One must remem-ber the rule of thumb “if it isn’t documented, it isn’t done.” The

Product Development


partnership team must identify all documents needed forfilings, technology transfer and ultimately, to support thePre-Approval Inspection. The Quality Department must en-sure that all documents are available, are approved, andsecured/controlled according to company and regulatory agen-cies' requirements. However, the Manufacturing Depart-ment, working with the development scientist, also musttake responsibility for ensuring the accuracy, security, andready access to such documentation.

The following are examples of documentation/reports to beconsidered:

• raw materials specification• in-process specifications• SOPs• development batch records• manufacturing batch records• raw data to support clinical batches• deviation reports• investigation reports• stability protocols• validation reports• stability data report• assay method development and transfer reports, and asso-

ciated training documents• manufacturing summary reports for submission• development history reports• final product specifications (rationale)• testing monographs• training records

In summary, the proposed illustration (Figure 2) identify thefactors and associated activities that teams require for bring-ing quality and manufacturing together into a winning part-nership. The effective quality partner should be prepared tounderstand the constraints of the manufacturing process,but also should be prepared to provide the proper guidanceand systems to facilitate the integration of the quality processinto product manufacturing. On the other hand, the manufac-turing partner should accept responsibility for ensuring thatmanufacturing resources also are focused on quality in addi-tion to maintaining production schedule and achieving costcontrol.

Our experience clearly showed us that effective communi-cation between Quality and Manufacturing, together withour management leadership, appropriate staff training, strongemphasis on documentation, clear goals, policies and proce-dures, could indeed help prevent problems and expedite thedevelopment process.

References1. Pharmaceutical Industry Primer - A Century of Promise.

PHRMA, Web site http://www.phrma.org/publications/publications/10.08.2001.528.cfm.

2. GMP/CMC Linkages Begin with Development Phase.The Gold Sheet, October 2002, page 11.

3. FDA Guideline on the Preparation of InvestigationalNew Drug Products (Human and Animal), Department ofHealth and Human Services, Public Health Service, Foodand Drug Administration, Center for Drug Evaluationand research, March 1991, reprinted in November 1992,posted on FDA Web site on March 1998.

4. 21 CFR Part 312, Investigational New Drug Application.

5. Guidance for Industry: Content and Format of Chemis-try, Manufacturing and Controls Information and Estab-lishment Description Information for a Vaccine or Re-lated Product - 1/5/1999.

6. FDA Guidance Document Concerning Use of Pilot Manu-facturing Facilities for the Development and Manufac-turing of Biological Products - 7/11/1995.

7. Points to Consider in the Production and Testing of NewDrugs and Biologicals, Produced by Recombinant DNATechnology - 4/10/1985.

8. Wagner, Carmen M., Bringing Quality Assurance andValidation into Research and Development - Points toConsider, European Journal of Parenteral Science, 1999,vol 4 (1): 12-17.

9. Wagner, Carmen M., and Kohn, Frank S., Utilizing Criti-cal GMP Drivers for Effective Technology Transfer, Sec-ond Annual Technology Transfer Conference, Atlanta,Georgia. Institute for International Research, February22-24, 1999.

10. Wagner, Carmen M., and Clontz, Lucia, Utilizing CriticalGMP Drivers for Effective Technology Transfer, Orlando,Florida, Institute for International Research, February23-25, 2000.

About the AuthorsCarmen M. Wagner, PhD, is President of Strategic Compli-ance International, Inc., in Cary, NC. She provides special-ized services in the area of regulatory compliance, qualityassurance, regulatory affairs, training, and technology trans-fer support for the biotechnology and vaccine industries. Dr.Wagner has held a variety of quality related positions in thehealth industry and has worked in the pharmaceutical indus-try for more than 20 years. Her work experience includestenures at E.I. DuPont, Johnson & Johnson, American Cy-anamid, and Wyeth. More recently, she was vice president,Quality and Technical Services at MERIX Bioscience, Durham,NC, and vice president, Regulatory Compliance and Technol-ogy Transfer at Serentec Inc., Raleigh, NC. While at Wyeth,Dr. Wagner worked closely with Dr. Kohn, director of manu-facturing, to develop a quality philosophy and quality sys-tems that worked well and lead to the successful launching ofnew vaccine products. Dr. Wagner is an active member ofPDA, ISPE, and ISCT.

Product Development


Strategic Compliance International, Inc., 2500 RegencyPkwy., Cary, NC 27511.

Frank S. Kohn, PhD, is currently Presidentof FSK Associates, a consultancy for thebiotechnology and vaccine industries. Priorto starting FSK Associates, Dr. Kohn heldtechnical and managerial positions at Wyeth,Schering Plough Corp., Armour Pharmaceu-tical, and Sanofi Inc. with experience in R&D,Manufacturing, Validation, and Quality. He

has more than 30 years of experience with pharmaceuticaland biological products. Dr. Kohn holds graduate degrees inenvironmental microbiology and operations management.He is certified as a Specialist Microbiologist (SM) by theAmerican Academy of Microbiology. He is a frequent confer-ence speaker and has given more than 150 technical semi-nars and conferences. He also has published in the area oftechnology transfer, cleaning validation, and microbiology inthe US and Europe. He is a member of the ISPE Publications/Internet Committee, current Chair for the Vaccine InterestGroup of PDA, and a member of the Technical Council of PDA.

FSK Associates, Inc., 1899 N. Twin Lakes Rd., Manson, IA50563.

Overencapsulation


Blinding Clinical Supplies UtilizingOverencapsulation

by Robert G. Myers and Colleen M. Cratty

This articlediscusses itemsthat need to beconsideredwhen using theoverencapsulationprocess as ablinding optionfrom componentselection andcomparatorpurchase to theavailability oftrainedpersonnel. During the preparation of clinical stud-

ies, the method for visually blindingthe dosage form is a decision that needsto be made early on in the process.

While there are many blinding options avail-able today, i.e., deprinting, mill and fill, manu-facture of generic drug, and overencapsulationto name a few, the overencapsulation methodstill seems to be the most popular method.Perhaps overencapsulation is not the least com-plex, but it is the most commonly chosen optionfor blinding clinical supplies today.

The following paragraphs detail some of theitems that need to be taken into considerationwhen utilizing the overencapsulation processas a blinding option. There are many items thatneed to be addressed to ensure that the opera-tion runs smoothly, from component selectionand comparator purchase to the availability oftrained personnel.

Overencapsulation is basically hiding an-other dosage form, tablet, or capsule inside acapsule shell. It is important to select theappropriate components that will be needed to

support the overencapsulation of the tablet orcapsule unit. Once the unit has been identified,the first thing to determine is what size capsuleshell will need to be utilized to properly blindeach unit. Although it is not completely neces-sary, it is recommended that the unit that isbeing encapsulated does not protrude abovethe body of the capsule shell when inserted. Ifthe unit does not “sit” properly inside the bodyshell, and backfilling is required, it may be-come necessary to backfill the capsule in amanner that will produce a considerable amountof backfill as waste.

There are various size capsule shells avail-able for blinding and perhaps the most popularis the DB CAPS™ capsule shell.1 This styleshell is typically shorter and larger in diameterthan the standard capsule shell sizes. An addi-tional feature of this style is the double layer ofshell that is created upon closing the capsule.This is due to the walls of each piece of the shellbeing almost identical in length. There is littlearea for the patient to grab a hold of at eitherend of the capsule, making it very difficult for

the patient to pull the cap-sule apart. These two fea-tures not only make the cap-sule more user friendly froma manufacturing stand point,but assist in keeping the drugblinded at the patient level.

While the DB CAPSTM

style shell is the most fre-quently used, it is not un-usual to see the standardsize capsule shells used aswell. Size 0 and size 00 cap-sules are most commonlyused for overencapsulationin this case. In some in-stances, the larger size 000,and smaller size 1, 2, and 3

Figure 1. Light frombeneath the capsule ringilluminates the shell thatdoes not contain theunit to be encapsulated.Here, the unit beingencapsulated is anothercapsule.



Overencapsulation


capsules also are chosen for blinding. Generally, a study willinvolve the breaking of tablets to fit them into the smallercapsule sizes. When tablets are broken, it is critical to ensurethat all of the tablet fragments are collected and accuratelyplaced into each associated capsule shell. If all of the frag-ments are not collected, the final dose of the blinded tabletcould be altered. Other matters to consider when choosing acapsule size is your study population. Is the study gearedtoward children or the geriatric population both of whom mayhave difficulty swallowing larger size capsule shells?

Besides the appropriate capsule size, capsule color is anextremely important detail that requires a lot of insight. It iscritical to choose a color that will completely hide yourenclosed unit. One that does not show any shadowing or airpockets due to the backfill encompassing the unit, or allow forany printing or coloring of the encapsulated tablet or capsuleto be seen, is the color that should be utilized. These aregenerally opaque capsules in nature and are usually not thesame color or shade of the unit being blinded, but ratherslightly darker or more opaque in color.

Not only do we need to ensure that the capsule color willeffectively blind the enclosed unit, but also that the color dyesand pigments used in the color formulation are acceptedwherever the study is being conducted. Many countries haverestrictions on particular colors. This needs to be researchedprior to selecting a color. There are several colors that areaccepted worldwide and capsule vendors should be able toprovide any information with these selections. Capsule ven-dors are capable of producing capsule colors in any imagin-able shade.

Once the color and size of the capsule shell have beenselected, one must be aware of the lead times involved withordering capsule shells. It is possible to have lead times of twoto three months depending on size and color requirements.Vendors normally stock a few colors in quantities of a fewmillion; however, it would be impossible for them to stockevery color and size combination conceivable. This is usuallythe main obstacle, which can delay the start of most studies.Upon receiving the capsule shells, storage becomes an impor-tant issue as well. Be sure that the capsules are stored underthe manufacturer’s recommended conditions. Generally, ifcapsules are stored for more than two years, it is not a badpractice to replace them with a fresh supply, especially if onecannot store them at the recommended temperature andhumidity conditions. Extended periods of storage can createbrittle and distorted capsules. This creates its own difficul-ties once encapsulation begins.

