best practices for qualification success: a statistical ...€¦ · merous project costs associated...

IQ and OQ Analysis

JANUARY/FEBRUARY 2007 PHARMACEUTICAL ENGINEERING On-Line Exclusive 1www.ispe.org/PE_Online_Exclusive

©Cop

yrig

ht IS

PE 2

007

This articlepresents astatisticalanalysis ofplanning andexecutionstrategies forInstallationQualification(IQ) andOperationalQualification(OQ) activitiescarried out in 50construction/modificationprojects.

Best Practices for QualificationSuccess: A Statistical Analysis ofCharacteristics and Practices thatDrive IQ and OQ Cost and Schedule

by Allison Aschman and Gordon Lawrence

On-Line Exclusive Article

PHARMACEUTICAL ENGINEERING®

The Official Magazine of ISPE

January/February 2007, Vol. 27 No. 1

Introduction

Installation Qualification (IQ) and Opera-tional Qualification (OQ)1 are frequentlyon the critical path of activities in theconstruction or modification of pharma-

ceutical production facilities. Any delay duringthe IQ/OQ phase is a major problem if it pre-vents product from being delivered to meetmarket demand and/or regulatory approval.

This article outlines the research that wasconducted on how pharmaceutical companiescarry out IQ and OQ activities. The researchresulted in three major deliverables:

• statistical models that can be used to bench-mark cost and schedule performance in theexecution of IQ/OQ

• a list of key project characteristics that af-fect qualification cost and schedule

• a list of key Qualification Best Practices thatdrive qualification success and minimizequalification failure

MethodologyIn this study, using ordinary least squaresstatistical regression methods, “uncontrollable”project characteristics and “controllable” projectpractices were reviewed for statistically signifi-cant links between them and what IQ/OQ costsand schedules were achieved.

Uncontrollable project characteristics arethose which are inherent in the project scopeand cannot be altered by the behavior of theproject team. An example is the type of facility

being built (e.g., abulk active pharma-ceutical ingredientfacility). Control-lable project prac-tices are those ac-tions that the com-pany managementand the project teamcan choose to do ornot to do. For ex-ample, the teamcould choose to de-velop a schedule ofqualification activi-ties during basic de-sign instead of wait-ing until later in theproject.

Key Project Dataset(n = 50)Characteristic

Facility Type Bulk chemical active pharmaceutical ingredient: 21%Biological: 35%Oral dosage/Non-parenteral secondary: 25%Sterile formulation and finishing: 11%Other (incl. packaging facilities, pharmaceutical device facilities, etc.): 8%

Facility Scale Production: 90%Pilot Plant: 10%

Project Type Standalone (Greenfield or colocated): 57%Expansions: 33%Revamps/revisions/modifications of existing facilities: 10%

Geographical Location North America: 49%Europe: 35%Puerto Rico: 8%Singapore: 8%

Product Description Prescription: 94%Over the Counter: 6%

Project Engineering Mean: $50 millionand Construction Median: $25 millionCost (US$ millions) Range: $2.5-$191 million

Table A. Characteristicsof the research set.

IQ and OQ Analysis

2 PHARMACEUTICAL ENGINEERING On-Line Exclusive JANUARY/FEBRUARY 2007 www.ispe.org/PE_Online_Exclusive

©Copyright IS

PE 2007

By seeking statistically significant links between projectcharacteristics or practices and cost or schedule outcomes, anattempt was made to uncover those characteristics and prac-tices that “drive” IQ/OQ cost and schedule.

Project DatasetFor the purposes of this study, data on 50 projects wascollected from nine major pharmaceutical companies. Thechoice of “major” or what is often referred to as “big pharma”companies was not a deliberate effort to exclude medium orsmall companies. It simply reflects the fact that the data wascollected from companies that were interested in participat-ing in the research, and all interested parties happened to bemajor companies. The companies are not named because theywish to retain their anonymity. Data was collected using acustomized questionnaire.2 Some key characteristics of theset are shown in Table A. In order to be able to conduct an“apples-to-apples” comparison, the data for all 50 projectswas “normalized” to a common cost baseline (currency, loca-tion, and year). In addition to asking project teams to com-plete a detailed questionnaire, each team was interviewed toverify the reliability of the data provided. In addition, eachcompany chose to participate in the study in order to receiveadvance notice of results that would help them improve theirestimating, planning, and execution of Commissioning andQualification activities. Consequently, it was in the interestsof each company to provide data that was as accurate aspossible.

Analysis of Qualification CostCost Data LimitationsWhen the project teams were interviewed in the course of thedata collection activities, it was discovered that they used awide variety of approaches in the recording of qualificationcosts. Variations in accounting were seen across companiesand from project to project within companies. Project teamsrarely recorded charges and hours for individual qualifica-tion activities, and owner participation in qualification wasoften not tracked or charged to projects. In some cases,qualification charges were “hidden” within equipment/ven-dor costs, project management costs, or in other capital costsfor the project. Every effort was made to trace the costs.Where these costs were expensed and traceable to a specificaccount, they were collected and included. The methodology

was as follows: With the active advice and assistance of thestudy participants – it was decided what costs should betracked and allocated as “qualification cost.” There are nu-merous project costs associated with qualification that mayor may not be captured, for example, owner costs associatedwith approving protocols - which is often captured in theproject management account. Once the decision has beenmade as to what costs will be allocated to this account, theprocess of normalizing from project-to-project, company-to-company is more straightforward. Therefore, while projectcost data can be “messy,” the methodology used “forced” thismessy data into defined “buckets” - a defined work break-down structure that could then be analyzed in an apple-to-apples manner).

While “messy” cost information increases the variance inthe data and the analyses, much of the error was essentiallyrandomized across companies. Therefore, statistically sig-nificant industry outcome benchmarks still can be developed.

However, a major conclusion of this study is that themajority of project systems do not actually know exactly howmuch money is being spent for the completion of all theactivities making up the total qualification effort.

IQ/OQ Cost ModelGiven the limitations of the available cost data, the costanalysis for this study focuses on a single point of interest: thetotal cost required to complete IQ/OQ; (i.e., the cost to de-velop, write, and execute IQ/OQ protocols. All costs accruedby the owner, including internal and external (contractor/consultant) costs). While some commissioning and PQ costdata was available for some projects, the majority of projectswere able to provide data for IQ/OQ only. Moreover, in themajority of cases, the IQ/OQ costs were not broken out fromeach other.

Project Size as an “Uncontrollable”CharacteristicIQ/OQ cost should be a function of basic project characteris-tics such as project size, project type, and facility type. It isreasonable to hypothesize that the size and/or complexity ofa project will affect the IQ/OQ cost for a project. On a basiclevel, the total project size indicates the volume of work thatwill be required to complete IQ/OQ, which would then bedirectly related to IQ/OQ costs.

There are several potential measures of project size andcomplexity, including total project cost, major equipmentcosts, facility capacity (in terms of product count per year,etc.), or major equipment count. For this study, the total costof engineering, materials, equipment, and construction (inother words the total installed cost – TIC)3 serves as a proxyfor the project size or complexity.

Table B shows the regression relationship between thenatural log of the TIC and the natural log of the IQ/OQ cost.(Regression relationships presented in this study use student’st-test. In the tables illustrating the regression relationships,the “t-score” and “probability” (p>t) associated with eachregression will be included. Both the t-score and probability

Table B. Relationship between key characteristics and practicesand IQ/OQ cost.

Key Drivers t-score P>t

ln (TIC) 8.54 0.00

Biological Facility(Biological facilities vs. API, Oral, 3.19 0.00Sterile, and Other)

Pilot Plant Facility(Pilot Plants vs. all other facilities) -2.10 0.04

Stand-alone Project(Stand-alone projects vs. Revamp 2.01 0.05and Expansion Projects)

Qualification Schedule Definition (Critical Path Method -3.03 0.01or Milestone schedules vs. No schedule planning/end-date only)

IQ and OQ Analysis


©Cop

yrig

ht IS

PE 2

007

provide an indication of the statistical significance of anindependent variable in the regression. For example, inTable B, the large t-score associated with ln(TIC) indicatesthat this independent variable is a statistically significantdriver of ln(IQ/OQ cost), which is the dependent variable inour regression. (t-score varies with number of observationsand has to be viewed in parallel with p>t, but generally, a ruleof thumb is that if the absolute value (i.e., plus or minus) ofthe t-score is greater than 2.00, then the variable is a statis-tically significant driver.) The probability (p>t) also expressesthe “confidence” with which a variable is regarded to besignificant in the regression. The general cut-off for theanalyses in this study is 90 percent confidence, or p>t=0.10.(i.e., there is 90 percent confidence that the correlation is notdue to random chance). Note also that in cases where p>t isquoted as 0.00, this does not mean it is absolutely zero, butmerely that it is zero to two decimal places, or put anotherway, it is smaller than 0.005).

The correlation between TIC and IQ/OQ cost was found tobe very strong. Figure 1 graphically shows the strength of therelationship. Clearly, TIC is the single best predictor of IQ/OQ cost. (Note that the Figure uses log scales, hence thepresence of minus numbers).

Other Significant “Uncontrollable”CharacteristicsAfter controlling for size, other characteristics were exam-ined for their significance in “explaining” variance in IQ/OQcosts.

• Biological Facilities.4 For the facility types in the studydataset, a significant relationship was found betweenbiological facilities and higher IQ/OQ costs. For biologicalfacilities in the study, the average IQOQ cost as a percent-age of TIC was twice as large as the average percentage forthe other facility types.

• Pilot Plant Facilities. For the facility types in the studydataset, a significant relationship was found betweenpilot plant facilities and lower IQ/OQ costs. Although afirst glance at this result suggests that pilot plant facilitiesmay be less rigorous in their approach to IQ/OQ, hencedriving lower costs, the overall data collected for this studydoes not support that finding. In fact, the pilot plantfacilities, especially those making products to be used forclinical trial, appeared to perform the same set of qualifi-cation activities and apply the same set of quality stan-dards as the other projects in the study database. How-ever, the findings indicate that the qualification effort forthese facilities may be smaller, if not less rigorous, relativeto the total costs of equipment and materials being in-stalled than non-pilot facilities.

• Stand-alone Project Type. A significant relationshipwas found between stand-alone facilities and higher IQ/OQ costs. As with the relationships between facility typesand IQ/OQ cost, a first glance at project types suggests

that stand-alone facilities would require more expensiveIQ/OQ based on the requirements to execute qualificationon all new facilities, including air-handling, utility tie-ins,etc. In contrast, expansion and revamp projects, by defini-tion, are installed in existing facilities where it is likelythat certain qualification activities have been performedpreviously. It is interesting that the significance of therelationship between stand-alone project type and IQ/OQcost holds up even after controlling for total project size.

For the drivers of IQ/OQ cost described above, each one issignificant within a multilinear regression of all variables.Table B shows the regression relationship between the natu-ral log of the IQ/OQ cost and the key “uncontrollable “projectcharacteristics based on their fit within a multilinear regres-sion analysis.

Key “Controllable” DriversA “controllable” driver that was found to have a majorinfluence on IQ/OQ cost was qualification schedule defini-tion.

• Qualification Schedule Definition. After controllingfor project size, facility type, and project type, the followingkey project practice was seen as a significant driver of IQ/OQ cost: Qualification schedule definition at the time ofauthorization/detailed engineering start. The quality ofschedule planning for the projects in the study rangedfrom Critical Path Method (CPM) planning to milestoneschedules to schedules consisting of start and end datesonly. For this analysis, the quality of the overall projectschedule was broken out from the quality of the commis-sioning and qualification schedule. A statistically signifi-cant relationship was seen between the quality of thequalification schedule and IQ/OQ cost. For projects thathad prepared a schedule to a critical-path or milestonelevel, IQ/OQ costs were lower than for projects that wereworking toward end dates only or had no schedule at all.

Even when a large portion of the variance in IQ/OQ cost is

Figure 1. IQ/OQ cost is driven by facility Total Installed Cost (TIC).

IQ and OQ Analysis


©Copyright IS

PE 2007

accounted for by basic characteristics such as size, facilitytype, and project type (in other words, even when thesecharacteristics are “held constant”), the quality of the quali-fication schedule accounts for some remaining variance inthe data and is a statistically significant driver of IQ/OQ cost.

Table B shows the regression relationship between thenatural log of the IQ/OQ cost and the “controllable” qualifica-tion schedule driver based on its fit within a multilinearregression analysis.

Other IQ/OQ Cost DriversIn total, the “base” model, described by the variables above,explains approximately 80 percent of the variance in the costof IQ/OQ for pharmaceutical projects.

Several project characteristics and practices explain someof the remaining variance in IQ/OQ cost as shown in regres-sions against the residual of the base model. Three of thesefactors are significant within a 95 percent confidence leveland are described below:

• Vendor Qualification Support. For the purposes of thisanalysis, projects were classified into three categories: 1.projects for which there was no significant involvement ofvendors in the qualification effort, 2. projects for whichthere was some vendor participation in the execution of IQand/or OQ, 3. projects in which vendors executed entireprotocols and/or delivered “prequalified” equipment orskids to the project team. The analysis shows that greatervendor involvement correlates with lower IQ/OQ cost.

• Project Manager Assurance Team Development Is-sues

• Quality Assurance Team Development Issues. Theprojects in the study were classified as to whether, duringthe life of the project, they experienced any “issues” with thekey team members on the project. Issues include latearrival on the team, inability to participate as much asrequired, because of conflicting priorities, competing projects,etc., and turnover. Team issues related to the projectmanager and/or quality assurance representation werestatistical drivers of higher IQ/OQ cost.

Three remaining practice or characteristics variables werefound to correlate with IQ/OQ cost and were statisticallysignificant within a 90 percent level of confidence whenregressed against the “base” model:

• Attendance by Qualification Personnel at FactoryAcceptance Tests (FATs) and/or Site AcceptanceTest (SATs). The participation of qualification personnelcorresponds with lower IQ/OQ cost.

• Other Validation Group Requirements. If qualifica-tion personnel had to divide their time between the quali-fication work on the project and a requirement to performother tasks (such as writing Standard Operating Proce-

dures (SOPs) or batch records and/or performing technicalassessments), then this was found to increase total IQ/OQcosts.

• New Technology. Projects containing some aspect ofnew technology correlate with higher IQ/OQ cost.

Other Practices that Correlate with IQ/OQ CostPerformanceOther practices that correlated with IQ/OQ cost performanceincluded:

• Commissioning and Qualification Integration. Forthe purposes of IQ/OQ cost analysis, many of the projectscould be placed in two categories:

1. projects with no planned or executed integration ofcommissioning and qualification

2. projects in which the planning and execution of quali-fication included referrals to commissioning tests aspart of IQ and/or OQ execution

Not surprisingly, the data suggest that increased integrationof activities correlates with lower IQ/OQ cost.

• Impact Assessment. The occurrence and formality ofImpact Assessment activities correlate with lower IQ/OQcosts for the projects in this study.

• Project Team Status. Two issues surrounding projectteam status at the time of authorization correlate with IQ/OQ cost. First, projects in which there is a clear under-standing of the project objectives and target dates by allmembers of the project team have lower costs. Second,projects in which team roles and responsibilities are de-fined and understood by all members of the project teamalso have lower IQ/OQ costs.

• Retest. Not surprisingly, projects noting that changesand retest were required for qualification appear to havespent more on IQ/OQ costs.

• Approach to Automation Qualification. Projects rely-ing on a separate approach to Computer System Valida-tion (CSV), that is those project teams that wrote andexecuted separate IQ/OQ protocols for process equipmentand utilities versus automation, appear to spend more onIQ/OQ costs.

Analysis of Qualification ScheduleSchedule Data LimitationsThe IQ/OQ schedule represents a period of time during whichresources are allocated and plans, schedules, and controls areput in place for the qualification effort. The “meaning” of theIQ/OQ duration varied somewhat from project to project. Forsome projects, this duration included a significant portion of

IQ and OQ Analysis


©Cop

yrig

ht IS

PE 2

007

the protocol writing, review, and approval cycle; for others, itincluded only protocol execution and report writing. Thedefined start and end dates also varied slightly from projectto project as the “boundary lines” between OQ and PQ weresometimes blurred. (As with the cost data collection method-ology, the way this was resolved was to define what would beconsidered within the schedule boundaries. In this case, onlyprotocol execution and report writing were considered in theIQ/OQ schedule. This is another example of how “messy” datawas normalized across projects.) For the purposes of thisstudy, these overlaps were essentially randomized, and whilecontributing to the overall variability of the cost and scheduledata for these phases, they do not significantly impact theindustry outcome benchmarks.

Qualification Schedule ModelsThere are several ways to examine qualification scheduleperformance of projects in the study database. As a broadanalysis, the overall duration of IQ/OQ (from the start of IQto the end of OQ) is renewed. Another means of studyingqualification schedule performance is to examine the dura-tion from the end of construction (mechanical completion) ofa facility to the end of OQ.

Duration of IQ through OQTo examine the duration of IQ/OQ (IQ/OQ schedule) the“start” of IQ can be defined as the date when the first IQprotocol is executed and the “end” of OQ as the date when thelast OQ protocol is executed and/or the final OQ report iswritten.

An important reason to study the IQ/OQ schedule is therelationship seen between qualification cost and qualifica-tion schedule. The duration of IQ/OQ represents a period oftime when resources must be allocated to the effort, includingvalidation and quality assurance personnel, owners andcontractors, and possibly plant operations and maintenance.In addition, plans, schedules, and controls must be in placefor the duration of IQ/OQ. Therefore, the total IQ/OQ sched-ule in this analysis, was examined.

Duration of Mechanical Completion through theEnd of OQAnother means of studying qualification schedule perfor-mance is to examine the duration from mechanical comple-tion of a facility to the end of OQ (MC/OQ schedule). This isperhaps a more powerful way to consider schedule perfor-mance, as it is the best description of how much time is “lost”between the point when a facility is “ready” to make productfrom a mechanical basis and when it is “ready” to makeproduct from a regulatory standpoint (of course, the comple-tion of OQ is not actually the final step in the project processthat is required for the regulatory approval to make product.Performance Qualification (PQ) often must follow and oftenoverlaps with OQ, and all IQ, OQ, and PQ activities providethe basis for process qualification or validation. For thepurposes of the analyses in this study, we will stop at thecompletion of OQ and examine the later stages of qualifica-

tion/validation in future studies). Therefore, the MC/OQschedule was examined in more detail in this analysis.

As with IQ/OQ duration and IQ/OQ cost, a relationshipexists between MC/OQ schedule and IQ/OQ Cost. Although itis not obvious that one outcome is driving the other, thecorrelation is statistically significant.

The IQ/OQ Schedule ModelAs with IQ/OQ cost, IQ/OQ schedule correlates with TIC,which may be seen as a proxy for overall project size andcomplexity. However, the regression relationship is not asstrong as that seen for IQ/OQ cost, and other drivers arefound to be just as influential as size. Therefore, rather thanholding a single characteristic constant, as was done withproject size when examining IQ/OQ cost, project characteris-tics and practices can be examined directly against the IQ/OQschedule.

Table C shows the regression relationship between thenatural log of the IQ/OQ schedule [ln(IQ/OQ schedule)] andthe key project characteristics and practices that make up theIQ/OQ schedule model. The statistics presented in the tableare based on the fit of the independent variables within amultilinear regression analysis.

“Uncontrollable” Characteristics• Project Size. Greater TIC correlates with longer IQ/OQ

schedules.

• New Technology. For the IQ/OQ schedule analysis,projects were grouped into three categories: 1. projectscontaining no new technology, 2. projects containing someaspect of technology, process, or product that can beconsidered new to the company or site, and 3. thoseprojects incorporating technology new to the industry.Increasing “new technology” ratings correlate with in-creased IQ/OQ schedules.

Key “Controllable” Drivers• Percentage Overlap of IQ/OQ Schedules. Increased

schedule overlap increases the total IQ/OQ schedule. Thisis a very interesting finding given that one could reason-ably assume that overlapping IQ/OQ would lead to ashorter overall duration. In fact, overlapping all phaseswithin a project schedule is usually done to decrease the

Table C. Relationship between key characteristics and practicesand IQ/OQ schedule.

Key Drivers t-score P>t

Project Size: [ln(TIC)] 2.49 0.02

New Technology -2.10 0.04

Percentage Overlap of IQ and OQ Schedules 2.77 0.01(months that IQ and OQ overlap divided by months oftotal IQ/OQ duration)

Engineering Definition at Authorization 2.01 0.05

Qualification Schedule Definition (CPM or milestone -3.03 0.01schedules vs. no schedule planning/end-date only)

IQ and OQ Analysis


©Copyright IS

PE 2007

overall duration of the project. In the MC/OQ scheduleanalysis, some schedule overlaps do, indeed, drive fasterschedules, but the empirical data show that this is not truefor IQ/OQ.

The reasonable explanation for why increased overlapsdrive longer IQ/OQ schedules may be derived from anexamination of why overlaps “fail” to produce the desiredgoal (faster schedules) for any phases within a project. Theprimary reasons are lack of planning in turnover from onephase to the next and repeated efforts. For example, aqualification team may complete IQ for one system andthen start OQ. If upon review of the executed IQ protocolan error is discovered and retest is required, this can thenhave a subsequent effect on OQ, requiring retest in thisphase as well.

Because the correlation between percentage IQ/OQoverlap and IQ/OQ duration is so strong, project teamsmay want to carefully examine their strategies for over-lapping schedules and be sure that controls and contin-gency plans are in place so that the most efficient sched-ules are achieved.

• Engineering Status at Authorization. Our analysesfound that as the status of engineering definition im-proved, the IQ/OQ duration decreased for projects in thestudy dataset. Earlier research has determined that Engi-neering Status is a critical parameter in the Front-EndLoading (FEL) of projects and can be linked, along withother FEL activities, with improved absolute cost andschedule performance and predictability for pharmaceuti-cal projects.5 While Engineering Status appears to beimportant in and of itself as a driver of qualificationschedule performance, it also may be standing in as aproxy for overall early definition efforts that can impactprojects throughout project execution.

• Qualification Schedule Definition. For projects thathad prepared a schedule to a critical-path or milestonelevel, IQ/OQ schedules were shorter than for projects thatwere working toward end dates only or had no schedule atall.

Other IQ/OQ Schedule DriversThe IQ/OQ schedule variables mentioned above account forapproximately 70 percent of the variance in the IQ/OQschedules in the study database.

Other potential drivers of IQ/OQ schedule may be elimi-nated from the model, either because their relationship with

the dependent variable does not hold up when key character-istics are controlled or because of strong co-linearity withother, more significant schedule drivers. However, theirrelationship to IQ/OQ schedule is suggestive. In addition,there is some overlap with the practices that appeared sig-nificant in our cost analyses.

• Commissioning and Qualification Overlap. As dis-cussed for the IQ/OQ cost analysis, the data suggest thatincreased overlap of activities correlates with shorter IQ/OQ schedules.

• Other Validation Group Commitments. Projects inwhich personnel involved in qualification activities wererequired to perform other tasks, such as writing plantSOPs or batch records and/or performing technical assess-ments were found to have longer IQ/OQ schedules.

• Overlap of Personnel Writing and Executing Proto-cols. For the majority of projects in the study, the person-nel responsible for writing protocols also were responsiblefor executing protocols. For those projects in which therewas no overlap or overlaps were minimal, a correlationwith longer IQOQ durations was seen.

The MC/OQ Schedule ModelTable D shows the regression relationship between the natu-ral log of the MC/OQ schedule [ln(MC/OQ Schedule)] and thekey project characteristics and drivers that make up the MC/OQ Schedule Model. The statistics presented in the table arebased on the fit of the independent variables within amultilinear regression analysis.

“Uncontrollable” Characteristics• New Technology. As with IQ/OQ schedules, increasing

“new technology” ratings correlate with longer MC/OQschedules. Based on the regression analysis, it is clear thatthe extent to which projects incorporate new technology isa very strong driver of the MC/OQ duration for pharma-ceutical industry projects. New technology was deter-mined to be the only significant project characteristic inthe MC/OQ Schedule Model.

Interestingly, although total project size is a driver of both IQ/OQ cost and IQ/OQ schedule, this project characteristic is nota significant driver of MC/OQ schedule in a multilinearregression model. However, total project size correlates withMC/OQ duration in a simple regression analysis.

Key “Controllable” DriversThe following schedule overlaps were significant drivers ofMC/OQ duration in the MC/OQ schedule model:

• Percentage Overlap of Qualification with Construc-tion Schedule. Not surprisingly, increased overlap drivesdecreased MC/OQ schedule.

The correlation between percent overlap of construc-Table D. Relationship between key characteristics and drivers andMC/OQ schedule.

Drivers t-score P>t

New Technology 5.77 0.00

Percentage overlap of Qualification with -6.76 0.00Construction schedule

Percentage overlap of IQ and OQ schedules 3.51 0.00

IQ and OQ Analysis


©Cop

yrig

ht IS

PE 2

007

tion and qualification and MC/OQ duration is so strongthat this appears to be a relatively simple strategy forproject teams to decrease the time between when a facilityis mechanically ready and when it is “ready” from aregulatory standpoint. Obviously, increased overlaps re-quire increased coordination between construction andvalidation personnel in the field, as well as earlier plan-ning, scheduling, and resource allocation. For the projectsin the study database, it appears that project teams wereable to prepare for and execute these overlapped phasesand gain a schedule advantage.

Figure 2 shows the strength of the relationship be-tween percentage overlap of construction and qualifica-tion and MC/OQ schedule. This schedule overlap is thestrongest predictor of MC/OQ duration in the MC/OQmodel. (Note that the Figure uses a log scale for MC/OQschedule, hence the presence of minus numbers).

• Percentage Overlap of Installation Qualificationand Operational Qualification Schedule Phases.Increased overlap drives longer MC/OQ schedules. Thisresult is consistent with the relationship between IQ/OQoverlap and the overall IQ/OQ schedule.

Other MC/OQ Schedule DriversThe above variables account for approximately 75 percent ofthe variance in the MC/OQ schedules in the study database.

Other potential drivers of the MC/OQ schedule may beeliminated from the model, either because their relationshipwith the dependent variable does not hold up when keycharacteristics are controlled for, or because of strong co-linearity with other, more significant schedule drivers. How-ever, we want to note these drivers, whether characteristicsor practices, because their relationship to MCOQ durations issuggestive, within the limits of the database.

“Uncontrollable” Characteristics• Project Size. Greater TIC correlates with longer MC/OQ

schedules.

• Process Complexity. For this analysis, process complex-ity is measured by the number of steps in a processrequired to perform all chemical and physical operationsfor the manufacture of product. For projects in the studydatabase, our analysis shows that an increased number ofprocess steps correlates with longer MC/OQ durations.This is not an unexpected result; and it is assumed thatmore complex processes take longer to complete qualifica-tion requirements, especially if the qualification strategyis to use a system-by-system approach.

• CIP/SIP. Based on regression analysis of a categorical(1,0) variable representing projects that were required toinstall Cleaning-In-Place (CIP) and/or Sterilize-In-Place(SIP) systems, this project characteristic was found tocorrelate with increased MC/OQ duration.

“Controllable” Drivers• Commissioning and Qualification Overlap. The data

suggest that increased overlap of activities correlates withshorter MC/OQ schedules. For this analysis, it was notpossible to evaluate if there were schedule trade-offsbetween commissioning and IQ/OQ and/or MC/OQ. Thismay be examined in more detail in future analyses.

• Overlap of Commissioning and Qualification TeamPersonnel. In addition to the tasks required to completecommissioning and qualification, the personnel used tocomplete these phases varied among companies and amongprojects within companies. For the majority of projects,the same team was used to perform all activities under thecommissioning and qualification umbrella, or there wasactive communication and crossover between teams. Aminority of projects used different personnel to executethe activities described as commissioning versus qualifi-cation. Projects were placed into one of three categoriesdescribing the extent of team overlaps. Projects withincreased overlaps of commissioning and qualificationpersonnel achieved shorter MC/OQ durations.

• Engineering Status at Authorization. As described forthe IQ/OQ schedule, as the status of engineering defini-tion improved, so the MC/OQ duration decreased forprojects in the study dataset.

• Qualification Schedule Definition at the Time ofAuthorization/Detailed Engineering Start. Forprojects that had prepared a schedule to a critical-path ormilestone level, MC/OQ durations were shorter than forprojects that were working toward end dates only or had noschedule at all.

• Other Validation Group Commitments. Projects inwhich personnel involved in qualification activities wererequired to perform other tasks such as writing plant

Figure 2. MC/OQ schedule is correlated with the percentageoverlap of construction and qualification.

IQ and OQ Analysis


©Copyright IS

PE 2007

SOPs or batch records and/or performing technical assess-ments were found to have longer MC/OQ schedules.

• Existing Corporate or Site SOPs for Commissioningand Qualification. For projects in the study database,about two-thirds were able to use existing site and corpo-rate standard operating procedures outlining the require-ments and approach for commissioning and qualification.Projects were classified as to whether the majority ofactivities for qualification were covered by pre-existingSOPs. The remaining projects, approximately one-third ofour project sample, had either no existing SOPs, limitedSOPs, or there were major changes to the existing SOPssometime during project execution. Projects in which teamshad access to existing, unchanging, standard qualificationprocedures, completed faster MC/OQ.

• Preparation of FAT and SAT Protocols. The majorityof projects in the study database prepared or had access toFAT and SAT protocols, which were executed as part of theproject equipment procurement and construction effort.

This practice was shown to correlate with faster MC/OQschedules and may be an indication of overall planningand controls for the projects.

• Date When the Validation Master Plan (VMP) wasFinalized. Projects were assigned into one of four cat-egories for “sign-off” date of the VMP. VMPs were final-ized and signed off at 1. project authorization, 2. beforethe start of construction, 3. before mechanical comple-tion, or 4. following mechanical completion. About two-thirds of the projects were in categories two and three.The remaining projects were equally divided betweencategories one and four. A regression analysis of the MC/OQ duration against the date of VMP sign-off shows thatprojects with earlier VMP approvals demonstrate fasterMC/OQ schedules. This result is not unexpected, andmay be correlated with an overall level of planning for IQ/OQ execution that results in faster delivery of the quali-fication effort.

NOTE: The status of the VMP at the time of authoriza-tion or by the start of detailed engineering was examinedfor each project with the idea that the VMP may be used asa planning guide for other project definition activities anddeliverables, such as cost and schedule estimates. Al-though it was reasonably expected that the early status ofthe VMP might affect later qualification outcomes, forprojects in the study, VMP definition was not a significantdriver of cost or schedule.

• Changing Objectives. Projects were categorized as towhether the objectives for the project were changed duringproject execution. Although cost or schedule predictabilityoutcomes are expected to be affected by these changes,which are driven from outside the project team, it isinteresting to note that changing objectives also can bestatistically linked with overall longer MC/OQ durations.

ConclusionsCommissioning and qualification are significant phases inthe overall project delivery system. Yet surprisingly, a largenumber of projects systems:

• are not accurately capturing the true costs in terms oflabor and currency of the qualification effort

• do not put in the effort in the early, front-end phase of aproject that is necessary to sufficiently plan and definequalification activities and ensure the future success ofthe qualification phase

However, this study has been able to determine those plan-ning and execution practices that are found to be statisticallysignificant drivers of qualification success, thereby providingcompanies with a potential focus for their activities as theywork to refine and standardize the qualification process forcapital projects. In summary, those Best Practices are listedin Table E.

Practice Description

Qualification Better-defined schedules in terms of milestone-levelSchedule development or critical path analysis, resource loading,Definition at and integrated activities improve performance.Authorization

Commissioning Increased overlap between the Commissioning andand Qualification Qualification activities is associated with improvedOverlap Qualification cost and schedule performance.

IQ/OQ Overlap Increased overlap between IQ and OQ dampensschedule performance.

Project Team Specific project team attributes are clearly associatedPerformance with better performance. These attributes include

avoiding key member turnover, eliminating the lateassignment of key functions to the team, andavoiding multiple assignments with conflictingassignments. In addition, functionally integrated teams,well-defined project objectives, and documentedroles and responsibilities improve performance.

Front-End Better-defined capital projects are clearly associatedLoading with improved absolute and predictable performance.

These practices also apply to the execution ofCommissioning and Qualification activities. Inparticular, improved Engineering Status and ProjectExecution Planning are important elements forCommissioning and Qualification project performance.

Existing Standard Procedures that are already in place facilitateOperating performance. Companies without SOPs should considerProcedures (SOPs) developing them.for Commissioningand QualificationActivities

Reduce Other Projects with personnel who are dividing their timeValidation Group between activities outside of Commissioning andCommitments Qualification tasks and the Commissioning and

Qualification tasks themselves, tend toward lesseffective Qualification performance.

Vendor Vendor involvement improves Qualification costs withQualification a neutral affect on schedules.Support

Table E. Qualification best practices.

IQ and OQ Analysis


©Cop

yrig

ht IS

PE 2

007

References1. Installation Qualification is defined as: “The documented

verification that all aspects of a facility, utility, or equip-ment, that can affect product quality, adhere to approvedspecifications (e.g., construction, materials), and are cor-rectly installed.” As noted in the ISPE Baseline® Guide onCommissioning and Qualification, the activities that makeup IQ correspond to the inspection requirements as notedper Good Engineering Practices (GEP).

Operational Qualification is defined as: “The docu-mented verification that all aspects of a facility, utility, orequipment, that can affect product quality, operate asintended throughout all anticipated ranges.” As noted inthe ISPE Baseline® Guide on Commissioning and Qualifi-cation, the activities that make up OQ correspond to theSetting-to-Work, Regulation, and Testing requirementsas noted per GEP.

2. In developing the questionnaire, we would like to acknowl-edge the usefulness of the ISPE Baseline® Guide on Com-missioning and Qualification in developing the questions.ISPE Baseline® Pharmaceutical Engineering Guide, Vol-ume 5 - Commissioning and Qualification, InternationalSociety for Pharmaceutical Engineering (ISPE), FirstEdition, March 2001, www.ispe.org.

3. Project TIC includes design, engineering, procurement,construction costs up to mechanical completion. (In otherwords, it excludes commissioning, qualification, and vali-dation costs).

4. Biological Facility – A production facility that uses cellculture (mammalian or bacterial) as the production method.

5. Lawrence, G.R., “Pharmaceutical Capital Investment: Timeto Rethink Corporate Culture,” The Chemical Engineer(TCE), August 2004, p. 28-29.

AcknowledgementsThe authors would like to acknowledge the assistance of thefollowing colleagues in the development of this study: Ed-ward W. Merrow, Scott Burroughs, David Gottschlich, GobJuntima, and Kelly Sambuchi.

About the AuthorsAllison J. Aschman is Manager of the Phar-maceutical Division at Independent ProjectAnalysis (IPA) and is based at the head officein Virginia, USA. IPA is a consultancy spe-cializing in analytical research into the rootcauses of success and failure of capital in-vestment projects and project systems. Priorto joining IPA, Dr. Aschman was quality

manager and analytical chemistry team leader for a specialtychemicals company producing fine chemicals and intermedi-ates for consumer products, healthcare, agriculture, andcommodity Industries. She has a doctorate in analyticalchemistry from Duke University, Durham, North Carolinaand a bachelor’s degree in chemistry from Bloomsburg Uni-versity of Pennsylvania. She is a member of the ISPE Chesa-peake Bay Area Chapter and a member of the AmericanChemical Society (ACS). She can be contacted by telephone at+1-703-726 -5338 or by e-mail at: [email protected].

IPA, 44426 Atwater Dr., Suite 100, Ashburn, Virginia20147.

Gordon R. Lawrence is a Senior ProjectManagement Analyst at Independent ProjectAnalysis (IPA). He is based in the Europeanoffice of the company, in the Hague, theNetherlands, where he manages Europeanpharmaceutical work. Prior to joining IPA,he was a project manager with the generalservice contracting firm, Jacobs Engineer-

ing. He has degrees in chemical and biochemical engineeringand in business administration. He is a Chartered Engineerin the UK and in Europe, a Fellow of the UK Institution ofChemical Engineers (IChemE), and a Member of the Ameri-can Institute of Chemical Engineers (AIChE). He is a mem-ber of the ISPE Netherlands Affiliate and currently serves onthe ISPE Membership Services Committee. He can be con-tacted by telephone at +31-70-315-1080 or by e-mail at:[email protected].

IPA, Prinses Magrietplantsoen 32, 2595 BR The Hague,The Netherlands.

Permissible Exposure Limits


©Cop

yrig

ht IS

PE 2

007

This articlepresents thenew OSHAregulationregardingpermissibleexposure limitsto hexavalentchromium, acommon by-product ofweldingstainless steel,and how toensurecompliance. Thearticle explainsthe regulation,includingreasons for thestricter limit,and quotes theOSHA standard.The process forcalculating theamount offumes generatedin a facility isprovided, alongwith anexample.

OSHA Limits Stainless Welding FumeExposure to 5 µg/m3 – “What will ittake to be Compliant?”

by Ed Ravert





Introduction

The term “weld fume” is a bit misleadingin any discussion regarding methods tocontrol employee exposure. The reality? Weld fume is a dust generated

during the welding process making it both adust collection and employee protection issue.Generally, the emitted fumes will be 0.6 to 7.0percent of the welding materials used. Of thissmall percentage, approximately five percentwould include hexavalent chromium (Cr(VI))fumes. This article details OSHA’s new regula-tion1 for employee exposure to Cr(VI) weldfume when welding stainless steel, and themost effective dust collection devices availabletoday to meet both exposure control and airfiltration guidelines.

The New Hexavalent ChromiumExposure Regulation

OSHA’s new regulation for employee exposureto hexavalent chromium, a natural metal usedin the manufacture of stainless steel, has sig-nificantly reduced the permissible exposurelimit from 52 to five micrograms. The regula-tion2 also includes provisions relating to pre-ferred methods for controlling exposure, respi-ratory protection, protective work clothing andequipment, hygiene, medical surveillance, haz-ard communication, and recordkeeping. Thenew rule, effective 30 May 2006, does not re-quire the installation of engineered controls,including dust collection and air filtration equip-ment until 30 May 2010.

Why the TighterRestriction?

This standard appliesto all manufacturingprocesses where hexa-valent chromium ispresent. These com-pounds are widely usedin the chemical indus-try as ingredients,catalysts in pigments,metal plating, andchemical synthesis.And, while all types ofwelding could be af-fected, the highest con-centration will be thefumes generated inwelding stainless steel.The major health ef-fects associated withexposure to Cr(VI) in-

Figure 1. Portablecartridge weld fumecollector.



©Copyright IS

PE 2007

clude lung cancer, nasal septum, ulcerations and perfora-tions, skin ulcerations, and allergic and irritant contactdermatitis.

This new regulation provides for greater employee protec-tion against these health risks by lowering the PermissibleExposure Limit (PEL) for hexavalent chromium, and for allCr(VI) compounds, from 52 to a much more stringent fivemicrograms of Cr(VI) per cubic meter of air (5 µg/m3) as aneight-hour time-weighted average.

Quotes from the Regulation“OSHA concludes that engineering controls, such as LocalExhaust Ventilation (LEV), process control, and processmodification or substitution can be used to control exposuresin most operations.”2

“OSHA has determined that the primary controls most likelyto be effective in reducing employee exposure to Cr(VI) areLocal Exhaust Ventilation (LEV) and improving generaldilution ventilation. ...this includes installing duct work, atype of hood, and/or a collection system.”4

“Paragraph (f) of the final rule, Methods of Compliance,establishes which methods must be used by employers tocomply with the PEL. It requires that employers instituteeffective engineering and work practice controls as the pri-mary means to reduce and maintain employee exposures toCr(VI) to levels that are at or below the PEL ... Engineering

controls can be grouped into three main categories: 1. Substi-tution, 2. isolation, and 3. ventilation, both general andlocalized.”5

“Welding: The welding operations OSHA expects to triggerrequirements under the new Cr(VI) rule are those performedon stainless steel, as well as those performed on high-chrome-content carbon steel and those performed on carbon steel inconfined and enclosed spaces. ... OSHA has determined thatengineering and work practice controls are available to per-mit the vast majority (more than 95 percent) of weldingoperations on carbon steel in enclosed and confined spaces tocomply with a PEL of 5 µg/m3. ... OSHA has determined thatthe PEL of 5 µg/m3 also is feasible for all affected welding jobcategories on stainless steel. ... The two most common weld-ing processes, Shielded Metal Arc Welding (SMAW) and GasMetal Arc Welding (GMAW), ...may require the installationor improvement of LEV. ... There are ongoing efforts to reducethe use of SMAW and replace it with GMAW for both effi-ciency and health reasons. ... OSHA has revised its estimateof the percentage of SMAW welders that can switch to GMAWfrom 90 percent to 60 percent. ... For those stainless steelSMAW operations that cannot switch to GMAW and evensome GMAW operations, the installation or improvement ofLEV may be needed and can be used to reduce exposures.OSHA has found that LEV would permit most SMAW andGMAW operations to comply with a PEL of 5 µg/m3, OSHArecognizes that the supplemental use of respirators may stillbe necessary in some situations.”6

Do We Need New Dust Collection and AirFiltration Equipment?