Now that capsule details have been determined, the selec-tion of backfill material becomes the next critical step in theoverencapsulation process. Backfilling the capsules is re-quired to eliminate the rattle of the unit inside the capsuleshell so that the patient is not able to determine the presenceof another dose inside the capsule. If the rattle is not elimi-nated, the patient can possibly break the blind. In rare cases,backfill may not be used and both the placebo and the activedoses contain overencapsulated tablets for similar rattlebetween the doses.

When selecting a backfill material, it is best, but notrequired, to choose an excipient that is present in the dosageform of which you are blinding. This information can usuallybe found on the package insert as well as in the Physicians’Desk Reference.2 Dissolution profiles and stability work shouldbe conducted to verify that the material selected does notinterfere with or create any bioavailability issues in theoverencapsulated dosage form. The most commonly usedexcipients for backfilling are Microcrystalline Cellulose andLactose Monohydrate. These materials are used both inde-pendently of one another as well as combined in a blend. Insome cases, research has shown that the combination of thetwo may improve the dissolution results.3 Depending on thegrade of the material chosen, a lubricant, usually, Magne-sium Stearate, present usually less than 0.5%, is added aspart of the backfill formulation. Not all grades of these twomaterials require such lubrication and the choice of addingthe Magnesium Stearate is usually based on its presence inthe formulation of the unit being encapsulated. Lead timesare usually not an issue with regards to backfill when com-pared to those that may be encountered with ordering capsuleshells.

When running a trial and overencapsulating a commercialproduct, there are additional things to consider besides whatcapsule size the unit will fit in or what backfill formulationshould be utilized. Perhaps the most important thing youneed to keep in mind is who is going to order the comparator.It is important to protect the confidentiality of the companydoing the study, and therefore consideration needs to betaken because suppliers may become aware of what com-pounds the company has involved in such trials. When thecompany conducting the trial orders the comparator, this cancreate the potential for various outside parties to know whichcompounds are being considered. If a second party orders thecomparator, the potential for the manufacturer of the com-pound to know that a study is being conducted against theircompound, as well as confidentiality issues can practically beeliminated.

In conjunction with ordering supplies, be sure that theproper size change parts to run the capsule size on theequipment are on hand. Change part availability could be-come an issue. Typically, there are long lead times and theparts can become rather costly. It is advantageous to have anadequate supply of parts on hand for the machinery. If you arein the middle of a run and the equipment fails, the down timethat can be saved by having items on hand is immeasurable.In this article, a semi-automated capsule filling machine isutilized for the encapsulation process. Additional equipmentsuch as a loading ring and a light table will enhance theprocess making it easy for operators to determine and elimi-nate defects in the drug as well as the capsule shells. Theseitems are discussed in more details within the actual processbelow.

Personnel also can be an issue during the overencapsulationprocess. One must ensure that there are an adequate numberof trained operators available to support the project(s). Theoverencapsulation process can be quite complex when per-

Overencapsulation


formed properly, and requires a dedicated team of trainedpersonnel that are familiar with the common issues thatoccur during the process, from equipment problems to recog-nizing issues created with the materials being used.

From this point forward, we will assume that the equip-ment, room, and personnel are adequately prepared formanufacturing, the components for the project are releasedaccordingly, and all production records have been writtenand approved. With everything in place, we can start toremove the drug from its commercial package. As the drug isremoved, it is important to perform a 100% gross inspectionon the drug that is being overencapsulated. Although theseproducts are commercially manufactured, packaged, andreleased, there are occasions where you may need to rejectentire lots of product due to anomalies found in the product.Examples of these anomalies include everything from brokentablets, crushed capsules, and broken or missing inductionseals, to foreign materials compressed directly into tablets.Once these units are eliminated from the commercial lot, theoverencapsulation process is ready to begin.

When overencapsulating any drug product, the use of atablet or capsule loading ring will enhance the efficiency ofthe process and assist in ensuring that only one unit is placedinto each capsule shell at a time. There are a number ofdifferent style loading systems available. However, the de-sign of the loading ring must be carefully chosen. The loadingring is utilized by flooding the ring with product and thenmanually working one unit into each cavity of the loadingring. This ring is then placed on top of the lower portion of thecapsule ring that contains the bodies of the capsule shells.The drug product is then released into the bodies. If addi-tional units are required in each capsule shell, the process isthen repeated as required.

There are several things to consider when selecting theproper loading ring:

1. The ring should not be made of aluminum. Aluminum hasthe potential to leave black markings on tablets whentraveling across the surface of the ring.

2. The ring should be designed so that each cavity of the ringis size specific to the shape of the tablet or capsule. Eachcavity also should accommodate only one unit at a time.

3. If the loading ring utilizes offset holes to load the units intothe capsule shells, be cautious when working with capletor oval shaped tablets. The ends of the units can get stuckin the cavities and when the spring mechanism is trig-gered to align the holes, the ends of the tablets can bebroken and/or chipped. It is extremely difficult to tell if theentire single unit went into the same capsule shell. Youmay not even know that the tablet has been damaged.

4. Outsourcing of loading rings can sometimes take severalweeks, leading to delays in starting the project.

Prior to placing the loading ring onto the capsule bodies,utilizing a light table underneath the bodies can have manybenefits. The first advantage is that the light will drawimmediate attention to any cavity that is missing a capsule

shell. Second, if the capsule bodies contain any defects suchas pinholes due to a thin gelatin area, usually found on thecapsule ends, the light magnifies these holes and the capsulescan be removed prior to filling. This capsule defect is ex-tremely difficult to detect otherwise. If this defect is notdetected prior to filling, it could result in capsules leakingpowder out of the ends of the capsule shell. If this defect ispresent, it is usually not noticed until the product is packagedand/or distributed, long after the capsules are filled andclosed. Third, once the capsules are filled with the drugproduct, the light will illuminate any empty capsules withoutthe product – Figure 1. Even though the loading ring willrelease a unit into each shell, human error can still result inan empty capsule. It is highly recommended to perform anadditional 200% visual, documented inspection with the finalcheck being completed by the operator responsible for back-filling the capsules, totaling a 300% inspection.

With the units loaded into the capsule shell bodies, thecapsule ring is then transferred to the filling machine. Uponcompletion of filling the first set of capsules, several capsulesshould be checked prior to formally closing the capsules. Thisis to determine if any “rattle” or movement can be felt or heardfrom the encapsulated drug inside the capsule shell. To dothis, remove several capsules by hand, closing them as theyare removed from the ring. If there is noticeable movement,there are several routes to take to “lock” the drug in thecapsule shell. However, be aware that depending on thebackfill and the shape of the unit being overencapsulated,there is a possibility that the movement will not be com-pletely eliminated. If this point is reached, a decision needsto be made on how to proceed. Two options for increasing theamount of backfill present in the capsule are as follows:

1. The ring of capsules may be tapped to settle the backfillaround the drug and a second filling can be done to addmore backfill to the capsule.

2. The auger on the capsule filling equipment can be changedto assist in forcing additional backfill into the capsuleshell, as well as altering the fill settings on the equipment.

Once the capsules are filled, it is typically necessary to polishthe capsules to remove any residual backfill material fromthe outside of the capsule. Having an empty capsule elimina-tor attached to the discharge area on the polisher will elimi-nate any possibility of an empty shell making its way into thefinished product container. An empty capsule can occur if acapsule is crushed during closing or if a capsule does not closeproperly and opens during the polishing process. This ispossible due to the turbulence created inside the polisher.

In low humidity conditions, the residual backfill maybecome difficult to remove due to static. If this situationarises, determine the relative humidity in the room, and ifpossible, raise the humidity to at least 25%, or move theoperation into a room at a higher relative humidity withoutjeopardizing operating conditions.

At this point, the overencapsulation process is basicallycomplete. A final weight check and inspection to assure

Overencapsulation


proper closure and no visual defects are the closing segmentsof production. During the entire overencapsulation opera-tion, there are many quality checks at integral parts of theprocess:

• gross inspection of the drug to be encapsulated• inspection of empty capsules for pinholes,• three documented, visual checks for the presence of the

proper number of units in each capsule body• a formal check for drug movement inside the capsule at the

initiation of powder filling to determine the final fillingrequirements.

A final confirmation in the process is weighing a minimum of10 capsules from every ring to ensure that they are within thedesired range and that the operators are in control of thefilling operation. Samples may be pulled from each ring tocreate a mini-batch of the entire process. Retains and releasesamples are then pulled from this composite.

The finished product upon final inspection should beplaced into a properly lined, tared, and labeled container.Labels should be present on the inside as well as the outsideof the containers for identification purposes. Pre-numberedtamper seals should be placed on the containers and thenumber recorded within the batch documentation. The manu-facturing operation is now complete and the material can bereleased for further processing.

Overencapsulation involves many individual operationsthat can create a variety of complex situations. Additionalcomplexities arise when tablets need to be broken to fit intocapsule shells, or half of a tablet needs to be placed into acapsule shell. Even different doses may be combined into thesame capsule to meet specified dose requirements. Analyticalsupport becomes extremely important when creating a newdose or altering the original form.

With the potential of using any of these different sce-narios, overencapsulation remains the most sought aftermethod of blinding drugs for clinical trials. The key to success

during each of these operations is to remember that eachsegment of the process is extremely important. Each requiresproper planning and careful execution. From capsule colorand size selection to having a well-trained team dedicated tothe manufacturing process, taking the time to make sureeach segment is managed properly will result in minimalproblems related to the encapsulation portion of the study, aswell as curtail problems that could occur once the patientreceives the supplies.

References1. DB CAPS™is a registered trademark of CAPSUGEL.