Or, Can we Retrofit Existing?If your operation has a well designed and operating close-capture hooding system followed by a cartridge dust collectorthat exhausts the “cleaned” air outdoors, you should be incompliance with the new OSHA 5 µg/m3 requirement. How-ever, if the air is returned into the work place, there shouldbe a monitored HEPA after-filter.

Ambient air fume control systems, with a cartridge dustcollector, followed by a HEPA after-filter, can be effective inremoving welding fume emissions. But note that with anambient air system, it is probable that some contaminatedfumes can be pulled up past a worker’s breathing zone. If youare using other types of fume control equipment for Cr(VI)gasses, it may be necessary to replace them with cartridgecollectors and HEPA after-filters.

Weld Fume GenerationTo determine weld fume control needs, the first step is tocalculate how much is being generated. The various types ofwelding processes and the type of metal being welded producedifferent amounts of fumes. The weight of fumes generated isa percentage of the weight of deposited metal. This percent-age can be based on the length of weld electrode used or theweight of weld electrode used. Table A lists the typical ratiosto determine the amount of fume generated by welding

Figure 2. Swing arm weld fume collector.



©Cop

yrig

ht IS

PE 2

007

process and metal type. From this, the amount of Cubic Feetof air per Minute (CFM) needed to capture and remove thefume can be determined to ensure that the proper sized dustcollector is specified.

Calculation Example:If continuously GMAW on carbon steel at the rate of 10pounds per hour, the maximum rate of weld fume producedwould be 10 pounds × 0.009 (0.9%) or 0.09 pounds per hour.By comparison, if continuously GMAW on stainless steel atthe rate of 10 pounds per hour, the maximum rate of weldfume produced would be 10 pounds × 0.07 (0.7%) or 0.7pounds per hour that would need to be removed by a properlysized dust collector.

Best Dust Collection Control Approach forStainless Weld Fumes

Media filtration units, including cartridge dust collectors, areideally suited for collection of weld fume. Depending on thewelding process, source capture systems do the best job ofcapturing weld fume contaminants using the least amount ofCubic Feet of air per Minute (CFM). Source capture systemsinclude hoods, ducting, an air-cleaning device, and air mov-ing devices (fans). The air-cleaning device can include swingarm(s) and hood(s). With separate hoods and ducting, thecartridge collector is the preferred way to deal with stainlessweld fumes because the weld fume is captured before it canescape into the ambient air. If ambient air collection isdesired, the welder will be required to wear personal respira-tory protection at all times. Even then, and if the air is to bereturned into the work place, a monitored safety HEPA filtershould be used.

There are many factors that can make a source capturesystem impractical to collect stainless weld fume. Specifi-cally where:

• Work involves large parts and the welder has no fixedoperating position, making source capture difficult toimpossible.

• The welder is unable to use a hooded system. Some sourcecapture systems may require physical positioning by thewelder. If it is unlikely that the worker will perform thatpositioning, the system will be rendered ineffective.

• There are a large number of small weld fume producers in

a confined area.• Overhead cranes and process obstruction make ducting

installation impossible. However, properly designedunducted systems can keep the air cleaning units out ofthe craneways and still achieve effective results.

NOTE: While Electrostatic Precipitators (ESPs) are widelyused in the collection of weld fume in carbon steel welding,they are not ideally suited for collecting stainless steel weldfume as their overall efficiency is less than that of mediafiltration collectors.

Does the New Regulation also Apply toField Fabrication?

Yes it does, “anytime the welder is exposed to Cr(VI) fumes.”If you are doing stainless steel welding on piping, and if thewelding can be done in one place where a dust collectionsource capture hood can be brought to the task, personalsafety breathing devices would not be needed. But if astainless weld bead is being run 20-feet down the length ofpipe, and the dust collection hood cannot move with thewelder, then the Cr(VI) fumes will escape to the ambientatmosphere requiring the use of a personal safety breathingdevice.

Should source capture be impractical for your operation,the only practical answer is for your welders to wear appro-priate protection gear. Additionally, it also would be appro-priate for your welders to occasionally wear a monitor tocheck on compliance. Contact the American Welding Societyfor more information on personal safety monitors and protec-tive gear recommendations.

Dust Collection Design ConsiderationsTo select the appropriate media filtration solution for youroperation, the following considerations need to be addressed:

• size and shape of the space where welding operations areperformed

• containment generation rate and the desired steady-statecontainment levels

• required number of ambient air changes per hour

Table A. Weld fume generation ratios.

RangeWeight of Fumes/

Welding Process Metal Type Weight of Deposited Metal

FCAW (Flux Core) Carbon Steel 0.9 to 2.4%Stainless Steel

SMAW Carbon Steel 1.1 to 5.4%

SMAW Stainless 0.3 to 1.4%High Alloy

GMAW Carbon Steel 0.3 to 0.9%

GMAW Stainless Steel 0.6 to 7%

Figure 3. Downward flow cartridge weld fume dust collector withsource capture.



©Copyright IS

PE 2007

• existing ventilation and replacement air rates, plus allHVAC system air volumes and airflow patterns. Thisapplies more to ambient air than with the use of sourcecapture systems.

• airflow pattern continuity noting seasonal variations• appropriate filtration technology to match room layout,

containment generation, and designed steady state con-tainment levels

The Only Way to Know for Sureif You Are Compliant

The reality is that it is very likely that some Cr(VI) gasemissions will be included in the fumes generated duringwelding stainless steel. Generally, the emitted fumes will be0.6 to 7.0 percent of the weight of the welding materials used.Of this small percentage, approximately five percent wouldinclude Cr(VI) fumes. While OSHA does not mandate the useof personal monitors, and as effective as source capture mediafiltration dust collection and air filtration equipment is, theonly way to know with absolute certainty in our view is tomake sure your welders, and nearby workers, wear personalmonitors that can indicate the Cr(VI) fume level.

References1. Detailed information and a copy of the 287-page regula-

tion can be found at the OSHA Web site: http://www.osha.gov/SLTC/hexavalentchromium/index.html.

2. (Volume 71, Number 39, 10099-10385).3. (Volume 71, Number 39, 10334).4. (Volume 71, Number 39, 10262).5. (Volume 71, Number 39, 10345).6. (Volume 71, Number 39, 10262-10263).

About the AuthorWith more than 30 years of experience in theventilation industry, Ed Ravert is an expertat solving industrial ventilation and air pol-lution problems. His knowledge and experi-ence have earned him a position as an in-structor at the annual Industrial VentilationConferences, where he has trained more than600 professionals on how to evaluate the

needs of their customer and design a solution that fits thoseneeds. Ravert’s experience and his ability to break down verycomplex technical subjects serves him well as both a teacherand in his responsibilities as a liaison between United AirSpecialists’ (UAS) sales and engineering departments. Healso assists in the development of new products and conductstraining programs for UAS employees and sales representa-tives. During his career, Ravert has worked for several majorplayers in the air pollution control market, including MACEquipment, United McGill, Donaldson/Torit and Dustex. Hecan be contacted by telephone at: +1-800-252-4647 or byemail at: [email protected].

United Air Specialists Inc., 4440 Creek Rd., Cincinnati,Ohio 45242-2832.

Particle Shape Analysis


©Cop

yrig

ht IS

PE 2

007

This articleintroducesnumericalmethods tocharacterizeparticle shapeand applicationsin thepharmaceuticalindustry. Basictheory andcomputationalmethods areshown, followedby a discussionof selectedapplicationswithin thepharmaceuticalmanufacturingcycle, such asequipmentselection,process safety,processstandardization,and counterfeitproductdetection.

Numerical Methods for Particle ShapeAnalysis (PSA) and Applications toPharmaceutical Manufacturing

by David A. Vetter





Currently, many facilities handling pow-ders assume that as long as the mate-rial is chemically identical from onebatch to the next, then the batches are

the same. This assumption is likely derivedfrom liquids processing where if the material ischemically identical and reasonably well mixed;all other properties are well predicted. How-ever, for powders, nothing could be furtherfrom the truth: a concrete wall before and afterbeing mechanically demolished has the samechemistry, but very different physical proper-ties. The same is also true in powdered materi-als: graphite dust is an excellent lubricant,while diamond dust is an abrasive, yet both arepure carbon. Understanding and predictingbulk material behavior requires tools to helpunderstand changes in individual particle prop-erties that may affect the bulk material. Re-cent work in non-pharmaceutical areas hasprovided new tools to identify and characterizeindividual particles beyond ‘mean particle di-ameter’ of size classification. One such tool isParticle Shape Analysis (PSA). PSA can beparticularly useful in understanding any sur-face area dependent property, from electrostat-ics to particle flow to surface chemistry effects.

PSA has several potential applications in

pharmaceutical manufacturing. PSA is amethod to measure the size, shape, and textureof particles, typically anything that can beconveniently imaged. Particles can range insize from several inches mean diameter downto smoke or soot particles. The difference be-tween shape and texture is actually an arbi-trary choice, as numeric methods show. Theapplications for PSA in pharmaceutical manu-facturing include material property testing,quality control, material behavior prediction,and potentially detection of counterfeiting.

PSA requires an image of each particle inthe sample, and a statistically large enoughnumber of particles in a sample to be represen-tative of the bulk material. The number ofparticles in a sample to be statistically valid isoften surprisingly small; only 300 kernels canaccurately grade a truckload of field corn. Im-aging the particles can be done in either aCartesian coordinate system (X-Y) or a polarcoordinate system, and the particle must havea fairly clear definition from the background.Choice of the scanning coordinate system isentirely arbitrary, but the choice of lightingand background will aid definition. Most often,the scanning is performed using the planarcoordinates with the particle resting in the

‘natural’ or ‘lowest en-ergy’ position, i.e., theoutline of the particlewhile it is resting on asmooth surface and notleaning against any-thing else. While math-ematically the FourierTransform can be usedfor three dimensionalsystems (X-Y-Z), the

Figure 1. CartesianCoordinate System.



©Copyright IS

PE 2007

technology to scan and interpret an object in three dimen-sions is not nearly as well developed as the software tomanage the information after it is in the computer. (PixarAnimation is much further along at modeling three dimen-sional objects than most body/facial recognition software ormotion capture systems are at capturing fine detail) ParticleShape Analyzers are now commercially available incorpo-rating the microscope, a video capture device, computer, andsoftware as a turnkey system.

Use of a Fourier Transform will make the initial particleorientation and the choice of starting coordinate systemirrelevant. In Particle Characterization in Technology,1 J.Beddow proposed using a Fourier transform of the scannedparticle samples’ coordinates as a means to quantitativeanalyze particle shape. This mathematical method has sev-eral advantages in particle analysis: the computation is notdifficult in terms of computer power; it is possible to automateboth the particle scanning and the computation; the trans-formed ‘values’ do have some direct correlation to physicalproperties; the transform is independent on the startingorientation of the particle; and statistical analysis of theresults becomes possible. The mathematical process of theFourier Transform translates the series of ‘perimeter’ coordi-nates into a series of coefficients or factors.

The correlation works as follows: the first Fourier coeffi-cient represents mean particle diameter; the second coeffi-cient is the degree of linearity (rod like); the third coefficientis how triangular; the fourth how square the particle is, etc.As is the case with waveform building in electrical signals, acombination of coefficients can create almost any closedshape. Particles with ‘higher order’ coefficients (above what

might be considered ‘shape’) that have a relatively highervalue are particles that have extended surface area and maybe more reactive in any surface area dependent process suchas ignition/combustion or leaching/solvation (the reverse ofcrystallization). Systems that J.K. Beddow and A. F. Vetterstudied at length and found a very high degree of correlationbetween particle shape and other properties included grain(particle shape accurately measures the number of foreignparticles, number of broken kernels, and moisture content)and fleet engine maintenance (a drop of oil from the crank-case can accurately predict when an engine needs an over-haul and which engine parts have excessive wear). A deeperintroduction to the mathematical calculations and the con-cepts influencing different milling technologies on particleshape can be found at: http://www.kona.or.jp/search/20_188.pdf.

Mathematical Coordinate SystemsFigures 1 and 2 demonstrate the issues with scanning andnumerically analyzing particle shape. The same particleoutline is shown, but obviously a Cartesian coordinate basedtable will not readily show that the two outlines are the same.A Polar Coordinate system based table may give a hint thatthe two shapes are identical except for the rotation angle inFigure 2, but will not easily recognize small differences in sizeor texture. Polar coordinates will produce very differentnumerical values should the origin of the coordinate systemnot coincide with the centroid of the figure. Transforming thedata into Fourier Space eliminates the information related tothe particles original orientation, and converts the meandiameter into a single coefficient. Statistical analysis of theterms after the first coefficient (mean diameter) will rapidlyshow if the particles are similar in shape. Since crystallinesolids will tend to fracture along the same faces, the particlesof a solid will tend to have the same shape if processed(milled) by the same forces. This similarity in shape holdstrue even when the size is different (due to the nature ofcrystals). This is then the heart of Particle Shape Analysis.

In Figure 3, the first five terms of the Fourier Transform areshown as if each term were dominant. In a less regular shape,these terms combine the same way a more complex wave formcan be built up from simpler (shorter frequency) sine waves.

Figure 2. Polar Coordinate System.

Figure 3. Examples of shapes with a dominate term among the first few Fourier Transform terms.



©Cop

yrig

ht IS

PE 2

007

Figure 5. Example of a closed shape that is not appropriate forpolar coordinates.

One limitation of the Fourier Transform should be noted:a particle shape such as Figure 4 cannot be represented inFourier Space, as it has two separate, unconnected surfaceboundaries. One aspect of Fourier Space mathematics is thatit can not accommodate either open figures (they must be aclosed surface) or complex figures with multiple unconnectedsurfaces.

Finally, the choice of the initial scanning coordinate sys-tem is important if the particles are deeply convoluted suchas in Figure 5. A polar coordinate system cannot resolve theparticle edge when the length along angle a has multiplevalues. In this instance, a Cartesian coordinate scan would bepreferable, but may still present issues in both normal andFourier Space.

Process SafetyShape or texture is a variable for dust explosions not men-tioned in “Inert Milling Systems.”2 In “Numerical Modeling ofDust Explosions, The Influence of Particle Shape on Explo-sion Intensity,”3 it was shown that the particle shape ortexture plays an important role in flammable dust deflagra-tions.

Material handling methods (including milling) of solidsproduce dust particles of varying shape as well as size. Forpotentially flammable solids, the ratio of particle surfacearea to mass is an important determinant of the ease ofignition and flame propagation. A particle with a high surfacearea to mass ratio (either small particles or a particle deviat-ing substantially from a spherical or cubical shape) will havea much lower Minimum Ignition Energy (MIE) compared tothe same chemical material, but with a larger size and/ormore spherical particle shape. By analogy, if the particle sizeand shape resembles kindling rather than logs, the ease ofstarting a ‘fire’ changes dramatically. In pharmaceuticalmanufacturing, this is important in three areas: over-millingof product to a finer size than intended (process control),material property characteristics (especially if milling is nearthe end of the manufacturing process of the API), and choiceof milling equipment.

Some of the variables in milling in food grade or pharma-ceutical manufacturing are well known: desired particle sizeand range of acceptable particle size, power input required,process throughput rate, materials of construction relative toproduct hardness, dust collection, ease of cleaning betweenbatches, etc. Safety in milling potentially flammable materi-als or materials degraded by heat add more complicatingfactors, and can prompt the use of inert gas systems.

Equipment that does not yield predictable, uniform par-ticle size distributions from batch to batch is not undercontrol. Variation in particle size recovery from crystalliza-tion, centrifuge, or dust collection equipment will obviouslyimpact milling, but so will changes in starting moisturecontent, as well as grain size and growth during crystalliza-tion. Also critical to safety is the width of the particle sizedistribution: a mill that produces a higher percentage of‘fines’ is more prone to ignition. A change in the filter cloth ofa centrifuge leading to a higher percentage of fines before

milling may initially appear to be a way to increase productrecovery, but not after the mill catches fire or explodes.Process control and repeatability is vital to safety in flam-mable material milling.

Equally critical is the shape of the particle produced by themill: a mill that gradually abrades coarse particles (like ariver producing rounded pebbles) may well produce particleswith a much higher MIE than a mill that shatters particles,producing a much greater surface area. There are examplesof grain milling operations that run for years without seriousdust control measures and do not have dust explosions, whilein one case, a mill exploded only a week after extensivechanges and testing to reduce dust gave it a ‘clean bill ofhealth.’ Thus, the choice of milling technology can impactprocess safety.

Therefore, for safety, it is imperative that the testingprogram2 be performed on material produced by the equip-ment in question. Any changes in equipment OR upstreamparticle growth, collection, or handling processes may invali-

Figure 4. Example of a shape not appropriate for FourierTransform (not a single connected surface).



©Copyright IS

PE 2007

date the testing and lead to a false sense of security. Re-peating the PSA testing from time to time to measure millwear or unknown/unplanned process changes may be pru-dent for safety as well as to maintain a validated processstate.

Process Control/ValidationFrom the process validation standpoint, changes in millingtechnology may represent another challenge. Changes inparticle surface area of an Active Pharmaceutical Ingredient(API) could significantly change the rapidity of bio-uptake.Consider the difference of ease and speed of dissolvingconfectioner’s sugar vs. an equal mass of sugar that has beencrystallized on the walls of a container. The same amount ofwater and same ‘power’ input and time will not yield equalsuccess in cleaning a dirty humming bird feeder as dissolvingthe humming bird food, despite the fact that it is ‘the samematerial.’ This is due to the fact that solvation is a surfacearea dependent process. Changes in particle shape/surfacearea may lead to changes in solution (or blood) concentrationand Safety, Identity, Strength, Purity, or Quality (SISPQ)impact. There is certainly the potential for an API that is nothighly soluble in the stomach (or the sterile solution for aninjectable drug) to produce a substantially different endconcentration in the body due to a change in particle shape,and therefore, available surface area, changing the time thatit needs to dissolve. Again, particle size does not provide thecomplete measure of relevant material properties. Seewww.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=9090926&dopt=Abstract for a studyof particle shape and in-vitro effects of the same API. Inintermediate pharmaceutical manufacturing, changes inmilling technology may be less troublesome than at the APIstage, but can affect process cycle times. Therefore, PSAbecomes a potential tool for Quality Control/Quality Assur-ance testing.

This also leads to using PSA to guide the choice of millingtechnologies. Since a significant fraction of milling in apharmaceutical environment is near the end of API produc-tion, critical understanding all the ramifications of a pro-posed process change should be obvious. The normal concernsof throughput, power input, ease of cleaning, materials ofconstruction and traceability, seal integrity, and control/capture of foreign material from mill wear will still operate.The energy input in milling will ultimately end up as heat.Products that are heat sensitive or subject to atmosphericdegradation or pose a risk of flammability/explosion willlikely require a mill that can be inerted. APIs may require re-validation to prove that they retain the same bioavailabilityafter an equipment change. Easy control of sources of adul-teration (either carryover batch to batch or mill wear pro-ducing unwanted metal particles) also should be testable. Allof these issues are perhaps most obvious when a processmoves from a clinical trial/pilot plant scale to full scaleproduction; but relocating production to a new facility orreplacing old equipment should trigger testing for both pro-cess safety/control and potential SISPQ impacts. PSA may

represent an easier method to show that a new millingtechnology or new equipment will not have SISPQ impact.

Unfortunately, many facilities assume that if the materialis chemically the same and passes the correct screen size,their process is under control. As discussed above, there arenumerous cases where this may not necessarily be true. Thefirst improvement in control comes when the measurementincludes a statistical distribution of particle sizes, ratherthan a simple pass/fail test of maximum particle size. Muchless common and much more sophisticated is utilizing PSAand then correlating the particle shape to important param-eters (such as chemical reactivity or bioavailability). For apharmaceutical process, the correlation of particle shape toSISPQ would have to be determined material by material.