2. Physicians’ Desk Reference. Medical Economics Com-pany, Inc. Montvale, New Jersey.

3. Faust, Mary Beth, “Effect of Variations in Backfill on Disso-lution for an Over-Encapsulated Comparator Product,” Phar-maceutical Engineering, May/June 1999, pp. 48-54.

About the AuthorsRobert G. Myers is currently the Managerof Clinical Manufacturing for Fisher ClinicalServices Inc. He has 25 years of experience inthe areas of manufacturing of solid dosageforms, clinical manufacturing, sterile opera-tions, formulation, and development.

Colleen M. Cratty is currently a Managerin Client Services for Fisher Clinical Ser-vices Inc. Her experience is in the area ofproject management and manufacturing ofclinical supplies for solid dosage forms. Crattyholds a BS in chemical engineering from ThePennsylvania State University.

Fisher Clinical Services Inc., 7554 SchantzRd., Allentown, PA 18106.

Design Qualification


Enhanced Design Review/DesignQualification

by Robert E. Chew, PE

EnhancedDesign Review/DesignQualification isan investmentthat paysdividends notonly to enhanceregulatorycompliance, butalso to improveproject deliveryand streamlinequalificationefforts. European Regulatory Agencies, and now

the US FDA, have an expectation fordesign qualification for certain types ofpharmaceutical and biotech projects. De-

sign qualification is more than a standardowner engineering approval, and it is morethan evaluating the design against a genericchecklist of common GMP/GEP attributes andpractices. Design qualification is an examina-tion of the design against stated requirements.Design qualification can help focus IQ/OQ/PQefforts on those aspects of the design (fabrica-tion, installation, operational, performance)that could impact process performance andproduct quality. Enhanced design review/de-sign qualification should be conducted in amanner that helps focus and streamline theoverall project delivery process. Efforts ex-pended on enhanced design review/DQ early inthe project will pay dividends during the finalstages of the project.

BackgroundThe ISPE Commissioning and Qualification(C&Q) Baseline® Guide espouses the concept ofan Enhanced Design Review, which is definedas “a documented review of the design, at anappropriate stage in a project, for conformanceto operational and regulatory expectations.”1

The Guide further explains the concept as:

• structured review of the design of facilities,utilities, and equipment

• a smart way to prepare for IQ and OQactivities

• a method of examining the design for impactof the system, system complexity, and de-gree of novelty or familiarity

• a review to verify that design links construc-tion to the User Requirement Specification

• a review to verify that the owner gets whathe asked for

Annex 15 of the EC Working Party on Controlof Medicines and Inspections, Qualificationand Validation, states under Design Qualifica-tion:2

“The first element of the validation of newfacilities, systems or equipment could bedesign qualification (DQ).”3

“The compliance of the design with GMPshould be demonstrated and documented.”

The ICH Q7A Good Manufacturing PracticeGuidance for Active Pharmaceutical Ingredi-ents was adopted last year by CBER and CDER.4

This document states under Qualification §12.3:

“Design Qualification (DQ): documented veri-fication that the proposed design of the fa-cilities, equipment, or systems is suitablefor the intended purpose.”

The FDA has indicated that it will have anexpectation for design qualification to be per-formed on API (and bulk biotech) facilities.

Finally, IEEE has standards on softwarequality assurance which include the concept ofa Functional Audit. A Functional Audit is de-fined as verification that all requirements havebeen implemented in the design.

Importance of Requirementsvs Design

Ultimately, the facility, equipment, and sys-tems must meet the cGMP requirements. The





selected design is merely a particular method of achieving orimplementing those requirements. And, there may be manyways to meet a requirement. Hence, whether or not the designis met is secondary; if cGMP requirements are met, then thefacility, equipment, and systems are in fact qualified, whetheror not they met each and every design attribute contained indrawings and specifications.

In most cases, there are aspects of the design which arecritical to meeting a particular requirement. For example, inorder to achieve cleanroom conditions, HEPA filter efficiencyand integrity are important. In order to achieve low microbiallevels in water systems, a combination of design features isnecessary. Although it is only necessary to satisfy require-ments, it is prudent to develop a solid design which providesa high degree of assurance those requirements will ulti-mately be met.

Of course, to emphasize the importance of requirementsmeans that rigorous and thorough requirements definitionmust occur, these requirements must be documented andsupported by development data, and they must be main-tained under strict QA change control throughout the project.

What is Enhanced Design Review/Design Qualification?

The ISPE Commissioning and Qualification Baseline® Guide,ICH Q7A, and EC Annex 15 all explain Enhanced DesignReview (EDR) or Design Qualification (DQ) in terms ofreviewing the design for conformance to operational andregulatory expectations, GMPs, and suitability for the in-tended purpose, as well as the User Requirements andFunctional Design as outlined in the Guide. The ISPE Com-missioning and Qualification Baseline® Guide contains addi-tional guidance and suggestions on how to implement andintegrate an EDR within the overall design development andreview process.

Specific regulatory language, which should be used todefine and structure the EDR/ DQ process, includes:

• …first element of validation…(Annex 15); …proposeddesign…(Q7A): this indicates that EDR/DQ is performedbefore the design is purchased or implemented in the field.

• …documented verification…(Q7A); … and documented(Annex 15): this indicates EDR/DQ is a documented activ-ity.

• …verification… (Q7A); …demonstrated…(Annex 15): thisimplies an additional level of documentation detail overand above a simple sign-off that the design has beenreviewed and approved for implementation. It impliesreview by persons who can evaluate the design from atechnical point of view as adequately addressing eachrequirement.

• …compliance of the design with GMP…(Annex 15); …de-sign is suitable for the intended purpose...(Q7A): this indi-

cates that in order for EDR/DQ to proceed, the GMPrequirements and intended purpose, e.g., user require-ments, must have been previously defined.

Enhanced Design Review/Design Qualification Approach

As indicated above, the EDR/DQ cannot be only an engineer-ing judgment-based (hip-shot) review of a design with someform of QA participation and formal sign-off. It should becentered about a more structured analysis based on a founda-tion of well-defined requirements with corresponding docu-mentation. In addition, the EDR/ DQ process can includeassessment of component impact on product quality (asdescribed in the ISPE C&Q Guide).5 Outputs of this processcan form the basis of IQ/OQ protocols, inspections, andtesting, as well as identification of critical instruments,ranges, accuracies, etc. against Acceptance Criteria estab-lished in Functional Design.

Well-defined requirements are the foundation of any basisof design. However, pure requirements differ from the tradi-tional Basis of Design (BOD) packages issued by manyengineering firms. Differences include:

• User, process, safety, environmental, quality, and regula-tory requirements should be established long before com-mencing the design, whereas much of the engineeringBOD is the Preliminary Design or a skeleton thereof.

• The engineering BOD typically consists of text para-graphs, preliminary drawings, equipment lists, etc. Well-defined requirements consist of a set of statements thatcould form the basis of inspection and test acceptancecriteria.

• It should be straightforward to establish a process tomanage changes to individual requirements, whereasmanaging changes to a BOD package may be more difficultor complex.

A suggested Best Practice is to define a comprehensive set ofrequirements, first at the user level, and later at the moredetailed functional requirements level. Requirements shouldencompass process, quality, maintainability, capacity, gen-eral facility, operability, safety, environmental, and regula-tory topics. Requirements should be short, distinct state-ments with measurable goals that can later be verifiedthrough a commissioning or qualification program. Eachrequirement should be uniquely identified to facilitate changemanagement.

Requirements must be comprehensive. Each and everyrequirement relating to product safety, identity, strength,purity, and quality must be identified. Hence, QA must havea significant role in reviewing and approving the final set ofrequirements, and must be an approver of changes to anyrequirement that can affect the above product or processattributes, e.g., GMPs.



Given a comprehensive set of requirements that has beenapproved by QA and is under project change management,the EDR/DQ process then can be reduced to two key objec-tives:

1. Documented verification that the overall design appearsto address, by some means, each and every requirementaffecting the product and performance of the manufactur-ing process. Or, in the case of unknown product or multi-product manufacturing facility, the required equipment/system performance capabilities.

2. Identification (and documentation) of the critical indi-vidual physical components, attributes, and operationalfeatures that directly support meeting each requirement.

Figure 1 illustrates how the EDR/DQ and component impactassessment process can work within the context of the “V-model” which describes the relationship between require-ments, design, and IQ/OQ/PQ.

Objective #1 satisfies the requirements of Annex 15 andQ7A. A suggested method for accomplishing Objective #1 is topresent the requirements in the form of a spreadsheet ordatabase, and document the components of systems and theirparticular drawing(s) and specification(s) section(s) that ap-pear to meet the given requirement. This annotated spread-sheet can then be attached to a generic DQ protocol, either ona system by system basis or for the project as a whole, andformally approved by engineering, user, QA, and other af-fected groups.

Objective #2 may be used to streamline qualificationefforts and the design change control process. According tothe C&Q Guide, IQ, OQ, and PQ efforts are focused on “thoseattributes…that can affect product quality.” Objective #2seeks to identify the attributes (critical parameters anddirect impact components.) that can affect product quality(e.g., which support meeting a GMP quality requirement);hence those and ONLY those attributes need be included inany IQ, OQ, or PQ protocol, as appropriate. Other aspects ofthe design, which may, in the past, have been included in IQ/OQ/PQ protocols (and may have resulted in inconsequentialdeviations to be laboriously processed), can be excluded fromthose protocols and instead incorporated into non-GMP com-missioning efforts under control of Good Engineering Prac-tice (such items might include electrical checks, lubricationand alignment, continuity checks, utility supply checks, make/model/serial number data, nameplate data, purchase orders,IOM manuals, etc.).