Because each material is different, understanding how aparticular milling technology achieves size reduction is criti-cal when evaluating the appropriateness of a piece of equip-ment for the given process application. Unfortunately, due tothe potential impact of particle shape in a pharmaceuticalenvironment, there are not many generalizations that can besafely made. Mills that shatter particles (such as hammer orcage mills) are in general more sensitive to changes inproduct moisture than abrasive type mills (ball mills), andrequire greater process control upstream of the mill to avoidunacceptable process variability downstream. Crushing (suchas some roller mills) may yield more uniform particle sizewith fewer ‘fines,’ but also will produce particles with a verydifferent shape than either ‘shatter type’ or ‘abrasive type’mills. Even changing machine size (same manufacturer) orspeeds (same machine) can have a profound impact on theproduct. There is simply no substitute for thorough testingand a deep understanding of the process and material prop-erties involved. The testing must be made using the actualstarting material, on the same model of mill, at the samemachine speed, as the proposed equipment. The testing mustalso compare the range of particle sizes and a statisticalmeasurement of shape to the existing process, be it from aclinical trial or a previous production run. Average size orpercent passing a given screen size simply does not begin toprovide all the relevant information.

Predicting Material Handling BehaviorWith suitable testing, PSA can be used to predict bulk powdermaterial behavior. Understanding the ratio of surface area tomass will allow predicting of how easily a dust may ag-glomerate, or conversely how easily it can become entrainedin air. To a degree, the speed a material goes into solution canbe better predicted once variability of particle surface area iscontrolled. PSA can explain confusing results for other sur-face dependent properties (such as electrostatics) when pow-der flow is an issue. The usual suspects for problems inparticle mass flow are electrostatics and surface adhesion,due to solvent wetting (such as in a hydroscopic powder).Awareness of particle shape’s influence on powder flow canadd another avenue to investigate when ‘bridging’ in a bin isan issue.4 If particles have surfaces that can interlock (like achild’s toy “jacks”), particle flow can be difficult and bridging



©Cop

yrig

ht IS

PE 2

007

common. See also “Effect of particle shape of Active Pharma-ceutical Ingredients prepared by fluidized-bed jet-milling onCohesiveness.”5

Another property that PSA can easily identify is if theparticles are not uniform in all three axes. Most size screen-ing equipment is rated as withholding particles over themean diameter of the screen openings. Note that if theparticles have a high degree of rod-like or needle-like shape,they can be several times the screen opening size in length,but may still pass through the screen if the long axis happensto align perpendicular to the screen. Such particles willalmost always be imaged such that the shape is evident. Thiskind of shape can have profound implications for electrostat-ics and other mass flow conditions. Another implication is ifthe particles are glassy (mineral) and in the 10-micron range,they may behave very much like asbestos – easily suspendedin air, yet easily precipitated out in the lungs, and thendifficult or impossible to excrete or be biologically modified.

Potential Detection ofCounterfeit Pharmaceuticals

Theoretically, application of PSA to counterfeit detection isstraightforward. Each model of mill should produce a unique‘fingerprint’ of particle shape distributions from a particularmaterial. Therefore, a suspect powder sample that does notmatch the manufacturer’s ‘equipment standard’ or showsforeign (adulterating) material would be immediately provedas counterfeit, even if the sample is chemically identical tothe manufacturer’s standard. Some process changes affect-ing crystal growth history upstream of the mill (or potentiallytableting conditions after milling) may be detectable. (Pillsfrom different tablet presses disintegrated by the same equip-ment may yield different particle shapes, due to the changesin pressure, moisture content, and surface finish in thedifferent presses). As a side benefit, the counterfeit samplemay now be traceable. Should the counterfeit manufacturinglocation be found, PSA would become another tool to provethat this was the source of the counterfeit material (usefulsince many counterfeit operations may not have good enoughprocess control to prove a counterfeiting case based on ‘chemi-cal fingerprinting’ alone). This would be analogous to thestudies by A.F. Vetter at John Deere: with statistical datafrom a fleet of engines any sample that deviated from thestatistical norm indicated something was wrong. Usually itindicated a part in the engine was wearing abnormally, butit also inadvertently detected a mislabeled sample (wrongengine class) and on one occasion detected an aftermarketmodification that had not been reported (a different/unap-proved oil ring increased horsepower short term, but causeda change in cylinder wall erosion).

In each of the applications above, the first step in applyingPSA is to take samples and develop a statistical norm tocompare future samples against. Once the existing conditionis defined, troubleshooting problems, examining the poten-tial impact of proposed changes, or detecting deviationsbecomes possible.

In summation, PSA has numerous applications for processcontrol, process safety, quality assurance, material handlingtroubleshooting, and potentially counterfeit detection in phar-maceutical powder operations.

References1. Beddow, J., “Particle Characterization in Technology,”

CRC Press, Boca Raton, Florida, 1984.2. Jacobson, Ruby, “Inert Milling Systems,” Pharmaceutical

Engineering, Vol. 25, No. 3, May/June 2005.3. Ogle, R.A., Beddow, J., and Vetter, A.F., “Numerical Mod-

eling of Dust Explosions: The Influence of Particle Shapeon Explosion Intensity” Process Technical Program,International Powder Bulk Handling Process, Rosemont,Illinois 1983.

4. Stanford, M.K., DellaCorte, C., and Elyon, D., “ParticleMorphology Effects on Flow Characteristics of PS304Plasma Spray Coating Feedstock Powder Blend,” NASA,2002.

5. “Effect of Particle Shape of Active Pharmaceutical Ingre-dients Prepared by Fluidized-Bed Jet-milling on Cohe-siveness,” Journal Pharmaceutical Scientists, Vol. 94, No.5.

About the AuthorDavid A. Vetter graduated with a BS fromthe University of Iowa in 1990. He worked asa safety and environmental engineer at anelectrical utility company before becoming aProcess, Project, and Validation engineer asa consultant. He has emphasized work in thefood and pharmaceutical industries for thelast eight years. He is a licensed Professional

Engineer, and a member of the American Institute of Chemi-cal Engineering, and ISPE. Vetter is currently employed byJacobs Engineering Group in the small project engineeringoffice in Indianapolis. He can be reached by e-mail:[email protected]

Jacobs Engineering Group, Landmark Building, Suite500, 1099 N.Meridian, Indianapolis, Indiana 46204.

Research Article

54 Journal of Pharmaceutical Innovation | September/October 2006

Journal of Pharmaceutical Innovation©2006 ISPE. All rights reserved. | www.ispe.org/jpi

IntroductionOxyContin, the brand name for the narcotic pain relieverOxycodone-HCl (Purdue Pharma, http://www.pharma.com/),is categorized as a Schedule II drug under the ControlledSubstances Act owing to its propensity for abuse and de-pendency [1]. It is an opium-based pain reliever prescribedfor relief of moderate to severe pain; however, it exhibitsheroin-like effects lasting up to 12 h when abused. The il-licit diversion of pharmaceuticals such as OxyContin is apervasive problem across the USA [2]. Measures have beenimplemented to protect pharmaceuticals and to prevent theirillegal distribution. For example, electronically monitorednarcotic cabinets and use of IDenticards (smart-card tech-nology that can integrate electronic, magnetic stripe orbarcode technologies, and biometric readers to increase se-curity) have reduced diversion of drugs from hospitals andpharmacies. But it is now time for a second line of defense,in the form of well-secured, better-regulated pill dispensing

NIR spectrometry for thecharacterization of fuel componentsin a novel tamper-resistant pill bottleJoseph P. Medendorp1, Jason A. Fackler1, Tom Henninger2, Bill Dieter3, andRobert A. Lodder1

1Department of Pharmaceutical Sciences, College of Pharmacy, A123 ASTeCC Bldg. 0286, Lexington, KY 40536, USA2Electrical and Computer Engineering, 683 Anderson Hall 0046, Lexington, KY, 40536, USA3Center for Manufacturing, 413 Center for Robotics and Manufacturing Systems 0108, Lexington, KY 40536, USACorresponding author: Lodder, R.A. ([email protected]).

The purpose of this paper is twofold: (i) to present the Pill Safe, a noveldesign for a tamper-resistant prescription container, and (ii) to presentuse of near-infrared (NIR) spectrometry for characterization of fuel com-ponents and prediction of the burn characteristics of the fuel mixturesused to destroy tablets in the Pill Safe. In the safe, drug tablets are stackednext to fuel and attempts to force the mechanism or penetrate the bottlecause their instant destruction. The Pill Safe offers a second line of de-fense against the illicit distribution of dangerous prescription drugs.

systems, to help prevent drug diversion from dispensed pre-scriptions. Companies have attempted to respond to thisneed, and e-pill (http://www.epill.com/) has developed aMonitored Automatic Pill Dispenser (MD.2; http://www.epill.com/md2.html) that features voice alarms andreminders. To our knowledge, the MD.2 is the only auto-mated vault-like delivery system on the market. At a highretail price, the cost of dispensing new MD.2 bottles witheach monthly refill would be prohibitive. Therefore, the MD.2comes with a lock and key, and it is the responsibility of thepatient to refill their bottles – potential thieves need only toobtain the key to pilfer the MD.2 contents. Thus, an oppor-tunity lies in building an inexpensive and impenetrable con-tainer as a fail-safe, capable of scheduling and dispensingmedications such as OxyContin, and deterring those inter-ested in obtaining the drugs purely for abuse and illicit pur-poses.

Research Article

September/October 2006 | Journal of Pharmaceutical Innovation 55


Pill SafeIn response to the need for a better-protected pill bottle, themedicine dispenser presented in this research (conceived byRAMM, LLC, Bourbon County, KY USA) helps prevent thediversion of dangerous prescription drugs. The medicine dis-penser simply presents the patient with a button. When

pressed, it dispenses the medication through a small aper-ture if and only if the prescribed dosing period has passedsince the previous pill was dispensed. Meanwhile, the medi-cine dispenser monitors its outer shell for tampering, rap-idly destroying all of the pills contained within upon tamperdetection. Destroying tablets rapidly requires a very fast,reliable chemical reaction that proceeds in the presence ofinterferences. Because the tablets are stored for an extendedperiod in close proximity to the components of this reaction,the destructive reaction must not proceed even at a verylow rate in the absence of a tampering trigger. For thesereasons, a stable metallic aluminum-based fuel was chosenas the means of tablet destruction. Figure 1 gives a blockdiagram of the system; the mechanism shown in Figure 2houses and delivers the pills and destroys them at the di-rection of the microcontroller. The pills are stored in col-umns adjacent to fuel. In this configuration, hot exhaustgases from the burning fuel are directed toward the pills.The entire assembly is essentially in a miniature ventedThermos® bottle, enveloped in a protective shell with a loopprinted on the interior using conductive ink. Breaching theshell breaks the loop, signaling the microcontroller to ignitethe fuel. The entire system is powered by two alkaline AAbatteries and constructed from inexpensive parts, costingless than US $10.

The medicine dispenser was designed to satisfy threemain regulatory concerns: clinical, pharmaceutical, andpackaging. From the clinical aspect, it was necessary thatthe burn residue from the incinerated medicine dispenserwas completely destroyed and inedible. From the pharma-ceutical aspect, drug tablets stacked next to fuel rods neededto be stable over time, not subject to chemical degradationin the presence of fuel. And lastly, the packaging for themedicine dispenser needed to be safe and inaccessible tothose for whom the medicine is not intended. It also wasdesigned so the hot gases created upon ignition are brief induration and contained in an insulating vessel so there isno danger of igniting external fires.

Near-infrared (NIR) spectrometry was used in designof the fuel mixtures, for quantification of the fuel compo-nents, and for prediction of burn characteristics of the dif-ferent fuel mixtures. Monitoring of fuel components usingNIR is important because burn characteristics depend onthe identity and quantity of fuel components (e.g., when thefuel mixture is deficient in ammonium perchlorate, the burnduration increases significantly, giving would-be thieves cru-cial extra seconds to break into the medicine dispenser). Thefuel mixtures that were most effective in the Pill Safe werethose that ignited the fastest and burned most completely.In these experiments, the concentrations were known be-fore mixing the samples, but an in-process assay capable ofassessing fuel mixtures would enable rapid and accuratedetection of manufacturing and formulation inconsistencies.

Figure 1. Schematic illustration of the main components of the medi-cine dispenser. A microcontroller serves as the master coordinator.Based on the inputs and programmed dispensing instructions, themicrocontroller dispenses the tablet as needed. If the sensor is per-turbed, the microcontroller signals for tablet neutralization, triggeringthe ignition of the fuel components.

Dispensebutton

Tampersensor

Dispensingmechanism

Ignitor

Microcontroller

Figure 2. The medicine dispenser and its main components. Pow-ered entirely by two AA batteries, the microcontroller activates the DCmotor to dispense a tablet once per dosing period. A conductive loopwraps around the assembly. Disruption of this loop causes themicrocontroller to ignite the fuel columns and incinerate the tablets.Fully assembled, the Pill Safe is 11.5 cm tall, 10 cm wide and 8.5 cmlong.

Tablet columns

AAbatteries

DC motor

Fuel columns

Continued on page 56.

Research Article



NIR has previously been used with great success for thequantification of fuel components in both solid and liquidpropellant mixes [3,4]. NIR and Fourier transform infrared(FTIR) spectroscopy have been used for accurate and pre-cise quality-control analysis of fuel pre-mixes, and havemonitored antioxidant depletion over time [5]. In-processreaction information, such as intensity distribution, igni-tion processes, reaction temperatures, and the identity andconcentrations of reaction species have been studied usingthe spectral range from UV and visible to NIR and mid-infrared [6]. Safety considerations for sample analysis us-ing NIR for the study of fuel components, such as the ther-mal response to NIR exposure and the effects of Raman spec-troscopy on the mixtures [7], also were addressed.

TheoryAbsorbance in the NIR region of the electromagnetic spec-trum is primarily a result of overtones and combinations ofthe fundamental bands from the mid-infrared and far-in-frared regions. The bands are a result of anharmonic stretch-ing and bending of functional groups such as N–H, O–H, C–H, and C=O. In most cases, the molecular structures aresufficiently complex that the spectral features of interestare highly overlapping, and thus, not directly usable with-out multivariate data analysis (Introduction to NIR Tech-nology, Analytical Spectral Devices Inc., http://www.asdi.com/asd-600510_nir-introduction_revc.pdf). Theformula for the medicine dispenser fuel source used to de-stroy pills included aluminum dust with iron oxide catalyst(fuel), NH4ClO4 (oxidizer), bisphenol A/epichlorohydrin (cast-ing agent), and a polyamide resin (curing agent). The oxi-dizer, casting agent, and curing agent are the primary spec-troscopically active constituents in this mixture; therefore,their concentrations were used for building and testing theregression model. Rather than using absolute mass, the pre-diction model used the constituent percentages (by mass) ofthe total mixture.

Because NIR spectra are usually a linear combinationof pure component spectra, according to Beer’s Law, it istheoretically possible to determine the concentration of eachindividual component in overlapping spectra. However, noisefactors tend to complicate this procedure, such as sampleinhomogeneity, particle size differences, and temperaturedrift. For a linear calibration model, outlying samples mustbe identified as outliers and removed from the calibration,or the noise and variation must be incorporated into themodel. In this work, Hadi outlier detection was used to iden-tify the spectra that belonged in the calibration model [8].Outlier detection is generally approached by forming a cleansubset of data M (free from outliers), followed by testing thefit of the remaining points relative to the clean subset. Con-sider the regression in Equation 1:

Y = Xβββββ + εεεεε (Eqn 1)

where Y is an n × 1 vector of responses, X is an n × k matrixof responses and observations, βββββ is a vector of estimatedregression coefficients from fitting the model to M, and εεεεε isa matrix of errors. The best clean subset M is found by thedeletion of the variables that result in the largest reductionin the residual sum of squares (SSEM).

When building a calibration model from NIR spectra, itcan be helpful to visualize the total instrument response,rk, as the sum of two orthogonal components: the interfer-

= ⊥ences, rk, and the net analyte signal (NAS), rk, according toEquation 2 [9]. The superscript ‘=’ denotes the fact that theinterferences span the space occupied by the analyte, andthe ‘⊥’ denotes the fact that the NAS is orthogonal to theinterfering species.

= ⊥rk = rk + rk (Eqn 2)

=A linear combination of the interferences produces rk ; there-=fore, the signal orthogonal to rk belongs exclusively to the

analyte of interest [10]. A projection matrix is calculatedaccording to Equation 3:

⊥ +Pk = (I – R–kR–k) (Eqn 3)

where R–k is a matrix of samples without the analyte, I isthe identity matrix, and the superscript ‘+’ is the Moore–

⊥Penrose pseudoinverse. The NAS vector, rcal, can be calcu-lated from a calibration spectrum, rcal,by projection of thespectrum onto the null space of the rows of R with Equa-tion 4:

⊥ ⊥rcal = Pk rcal (Eqn 4)

The NAS vector is normalized to length one by Equation 5:

⊥

NAS rcalrk = _____ (Eqn 5)

⊥rcal

A linear regression is fit to this vector, and the regressioncoefficients are used for subsequent predictions.

One advantage of using the NAS approach is in thecalculation of so-called figures of merit. In severely over-lapping spectra, it has historically been difficult to quan-tify selectivity, sensitivity, and signal-to-noise (S/N) ratiobecause of the inability to distinguish between interfer-ences and the analyte of interest [10,11]. With the NAS,these quantities can be measured directly. Selectivity canbe calculated as the scalar degree of overlap, a, betweenthe NAS vector and the calibration spectrum according toEquation 6:

Research Article



⊥rcala = _____ (Eqn 6)

rcal

The selectivity is a measure from 0 to 1 indicating howunique the analyte of interest is compared with the inter-ferences. The sensitivity is a measure of how much theanalyte varies in response to a change in concentration. Thisquantity can be expressed as Equation 7:

NASsk = rk / ck (Eqn 7)

where ck is the concentration of the k-th analyte. Theoreti-cally, sensitivity should be the same for each concentrationand each NAS vector [12]. The S/N ratio can be expressed asEquation 8:

NASck rkS/N = ________ (Eqn 8)εεεεε

where εεεεε is the random instrumental error.To prove that the relationship between NIR spectra and

the fuel components is not a product of correlated initialconstituent concentrations, the mixture concentrations werecalculated by the following orthogonalization procedure. Arandom matrix A corresponding to I mixtures and J compo-nents per mixture was constructed with a random-numbergenerator. Singular value decomposition of A according toEquation 9 yields orthogonal principal component scores,U, between –1 and +1; S is a diagonal matrix of singularvalues and V is the matrix of eigenvectors (loadings) [13].

A = USV (Eqn 9)

A coefficient matrix, K, was constructed according to Equa-tion 10 such that each row contains 0 and 1.

jmin[Ui] + [Ui]j=1K = __________________ (Eqn 10)

max[Ui]

A final matrix of concentrations was constructed by multi-plication of K with the fuel ratios from the accepted fuelformula [16% Atomized aluminum powder (fuel), 69.8%ammonium perchlorate (oxidizer), 0.2% iron oxide powder(catalyst), 12% binder, and 2% epoxy curing agent; http://www.nasa.gov/returntoflight/system/system_SRB.html].In a search for the most descriptive multivariate model tocorrelate NIR spectra to constituent concentrations and burncharacteristics, the present research compared NAS regres-sion with principal component regression (PCR) [14], inter-val PCR (iPCR), and PCR-uninformative variable elimina-tion (PCR-UVE) [15,16]. For each of these models, principalcomponents were calculated by a singular value decomposi-tion of the raw spectra according to Equation 9.

The regression of U in Equation 11 indicates which ofthe components have the strongest correlation to a changein constituent concentration c, where a is the y-intercept, bis a vector of regression coefficients, and εεεεε is the residual.

c = a + bU + εεεεε (Eqn 11)

Equation 12 demonstrates how a leave-one-out cross vali-dation was used to predict the concentration of fuel compo-nents, where σ 2 is the variance, fi(Ui) is the prediction of themodel for the i-th pattern m in the training set, after it hasbeen trained on the m–1 other patterns.