There are several suggested methods by which Objective#2 may be accomplished, while at the same time meeting theC&Q Guide expectations for component level impact assess-ment and definition of critical instruments:

• Using the same mechanism as described for meetingObjective #1 (annotated spreadsheet), identify the critical

attributes, components, operational features, etc., thatappear to impact or support meeting each GMP require-ment, i.e., a list of critical components in direct impactsystems.

• Using drawings, instrument lists, equipment lists, andspecifications, annotate the product or process require-ment (by number(s) that the particular component sup-ports or impacts).

The DQ protocol also should include a list of the drawings,specifications, etc., that were reviewed as part of the designreview/qualification process. The completed protocol shouldthen be approved by QA. QA should participate in the designqualification process to ensure that the design meets quality-related requirements including critical parameters, sanitaryconstruction, cleanability, environmental controls, etc.

The DQ process including pre-requisite requirements defi-nition, as integrated with the overall project, might proceedas follows:

1. Users define requirements, typically as part of the concep-tual design phase.

2. Requirements baseline (rev 0) are approved and placedunder project change control, approximately the samepoint as conceptual design is complete. QA is key approverof requirements.

3. Requirements are added, modified, or deleted as the projectproceeds. QA approves any changes to GMP requirements.

4. As major equipment is ordered, the bid packages arereviewed to confirm the appropriate user requirementshave been included. This is the first step in EDR/DQ, andshould be documented. QA involvement is not required,but may be advisable at this point.

5. Functional requirements definition proceeds in parallelwith preliminary design.

6. Detail designs are received from equipment vendors andengineering design firms. Formal DQ is performed on thedetail design to confirm the designs appear to meet allrequirements, and to identify critical components. Re-quirements that cannot be met by the design should beevaluated, and acceptable solutions (which may meandeleting or changing a requirement) reviewed and ap-proved by the group which defined the original require-ment. Such decisions are documented in the DQ report andreflected in updates to the user requirements. QA isinvolved at this stage and approves the completed DQreport, along with the system owner/user. Ultimate re-sponsibility for ensuring the design incorporates all re-quirements, and for identifying critical components, restswith the manufacturing firm as represented through QAand owner/user.



Figure 1. Design qualification confirms that all requirements and general cGMPs have been included in the design.

7. Software detailed design is reviewed against the func-tional requirements in a similar manner.

8. The detail design is placed under project change manage-ment.

9. Qualification confirms that the critical components of thedesign meet specification, and that the user and functionalrequirements have been achieved in the delivered system.Given a thorough DQ with component impact assessment,generation of IQ/OQ protocols is much easier, faster, lessexpensive, and less controversial as to what should beincluded.

Controlling Changes to a “QA-Qualified/Approved” Design

Now that QA has participated in the EDR/DQ process, andhas presumably approved the design as a result, how shouldwe control changes to the design? QA, via EDR/DQ, hasconfirmed that the design appears to meet all requirementsand cGMPs. How do we assure that the final design imple-mentation also will accomplish this? One school of thought isthat since QA has approved it, then any changes thereafter,e.g., to any drawing, specification, or vendor design submittal

also must be approved by QA. This also assumes that thereare no errors in the design and that everything will gotogether without field changes. This is unrealistic. Imple-mentation of such a process is probably cumbersome andunnecessary.

The ISPE Water and Steam Systems Baseline® Guide,which has been favorably reviewed by the FDA, offers alter-natives and options regarding the design of high purity watersystems. In other words, there is no single design solution;many may be acceptable. The Agency’s comments to theWater and Steam Systems Guide stated that the degree towhich GEPs are included in any design is a risk assessmenton the part of the manufacturer;6 QA should be involved inthat risk assessment as part of EDR/DQ. In reviewing theISPE C&Q Guide, the Agency stated that the engineeringchange management system shall allow for QA review andapproval of changes that: (i) change the User RequirementsSpecification; (ii) fundamentally change the operational con-cept; (iii) alter a system’s potential impact on product quality.Hence, QA should be involved in approving all changes torequirements, and approving a selected subset of all designchanges.

Second, QA is involved in the following two key projectprocesses:



1. definition, approval, and modifications to the comprehen-sive set of requirements, especially those tagged as “GMP”requirements, critical parameters, and the like (changesto non-GMP items are controlled by good engineeringpractices, e.g., project change management)

2. approval of qualification and process validation protocolswhich should verify that each and every GMP requirementhas been met by the installed facility

The exact mechanism by which a given requirement is met issecondary, provided the requirement is ultimately satisfied.The mechanism (the “how”) is the design. Qualification pro-tocols will confirm that all requirements and critical designfeatures have been met.

Therefore, it is not necessary for QA to approve everydesign change. Instead, the onus should be placed upondesign engineers, given a completed and approved EDR/DQthat identifies critical aspects of the design, to recognizewhen a proposed design change might affect the ability tomeet a requirement. When this is the case, the designersshould request QA involvement in approving that particularchange (thus meeting FDA expectations as stated in theirreview of the C&Q guide). If the design group fails in thisregard, the QA approved qualification protocols should causeinspections and/or testing to identify any non-conformance torequirements or critical design features when those protocolsare executed.

Suggestion: If the requirements are maintained in a search-able database, and if the EDR/DQ process is conducted inaccordance with the best practices described above, then thedatabase could reference those drawings/specification sec-tions which meet each requirement. If a design engineerwants to change a drawing or specification section, he or shecould search the database for any listing of that particulardrawing or specification and would then be made aware of thepotential for the change to impact the ability to meet theparticular requirement. Identification of design changes thatcould have GMP impact then becomes easy with electronicaids.

A more comprehensive (and yet simple) solution would beto include QA on the distribution of every change. Since theproject is still under construction and no product is beingmanufactured, the change management process could allowfor QA notification in all cases, but not require QA pre-approval of changes; QA can be afforded the opportunity toquestion any change of their choice and make objections asthey see fit without implementing a pre-approval require-ment during the design and construction phases of projects.

Finally, the IQ/OQ/PQ process will examine the installed,operational version of the ultimate design including changes;this will demonstrate acceptable conformance to require-ments and will force evaluation of any deviations of criticalcomponents from specifications. This provides sufficient con-trols and quality assurance of the delivered system prior toproduct validation.

Timing of EDR/DQThe ISPE C&Q Guide places EDR/DQ after detailed designand prior to implementation in the field. However, there maybe reasons to conduct a preliminary EDR/DQ as part ofreviews of the Basis of Design and Preliminary Design. Thepreliminary EDR/DQ could be restricted to Objective #1,verification that every requirement has been factored intothe design at that stage of the design development. Thisprevents incomplete or inaccurate design packages frombeing issued to groups engaged in detailed design work. EDR/DQ should be structured around individual systems, majorequipment, support utility systems, HVAC, facility architec-tural, etc. And, the EDR/DQ may need to be revisited whensignificant field changes are to be implemented. Hence, EDR/DQ is an ongoing process that ends when the design isimplemented in the field. Once implemented, the remaining“Q” processes (IQ, OQ, PQ) take over to verify that the finaldesign, as implemented, meets requirements.

Case StudyA recent project to expand the fermentation capacity of anestablished process used the concepts of thorough user re-quirements definition followed by formal design qualifica-tion. User requirements were derived from sources including:

• voided production tickets from the existing process• existing fermentation requirements for the DCS• process flow diagrams for the existing system• P&IDs from the existing system• process qualification report for the existing system• control charts• periodic product quality evaluation that was done in 2000• key process, product, and user requirements (interviews)• regulatory requirements (OSHA, FDA, NDA, Regulatory

commitment documents)• [company] corporate standards• industry requirements (NIST, NFPA, ISO)

Interactive meetings were held where requirements weredisplayed and edited to incorporate comments. The final setof requirements were captured in a database (spreadsheet)and made available to the project team on a shared file serverwith read-only access.

Lessons learned: (1) There were no drawbacks to havingwell-defined user requirements. They formed the basis forthe entire project, through design, construction, and qualifi-cation. (2) There is a fine line between too general and toospecific. They must be specific enough to serve as a basis forinspections and testing, but not too specific such that they aredictating design. For example, the user requirement shouldnot specify the material of construction as 316SS, but ratherthat “materials of construction should be non-corrosive, non-absorptive, non-additive…” (3) Because of the tight linkbetween user requirements and design qualification, moretraining of participants on the DQ process prior to defininguser requirements would have been helpful.



Design qualification was an organized approach at evalu-ating the design. It was a systematic way to make sure theuser requirements were being met before the materials wereactually ordered. DQ caught several design errors andprompted discussions that the technical operations groupand the engineering group might not have had for monthsdown the road. DQ was conducted by completing worksheetsand archiving in a design notebook that will serve as areference of design decisions and design evaluation, as wellas a record that the design appeared to meet all require-ments.

Summary and ConclusionEnhanced design review/design qualification is both an emerg-ing regulatory requirement and a tool for project success. Ifstructured and executed properly, it can help assure thedesign is robust and complete, giving the project team and QAreviewers a high degree of confidence that every requirementhas been factored into the design. EDR/DQ can identify anddocument the critical, direct impact components, and opera-tional features. If these items are clearly identified with asound basis, then IQ, OQ, and PQ protocols can be stream-lined and focused, yielding a more effective and efficientqualification process. Finally, EDR/DQ cannot be performedwithout first formally defining requirements. Meeting re-quirements is the ultimate objective of the project; verifyingrequirements have been met is the role of qualification.

References1. ISPE, Commissioning and Qualification Baseline® Guide,

Volume 5, pgs 75-81.

2. European Commission, Working Party on Control of Medi-cines and Inspections, “Revised Draft Qualification andValidation,” pg. 5.

3. Note that in this context, validation is used as a broad termencompassing IQ, OQ, PQ, and process validation. Quali-fication, as defined in the ISPE C&Q Guide, includes IQ,OQ, and PQ and is focused on facilities, equipment, andsystems. And, by extension, DQ also would be included asa fourth element of qualification. Process validation is notpart of qualification.

4. Guidance for Industry, “Q7A Good Manufacturing Prac-tice Guide for Active Pharmaceutical Ingredients,” USFDA, August 2001, pg. 38.

5. ISPE Commissioning and Qualification Baseline® Guide,pgs 30-32.

6. ISPE Conference, Principles and Applications of Waterand Steam Systems Baseline® Guide.

About the AuthorRobert E. Chew, PE is Vice President ofCommissioning Agents, Inc. He has 12 yearsof experience in the pharmaceutical industryin areas of compliance, commissioning, andvalidation. He is a former Navy submarineofficer and is a registered professional engi-neer. He has a degree in chemical engineer-ing from Case Western Reserve University.

He can be contacted at: [email protected], or phone:1-317/ 710-1530.

Commissioning Agents, Inc., 1515 Girls School Rd., India-napolis, IN 46214.

Human Insulin Interaction


Human Insulin Interaction withSoybean Powder

by Antoine Al-Achi, Thomas J. Clark, III,Robert Greenwood, and Shylock Sipho Mafu

Thecharacterizationof soybeanpowder and theadsorptionprofile of insulinon the surfaceof the powderare presented inthis article.

The objective of this study was to eluci-date the mode of interaction betweenhuman insulin and soybean powder.Soybean has been shown to contain

substances that have anti-trypsin and anti-chymotrypsin activities, which may potentiallyprotect insulin from degradation. Soybean pow-der had an average size particle of dvs = 45.5 mm(an average diameter for a sphere having thesame volume and surface area as the particles)and dvn = 31.6 µm (an average diameter for asphere having the same volume and number ofparticles per unit weight as that of the testedpowder). The powder (mean ± s.d., n = 6) had abulk volume of 44.5 ± 0.55 ml and a bulkdensity of 0.625 ± 0.0085 g/ml. The angle ofrepose of the powder was 38.33 ± 1.106° and itsCarr index was 26.72 ± 1.06%. The findingsfrom this study showed that the mode of inter-

action between human insulin and soybeanpowder is of an adsorption type. Langmuir andFreundlich isotherms, two commonly usedmodels for adsorption, showed that humaninsulin was able to adsorb on the surface of theparticles, interacting weakly with its adsorp-tion sites. Thus, soybean particles can act as aphysical carrier for insulin, yet insulin is easilyreleased from the carrier due to its weak bond-ing with adsorption sites.

IntroductionHuman insulin is a hormonal drug that is aproduct of biotechnology. It contains 51 aminoacids with a molecular weight of 5,808 daltons.Endogenous insulin is secreted by β cells of thepancreas which is then delivered to the circula-tion via the portal vein. The entire humanpancreas contains at any given time about 8 mg

Figure 1. Langmuirisotherm: C/V = 0.0704+ 0.0094 C (r = 0.87,p < 0.0001). V is theamount of insulin (U)adsorbed per 1 g ofsoybean powder at anygiven equilibrium insulinconcentration C.





of insulin (about 200 biological units). Human insulin, beinga peptide, cannot be given orally due to its degradation byproteolytic enzymes such as pepsin, trypsin, and chymot-rypsin. Its main mode of administration has been theparenteral route.

Soybean (Glycine max), also commonly known as soy orsoya bean, is a leguminous plant that is important in Asianculture. Soybean seeds are almost spherical in shape andyellow. Some varieties of the seeds are found to be brown orgreen in color. Both the seeds and the plant itself are referredto by the same name, soybean.

There are two main components of the soybean seed:protein and oil. The protein accounts for about 40% while theoil is about 20%. Soybean contains various kinds of proteinsthat were shown to inhibit trypsin and chymotrypsin pro-teolytic activities. The whole soybean contains between 16.7to 27.2 mg of trypsin inhibitors per gram.2 The major inhibi-tors are Bowman-Birk (up to 4.9 mg/g, molecular weight 7848daltons) and Kunitz (1.1 - 19 mg/g, molecular weight 20083daltons).2,3 Other major constituents in soybean are phyticacid (1.0 - 2.3 g/100 g dry matter), saponin (0.09 - 0.53 g/100g dry matter), and isoflavone (1200 - 4200 µg/g).2 Lyophilizedsoybean powder obtained from Sigma (St. Louis, MO) con-tains approximately 80% protein. One milligram of thispowder inhibits 3-5 mg of trypsin and 2-5 mg of chymotrypsin.Crude soybean soluble powder (1 mg) inhibits about 1 mg oftrypsin.

The main objective of this project focuses on elucidatingthe mode of interaction between human insulin and soybeanpowder. This study is an important step in examining amethod to potentially protect insulin from degradation ifgiven orally.

Materials and MethodsMaterialsSoybean (Glycine max) was purchased from a local store inNorth Carolina. Human insulin (Humulin R, 100 U/ml, Lilly)was obtained from NC Mutual, North Carolina. Avicel wasobtained from FMC Corporation (Newark, DE). All otherchemicals were from Sigma, St. Louis, MO, except for aceto-nitrile which was from Fisher Scientific (Pittsburgh, PA). Allreagents were of analytical grade, except those used in theHPLC assay were HPLC grade.

Methods1. Preparation of soybean powder: a total of 2 kg of soybean

were ground in a coffee grinder (Mr. Coffee, SunbeanProducts, Hatiesburg, MS) in 57 ± 3.59 g weights for 50seconds to yield a fine “expresso” powder. The resultingpowders after each grinding procedure were combinedand blended together using a V-blender set at 25 r.p.m. for20 minutes. The resulting soybean powder mixture wasstored at 4°C until further experimentation.

2. Characterization of soybean powder: the soybean powderwas subjected to various tests in order to establish itscharacteristics.4 Also, as reference points, Avicel, calciumcarbonate, and lactose were subjected to the same tests asthose applied to soybean powder.

2A.Average particle diameter: 10 mg of powder were dis-persed in 10 ml of light mineral oil. One drop of theresulting dispersion was then examined under light mi-croscopy using a reticle (10 µm/division) at 10x lens. Sixdifferent samples from the powder were examined; thehorizontal width of 100 particles from each sample wasrecorded, and the volume-surface diameter (dvs) and thevolume-number diameter (dvn) were estimated:

dvs = Σ di ni3/Σ di n2 (1)

dvn = [Σ di ni3/Σ ni]1/3 (2)

where di is the midpoint of the interval i and ni is thenumber of particles within the interval i. Since the par-ticles within the powder vary in shape, size, and form,methods have developed to relate the average diameter ofparticles in a powder to that of a sphere. These equivalentdiameters describe diameters for a sphere having thesame volume and surface as that for the particles (dvs) orthe same volume and number of particles per unit weightas that of the tested powder particles (dvn).

2B. Bulk volume and bulk density determination: soy-bean powder was introduced inside a 100-ml graduatedcylinder to the 50 ml mark. The cylinder was then droppedthree times on a bench-top from a distance of about 1 inchheight at 2-second intervals. The tapped final volume ofthe powder was tapped bulk volume (Vb) in milliliters.

Powder dvs (µm)a dvn (µm)b

Soybean 45.5 ± 12.8c 31.6 ± 5.9

Avicel 25.3 ± 8.4 14.1 ± 7.6

Calcium Carbonate 18.9 ± 4.0 16.0 ± 2.8

Lactose 29.6 ± 5.2 19.7 ± 2.9a Volume-surface mean.b Volume-number mean.c Mean ± S.D. of 6 observations.

Table A. Average particle size of soybean powder as compared tothat of Avicel, calcium carbonate, and lactose.

Table B. Bulk volume and bulk density for soybean powdercompared to those of Avicel, calcium carbonate, and lactose.

Powder Bulk Volume (ml) Bulk Density (g/ml)

Soybean 44.5 ± 0.55a 0.629 ± 0.0085

Avicel 47.8 ± 0.75 0.416 ± 0.0064


Lactose 45.0 ± 0.89 0.620 ± 0.0130a Mean ± S.D. of 6 observations.



The tapped bulk density (g/ml) of the powder was calcu-lated from:

Tapped ρb = Weight (g)/Vb (3)

The bulkiness (ml g-1) of the powder was estimated bytaking the inverse of the tapped ρb.

2C. Angle of repose: glass, stemmed funnel was placedinto a clamp of a ring-stand, 5.7-cm off the bench-top. Apaper was centered and placed underneath the funnel’sstem. Soybean powder was poured through the funnel,making a mound until it touched the funnel’s stem. Acircle around the mound was then drawn, and its diam-eter was measured using a ruler. The angle of repose (θ)was estimated from

tan θ = 2h/D (4)

where tan θ is the coefficient of friction, h is the height ofthe mound, and D is the diameter of the mound.

2D. Carr index: soybean powder was introduced into a100-ml graduated cylinder to the 100 ml mark. Thecylinder was then dropped 75 times on a bench-top froma distance of about 1 inch at 2-second intervals. Thetapped volume was recorded in milliliters. Carr index wascalculated from:5

Carr Index = [(Initial volume - Final volume)/Initial volume] × 100 (5)

3. HPLC assay for human insulin: human insulin concen-tration was determined using a HPLC assay with the

Powder tan θθθθθa θθθθθ (o)

Soybean 0.79 ± 0.03b 38.33 ± 1.106

Avicel 0.68 ± 0.03 34.45 ± 0.948


Lactose 1.02 ± 0.03 45.63 ± 0.918aThe Coefficient of Friction.bMean ± S.D. of 6 observations.

Table C. Angle of repose (θ) of soybean powder compared to thatof Avicel, calcium carbonate, and lactose.

Powder Carr Index (%)

Soybean 26.72 ± 1.06a

Avicel 16.00 ± 0.63

Calcium Carbonate 32.83 ± 3.82

Lactose 25.80 ± 2.23aMean ± S.D. of 6 observations.