1 i=m

σ 2 = ____ Σ (ci – fi(Ui))2 (Eqn 12)LOO m i=1

Mixture Al NH4ClO4 Fe2O3 Epoxy resin Curing agent Ignition Burn (%) (%) (%) (%) (%) time(s) duration(s)

1 13.34 58.03 0.37 21.46 6.81 2.0 14.02 10.02 60.52 0.40 21.09 7.97 1.5 11.03 16.69 60.47 0.51 16.96 5.36 2.0 15.04 9.40 69.57 0.33 15.46 5.24 0.5 9.05 19.23 51.70 0.55 21.53 7.00 4.0 18.06 17.07 44.82 0.45 28.06 9.59 3.5 26.07 21.33 36.64 0.39 31.31 10.33 4.0 40.08 13.13 56.60 0.24 22.89 7.14 2.0 16.09 16.65 69.92 0.27 9.83 3.33 1.0 6.010 12.38 63.17 0.29 18.22 5.93 1.5 12.011 11.02 59.31 0.41 20.06 9.2 1.5 4.0

Table 1. Constituent percentages, ignition time, and burn duration for each fuel mixture.


Research Article



In the case of a two-component system, it is a simple matterto observe a linear change in the analyte concentration. Inthe presence of two system components, theoretically, theconcentration change can be modeled by one principal com-ponent. The loading corresponding to this principal compo-nent accurately reflects the contribution of each wavelengthto the overall classification. However, when multiple sys-tem components change simultaneously, multiple principalcomponents are needed for the prediction model.

Interval PCR performs in similar way to the aforemen-tioned analysis, but rather than using the full spectrum, ituses smaller subsets of variables [16]. For example, whenthe experimenter specifies an interval (In) of 100 wave-lengths, the algorithm performs PCR followed by principalcomponent selection and cross-validation on intervals of 100.With a moving boxcar, all wavelengths are paired with allother wavelengths inside of ±In. For example, after the algo-rithm analyzes 2101–2200 nm, the next iteration analyzes2102–2201 nm, and so on. At the final wavelength, the firstIn wavelengths are added to the end for the final iterations.In this manner, each wavelength is included in a new model2 × In times. The goal of interval selection is the minimiza-tion of the standard error of performance in Equation 13,

which indicates the interval with the highest correlation tothe change in drug concentration:

⎛ εεεεε2 ⎞1/2

⎜ _____ ⎟⎝ N – 1 ⎠

SEP = ______________________ (Eqn 13)max(c) – min(c)

where εεεεε is the residual, N is the number of spectra, and c isa concentration vector.

Experimental methodsFuel preparationEleven different mixtures of ammonium perchlorate(NH4ClO4), aluminum dust (Al), iron oxide (Fe2O3), castingresin (Bisphenol A/Epichlorohydrin), and polyamide curingagent (Versamid 140; Firefox Enterprises, http://www.firefox-fx.com/) were constructed such that the percent-age of each component did not correlate with those of theother components. Of the 11 fuel mixtures used, NH4ClO4

was 36.64 to 69.92% of the total mixture weight, Al was9.40 to 21.33%, Fe2O3 was 0.24 to 0.51%, casting resin was9.83 to 31.31%, and curing agent was 3.33 to 10.33%. Con-

Figure 3. Pure-component NIR spectra. The spectrum for epoxy resinis in blue, that for curing agent is in green, and that for NH4ClO4 is inred.

1200 1400 1600 1800 2000 2200 2400Wavelength (nm)

Abs

orba

nce

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1200 1400 1600 1800 2000 2200 2400Wavelength (nm)

Sec

ond

deriv

ativ

e (x

10–3

)

–6

–4

–2

0

2

4

Figure 4. Second-derivative NIR spectra. The spectrum for epoxy resinis in blue, that for curing agent is in green, and that for NH4ClO4 is inred.

NIR NH4ClO4 Epoxy resin Curing agent Ignition time Burn durationperformance iPCR NAS iPCR NAS iPCR NAS iPCR NAS iPCR NAS

r2 0.992 0.983 0.969 0.997 0.987 0.996 0.959 0.993 0.917 0.989RMSECV (%) 2.58 4.54 4.33 2.31 2.94 4.27 5.93 2.20 6.98 2.60Position (nm)b 1566–1966 1324–1724 1633–2033 1750–2150 2088–2488

aNIR performance as values of r2 and root-mean-square error of cross-validation (RMSECV) were predicted using intervalprincipal component regression (iPCR) and net analyte signal (NAS).bThe wavelength regions were selected and used with iPCR, whereas NAS used the full spectrum.

Table 2. Predicted results for NH4ClO4, epoxy resin, curing agent, ignition times, and burn durations.a

Research Article



stituent ratios are reported in Table 1. Epoxy resin and cur-ing agent were heated to 35°C on a hot plate. Powders weremixed together and stirred to ensure a homogenous mix-ture. Epoxy resin and curing agents were then added to thepowder mixtures and stirred vigorously for three minutes.The resulting fuel was poured into a ceramic mold so it couldeasily be inserted into the medicine dispenser prototype.

NIR data collectionNIR spectra were collected in reflectance mode from 1100 to2500 nm in 2 nm steps using a customized scanning spec-trometer, as described previously [17]. To eliminate roomnoise, samples were scanned inside the instrument drawer.Each of the mixtures was scanned six times, all in randomorder to eliminate the effects of drift. All NIR data wereexported to Matlab 7.0.1 (The Mathworks Company, http://www.mathworks.com/) for processing and analysis.

Ignition testsTo identify the burn characteristics, fuel samples were ig-nited quickly using a propane torch. With the torch andsamples positioned reproducibly, ignitions were recorded ondigital video and ignition start times and durations werenoted (Table 1).

Analysis of NIR spectraAlgorithms for NAS, Hadi outlier detection, PCR, and iPCRwere all written by the authors for Matlab 7.0.1. NAS, PCR,and iPCR were all used to predict the concentrations of eachfuel component from the collected NIR spectra. These meth-ods enabled the identification of the statistically significantregions of the NIR spectrum for the measurement of eachcomponent. Using the same methods, NIR spectra were thenused to predict the ignition times and burn durations. Thetwo most effective methods for calibration and prediction ofburn characteristics were iPCR and NAS.

Extent of incinerationThe extent of OxyContin incineration was determined byhigh-performance liquid chromatography (HPLC). The pillbottle was loaded with fuel mixture number 1 (Table 1). Burnresidue from the pill bottle stacked with 20 mg OxyContintablets (Purdue Pharma, http://www.pharma.com/) wasground with a mortar and pestle, washed with 200 ml ofmobile phase, filtered twice through a 0.2 μm filter, and achromatogram was collected to measure the remaining con-centration of OxyContin. The HPLC assay was conductedusing a Waters 717plus Autosampler, Waters 1525 Pump,and Waters 2487 Dual Wavelength Absorbance Detectorwith Waters Breeze v3.30 Software (http://www.waters.com/).

Figure 5. A plot of principal component (PC) scores indicates whether a mixture is deficient in a particular fuel component. PC score 1 correlatesvery highly with the presence of curing agent, PC2 with the presence of epoxy resin, and PC3 with the presence of NH4ClO4. The mixture withideal burn (fastest ignition and longest burn) was mixture number 1; therefore, projection of PC scores onto this plot can be used as a map for theprediction of fuel burn characteristics.


Research Article



A Waters μBondapak C18 column (300 × 3.9 mm) was usedwith the UV/VIS Detector set at a wavelength of 206 nm forOxyContin standards. The mobile phase used for OxyContinwas 0.005 M 1-hexanesulfonate, methanol, phosphoric acid,and triethylamine (900:100:5:2). The flow rate of the mobilephase was at 1.5 ml min–1 with 10 ml injections of the sample.Standards analyzed exhibited excellent linearity over theconcentration range employed. The retention time forOxyContin was 22.22 ± 0.211 min. The sensitivity of theassay was 25 ng ml–1.

Results and discussionHPLC analysis indicated that following tablet incinerationusing the most effective fuel mixture, the burn residue con-tained 5.58% by mass of the initial OxyContin. It must benoted that this 5.58% was recovered only by grinding, wash-ing, extracting, filtering, purifying, and chromatographicallyseparating the burn residue. With such an extensive extrac-tion procedure and such a small yield, it is likely that would-be thieves would be sufficiently discouraged from a recov-ery attempt. The best fuel mixture required ~14 s to con-sume the contents of the safe completely, making it nearlyimpossible to break into the bottle and remove the contentsbefore the burn was complete. The medicine dispenser isdesigned so that the flame is contained, and there is no dan-ger of igniting external fires. The fuel formula contains anoxidizer; therefore, there is no need for atmospheric oxygento sustain the reaction that destroys the tablets.

The most effective chemometric methods for the mea-surement of fuel components and prediction of ignition char-acteristics from the NIR spectra were NAS regression andiPCR. The fuel components had no significant correlation toeach other; therefore, the predictive ability of the NIR wasnot a product of correlated concentrations. Additionally, theNAS was able to detect the portion of the signal that wasunique to the analyte of interest; thus, there were two mea-sures of certainty that the strong correlation was not anartifact. Using NAS, the NIR measurements over the re-spective concentration ranges resulted in r2 = 0.983 and root-mean-square error of cross-validation (RMSECV) = 4.54%for NH4ClO4, r2 = 0.997 and RMSECV = 2.31% for the epoxyresin, and r2 = 0.996 and RMSECV = 4.27% for the curingagent. NAS predicted ignition times with r2 = 0.993 andRMSECV = 2.20% and burn durations with r2 = 0.989 andRMSECV = 2.60%. Because this experiment was performedusing cross validation, no external validation set was used.Consequently, all spectra satisfied the Hadi criterion forclass membership; therefore, no spectra were discarded asoutliers in the NAS calibration. The NAS calibration andfigures of merit were calculated from the unsmoothed, un-processed full spectrum data. Interval PCR data performedcomparably to NAS data for the measurement of fuel con-centrations, but not as well for the prediction of burn char-

acteristics. See Table 2 for comparison of iPCR and NASperformance statistics.

Pure-component and second-derivative NIR spectrafrom NH4ClO4, epoxy resin, and curing agent are shown inFigures 3 and 4. It is apparent from these figures that thespectra were sufficiently distinct from each other, makingit relatively easy to identify the NAS for each component.Iron oxide and aluminum had no distinguishing spectral fea-tures in the NIR; therefore, they were not quantified in thisexperiment. Of course, other fuel formulas could be used. Amixture with five components, each with its own unique NIRchromophore, might make the most sense from a quality-control standpoint. However, iron oxide can be determinedin mixtures of aluminum dust using visible light spectrom-etry. Many NIR spectrometers are capable of collecting vis-ible and NIR spectra simultaneously.

The figures of merit calculated from the NAS vector re-sulted in: selectivity = 0.014, sensitivity = 5.874, and S/N =39.941 for NH4ClO4; selectivity = 0.005, sensitivity = 6.467,and S/N = 50.76 for the casting agent; and selectivity = 0.022,sensitivity = 4.097, and S/N = 35.32 for the curing agent.Selectivity is a unitless quantity, and sensitivity is given inunits of signal per concentration. The limit of detection (LOD)is defined as the concentration at which S/N = 3 [10]. The S/N ratios were calculated separately for each individual con-centration, and the theoretical LOD was extrapolated fromthe linear plot of concentration versus S/N. The LOD was0.93% for NH4ClO4, 1.37% for casting agent, and 1.36% forcuring agent. Precision was measured by calculating therelative standard deviation (RSD) of the predicted concen-trations from the NAS vector. The RSD was 0.38% forNH4ClO4, 0.32% for casting agent and 0.70% for curing agent.Bias was calculated by subtraction of the measured NIRvalues from the reference gravimetric values. Although someof the individual measurements were offset slightly high orslightly low, there was no net bias for the calibration.

Figure 5 is a plot of principal component ellipses calcu-lated from the fuel mixtures, demonstrating how well NIRwas able to separate them. Ellipses are drawn six standarddeviations from their cluster centers. NH4ClO4 demonstrateda high correlation to both ignition time and burn duration,suggesting that it was largely responsible for the reactionrates. High correlations were demonstrated by principalcomponent (PC1) to the concentration of curing agent, PC2to the concentration of casting agent, and PC3 to concentra-tion of NH4ClO4. Mixture 1 demonstrated the best burn char-acteristics for the purpose of this research (i.e., fastest igni-tion and longest burn). The cluster location from mixture 1can quickly be compared with other fuel mixtures, enablingthis figure to act as a map to identify components in which aparticular mixture might be deficient. This figure also canbe used to predict the ignition time and burn duration ofeach mixture depending on its constituent concentrations.

Research Article



NIR spectroscopy is a versatile, nondestructive, andrapid method of analysis. It has been applied to both liquidand solid propellants with no sample preparation required[3–6]. As is evident from the PC clusters in Figure 5, it isalso highly reproducible. Even when samples are very closeto each other in concentration, they still cluster in distinctlydifferent regions of a PC plot. This effect can be quantifiedby the bootstrap error-adjusted single-sample technique(BEST) [16]. According to the BEST, multidimensional stan-dard deviations (MSD) greater than three indicate that clus-ters belong to different populations. When MSDs are lessthan three, clusters are considered inseparable. The aver-age MSD separation between fuel mixtures was 431.36, andthe average MSD separation calculated between repeat scansof the same mixture was 1.69, demonstrating the high de-gree of reproducibility from scan to scan. NIR analysis isalso a rapid technique. NIR spectrometers can now collectthousands of spectra per second with high resolution [18],and NIR can effectively function as an in-process assay forthe quantification of fuel mixtures as they are cast into themedicine dispenser model. From the 200 NIR spectra col-lected for this experiment, there were no aberrant scans,demonstrating that NIR is also a robust assay.

ConclusionThis research presents a novel design for a medicine dis-penser, intended to provide a second line of defense againstthe diversion of prescription drugs. NIR spectroscopy wasable to quantify rapidly and accurately the fuel componentsin solid fuel mixtures, and to identify burn characteristicsof the different mixtures. This research suggests that PillSafe manufacturing and the fuel analysis need not be toochallenging or expensive, making it a valuable product foruse in the pharmaceutical industry.

AcknowledgementsThis research was supported in part by the US FDA throughcontract 200406251503 and through KSEF-148–502–03–61.

JPI

References1. National Drug Intelligence Center (2002) Information Brief: Prescrip-

tion Drug Abuse and Youth. (http://www.usdoj.gov/ndic/pubs1/1765).2. The New York Times (2004) Pill thefts alter the look of rural drugstores.

The New York Times (http://query.nytimes.com/gst/health/article-page.html?res=9C03E1D6163BF935A35754C0A9629C8B63).

3. Wang, J. et al. (2004) Application of near infrared spectroscopy in liquidpropellant analysis. Huaxue Tongbao 67, w023/1–w023/4.

4. Judge, M.D. (2004) The application of near-infrared spectroscopy for thequality control analysis of rocket propellant fuel pre-mixes. Talanta 62,675-679.

5. Rohe, T. et al. (1996) Near infrared-transmission spectroscopy on pro-pellants and explosives. Int. Annu. Conf. ICT 27, 85.1–85.10 (http://www.ict.fraunhofer.de/english/events/anconf/a27general.html).

6. Weiser, V. and Eisenreich, N. (2005) Fast emission spectroscopy for abetter understanding of pyrotechnic combustion behavior. PropellantsExplos. Pyrotech. 30, 67-78.

7. Harvey, S.D. et al. (2003) Safety considerations for sample analysis us-ing a near-infrared (785 nm) Raman laser source. App. Spec. 57, 580-587.

8. Hadi, A.S. and Simonoff, J.S. (1993) Procedures for the identification ofmultiple outliers in linear models. J. Am. Stat. Assoc. 88, 1264-1272.

9. Boelens, H.F. et al. (2004) Performance optimization of spectroscopicprocess analyzers. Anal. Chem. 76, 2656-2663.

10. Lorber, A. (1986) Error Propagation and figures of merit for quantifica-tion by solving matrix equations. Anal. Chem. 58, 1167-1172.

11. Booksh, K.S. and Kowalski, B.R. (1994) Theory of analytical chemistry.Anal. Chem. 66, 782a-791a.

12. Lorber, A. et al. (1997) Net analyte signal calculation in multivariatecalibration. Anal. Chem. 69, 1620-1626.

13. Jolliffe, I.T. (2002) Principal Component Analysis. Springer.14. Leardi, R. and Norgaard, L. (2004) Sequential application of backward

interval partial least squares and genetic algorithms for the selection ofrelevant spectra regions. J. Chemometr. 18, 486-497.

15. Noord, O.E. (1996) Elimination of uninformative variables for multi-variate calibration. Anal. Chem. 68, 3851-3858.

16. Lodder, R.A. and Hieftje, G. (1988) Detection of subpopulations in near-infrared reflectance analysis. App Spec. 42, 1500-1512.

17. Yokel, R.A., et al. (2005) 26Al-containing acidic and basic sodium alumi-num phosphate preparation and use in studies of oral aluminumbioavailability from foods utilizing 26Al as an aluminum tracer. Nucl.Instrum. Methods Phys. Res. B 229, 471-478.

18. Klahn, T. et al. (2004) A high speed NIR-spectroscope to investigate ig-nition and combustion of propellants and pyrotechnics. Int. Annu. Conf.ICT 35, 161/1–161/9 (http://www.ict.fraunhofer.de/english/presse/anre2004_05.pdf).

JANUARY/FEBRUARY 2007 PHARMACEUTICAL ENGINEERING 1

Aseptic Filling

©Copyright ISPE 2007

This articlereviews theconsiderationsinvolved in theselection ofappropriatetechnologies forasepticprocessing. Itcompares andcontrastsconventionalcleanroomoperations,RestrictedAccess BarrierSystems(RABS), andisolationtechnology withrespect to theadvantages/disadvantagesof each for usein aseptic filling.It evaluates allaspects of theselectionprocess,beginning withproject timelines, facilityconcerns,operation, andcompliance.

Choosing Technologies for AsepticFilling: “Back to the Future, Forwardto the Past?”

by James Agalloco, James Akers, and Russell Madsen

Introduction

Designing a facility for the aseptic pro-duction of sterile products in their fi-nal containers involves considerationof many different factors. Unlike some

non-sterile product or active pharmaceuticalingredient facilities that can sometimes beadapted for the production of different prod-ucts, parenteral facilities can be difficult tomodify to fit changing circumstances. Thechoices made regarding facility arrangement,HVAC system, and equipment selection duringinitial facility design can only be altered atconsiderable expense with significant down-time once the facility has been placed in opera-tion. To ensure success, it is essential that thedesign meet specified operational requirementsand constantly evolving regulatory expecta-tions. It is important to consider that a firmmay be living with initial design decisions for along time since the useful lifetime of a facilitymay exceed 15 years. With the introduction ofnew aseptic processing technologies over thelast decade, those designing and constructingaseptic processing facilities are now faced withseveral technology alternatives.

This article will focus on which environmen-tal technology affords the best facility designchoice considering production utility, flexibil-ity, and longevity in the rapidly evolving asep-tic processing landscape. Among the contami-nation control design alternatives that can beconsidered are:

• Conventional cleanroom facilities with vari-ous types of barriers (rigid or flexible)1

• Restricted Access Barrier Systems (RABS)2

• Isolators (closed or open)3

In recently reviewed projects, firms have con-sidered elements of all three of the alternativeslisted above. For example, sterile compoundingmay be done in isolators, while filling may bedone in a conventional cleanroom using somevariety of barrier technology. Modern sterileproducts facilities are typically outfitted withequipment that includes barriers of one type oranother. Additionally, modern aseptic process-ing equipment can include a substantial amountof automation that may accomplish the sameobjective as barrier systems: the reduction ofpotential risk from human-borne contamina-tion.

There are other production technologiesavailable (i.e., blow-fill-seal, closed-vial fill-ing), but these do not provide the means to fillthe full range of product formulations or con-tainer presentations.4,5 Their selection, whileoffering substantial advantages, may restricttheir application to special circumstances. Inthe case of blow-fill-seal, the equipment con-tains its own air handling system and the needfor human intervention is restricted by design.Environmental considerations for closed vialfilling are still under discussion. The core facil-ity technologies listed above are commonly se-lected since they offer the greatest flexibility inuse over what is expected to be a lengthy opera-tional life.