Table D. The Carr index of soybean powder compared to Avicel,calcium carbonate, and lactose.

Figure 2. Langmuir isotherm: x experimental data and ςfitted curve: Y = θ = (b C)/(1 + b C). Where θ is the fraction of the surfacecovered, b is fraction of the adsorption rate constant to desorption rate constant, and C is equilibrium concentration.



Figure 3. Freundlich isotherm model: log V = 1.32 + 0.37 log C(r = 0.75, p = 0.002). V is the amount of insulin (U) adsorbedper 1 g of soybeans powder at any given equilibrium insulinconcentration C.

following specifications: Protein C4 column; ConstaMetric4100 solvent delivery system; Waters Lambda Max Model481 LC Spectrophotometer; Waters 712 WISP; Waters740 Data Module; flow rate: 1 ml/min; wavelength: 215nm; injection volumes: 5-20 µl; and a mobile phase ofacetonitrile, water, trifluroacetic acid, and hexanesulfonicacid-sodium salt (30:70:0.1:0.1). The absorbance of hu-man insulin was linear over a range of 0.5-50 U/ml (r =0.99).

4. Incubation of human insulin with soybean powder: 1 mlof human insulin solution (100 U/ml) was mixed withvarying amounts 5, 10, 50, 100, and 200 mg of soybeanpowder dispersed in 1 ml of water. The mixture was thenincubated for 1 hour at 37°C. Following the incubationperiod, the mixture was centrifuged, and the amount ofinsulin remaining in the supernatant was determined byHPLC.

In a second set of experiments, varying concentrations(10, 30, 50, 80 U/ml) of insulin were used with a fixedamount of soybean powder (200 mg). 1 ml of insulinsolution was mixed with a dispersion of 200 mg soybeanpowder in 1 ml of water. The mixture was then incubatedfor 1 hour at 37°C. The dispersion was then centrifugedand the concentration of insulin in the supernatant wasdetermined by HPLC.

Results and DiscussionThe average particle size, dvs, is the average diameter corre-sponding to a sphere with the same volume and surface areaof those of soybean particles. Similarly, the average dvn

corresponds to a sphere’s diameter with the same volume andnumber of particles as those of soybean particles.4 Table Ashows the average particle size of soybean powder as com-pared to those of Avicel, calcium carbonate, and lactose.Soybean particles are significantly larger (dvs = 45.5 mm anddvn = 31.6 µm) than any of those other powders investigated

in this study. In theory, given everything else is the same, thelarger the particles, the better is the flowability of a powder.The bulk volume of soybean powder was similar in magnitudeto that of the other powders - Table B. However, its bulkdensity (ρb = 0.63 g/ml) was similar to that of lactose, andsignificantly higher than that of Avicel or calcium carbonate- Table B. The coefficient of friction (tan θ) of soybean powderwas closer in magnitude to Avicel than the other two powders- Table C. This is significant, since Avicel is known to be a goodflowing powder. The Carr index (CI) (Table D), also known asthe compressibility index, is a measure of the powder’s abilityto compress into a compact mass under pressure. It is also anindication of the powder’s flowability; the smaller the CI thebetter the flowability. Soybean powder’s CI value is similar tothat of lactose; its value is better than that of calciumcarbonate, but worse than that of Avicel. Powder flowabilityis an important parameter in tableting and capsuling dosageforms, where the powder’s ability to flow from one compart-ment of a tablet press, e.g., the hopper, to another is ex-tremely important during the tableting process.

When human insulin was incubated with soybean powder,some of the insulin was found to be associated with powder.To examine the type of association between insulin andsoybean, two models were selected. Freundlich isothermmodel was originally derived for describing the adsorption ofgas molecules on the surface of solid. Similarly, Langmuirisotherm was used to describe the adsorption of the gasmolecules forming a monolayer on the surface. These twomodels are traditionally used in pharmaceutical applicationsto describe the adsorption of drug molecules on the surface ofsolid materials. Parameters obtained from these two modelscan explain in part the mode of interaction between insulinwith its binding sites on soybean. The data was fitted to aLangmuir isotherm (Figures 1 and 2):

C/V = 1/(b Vm) + (1/Vm) C (6)

where C is the equilibrium concentration of insulin, V is theamount of insulin (U) adsorbed per 1 g of soybean powder atany given C, b is the ratio of the adsorption rate constant tothe desorption rate constant, and Vm is the maximum amountof insulin adsorbed per 1 g of powder. Based on this model, theb value for human insulin adsorption on soybean powder’ssurface was 0.13, indicating that the rate of desorption isfaster than that of adsorption, which implies that the strengthof interaction of insulin with the adsorption sites on thesurface of the particles is relatively weak. The amountadsorbed reaches a maximum value Vm of 106.0 U/g.

The data was also fitted to a Freundlich isotherm model(Figure 3):

log V = log k + (1/n) log C (7)

where k is the amount of insulin (U) adsorbed per 1 g ofsoybean powder at an equilibrium insulin concentrationequal to unity (1 U/ml), and n is a constant. The value of theslope (1/n) reflects the strength of the interaction between



insulin molecules and the adsorption sites on the surface ofsoybean particles, and V and C are as defined above. Thehigher the value of the slope, the stronger is the interactionbetween insulin and the adsorption sites. This model showedthat the value of k was 20.9 U/g and n = 2.7. This also leadsto the conclusion that the strength of the interaction betweeninsulin molecules and adsorption sites on the surface wasrelatively weak (implies a negligible affinity less steric inter-actions). This weak interaction is favorable since soybeanpowder acts as a carrier for insulin and is able to adsorb thehormone, but yet capable of releasing it when needed. Fur-ther studies are in progress to examine the interactionbetween the soybean-insulin complex and the various diges-tive enzymes in vitro.

ConclusionThis study demonstrated that human insulin adsorbs on thesurface of soybean powder particles. This adsorption is aphysical one since it was reversible. The strength of theinteraction between insulin and the adsorption sites wasweak. This may allow soybean powder to act as a carriersystem for insulin easily capable of releasing the hormone.

References1. Karam, J.H., Pancreatic Hormones and Antidiabetic Drugs;

in Basic and Clinical Pharmacology, Katzung, B.G., (edi-tor), Lange Medical Books, McGraw-Hill, New York, 1998,p. 684-705.

2. Anderson, R.L., and Wolf, W.J., Compositional Changes inTrypsin Inhibitors, Phytic Acid, Saponin, and IsoflavonesRelated to Soybean Processing. J. Nutr. 125:581S-588S,1995.

3. Rackis, J.J., Wolf, W.J., and Baker, E.C., Protease Inhibi-tors in Plant Foods: Content and Inactivation. Adv. inExper. Med. & Biol. 199:299-347, 1986.

4. Martin A., Bustamante P., and Chun, A.H.C (editors),Micrometrics; in Physical Pharmacy: Physical ChemicalPrinciples in the Pharmaceutical Sciences. Fourth Edi-tion, Lea & Febiger, Philadelphia, 1993, p. 423-432, 446-448.

5. Staniforth, J.N., Powder Flow in Pharmaceutics: TheScience of Dosage Form Design; Aulton, M.E. (editor),Churchill Livingston, Edinburgh, 1988, p. 612.

About the AuthorsAntoine Al-Achi is an Associate Professor of Pharmaceuticsat Campbell University School of Pharmacy. He received hisPhD form Northeastern University in Boston, Massachu-setts. His area of research interest is in the drug deliverysystems for peptides and proteins. Before joining Campbell,Al-Achi had his postdoctoral training at Dana Farber CancerInstitute in Boston, MA. He is a member of Sigma Xi,Controlled Release Society, and the American Society ofClinical Pathologists.

Thomas J. Clark III received BS in pharmaceutical sciencesfrom Campbell University School of Pharmacy. He works atEckerd as a pharmacy intern and head technician.

Robert B. Greenwood is an Associate Professor ofBiopharmaceutics/Pharmacokinetics at Campbell Univer-sity School of Pharmacy. He received his BS in pharmacy andPhD from University of North Carolina. Dr. Greenwood hasbroad research interest in biopharmaceutics, pharmacoki-netics, drug and additive stability, and dosage formulationand design. He is a member of AACP, NCAP, and APhA.

Shylock Sipho Mafu received his MS in industrial phar-macy from Campbell University School of Pharmacy. He is amember of ISPE and Student National Pharmaceutical Asso-ciation (SNPhA).

Campbell University School of Pharmacy, PO Box 1090,Buies Creek, NC 27506.

Incubator Design


Incubator Design for Optimal HeatTransfer and Temperature Control inPlastic Bioreactors

by William Adams, Colette Ranucci, Sara Diffenbach,Kim Dezura, Charles Goochee, Abraham Shamir, andScott Reynolds

A model wasdeveloped todescribe thethermalresponse oflarge plasticbioreactors inincubators withvariable airspeed, and thenused to providea general guidefor incubatordesign.

Figure 1. Isometric viewof incubator layout andtypical load pattern.

designed to control temperature using forcedair circulation, and may range in size from asmall tabletop unit to full walk-in rooms ca-pable of storing thousands of reactors. Toachieve process consistency and control, thetime required for warming to optimal tempera-ture must be satisfactory relative to the timescale of biologically significant events in thereactor, such as cell settling and attachment,initiation of growth, or viral transmission.

In recent years, the design of bioreactors hasadvanced in two ways. Firstly, to provide greater

cell culture surface area withina given manufacturing foot-print, and secondly to reducethe number of container open-ings and thereby provide in-creased sterility assurance. Inso doing, however, it is appar-ent that the heat transfer nec-essary to provide the desiredwarming rates has become abigger challenge. Table Ashows the impact of design onthe ratio of heat transfer sur-face area to cell culture areafor several commercially avail-able bioreactors, including the40-tray Nunc Cell Factories(NCFs) reactors studied in thiswork.

These design properties sug-gest that heat transfer into the40-tray units will be ~10 timesslower than into a T-flask if allother factors are equivalent.

Culture of mammalian cells for the manu-facture of vaccines for human use hastraditionally been performed in smallplastic bioreactors such as T-flasks and

roller bottles. Generally, the reactors are ma-nipulated at ambient temperature for processoperations such as cell plant, refeed, or infec-tion. Subsequently, the reactors are transferredinto incubators which warm and then maintainthe cells at the optimal temperature for cellgrowth or viral propagation (e.g., ~37°C forhuman cell lines). The incubators are generally



Incubator Design


While the impact of warming rates has been shown to beimportant in some applications of cell and virus cultivation,heat transfer and temperature control of the largest reactorlisted in Table A has not been well understood. Standardincubator designs generally promote heat transfer by re-circulating air controlled at a specific temperature set point.Air speed and direction have essentially been driven bycGMP design factors rather than rigorous thermal perfor-mance analysis. To provide a class 10,000 environment, forexample, walk-in incubators would generally be designed toprovide HEPA-filtered airflow from the ceiling into the roomwith low wall returns to the HVAC system, typically at aminimum of 35 air changes per hour. Achieving a consistentwarming time among bioreactors having very different sur-face area to volume ratios; however, will require differentspecific incubator airflow conditions.

The focus of this work was to analyze the transient ther-mal characteristics of 40-tray NCFs to determine the depen-dence of warming rates on airflow conditions in incubators.The results of this study were then generalized to facilitateincubator design and operation to enable optimal bioreactorperformance and process robustness. More specifically, thisarticle describes an experimental approach to assess thethermal response of 40-tray NCFs in walk-in incubatorsalong with the development and use of Computational FluidDynamics (CFD) to model the thermal behavior, consideringboth steady state (fixed temperature) and transient (warm-ing) modes. The utility of the model was to:

1. calculate air velocity and direction around the NCFs as afunction of incubator design features

2. provide the lumped transient thermal model parametersfor the NCFs that produce the same predicted transientresponse as measured data

3. define the impact of air flow conditions on NCF warmingrates in a general way to facilitate future incubator designoptimization

Experimental and Computational Methods

Temperature Mapping in NCFsThe 40-tray polystyrene NCFs were positioned in sets of fouron the stainless steel carts used for automated cell culturemanipulations. Small holes were drilled in the sidewalls ofselected trays to allow for the insertion of thermocouples. Thetubing assemblies for vent and plant operations were placedinto the applicable ports of each NCF, and the NCFs werefilled with 0.33 mL/cm2 of Water-For-Injection (WFI) (equiva-lent to 210 ml per tray). Thermocouples were inserted mid-way into the selected trays, manipulated until the end of eachthermocouple was submerged in water, and subsequentlytaped in place to prevent inadvertent displacement duringthe mapping study. The temperature of up to 20 thermo-couples available for use per mapping was monitored andrecorded using standard data logging software. The thermo-couples were calibrated pre- and post-use to ensure accuratedata acquisition. The NCF carts were maintained at roomtemperature in an attempt to provide a uniform temperaturefor all trays. To start the experimental mappings, the cartscontaining the NCFs were wheeled into the 37°C incubator,and temperature data was then collected at five-minuteintervals from each of the thermocouples. Temperature map-ping studies were typically continued until all trays reachedwithin 0.5°C of the 37°C incubator control set point. The datawas transferred to spreadsheets for numerical and graphicalanalysis.

Figure 2. NCF load pattern and numbering key plan.

Bioreactor Type Cell Growth Area Heat Transfer Area Ratiocm2 cm2

T-flask 175 546 3.1

Roller Bottle 850 1,050 1.2

NCF (2-tray) 1,264 1,741 1.4

NCF (40-tray) 25,280 8,718 0.36

Table A. Relative heat transfer considerations.

Figure 3. NCF thermal response at low air speed. Average local airspeed ~28 FPM. Side wall air circulation units providing 0 FPM atsupply slot inlets (off).

Incubator Design


Incubator Design FeaturesTwo different-sized walk-in incubators were evaluated in thestudy. The first incubator was 9 ft wide by 9 ft long withoverhead HEPA-filtered air supplied at ~1240 CFM (cubicfeet per minute) at 37°C. Typically, five NCF carts, eachholding four 40-tray NCFs, were placed in the incubators.Two low wall air returns were located along one wall. Thesecond incubator measured 20.6 ft long by 9.8 ft wide withoverhead HEPA-filtered air supplied at ~2,400 CFM and37°C, and with low wall returns positioned at each corner ofthe room (four total). The bulk average air velocity in themiddle of the incubators resulting from these supply andreturn arrangements was about 15 feet per minute (FPM).Both incubators had a stainless steel interior finish with wellinsulated wall panels. To provide for a wide range of airvelocities in the vicinity of the NCFs, additional wall-mountedair recirculation units were installed. The fans in these unitsdrew air in at an elevation of ~7 ft and drove it through a bankof 2 ft by 2 ft HEPA filters positioned at the level of the NCFs(approximately 1.1 ft to 3.1 ft above the floor). The flow fromthe filters was forced through slots about 0.08 ft wide by 1.8ft tall (positioned about 1.2 ft apart), through which thevelocity was increased. The output of the fans was designedto provide an air speed of up to ~1,000 ft per minute (FPM) atthe slot outlet. By adjusting the output of the fan-poweredrecirculation units, the air speed in the vicinity of the NCFscould be significantly varied. Volumetric air flow rates weremeasured using standard portable flow hoods, and local airvelocities were measured using a hand-held hot wire an-emometer.

Computational Fluid Dynamics ModelingThe two incubator designs were modeled using Computa-tional Fluid Dynamics (CFD). Large-scale three-dimensionalCFD models were created from the physical domains de-

scribed in the previous section. These models were built usingcommercially available general purpose CFD software. Pa-rameters used for the model included the following: anunstructured solver, k-ε RNG turbulence capabilities, and adomain descretized with tetrahedral cells throughout. Ap-proximately 620,000 computational cells (small incubator)and 910,000 cells (large incubator) were employed to solve forthe three velocity directions, two turbulence terms, onepressure term, and one energy term.

Experimental ResultsMore than 30 distinct temperature mappings of NCF warm-ing were performed covering various NCF locations, airflowspeeds, and directional patterns. An isometric view of thesmall incubator with five NCF carts is shown in Figure 1. Atypical experimental load pattern along with the numberingscheme for carts and specific NCFs on each cart is indicatedin Figure 2. The bulk average thermal performance of the 40-tray NCFs on the carts in low air flow conditions is shown inFigure 3 which shows the temperature versus time profilesfor the top, middle, and bottom trays (numbers 1, 20, and 40respectively). The data was taken from NCF #2 among thefour on cart #1 (identifying nomenclature Cart1-N2). Alltrays exhibit the expected first order dependence on thedifference between the tray and incubator temperature. Theresponse is notably asymmetric; however, with the top traywarming the fastest, the bottom tray notably slower, and themiddle tray warming the slowest following an initial lagphase. This pattern in the relative thermal response for thedifferent trays was observed in all studies although thespecific response times and magnitude of disparity betweentrays was influenced by airflow and heat transfer conditions.

To summarize the performance in a way most meaningfulto the cell and virus cultivation, we evaluated the time toreach a temperature within 1°C of the incubator controlpoint. Referring to the data in Figure 3, the performance issummarized in Table B.

For this experiment, the incubator temperature was 36.7°C,and the times listed above therefore represent the time toachieve 35.7°C. Within a group of experimentally mappedNCFs, the response time of comparable trays could vary by afew hours. The standard deviation (σ) in time required toreach within 1°C of the control temperature ranged from~three hours for top trays to ~six hours for the slowestresponding middle trays.

Representative performance within the incubator at higherair speed conditions is shown in Figure 4 which plots thetemperature versus time profiles for the top, middle, andbottom trays from Cart 5-N4 in the load pattern. In this case,the sidewall air circulation units were providing ~550 FPM at

NCF Number and Layer Time to Within 1°C (hours)

Cart5-N4-Top 5.0

Cart5-N4-Middle 9.0

Cart5-N4-Bottom 6.5

Table C. Thermal response time for trays at higher air speed.

NCF Number and Layer Time to Within 1°C (hours)

Cart1-N2-Top 7.1

Cart1-N2-Middle 19.0

Cart1-N2-Bottom 14.0

Table B. Thermal response time for trays at low air speed.

Figure 4. NCF thermal response at higher air speed. Average localair speed ~40 FPM. Side wall air circulation units providing ~550FPM at supply slot inlets.

Incubator Design


Figure 5. Plan view of velocity vectors in FPM - small incubator at 2.25 ft above floor.

the supply slot outlets. Although the NCF trays warm morerapidly in the higher air flow conditions, the asymmetrictrend in response between trays remains prevalent. Theresponse times can be summarized in Table C.

The faster response and reduced absolute deviation be-tween layers would be expected to provide improved consis-tency and control of the cell culture.

The asymmetric thermal response between top and bot-tom layers in the 40 tray NCFs is in contrast to performancein the smaller two tray and 10 tray NCFs. In the 10-traybioreactor, for example, both top and bottom trays lead theresponse of the middle tray by an equivalent margin. Appar-ently the cart used to support the 40 tray NCFs provides a

significant additional heat capacity and thermal resistance.Interestingly, there was not any significant difference be-tween the NCFs at different positions on the carts. That is theperformance was essentially the same between the outsideand inside positions (i.e. positions #1 through #4 are allsimilar). This was determined through analysis of varianceapplied to all the data from studies covering all positions inthe load patterns. Warming rates were not significantlyaffected by cart location within the incubators studied, mostlikely due to the similar proximity of carts to airflow sources.

Modeling and Discussion of ResultsThe CFD models were first used to calculate air velocities

"Therefore, CFD was deemed as the more accurate andconvenient means to solve the transient response problem as long as

suitable values for R and C could be estimated."