A general description of the more prevalentenvironments for aseptic processing is useful tounderstand the unique elements of each. Eachof these incorporates “state-of-the-art” elementsas would be found in a new facility designed forhigh-throughput aseptic filling. The focus ofthe descriptions is on the filling environment,as the formulation areas for each would likelybe essentially identical.

Reprinted from




2 PHARMACEUTICAL ENGINEERING JANUARY/FEBRUARY 2007

Aseptic Filling


Conventional FacilityThe core of the aseptic area is curtained or partitioned-off todefine the aseptic processing area in which ISO 5 (traditionalFED-STD 209E Class 100 unidirectional air flow) conditionsare specified.6,7 Surrounding this is an environment of slightlylower class, often ISO 6 or 7 (Class 1,000 to 10,000) in whichthe personnel work the majority of the time that they are inthe aseptic processing area. EU expectations are for the ISO5 unidirectional airflow aseptic filling environment to beaccessed from an ISO 5 background environment (whenclassified under static conditions) in which unidirectionalflow is not required.8 Several firms have addressed this bydesigning a background environment that meets ISO 5 underdynamic conditions by providing unidirectional air supplythroughout the filling room. Other facility designs havesucceeded with ISO 6 or 7 both of which can be compatiblewith EU Grade B requirements depending upon how classi-fication is accomplished. Sterilizers are unloaded in thisarea, and personnel transfer sterilized items to the ISO 5environment. Glass containers are typically fed directly tothe aseptic processing environment using continuous-feeddry-heat tunnels. Sanitization of the facility using a defineddisinfection program is performed at routinely scheduledintervals to meet current regulatory and cleanroom practiceexpectations. Personnel working in the aseptic area wearsterilized garments over their entire bodies to minimize therelease of contaminants. Operators access the ISO 5 environ-ment using gloved hands and frequently some part of thegown also will enter this critical zone. The human activitiesconducted in ISO 5 environments are those required toperform the set-up, environmental monitoring, and routineand non-routine interventions.

RABS FacilityThe core of the aseptic area includes an isolator-like structurethat defines the ISO 5 aseptic processing area in whichunidirectional airflow conditions are generally specified.Surrounding this is an ISO 6 or 7 environment in which thepersonnel are located during work. Full adherence to EUexpectations would mandate a “Grade B” (which is ISO 7under dynamic conditions or ISO 5 at rest) backgroundenvironment as described for cleanrooms (see ConventionalFacility above). Sterilizers are unloaded in this area andpersonnel transfer sterile items into the ISO 5 environmentusing Rapid Transfer Port (RTP) systems or transfer hatches.9

Containers are fed directly to the ISO 5 environment usingcontinuous-feed tunnels. Sanitization of the facility and RABSinternal surfaces using a defined disinfection program isperformed on a routine basis to maintain control over micro-bial contamination. An environmental monitoring program,confirming the microbial and non-viable particle conditionsmeet the desired state, is required for the RABS and thesurrounding room. Personnel working in the aseptic areawear sterilized garments over their entire bodies to minimizethe release of contaminants. The operators use gloves on theRABS in conjunction with the RTP to perform set-up, envi-ronmental monitoring, and any required interventions.

Because of the need for flexibility or due to product consid-erations it may be necessary to allow gowned operator accesswithout the use of RABS features. Thus, RABS could operatein at least two different states of control, one in which thereare no direct interventions and another in which direct, open-door interventions are permitted in the same manner as in aconventional cleanroom. In the latter instance, the userwould place restrictions on the frequency with which a doorcould be opened and define the circumstances of the interven-tions. Operated in that manner, the RABS becomes simply abarrier as might be used in a conventional cleanroom torestrict access.

Isolator FacilityHigh throughput isolator design has an open isolator (pro-tected by air overpressure) in a surrounding room of ISO 8(Class 100,000). The isolator provides an ISO 5 environmentwithin which the filling process is performed. The pressuredifferential between the isolator and the surrounding room iscontinuously monitored and alarmed. Container feed is froma continuous tunnel. Isolators allow components to be fed tothe supply hoppers either directly from other isolators inwhich bags of sterile supplies are surface decontaminated, orthrough a Rapid Transfer Port (RTP), and discharge to thesurrounding room through an opening (often called a mousehole) with sufficient air flow to effect an air overspill seal.Decontamination of the isolator enclosure and enclosed sur-faces is accomplished using an automated system (using H2O2

or other sporicidal agents) that can be validated to provide afour-to-six-log reduction of a highly resistant biological indi-cator as stipulated in FDA’s 2004 aseptic processing guid-ance.10 Control of the environment is focused on the isolatorenclosure and contained equipment, as the surrounding ISO8 room requires a substantially lower degree of control.Personnel may wear uniforms with hair/beard covering andgloves when using the isolator gloves. The operators use theisolator gloves to perform set-up, environmental monitoring,and any interventions. Quite often isolator-based filling equip-ment is sterilized and cleaned-in-place and interventions forroutine tasks such as replenishment of supplies may beaccomplished using automation.

Evaluation CriteriaThe impact of the decision process is of such magnitude thatconsiderable deliberation should be given to each of therelevant factors. The evaluation process should address eachof the following: facility lead time; initial facility cost; quali-fication duration; qualification obstacles; operating cost; op-erational hurdles; line preparation; environmental sanitiza-tion; line operation; environmental monitoring; process simu-lation; impact of personnel; cleaning; change over; suitabilityfor campaign usage; system maintenance; complexity; nov-elty; regulatory expectations; industry perspectives and in-tangibles.


Aseptic Filling


Facility Lead Time - The time period required to bring thefacility into full operation exclusive of the qualification/validation period.

Initial Facility Cost - The installed cost for the facility,including infrastructure, facility systems, and process equip-ment.

Qualification Duration - The time required to complete thequalification of the facility, including environmental controlsystems or isolator decontamination.

Qualification Obstacles - Unique problems associatedwith the technology choices that hinder the completion of thequalification/validation effort.

Operating Cost - Cost to operate the facility, includingutilities, personnel, sanitization materials and labor, envi-ronmental monitoring (including cost of testing), gowningmaterials, and consumables.

Operational Hurdles - Difficulties associated with the useof technology for aseptic filling.

Line Preparation - Those activities required to ready theline for filling.

Environmental Sanitization - The methods used for disin-fection/sanitization of the processing environment or thedecontamination of the isolator.

Line Operation - Impact on the routine use of the line.

Environmental Monitoring - The requirements for peri-odic sampling of the processing environment.

Process Simulation - Media fill execution as requiredinitially and periodically.

Impact of Personnel - The extent to which personnelpractices can influence aseptic operations.

Cleaning - Methods for cleaning product contact and non-product contact parts.

Change Over - Conversion of the line from one size/formatto another.

Suitability for Campaign Usage - Considerations relativeto the repetitive use of the line for the same product format.

System Maintenance - Impact of system maintenanceduring the aseptic filling operation.

Complexity - The overall complexity of the facility and itsequipment.

Containment Utility - The potential for utilization of thefacility for the aseptic processing of potent compounds.

Regulatory Views - The perception of the technology fromthe perspective of the inspector/reviewer.

Industry Perspectives - The end users’ views of the tech-nology.

Intangibles - Factors not specifically addressed in the priorelements.

Overall AssessmentRABS designs have features in common with conventional,manned cleanroom technology and with isolation technology.Table A highlights some of the common features of thesesystems. Passive RABS are substantially closer in bothconcept and capability to the more evolved cleanroom designsthat have utilized rigid and semi-rigid barriers between theoperator and the critical aseptic fill zone. Active RABS arecloser in design to open isolators, the primary differencesbeing the mode of sanitization-decontamination and theaerodynamic separation mechanisms, i.e., air overspill forRABS and defined positive pressure differentials for isola-tors.

As first conceived, RABS were intended to offer the perfor-mance advantages of isolators at less cost and with fewerqualification/validation hurdles. The recent increased inter-est in RABS as a concept appears to have been an outgrowthof less than fully successful isolator projects, stimulated bythe hope that the technology would mitigate the more diffi-cult aspects of isolation technology design and validation,which in the past had contributed to rather long projecttimelines. However, in recent years, the isolator design,validation, and regulatory compliance issues have been mostlyresolved. While there are benefits to some RABS conceptsrelative to cleanrooms, RABS can realize only some of thebenefits of isolation and dramatically reduce process designand validation efforts relative to isolators. Over the lastseveral years, isolator technology has been implementedsuccessfully many times and the regulatory authorities willcontinue to favor it as the aseptic processing approach thatcomes closest to fully eliminating risk from human contami-nation. The number of isolators in use each year continues togrow, and they have performed very well.11

There are several multi-national firms that have had suchsuccess with isolators that they presently operate more than20 units worldwide. However, success has not been limited tolarge and well-financed projects; each of us has worked withsmaller firms whose first (and only) aseptic fill operation isisolator-based. Projects that utilize a clear statement of userrequirements (and those where those requirements are real-istic rather than arbitrary) have consistently delivered suc-cessful outcomes.

The key issue is whether or not RABS and isolator perfor-mance can be considered equivalent. If that were the case,then the less expensive RABS design should prevail withisolators reserved only for those applications where a RABSis inadequate for the task because of operational consider-ations. Of course, the previous statement assumes thatRABS are truly less expensive than isolators, but when fullfacility and operational costs are factored in, this is far fromcertain.12


Aseptic Filling


Table A. Aseptic Filling Technology Comparison.

Project Concern Conventional RABS Isolator

Facility Lead Time • Building infrastructure time consuming due to • Building infrastructure time consuming, because clean • Lead time for isolator equipment is typically shortersignificant classified clean space requirements space requirements are effectively the same as than that for an ISO 5 controlled environment facility.

• Permitting/site activities more complex conventional • Equipment more complex, but facility simpler• More project elements/vendors • Facility activities more complex

• More project elements and vendors involved

Initial Facility Cost • Costs well defined historically • Costs higher than conventional • Isolator equipment is more expensiveCosts for architecture and engineering as Costs for architecture and engineering as • Is higher cost justified or is it added vendorwell as HVAC systems can be substantial well as HVAC systems can be substantial margin?

• Large footprint of higher classified environments • More equipment related costs • Costs less certain at project start• Large footprint of higher classified environments

Qualification Duration • 6 to 9 months typical • 6 to 9 months typical • 6 to 9 months typical• No new issues • Could theoretically avoid some of the more controversial • Longer periods are a reflection of excess requirements

aspects of isolators, rather than any insurmountable technical hurdle• May result in new requirements that have their own

controversies

Qualification Obstacles • Issues well established and easy to resolve • New issues require new solutions • New issues require new solutions

Operating Cost • Baseline • Higher than conventional • 25-30% of cleanrooms, mostly related to HVAC• Few opportunities for cost reduction that are exclusive operating costs

• Other savings in gowns, supplies, labor utilization, EM

Operational Hurdles • None • Less complex than isolators • Requires many new elements• Simpler adaptation from earlier operating modes • Changes to old paradigms necessary

Line Preparation • Hands on by operators • Hands on by operators • Hands on by operators• Much less secure due to operator proximity • Slow due to access constraints • Slow due to access constraints

• Less secure due to operator proximity • Safer because of isolator

Environmental • Disinfection/sanitization performed by gowned personnel • Disinfection/sanitization performed by gowned personnel • Reproducible decontamination with a sporicidal agentSanitization • Reproducibility and validation uncertain • Reproducibility and validation uncertain • Requirements for decontamination well defined

Access for anti-microbial treatment requirements and generally harmonizednot defined and still subject to discussion • Can be validated

Line Operation • Superior operator access • Restricted access slows interventions • Restricted access slows interventions• More risk of contamination • Reduced risk of contamination relative to cleanroom, • Risk of contamination lower than for either

less capable than isolators conventional or RABS• Open door interventions are risky

Environmental • Methods well defined • Methods well defined, but some adaptation may be • Methods well defined but more adaptationMonitoring • Performance generally excellent but with occasional required may be required

excursions • Performance could be only slightly below that of • Performance excellentisolators, but currently there is only asmall experience base

Process Simulation • Performance excellent • Performance excellent, but currently the • Performance excellentexperience base is very limited

Impact of Personnel • Familiar environment • Environment separation substantially less effective • Essentially removed from critical area• Worker protection minimal than isolators • Closed isolator enhances worker safety with potent

• Worker protection limited compounds

Cleaning • Largely COP for product contact, CIP possible • Largely COP for product contact, CIP possible • Largely COP for product contact, complete CIP possible• Manual cleaning of non-product contact surfaces • Manual cleaning of non-product contact surfaces • Manual cleaning of non-product contact surfaces• Difficult issue when handling potent compounds • Difficult issue when handling potent compounds • Operator access slows cleaning when closed

Change Over • Size, component and product change all relatively easy • Size, component change easy • Size, component change relatively easy• Greater risk of microbial contamination • Product change requires internal cleaning • Product change requires internal cleaning

• Greater risk of microbial contamination

Suitability for • Considered easier to validate • Campaign potential unknown • Conceptually easyCampaign Usage • Can potential product carryover be ignored? • Considered difficult to validate by some

• Product carryover a concern

System Maintenance • Dark side maintenance not widely available • Dark side maintenance possible • Dark side maintenance during run easy

Complexity • Systems are least complex • Systems are less complex than isolators • More controls, equipment and instrumentation required.• Decontamination adds extra elements

Containment Utility • Limited given greatest potential for worker exposure • Improved due to greater worker separation, but not • Near perfect for closed isolators due to presence ofabsolute complete physical separation of the environments

Novelty • None • New technology with its own limitations • Some firms have almost no isolators• Some firms already have substantial experience

Regulatory Views • Proven technology • Well accepted proven technology • Considered superior, but cautions still raised• Recognized as less capable and becoming obsolete • Recognized as less capable than isolators • More rapid implementation outside US.

Industry Perspectives • Proven technology with known limitations • Largely proven technology with known limitations • Embraced by some, avoided by others• Few uncertainties • Unproven technology, can lead to uncontrollable project

costs

Intangibles • Easy to implement • Easy to implement • More capable once fully operational• Not the latest nor the safest approach • Not the latest nor the safest approach • Now fairly well established technology.• Mature design may limit useful lifetime.Will it be GMP in • Further advances may not be possible • Further advances in design are underway

10 years time?

Note 1: The entries presented in the table reflect the opinions of the authors, drawing upon our 40 years of experience as consultants, and nearly 100 years working with sterile products technology.Note 2: Shaded boxes indicate where essentially the same situation prevails with more than one technology.


Aseptic Filling


The primary technical differences between RABS andisolator designs lie in the following areas:

• Isolator decontamination is performed by an automatedsystem and can be validated to provide multiple-log sporereduction. RABS are manually disinfected by gownedpersonnel.

• RABS do not provide a complete physical separation be-tween the internal and external environments. Isolatorsmaintain a continuous pressure differential with respectto the environment or are closed.

• Passive RABS must be positioned in an ISO 5 environment,while isolators have been successfully operated in an ISO8 environment.13

• RABS are poorly suited for containment applications, anincreasingly important concern as drugs become increas-ingly potent. Closed isolators provide the highest degree ofseparation currently available, providing worker protec-tion while shielding the aseptic zone from worker-bornecontamination.

It seems to us that these differences are meaningful enoughto make the choice relatively easy. The ability to reliablydecontaminate isolators using an automated system is amajor advantage. This eliminates the vagaries and uncer-tainties of manual disinfection that are time-consuming, lessthan fully effective, and hazardous to personnel. The effortsrequired to initially validate isolator decontamination maybe greater (we have not experienced any substantial diffi-culty in our recent efforts, but we have heard this regardingprojects with which we have not been involved), neverthelessthe increased reliability and robustness of isolation decon-tamination is difficult to ignore. The majority of isolatorshave been installed in ISO 8 (Class 100,000) environmentswhere the operating requirements are relatively easy. Pas-sive RABS designs, even those without “open door” interven-tions, should be in ISO 5 environments since there is nopressure differential between the interior and exterior of theenclosure. An ISO 5 background for the RABS mandatesperiodic sanitization, aseptic gowning, and environmentalmonitoring in the background environment at a level com-mensurate with manned cleanrooms. ISO 7 environmentsmay be suitable for active RABS designs operated without“open door” interventions. For containment applications, thesuperiority of closed isolators to all other technologies isunquestioned.

RABS is an interesting offshoot in the continued evolutionof aseptic processing technology. Unfortunately, it can some-times be a “Back to the Future, Forward to the Past” ap-proach. It seems inappropriate to remove what are some ofthe more beneficial design features in isolators and revert towhat is essentially a sophisticated barrier system of the typethat has been used for many years. Isolators represent thefuture of aseptic processing in our industry; we owe ourselves

and the patients who take the drugs we produce the bestavailable technology we can provide.

References1. ISPE Baseline® Pharmaceutical Engineering Guide, Volume 3

- Sterile Manufacturing Facilities, International Society forPharmaceutical Engineering (ISPE), First Edition, January1999, www.ispe.org.

2. ISPE RABS Definition.3. Agalloco, J. P., J. E. Akers, et al. (2001), “Technical Report No.

34, Design and Validation of Isolator Systems for the Manufac-turing and Testing of Health Care Products,” PDA J. Pharma-ceutical Science and Technology, 55 (5, Supplement).

4. www.medinstill.com.5. Verjans, B., J. Thilly, et al. (2005). “A New Concept in Aseptic

Filling: Closed-Vial Technology,” Pharmaceutical TechnologyAseptic Processing Supplement: s24-s29.

6. ISO 14644-1.7. Anonymous (1992). FED-STD-209E. Airborne Particulate

Cleanliness Classes in Cleanrooms and Clean Zones, U. S.General Services Administration.

8. Annex 1.9. ISPE, Restricted Access Barrier Systems (RABS) for Aseptic

Processing Definition, 16 August 2005.10. FDA (2004), Guidance for Industry, Sterile Products Produced

by Aseptic Processing - Current Good Manufacturing Practice.11. Lysford, J. and Porter, M. “Barrier Isolation History and

Trends,” ISPE Washington Conference: Barrier Isolation Tech-nology, 2 June 2004.

12. Agalloco, J., Akers, J., and Madsen, R. Private communica-tions with clients.

13. Some proponents of RABS have suggested that the units canbe opened to the surrounding environment in the midst of theaseptic process to perform an intervention, followed by disin-fection and a resumption of the process. This seems to be a riskypractice that should be avoided. We have not considered thispractice, and restricted our discussion to RABS that areoperated without “open door” interventions of any kind. If“open door” interventions are contemplated in the operation ofthe RABS, then conformance to the EU Annex 1 backgroundrequirements of ISO 5 should be maintained at all times(including when the RABS is closed).

About the AuthorsJim Agalloco is President of Agalloco & As-sociates, a consulting firm to the pharmaceu-tical and biotechnology industry. He was pre-viously employed at Bristol-Myers Squibb,Pfizer, and Merck. Agalloco holds a BS and MSin chemical engineering and an MBA in phar-maceutical studies. He is a past president anddirector of PDA. He is a member of USP’s

Microbiology and Sterility Assurance Expert Committee for2005-2010. He is a member of the editorial advisory board’s ofPharmaceutical Technology and Pharmaceutical Manufactur-ing. He is a frequent author and lecturer on sterilization,aseptic processing, and process validation. He can be contacted


Aseptic Filling


by phone at: +1-908-874-7558 or by e-mail: [email protected] & Associates, PO Box 899, Belle Mead, New Jersey

08502-0899.

James Akers, PhD, is President of AkersKennedy & Associates, which offer validationand regulatory services to the pharmaceuticaland biotechnology industry. While in indus-try, he has held positions of increasing respon-sibility in Quality Control/Quality Assurance,eventually serving as director - Quality Assur-ance/Quality Control/Technical Services at

Praxis Laboratories. Akers received a BA in biology with aminor in chemistry from the University of Kansas and receivedhis PhD in medical microbiology from University of KansasSchool of Medicine, Kansas City, Kansas. Akers is a pastpresident of PDA and served as an officer or director from 1986to 1998. He has served as a chairman and member of numerousPDA committees and remains extremely active in PDA’s tech-nical efforts. He also has been a member of the USP’s ExpertAdvisory Panel on Microbiology since 1990. He has authored orco-authored more than 75 technical papers and lectured exten-sively on process validation at industry meetings, around theworld. He can be reached by phone at: +1-816-822-7444 or by e-mail at: [email protected].