Incubator Design


Table D. Average time to within 1°C of incubator temperature.

Air Velocity Bottom trays Middle trays Top traysAvg per slot (FPM) (hrs) (hrs) (hrs)

0 15.5 16.5 9.5

500 6.5 8.5 6.3

1000 4.8 6.8 4.8

throughout the incubators as a function of air speed settingsfor the wall-mounted recirculation units. Figure 5 shows aplan view of the predicted velocities at a distance of 2.25 feetabove the floor. Upon entry into the room from the sidewallslots, the velocity vectors show a decline from the initial~1,000 FPM with distance into the room and from impactwith the cart obstructions. The areas in between NCFs on anygiven cart are slightly less than an inch apart and thereforeretain relatively low air speeds for the orientation shown.Figure 6 shows a side-view of the same case, highlighting thehigh local velocities around the NCFs, the relatively lowervelocities elsewhere, and the overall air circulation patternsthroughout the room.

It was generally desired to isolate a closed-form functionrelating temperature rise time to various known physicalparameters and boundary conditions. By identifying thisfunction, a separate CFD simulation would not be required

each time an incubator was designed. From standard tran-sient heat transfer relationships, it was felt that the followingequation could be used as a first order approximation:

Tnunc = (Ti - T∞) e –t/RC + T∞ (1)where:

Tnunc = Transient Nunc Cell Factory temperatureTi = Initial Nunc temperature (nominally 18°C)

Figure 6. Side view of velocity vectors in FPM - small incubator.

Incubator Design


Figure 8. Correlation between NCF warming time and localaverage air velocity.

Figure 7. Comparison of model-predicted temperature rise toexperimental measurements of a similar incubator layout at highairflow conditions (wall units at ~550 FPM).

T∞ = Incubator bulk air temperature (37°C)t = Time (seconds)R = Thermal Resistance (1/hA)C = Thermal Capacitance (ρCpV)Cp = Specific Heatρ = DensityV = volumeh = Heat transfer coefficientA = Surface area

Initially, it was believed that the parameters of equivalentthermal resistance and thermal capacitance could be directlydetermined using calculated values for surface areas, vol-umes, densities, conductivities, specific heats, and heat trans-fer coefficients for the carts and bioreactors. However, manyattempts at solving the function revealed that the relation-ship was more complex with a strong dependence on otherfactors such as local airflow velocities, cart orientation, etc.Therefore, CFD was deemed as the more accurate and conve-nient means to solve the transient response problem as longas suitable values for R and C could be estimated. Along theselines, several transient CFD models were run using closed-form thermal property calculations (to calculate values for Rand C), but all such simulations failed to adequately describethe measured transient response of the Nuncs.

After the above technique failed to produce the desiredresults, an attempt to iteratively “back-calculate” the R and Cparameters was made as follows. From the steady state CFDsimulation, the heat transfer coefficients were calculated aswere the bulk air temperature and the initial NCF layertemperatures. A two-variable least square fit for R and C wasmade using a power function developed from the experimentaldata acquired for various air flow rates. This equation used theknown variables of temperature and time to provide the bestfit values for R and C. These values were then used in a CFDtransient model to predict the comprehensive temperatureresponse behavior of the Nuncs. This procedure was furtherrepeated until the calculated thermal resistance and capaci-tance provided a good emulation of the experimental data forthe low air velocity case. Model predictions for the time to reachwithin 1°C of the incubator control point agreed with experi-mental data to within two hours for all such cases.

The model was then challenged to predict warming ratesunder distinctly different air flow conditions. Such an assess-ment is exhibited in Figure 7 which shows both experimentaland predicted thermal response curves for the case in which airflow from the side wall units provides ~550 FPM at the HEPAfilter outlet slots. There are three sets of data shown for thisfaster warming case, corresponding to three representativetray mapping locations among the overall load pattern of fivecarts containing 20 NCFs. In Figure 7, the dashed lines showthe model-predicted temperatures relative to the actual ex-perimental values. Agreement is good in general, the predictedtimes to reach within 1°C are accurate to within two hours.

The model was used similarly to predict warming rates atthree different air flow conditions. The thermal response ofthe NCFs as a function of the nominal side wall air velocitycan be summarized in Table D.

The times listed are average response times for all samplepoints on the indicated tray location (the model allowed for atotal of 11 such points). Table D shows that air speed has lessimpact on the top trays which have relatively high exposedsurface area. The bottom or middle trays have smaller ex-posed surface areas which leads to a stronger dependence onchanges to the heat transfer coefficient.

Incubator Design GuideAn important potential use of this heat transfer analysis andNCF thermal response modeling is to facilitate incubatordesign. Ideally, incubators could be designed for reliable,predictable performance without the need for developing andrunning a detailed computer model of thermal response forNCFs (or other bioreactors) in each specific incubator geom-etry and load pattern. Along those lines, we sought to gener-alize the model results to enable prediction of NCF warmingrates as a function of local average air speed.

More specifically, we developed a correlation for the NCFwarming time as a function of the average air velocity whichcan be experimentally measured and is a physically intuitiveparameter. Average air velocities were calculated by the CFDsimulation program for the different side-wall velocity cases.The average was calculated by sampling 120 point locations

Incubator Design


around the NCF carts. These points were regularly spaced, 2inches away from the carts, covering the top, middle, andbottom zones of each NCF, along both the long and short sidesof the cart. The resulting data for all tray locations was thenplotted in Figure 8 as time to achieve 36°C (average time forall trays) versus the average local air speed in FPM. The plotclearly shows a steep decline in warming time between lowair speed of ~28 FPM and the more moderate speed of ~48FPM with a slower decline expected beyond that. The steplike appearance in the middle of the plot is believed to bedriven by a transition from the laminar to the turbulent flowregime. Figure 8 suggests that reasonably rapid thermalresponse as well as effective temperature consistency andcontrol will be provided from an incubator environmentdesigned to provide average local air speeds of at least 35FPM. Note that much higher local point velocities may beneeded to achieve an average surrounding velocity of 35 FPMbecause other areas may be relatively stagnant.

Higher air speeds generally dictate larger air handlersand proportionally greater surface area of HEPA filters sincethe absolute air speed through the filters should be less than~100 FPM. As a result, higher incubator air flows will alwaysbe more costly to install and operate. The most economicalapproach to achieve higher local air velocity is to position thedominant room air flow drivers, such as the main supplyinlets and/or returns, as close to the bioreactors as is practi-cal. Higher local velocities may require the use of ductwork toprovide post-filtration air flow convergence to achieve thedesired minimum air speed at important locations in the loadpattern. Reasonably rapid and consistent bioreactor warm-ing rates can be achieved through such incubator designapproaches.

ConclusionIncubator air speed was observed to have a significant impacton the warming rate of large plastic bioreactors commonlyused for cell and virus cultivation. The relationship betweenthermal response and air velocity was well correlated throughthe use of a model utilizing computational flow dynamics.Application of the model indicated that average local veloci-ties in excess of 35 feet per minute were required to achievetimely and consistent response.

Credits1. 40-tray NCF bioreactors were Nunc Cell Factories pro-

cured from Nalge Nunc, Intl.

2. The CFD simulation software used was Fluent®, fromFluent, Inc.

About the AuthorsWilliam Adams received his PhD in chemical engineeringfrom Cornell University with a minor in computer science. Hejoined Merck in 1990, and has progressed through positions

in Technical Service, Operations Management, and is cur-rently Associate Director in Vaccine Process Engineering.His current principal focus is on process equipment andfacility design for live virus vaccines.

Merck & Co., Inc., WP60T-2, PO Box 4, Sumneytown Pike,West Point, PA 19486.

Colette Ranucci joined the Bioprocess Research and Devel-opment department within Merck Research Laboratories in1992, shortly after receiving BS/BA degrees in chemicalengineering and German. After working in the area of vaccineprocess development for several years, she returned to RutgersUniversity to complete the PhD program in chemical engi-neering. Since rejoining Merck in 2000, her primary respon-sibilities have included process development, scale-up, andprocess validation related activities. She is an active partici-pant of the BIOT division of ACS, serving as the membershipchair.

Merck & Co., Inc., WP17-201, PO Box 4, SumneytownPike, West Point, PA 19486.

Sara Diffenbach received a BS in chemical engineeringfrom the Pennsylvania State University in 2000 and joinedMerck & Co., Inc., as a staff engineer in the Viral VaccineTechnology and Engineering group, focusing on technologytransfer and facility start-up. She recently joined Pfizer, Inc.,as a Production Leader in the manufacturing of personal careproducts.

Pfizer, Inc., 400 W. Lincoln Ave., Lititz, PA 17543.

Kim Dezura is a Senior Process Engineer at Merck.Merck & Co., Inc., WP62-7, PO Box 4, Sumneytown Pike,

West Point, PA 19486.

Charles Goochee is a Director of Bioprocess R&D at Merck.Merck & Co., WP17-201, PO Box 4, Sumneytown Pike,

West Point, PA 19486.

Abraham Shamir joined Merck in 1989 and has held posi-tions of increasing responsibility in technical operations andprocess engineering. After receiving a BS in biochemistry anda PhD in chemical engineering from Columbia University, hejoined Schering-Plough where he progressed to the position ofmanager for large-scale recombinant protein purification. Hewent on to join BioTechnology General as director of proteinpurification - Process Development and Manufacturing.

Merck & Co., Inc., WP60T-2, PO Box 4, Sumneytown Pike,West Point, PA 19486.

Scott Reynolds is a Senior Engineer at Computer AidedEngineering Solutions.

Computer Aided Engineering Solutions, Division ofBearsch Compeau Knudson Architects and Engineers, PC, 41Chenango St., Binghampton, NY 13901.

a modular reprinted from the official journal of ispe ... · pharmaceutical production facility...

Documents