Akers Kennedy & Associates, Inc., P.O. Box 22562, KansasCity, Missouri 64113-0562.

Russell E. Madsen is President of TheWilliamsburg Group, LLC, Gaithersburg,Maryland, engaged in pharmaceutical con-sulting in the areas of cGMP compliance andauditing, quality systems, design review, asep-tic processing and sterilization technology,sterile filtration, due diligence evaluation,process validation, regulatory liaison, and

general technical services. Most recently, he has served PDA asacting president and was senior VP of Science and Technologyresponsible for the overall scientific, technical, and regulatoryaffairs activities of the association. Before joining PDA, he wasemployed by Bristol-Myers Squibb Company as director ofTechnical Services, providing technical and general consultingservices to Bristol-Myers Squibb pharmaceutical, medical de-vice, nutritional, and cosmetic operations, worldwide. He cur-rently serves as a member of the USP Parenteral Products-Industrial Expert Committee and is a member of Pharmaceu-tical Technology’s Editorial Advisory Board. He also is amember of PDA, and is Senior Technical Documents Advisor toISPE. Madsen holds a BS from St. Lawrence University and aMS in chemistry from Rensselaer Polytechnic Institute. He canbe contacted by phone at: +1-301-869-5016 or by e-mail at:madsen@ thewilliamsburggroup.com.

The Williamsburg Group, LLC., 18907 Lindenhouse Rd.,Gaithersburg, Maryland 20879.


Advanced Aseptic Processing


This articlecompares theuse ofRestrictedAccess BarrierSystems (RABS)and IsolatorSystems forproduct final filloperations.

Advanced Aseptic Processing: RABSand Isolator Operations

by Eric A. Isberg

Advanced Aseptic Processing (AAP) is aterm referenced in the recently pub-lished ISPE RABS definition1 to coverthe spectrum of Restricted Access Bar-

rier Systems (RABS) and isolator systems. Ingeneral, AAP techniques are physical barriermethods of product protection and contain-ment that are used during manufacturing op-erations to separate (primarily) operators fromthe process. These methods are most often usedduring open processes or other critical processsteps to ensure product is not exposed to viableorganisms and particulate contamination.While there are many methods to choose from,there is no argument that AAP techniques arewidely used.

Complete and absolute ingress control isessential to AAP operation, and is essential toensure an improvement over traditional opencleanroom processing. Operations within equip-ment like Laminar Flow Hoods (LFHs) orBiosafety Cabinets (BSCs) do not fit within thedefinition of AAP, as this equipment only pro-vides partial separation (i.e., the hands andarms of the operators are not physically sepa-rated from the process when using this equip-ment). Curtained cleanroom areas also are notin this definition, as curtains provide little realbarrier from ingress. Only complete, rigid wallenclosures fall within the scope of AAP.

The focus of this article is to compare the useof RABS and isolators for product final filloperations. It will give a very general overviewof both systems. It will then describe key me-chanical and operational areas in which thetwo systems differ, while at the same timehighlighting several key operational areaswhere they do not. Finally, it will propose thateither system, if operated correctly accordingto approved procedures, will work successfullyto ensure an improvement over traditionalopen cleanroom processing.

RABS and Isolator OverviewThe ISPE RABS definition1 describes the com-mon characteristics of a RABS system. Thesystem has an ISO Class 5 environment2 withunidirectional airflow enclosed in a rigid wallenclosure with glove port access where neces-sary. The interior of the enclosure is manuallysanitized with sterilized equipment and partsintroduced using aseptic procedures which caninclude transfer systems. Because the systemis open to the surrounding room, it is commonlylocated in an ISO Class 7 or better environ-ment.2 All product or process contact partswithin a RABS are sterilized or Steamed-In-Place (SIP) prior to use. Although doors can beopened, this only happens rarely, after whichappropriate line clearance and cleaning mustoccur per procedures.

Closed RABS, like isolators, fully encloseand seal the work area, and supply the interiorwith HEPA filtered air that is returned thoughsealed ductwork. The main difference betweenclosed RABS and isolators is that closed RABShave no automated bio-decontamination cycleusing H2O2 vapor or another sanitant. Theinterior of the closed RABS unit is bio-decon-taminated manually using cleaning solutions.One purpose for closed RABS units is for highlypotent compounds, where personnel protectionis the purpose for product containment. In thiscase, they are designed as containment RABS,which require special leak tightness require-ments, air filtration systems, and decontami-nation processes for safe operation.

Isolators are fully enclosed and sealed unitswith HEPA filtered air supplied in a unidirec-tional manner to the ISO Class 5 interior.2

Completely closed isolators may be suppliedwith either turbulent air or unidirectional air.Air is typically recirculated by returning it tothe air handlers though sealed ductwork. Thechamber is bio-decontaminated via an auto-

Reprinted from







Figure 2. Basic Isolator airflow diagram.

mated cycle using H2O2 or another sanitant. All access isthrough glove ports and sterile transfer systems. All itemsentering the system after bio-decontamination are pre-ster-ilized. Because they are sealed, isolators are commonly lo-cated in ISO Class 8 environments.2

Air Handling DifferencesRABS systems operate in a similar fashion as LFHs in thatthey are fed clean air from fan units through HEPA filters andthe air vents from the unit into the surrounding room - Figure1. The air is unidirectional via diffuser panels and multiplefan/HEPA filter locations. Therefore, the air handling re-quirements are relatively uncomplicated. Pressure balanc-ing involving supply and return ductwork and return fans arenot required. The exception to this is closed RABS, which canhave a pressure differential to the outside room and thereforebehave like an isolator in regard to basic air handling require-ments.

Basic isolator air handling requirements are more compli-cated than RABS - Figure 2. Air is re-circulated so return fansand ductwork are required. In order to maintain positivepressure, a large part of the air handling unit, including thereturn ductwork, must be leak tight. Leak testing of thesystem, which involves recording pressure decay over time, isrequired to prove the unit is sealed prior to sanitization andoperation. Ductwork that leads outside of the cleanroom, andoutside of the building, is often required to safely vent H2O2

vapor after sanitization where this chemical is used.The bio-decontamination cycles for isolator units are me-

chanically complicated. For example, before injection of the

H2O2 vapor (when used), the chamber and air handlingductwork must be conditioned. The purpose of the condition-ing is to ensure a sufficient concentration of H2O2 vapor isinjected into the system and that the H2O2 stays in vapor formduring the cycle. Conditioning consists of heating the cham-ber and ductwork and lowering the humidity of the air in thesystem for low humidity bio-decontamination systems. Highhumidity bio-decontamination systems do not require hu-midity control during conditioning. After the bio-decontami-nation cycle is complete, out gassing of the H2O2 vapor isrequired to bring the concentration down to levels that areboth safe for personnel, yet not high enough to affect theproduct being filled. Heating, Ventilation, and Air Condition-ing (HVAC) systems are often required to perform thesefunctions.

For smaller systems, the conditioning, bio-decontamina-tion, and aeration events can be performed using self-con-tained, external bio-decontamination systems. However, forlarger isolator systems, multiple external bio-decontamina-tion systems are often necessary, along with the HVACsystem. These systems must be tied together to ‘act as one’during bio-decontamination. One alternative is to have anintegrated HVAC and bio-decontamination system. The ad-vantage of this is that air handling, bio-decontamination,aeration, and overall pressure balance can be controlled bythe same unit.

Cleaning and Bio-DecontaminationDifferences

The interior of isolators are bio-decontaminated using anautomatic sequence which most often includes injection ofH2O2 vapor as the sanitant. These cycles are very consistentand lead to a validatable bio-decontamination method. How-ever, manual cleaning of the interior is still required on a

Figure 1. Basic RABS airflow diagram.




regular basis. Direct contact cleaning is required to removesurface contaminants and to reduce the likelihood of biofilmformation. Also, manual cleaning is often required wheneverthe chamber is opened for parts changeover and other inva-sive events. Procedures should dictate when isolators aremanually cleaned.

As described in the ISPE RABS definition,1 the bio-decon-tamination of RABS units is not automatic. Manual sprayand wipedown methods must be employed. The difficulty liesin performing consistent, repeatable, and complete bio-de-contamination using manual methods. Validation of theeffectiveness of the cleaning and bio-decontamination solu-tions is an important step in justifying manual cleaningprocesses. In many cases, companies are trying to move awayfrom manual cleaning and bio-decontamination because ofconsistency issues and the difficulty of validating manualmethods. Alternatively, some companies offer periodiccleanroom bio-decontamination services using automatedequipment that can be used for RABS systems.

Environmental MonitoringEnvironmental monitoring for both viable and non-viableparticulates is key to determining classification levels incleanroom space. Ongoing monitoring is required to showthat the system maintains an ISO Class 5 environment overtime.2

Isolator particulate and microbiological monitoring canonly be achieved via built-in sampling ports or by transfer-ring pre-sterilized sampling devices and sampling plates intothe isolator in a sterile manner. Planning for environmentalmonitoring, including sampling methods, location of sam-pling devices, and frequency of sampling, must be part of theisolator system design considerations.

Environmental monitoring via built-in sampling ports orby transferring pre-sterilized sampling devices and samplingplates also can be used for RABS. However, RABS units oftenhave openings near floor level for air to flow out of the interiorof the chamber. Therefore, there is the option of using por-table sampling devices that have sampling probes that areinserted into these openings.

Personnel GowningOperators must gown according to the classification of thearea surrounding the AAP system. In the case of isolators,the gowning is often for an ISO 8 area.2 ISO 8 gowning isoften comprised of items like a plant uniform or jumpsuit,lab coat, head cover, and shoe covers. Single gloves are oftenused as a precaution during work using the isolator gloveports.

For RABS, the gowning must be for the ISO 7 or betterenvironment2 where the equipment is located. This can in-clude the addition of full, sterile one-piece suits, sterile facemasks, sterile head and shoe covers, goggles, and multiplelayers of gloves. The personnel must still use glove portswhen performing work within the RABS. Obviously, person-nel comfort is a factor in RABS operation.

Glove Replacement and TestingOne area in which RABS and isolators do not differ is in theway that glove-port gloves and gauntlets are controlled. Theymust be sterilized prior to use, either by bio-decontaminationor sterilization processes. The gloves also require inspectionprior to use, and periodic replacement is necessary to ensureeffectiveness. A system of glove integrity testing and/ormechanical inspection may be required to provide evidencethat leaks do not exist, as leaks can compromise the ISO Class5 environments within the barrier system.

Built-in, automated glove testing systems are availablewith some RABS and isolator systems. These automatedsystems have the advantage of speed, as glove ports can betested simultaneously, and of accuracy compared to manualmethods.

Procedural RequirementsA “paradigm shift” is required by an organization to go fromtraditional open cleanroom processing to RABS and/or isola-tor use. All unit operations, including manufacturing, main-tenance, quality, validation, and other functions, must pre-pare for this shift. Because processes will change and newprocesses will be required, Standard Operating Procedures(SOPs) must be created to govern them.

Adherence to new and revised SOPs is essential to ensureRABS and isolators are operated successfully. SOP require-ments are more critical to RABS operation because there aremore manual operations, such as the new bio-decontamina-tion processes that will be required. However, if SOPs arewritten well and are followed by all personnel, either AAPmethod can be performed successfully.

ConclusionsRestricted Access Barrier Systems (RABS) and isolator sys-tems are two methods of Advanced Aseptic Processing (AAP)that ensure an improved processing environment forcleanroom operations. Knowing some of the key mechanicaland operational areas in which the two systems differ willincrease awareness of the systems in the industry, and willhelp create a more detained definition of each.

These two AAP methods may, on first glance, appear to bevery similar. Both methods provide ISO Class 5 cleanroomspace and fully separate the operators from the process.However, of the two systems, only isolators are widely ac-cepted within the industry for use in product fill operations.This is because RABS has been used in the past to label a widevariety of containment systems. Hopefully, the new ISPERABS definition, along with detailed AAP system compari-sons like this one, will help by increasing knowledge of howeach system can be used successfully to ensure an improve-ment over traditional open cleanroom processing.

References1. Restricted Access Barrier Systems (RABS) for Aseptic

Processing, ISPE Definition, 16 August 2005.




2. International Standard ISO 14644-1:1999(E) Cleanroomsand Associated Environments – Part 1: Classification ofair cleanliness.

About the AuthorEric Isberg is Product Manager in liquidpharmaceuticals for Bosch Packaging Tech-nology, specializing in advanced aseptic pro-cessing for product filling operations and infilled product inspection equipment. He hasmore than 14 years of experience working inthe areas of biopharmaceutical and pharma-ceutical processing. Previously, he worked in

the biopharmaceutical industry at PDL Biopharma and BiogenIdec in the areas of equipment and process validation, manu-facturing, process development, and technical services. Heholds a BA in biology from Gustavus Adolphus College. He isa member of ISPE, PDA, and the American Glovebox Society.He can be contacted by phone at: +1-763-493-6143 or by e-mail at: eric.isberg@boschpackaging. com.

Bosch Packaging Technology, 8700 Wyoming Ave. N.,Brooklyn Park, Minnesota 55445.


Decontamination Methods


This articlereviews thephysicalchemistry of thesafe productionof hydrogenperoxide vaporand how tomaintain aconstantconcentration ofthe vapor in achamber. Thisarticle alsoreviews thecalculations andprocedures usedto obtain themaximumconcentration ofvapor withoutallowingcondensation ofliquid to occur.

The Physical Chemistry ofDecontamination with GaseousHydrogen Peroxide

by Carl Hultman, Aaron Hill, and Gerald McDonnell

Introduction

The decontamination of surfaces con-taminated with microorganisms withincritical, enclosed areas is an importantconsideration to pharmaceutical, re-

search, and other facilities. A safe and reliableway to decontaminate hard surfaces (includingthose in isolators, laminar flow cabinets, andcleanrooms) involves exposing the surfaces togas-phase hydrogen peroxide.1,2,3 Gas phasehydrogen peroxide is a known rapid, broad-spectrum antimicrobial which, as part of acontrolled process, can allow for reproduciblearea decontamination. Hydrogen peroxide alsohas a rather safe profile, both from a user andenvironmental perspective, in comparison totraditional fumigation methods that have used

formaldehyde, ethylene oxide, or propyleneoxide gas. An understanding of the physicalchemistry behind the generation and control ofgaseous hydrogen peroxide is important to op-timize the safety, efficacy, and reproducibilityof a given decontamination process. This ar-ticle discusses the physical chemistry behind atypical process and the procedures needed toachieve optimal decontamination using gas-eous hydrogen peroxide.

Liquid and GaseousHydrogen Peroxide

Hydrogen peroxide (H2O2) is widely used as anantiseptic, disinfectant, and sterilant.4 It is adesirable biocide because it demonstrates broadspectrum antimicrobial activity, has low toxic-

Figure 1. Hydrogenperoxide vaporizationprocesses.

Reprinted from








ity, and breaks down into water and oxygen in the environ-ment. For example, liquid hydrogen peroxide solution is useddirectly on the skin at up to 6% w/v and at higher concentra-tions as a general surface disinfectant. Pure hydrogen perox-ide exists as a liquid at room temperature (25°C) and atmo-spheric pressure (101.35kPa). For antimicrobial applica-tions, greater efficacy is observed as the concentration isincreased, which is particularly important to achieve spori-cidal activity; however, at high concentrations, liquid hydro-gen peroxide is unstable/reactive and may be explosive orundergo spontaneous combustion depending on how it ishandled.5,6 Therefore, liquid preparations are used at lowerconcentrations diluted in water (generally 3 to 59% by weight)and often in synergistic formulations with other biocides(including peracetic acid). The antimicrobial activity of hy-drogen peroxide is dramatically increased when in a gaseousphase. For example, the efficacy of hydrogen peroxide againstbacterial spores has been shown to be similar at a gaseousconcentration of 1mg/L in comparison to ~400mg/L in liquid.5

To generate the gas, it is necessary to heat liquid hydrogenperoxide, generally by flash vaporization on a hot surface thatwill be discussed in more detail. Under these gaseous condi-tions, even lower concentrations (~0.1mg/L) of peroxide arerapidly antimicrobial and may be used to achieve sporicidalactivity.

A further consideration in the comparison of liquid andgaseous peroxide is material compatibility; attention shouldbe focused on two factors when considering surface contact:

1. the effect of the material on the decomposition rate of theperoxide (therefore loss of activity)

2. the effect of hydrogen peroxide on the material itself.

With liquid hydrogen peroxide (of about 45% by weight ormore), the possibility of forming detonatable mixtures onreaction with organic substances may exist. Commonly usedmaterials have been categorized into four different classesaccording to the suitability of exposing these materials toliquid concentrations equal to or higher than 90% by weightof peroxide in water.7 Of note, Class 4 materials may causedecomposition of hydrogen peroxide, and particularly withconcentrated liquid peroxide, cause damage to the surface orform explosive mixtures. Some of these Class 4 materialsinclude wood products and polymers such as neoprene, bunarubber, silicone rubber, and tygon. These materials mayundergo degradation and/or cause spontaneous combustion.Other Class 4 metals, including copper, lead, magnesium

alloys, and stellite #6, act as catalysts that accelerate thedecomposition of peroxide and also may encourage combus-tion when in contact with other materials. Lubricants such assilicones, paraffin, and aroclors also are Class 4 materials.Highly concentrated solutions of peroxide also may degradecertain electrical components or other materials; an exampleis that liquid hydrogen peroxide can attack the plasticizers inPolyVinyl Chloride (PVC), thus making the surface brittleafter extended exposure. Additionally, certain grades of stain-less steel, such as 304 grade, can show slight corrosion ordiscoloration for liquid peroxide solutions ranging from 10 to100% by weight.

Hydrogen PeroxideEvaporation and Condensation

Hydrogen peroxide vapor can be conveniently produced fromhydrogen peroxide/water solutions by two distinct processes,evaporation and flash vaporization - Figure 1. It is importantto note that both processes will produce different concentra-tions of hydrogen peroxide vapor in a given volume. Whenliquid hydrogen peroxide is allowed to evaporate into a dry,enclosed space, the concentration of hydrogen peroxide in thevapor state is much lower than the concentration in liquidstate - Figure 2. This occurs because the water leaves themixture at a higher rate than the peroxide due to the highervapor pressure of water. For instance, if a 35% (by mass)hydrogen peroxide/65% water mixture evaporates into a dryenclosed space at 25°C, the resulting gas will consist of 2.15%(by mass) hydrogen peroxide and 97.85% water at saturation.Saturation refers to the point at which the air can not holdfurther peroxide/water and at which point condensation (orprecipitation) or water/peroxide will occur. The concentra-tion at which this occurs (referred to as the ‘dew’ point) can bepredicted based on the peroxide/water concentrations, andthe temperature of the gas - Figure 2. A detailed discussion ofthis phenomenon is discussed in the literature.5

The evaporation rate can be increased by applying anenergy source, for example heat can be applied to boil theperoxide/water solution. Since water has a lower boilingpoint than hydrogen peroxide, the water will vaporize at ahigher rate, giving a higher concentration of water in thegaseous state and an increased concentration of hydrogenperoxide in liquid state. With time, the higher peroxide liquidconcentration may become unsuitable for materials in con-tact with the liquid and can pose a safety risk.

The equilibrium concentration of peroxide in the vaporover a liquid is always lower than the concentration in theliquid. Table 1 shows the concentration of peroxide in thevapor over liquid solutions at different peroxide concentra-tions at 25°C. The equilibrium concentration of peroxide inthe vapor is always lower because the water escapes theliquid at a higher rate than the peroxide. The reverse is truewhen peroxide/water vapor condenses to the liquid state. Asshown in Table A, the equilibrium concentration of the liquidformed when a vapor of 35% by weight peroxide condenses isabout 77.8% by weight peroxide. This is because the peroxidein the vapor has a greater desire to enter the liquid state than

Hydrogen Peroxide Weight (%w/v) t = 25 C

Vapor Liquid1.87 32.18.0 55.7

24.1 73.935 77.8

56.4 88.3

Table A. Equilibrium concentrations of hydrogen peroxide vaporthat will form (evaporate) over liquid hydrogen peroxide (seeFigure 1, Evaporation).




the water in the vapor. Therefore, the peroxide condenses ata higher rate than the water, causing the higher concentra-tion of peroxide in the resulting liquid.

Flash vaporization is a distinct process which can beachieved by applying energy, e.g., by direct application of aperoxide/water mixture to a hot surface - Figure 1. Flashvaporization forces the water and hydrogen peroxide in theliquid solution to evaporate simultaneously, thereby produc-ing gaseous concentrations of water and peroxide at approxi-mately the same concentration as the starting liquid mix-ture. The gas concentration will stay constant as long ascondensation does not occur. Table B shows the concentra-tions of hydrogen peroxide that will saturate the vapor andcause condensation. If the gas concentration is increasedabove saturation or if cooled to below the dew point, conden-sation will occur. For example, when condensation occurswith a gas at ~35% by weight peroxide it will condense as aliquid at ~77.8% by weight peroxide - Figure 2.

Physical ChemistryThe vapor pressures of gases over multi-component liquidsolutions may be calculated using the following equations:5,8,9

Pp(gas) = Xp(liq)YpPop (1)

Pw(gas) = Xw(liq)YwPow (2)

Yw(liquid) = exp[{(1-Xw)2/RT}(B0 + B1 (1 – 4Xw) (3)+ B2(1 – 2Xw) (1 – 6Xw)]

Yp(liquid) = exp[{Xw2/RT}(B0 + B1(3 – 4Xw) (4)

+ B2(1 – 2Xw) (5 – 6Xw)]

where:- Pp(gas) is the vapor pressure of the peroxide in the vapor in

atmospheres- Pw(gas) is the vapor pressure of water in the vapor in atmo-

spheres- Xp and Xw are the mole fractions of peroxide and water

respectively in the liquid- Yp and Yw are the activity coefficients for peroxide and

water respectively in liquid solution

Pop and Po

w are the equilibrium vapor pressures in atmo-spheres of pure peroxide and water respectively at the tem-perature of interest. B0, B1, and B2 are empirically deter-mined constants for hydrogen peroxide with the values shownbelow.

B0 = -752 + 0.97t t in degrees centigradeB1 = 85B2 = 13

Vapor pressure data may be converted into gas phase concen-tration units of mg gas per liter using the ideal gas equationas shown below.

mg/liter = P(Mol Wt)(1000 mg/g)/RT (5)

The concentration in mg/liter calculated from equation five isthe concentration at a given temperature that will causecondensation, where, P is pressure in atmospheres, Mol Wtis the molecular weight of the gas of interest in grams/mole,R is the ideal gas constant 0.082 latm/deg mole, and T istemperature (°K).

Achieving Optimal Concentration ofPeroxide in Vapor Form

As more vapor at a composition of 35% by weight peroxide isintroduced into an enclosed chamber, the pressure and con-centrations of the peroxide/water vapor in the chamber willincrease. Eventually, a high enough concentration (expressedin mg/liter) will be reached at the operating temperature tocause the undesirable condensation of liquid in the chamber.The maximum allowable concentration of peroxide in thevapor that will not cause condensation may be calculatedusing Equations one through five. One of the complicationsassociated with condensation is that undesirable high con-centrations of peroxide liquid solutions will be formed onsurfaces. Also, as liquid condenses, it may not uniformlycover all solid surfaces which may lead to non-uniform disin-fection/sterilization of surfaces. Other complications associ-ated with condensation will be elaborated on in a latersection. The gaseous state for hydrogen peroxide when usedfor decontamination is advantageous because:

Figure 2. A representation of the Hydrogen Peroxide/Water PhaseDiagram. The upper solid line indicates the point at which the gasmixture is formed (note higher concentrations at highertemperatures) and the lower solid line when liquid is present. LineA shows the conversion of 35% liquid peroxide/water mixture (at25oC) to vapor at ~2% peroxide. Line B shows the condensationof 35% gaseous peroxide/water mixture to liquid at ~78%.





Figure 3. Dropwise condensation on contaminated metal surface.

1. Gas will have uniform contact with all exposed surfaces,thus assuring that all surfaces are uniformly decontami-nated.

2. Gas will have uniform contact with surfaces with complextopographies. Examples include horizontal or verticalsurfaces, cracks, and complex curvatures.

3. Gas may be safely maintained in the chamber.

4. Gas may be efficiently and quickly removed from thechamber at the end of a given decontamination time thusdecreasing cycle time.

As the antimicrobial efficacy increases with increased perox-ide concentration, it is desirable to get the peroxide concen-tration in the gas as high as possible in the chamber withouthaving condensation occur. It is possible to calculate themaximum pressure (and/or concentration) of peroxide (or mg/L of peroxide in the chamber) that will initiate condensationusing Equations one through five by knowing:

1. the concentration of the flash vaporized gas being used tofill the chamber

2. the total pressure in the chamber

3. the humidity in the carrier gas (e.g., air used to circulategas into and out of the chamber showing the design of thedecontamination system - Figure 4) and the decontamina-tion chamber

4. the chamber temperature5,8,9

As these equations can often be time consuming to do byhand, computer programs have been developed to calculatethe optimum decontamination conditions for any given en-closed volume.10 Table B shows examples of the maximumperoxide concentration allowed in the chamber to preventcondensation of liquid at various temperatures at two differ-ent humidity levels for the carrier gas. Calculations assumedflash vaporization of 35% by weight peroxide. Note thatincreasing the relative humidity in the carrier gas decreasesthe maximum concentration of peroxide that may be achievedin the decontamination chamber.

Why Condensation Should be AvoidedAs described above, high concentration of peroxide in theliquid state (as occurs with condensation) can be antimicro-bial, but also poses some further disadvantages. The first ismaterial compatibility. As shown in Figure 1 and Table A, theconcentration of a liquid formed when 35% by weight gascondenses will be about 78% by weight peroxide. This ishigher than the recommended maximum concentration of45% by weight peroxide to assure suitable interactions withother materials. As discussed above, this peroxide concentra-tion or higher may not only cause spontaneous combustionand also may accelerate decomposition of peroxide or causedegradation of materials. Incompatible materials can in-clude painted surfaces and electronics. Peroxide condensa-tion also can decrease the useful life of chamber materials inparticular, those used in flexible-walled isolators. Once thecondensate is formed it will eventually have to be evaporatedfrom the chamber as peroxide is removed at the end of adecontamination phase. The slow evaporation will cause theperoxide concentration in the liquid to reach even higherunsuitable concentrations regarding safety and degradationof materials.

Decontamination reproducibility is a further concern. Whencondensation occurs, the surfaces in the chamber may not getuniformly exposed, depending on how condensation occurs.Condensation can occur either in a drop wise or film form,depending on the nature of the contact surface. The determin-ing property is the surface tension of the solid surface. Ingeneral, film condensation occurs if the surface tension of thesolid is at least 10 dynes/cm higher than the surface tensionof the liquid condensing. The surface tension of liquid perox-ide/water solutions range from 73 dynes/cm (pure water) to

Temperature(oC) Maximum Hydrogen Peroxide Concentration(no condensation) (mg/L)

0% Relative Humidity 10% Relative HumidityCarrier Gas Carrier Gas

0 0.35 0.2810 0.77 0.6320 1.56 1.2830 3.01 2.5040 5.49 4.6050 9.60 8.1160 16.13 13.68

Table B. Maximum peroxide vapor concentration at varioustemperatures.




Figure 4. Typical Vaporized Hydrogen Peroxide (VHP)decontamination system. Vapor is produced by flash vaporizationof liquid hydrogen peroxide and blown into a volume to bedecontaminated. The vapor concentration is held constant by aconstant flow of gas through a destroyer (e.g., x), air blower.

about 80 dynes/cm (pure peroxide). The surface tension for78% by weight peroxide liquid is about 78 dynes/cm. There-fore, drop-wise condensation will occur on solid surfaces withsurface tensions less than about 88 dynes/cm when gas at35% by weight peroxide vapor is condensing. Recall that 35%by weight peroxide vapor condenses as 78% peroxide in theliquid. Polymer materials surface tensions typically rangefrom 20 to 45 dynes/cm or lower. Therefore, drop wise conden-sation should occur on most polymer materials. Clean metaland glass surfaces typically have surface tensions that arehigher than 200 dynes/cm. Contamination on a metal or glasssurface may dramatically drop the surface tension allowingfor drop wise condensation. Figure 3 shows how a contami-nant on a metal surface causes drop wise condensation. Thiswill not allow the entire surface to be exposed to liquidresulting in a non-uniform surface exposure to peroxide.Even if film wise condensation occurs, there is a possibility ofnon-uniform liquid exposure on the surface due to developingliquid flow patterns on vertically sloped surfaces.

A further complication associated with condensation ofliquid is the additional time required to remove peroxide fromthe chamber at the end of an exposure phase. Gas can quicklybe removed from the chamber by purging with an inert gas(such as air). Additional time is necessary to remove peroxidein the liquid form due to the time required to evaporate theliquid from all surfaces. If a material is permeable to liquidwater or peroxide, additional time will be needed to removeany liquid that absorbed into the solid surface. Examplesinclude porous materials and certain polymers, which arepermeable to liquid water and/or peroxide.

Finally, condensed peroxide at high concentration mayhave un-suitable (even violent) reactions with certain mate-rials in the decontamination chamber. Special safety precau-tions should be in place to handle any standing liquid hydro-gen peroxide condensate that can be violently reactive andwill cause severe burns.

Decomposition of Hydrogen PeroxideHydrogen peroxide, especially in gaseous form or impureliquid solutions, is not a very stable compound. In particular,it spontaneously decomposes to form oxygen and water asshown below:

H2O2 ® ½ O2 + H2O

Therefore, the concentration of peroxide gas in the steriliza-tion chamber will steadily decrease over time, depending onthe chamber area, material of construction, contents, etc. Asimple procedure to maintain a constant peroxide concentra-tion involves continually circulating the gas in the steriliza-tion chamber through a system that regenerates fresh perox-ide vapor as shown in Figure 4. This design is successfullyused in gaseous hydrogen peroxide decontamination sys-tems. Fresh peroxide gas produced from a flash vaporizer isintroduced into the chamber as gas in the chamber is re-moved, thereby maintaining a consistent peroxide vaporconcentration. Further, the removed peroxide gas can be

decomposed to water and oxygen in a destroyer and the waterremoved in a dryer. Drying the carrier gas stream is impor-tant as the water content in the carrier gas will affect thepoint at which condensation will occur as shown in Table B.For example, at 25°C, the maximum allowed concentration ofperoxide vapor drops from 2.184 mg/liter to 1.805 mg/liter asthe moisture content in the carrier gas goes from 0% RH upto 10% RH. This is a 17.4% drop in the maximum allowedperoxide concentration that may be introduced into the de-contamination chamber.

Measuring and ControllingPeroxide Gas Concentration

As shown above, Equations one through five may be used tocalculate the maximum level of hydrogen peroxide that maybe achieved in a chamber without causing condensation.Once the maximum peroxide vapor concentration is known,it is possible to develop an operating cycle that gives themaximum concentration of peroxide while preventing con-densation. The controlled variables for the cycle will includeair flow rate, rate of injecting liquid peroxide into the flashvaporizer, concentration of the liquid to be flash vaporized,and the chamber temperature, volume, and humidity level.The peroxide and water concentrations during a cycle can bemonitored using either infrared spectroscopy or electrochemi-cal methods.10,11 It is desirable to run the peroxide concentra-tion as high as possible, but at the same time reducing therisk of producing condensation (typically targeting less than90% of the saturation level). It should be noted that thesedetector systems can only be used to monitor gas concentra-tions, where condensation of liquid will cause the sensor tofail. Overall, higher levels of peroxide vapor concentrationwill result in higher kill rates and therefore shorter cycletimes, but it is recommended that this should be balanced byensuring that condensation does not occur. Once the cycle isdeveloped for a given volume and temperature, it may not be





necessary to monitor peroxide and water concentration aslong as the carrier gas flow and liquid peroxide injection rateare held constant.

In conclusion, hydrogen peroxide in the vapor phase offersan effective, controllable, fast, and safe means of decontami-nating surfaces within a closed chamber of any size. Correctcontrol of decontamination cycles is important to ensure thatoptimal and safe cycles are used for validated, repeatableapplications.

References1. Akers, J.E., Agalloco J.P., Kennedy C.M., “Experience in

the Design and Use of Isolator Systems for SterilityTesting,” PDA J Pharm Sci Technol., 1995, 49(3):140-4.

2. Klapes, N.A., and D. Vesley, “Vapor-Phase HydrogenPeroxide as a Surface Decontaminant and Sterilant,”Appl. Environ. Microbiol., 1990, 56: 503-506.

3. Krause, J., G. McDonnell, and H. Riedesel,“Biodecontamination of Animal Rooms and Heat-Sensi-tive Equipment with Vaporized Hydrogen Peroxide,”Contemp. Topics Lab Anim Sci., 2001, 40: 18-21.

4. McDonnell, G., and Russell, A.D., “Antiseptics and Disin-fectants: Activity, Action, and Resistance,” Clin. Microbiol.Rev., 1999, 12: 147-179.

5. Schumb, W.C., Hydrogen Peroxide, monograph series,American Chemical Society, Reinhold Pub. Corp., 1955

6. Block, S.S., 1991. Peroxygen Compounds. p. 167-181. InS.S. Block (ed.). Disinfection, Sterilization, and Preserva-tion. 4th ed. Lea and Febiger, Philadelphia, PA.

7. Davis, Jr., N.S. and J. Keefe, Jr., J. Am. Rocket Society, p.63, March-April 1952 Buffalo.

8. Scatchard, G., G. Kavanagh, and L. Ticknor, J., Am.Chem. Soc., 74, 3715, 1952.

9. Pitzer, K. and L. Brewer, Thermodynamics, McGraw Hill,1961, Chapter 16 - 20.

10. Day, R. and A. Underwood, Quantitative Analysis, McGrawHill, 1961, chapter 14.

11. Adams, D., G. Brown, C. Fritz, and T. Todd, “Calibrationof a Near-Infrared (NIR) H2O2 Vapor Monitor,” Pharma-ceutical Engineering, Vol. 18, No. 4, 1998, p. 66-82.

12. Electrochemical Co., “Engineering Materials for Use with90% Hydrogen Peroxide,” Part 1, 1 December 1950.

About the AuthorsDr. Carl Hultman received his BS in chem-istry from the University of Wisconsin andhis PhD in physical chemistry from the Penn-sylvania State University. He is currently aprofessor of chemistry at Gannon Universityin Erie, Pennsylvania. Dr. Hultman has pub-lished papers and received a patent for workin surface chemistry. His research interests

include sublimation of solids, molten salt electrochemistry,and synthesis of nano-particle metal oxides. He is a memberof the American Chemical Society. He can be contacted byphone at 1+814-825-5144 or by e-mail at: [email protected].

Gannon University, Department of Chemistry, UniversitySquare, Erie, Pennsylvania 16541.

Aaron Hill received his BS and MS degreesin mechanical engineering from the GannonUniversity in Erie, Pennsylvania. He is cur-rently an engineering manager at Steris Cor-poration at their Erie, Pennsylvania facility.Hill has received four patents pertaining tovaporized hydrogen peroxide equipment. Hewas the project leader for several new vapor-

ized hydrogen peroxide decontamination systems developedat Steris Corporation. He is a member of the AmericanSociety of Mechanical Engineers. He can be contacted byphone at: +1-814-870-8109 x8113 or by e-mail at:[email protected].

Steris Corporation, 2424 W. 23rd St., Erie, Pennsylvania16506.

Dr. Gerald McDonnell has a BSc (Hons.) inmedical laboratory sciences from the Univer-sity of Ulster and a PhD in microbial geneticsfrom Trinity College Dublin. He is currentlya Senior Director, Technical Affairs for SterisCorporation, based at their European head-quarters in Basingstoke, UK. Dr. McDonnellis widely published in the decontamination

and sterilization area. His research interests include infec-tion prevention, decontamination microbiology, emergingpathogens, and the mode of action/resistance to biocides. Dr.McDonnell is the current editor of the Central SterilisationClub (UK), and a member of the American Society of Micro-biology, Society for Applied Microbiology, Council ofHealthcare Advisors, European Biological Safety Associa-tion, and AAMI. He can be contacted by phone at: +44-1256-866560 or by e-mail at: [email protected].

STERIS Limited, STERIS House Jays Close, Viables,Basingstoke, Hampshire RG22 4AX, UK.


Lyophilizer Loading Systems


This articlediscusses thebasics ofAutomatedLoading andUnloadingSystems(ALUS),including thedifferencesbetween fixedand flexiblesystems, as wellas covers theevolution of thetechnology,challengesassociated withthe use of suchsystems,industry growth,and recenttechnologytrends.

Automated Loading and UnloadingSystems for Lyophilizers

by Peter Heyman

Introduction

In recent years, significant focus has beenplaced on the development and continuousimprovement of fully Automated Loadingand Unloading Systems (ALUS) for lyo-

philizers since their introduction in the late1980s. Steady increases in targeted sterilityassurance level, industry adoption of barrierisolation technology, and the need for potentproduct containment have driven this evolu-tion. Both fixed and flexible loading systemshave been tailored to meet manufacturing re-quirements, and the lyophilizers themselveshave been upgraded to properly interface withthese systems.

Reasons for Installing an ALUSThe two primary motivating factors for incor-

porating an ALUS on a filling/lyophilizationline relate to product and operator protection.

Considerable energy has been expended overrecent years in achieving higher and higherlevels of product sterility assurance, throughthe use of steam-in-place, closed tank and pip-ing systems, in-line final sterile filtration, au-tomated checkweighing, and the separation ofoperators from the filling process using Re-stricted Access Barrier Systems (RABS) andisolators.

As the vials to be processed are only par-tially stoppered, the product inside the vials isexposed to the external environment in thelyophilizer aisle during transfer to and loadinginto the freeze dryer. In the traditional process,operators hand carry individual trays with filledvials to the lyophilizer or load trays onto a cart

which is then moved to the cham-ber. The lyophilizer doors areopened wide, exposing the cham-ber to the room environment, andthe operator loads the trays one-by-one on the (usually) pre-cooledshelves. The operator is in closeproximity to the open vials duringthis transfer, thereby increasingthe opportunity for contamination.An ALUS system automates thisprocess, for the most part, remov-ing the operator from the immedi-ate area. An example of one typeof ALUS system is shown in Fig-ure 1. With minimal human pres-ence, better control can beachieved over the bioburden inthe lyophilizer loading corridor,which can be further enhanced ifthe loading system is built withits own RABS or isolator protect-ing the vials throughout the trans-fer process.

Figure 1. Flexible loadingsystem: transfer cart onrails with dedicatedlaminar flow.

Reprinted from







Loading of the lyophilizer through a slot door (also knownas “pizza door”) - Figure 2, which is either opened intermit-tently or throughout the loading process, decreases the like-lihood of introducing particulate to the interior of the cham-ber compared to the traditional, full open door scenario. Anadded benefit is reduced water vapor condensation and frostformation on the cold shelves.

In the case of cytotoxic and other potent entities, an ALUSprovides improved operator protection by distancing him orher from the product. This separation can be further en-hanced by including a RABS or isolator.

Another benefit of an ALUS is improved manufacturingefficiency through the reduction of personnel in the asepticcore, increased flexibility in timing of loading and unloadingof the lyophilizer, and the potential for unloading multiplemachines simultaneously.

The size of the lyophilizer corridor may be reduced whenusing an ALUS (this is project-dependent), and the use of aRABS or isolator should permit the corridor to be designedand operated to a less stringent environmental classification,thereby reducing utility and maintenance costs.

Types of SystemsThe first fully automated ALUS were installed in the late1980s. Since then, approximately 100 ALUS have been madeoperational, most within the last 10 years, with at least 35systems being built since 2003. A greater proportion of theunits have been placed in the US (~60%) than in Europe(~30%) or Japan (~10%).

ALUS can either be “fixed,” conveyor based and mounteddirectly at the freeze dryer, or “flexible,” employing a LoadingAccumulation Table (LAT) connected to the filling line, aTransfer cart (T-cart) to transport each shelf load of vials toand from multiple lyophilizers, and an Unloading Accumula-tion Table (UAT) connected by conveyor to a capping machine- Figure 3.

Hybrid or mixed systems also have been used with flexible

loading and fixed unloading, or vice versa, typically associ-ated with a pass-through style dryer. Fixed systems arenormally selected for operations involving one or two lyo-philizers, while flexible systems make more economic sensewhen using four or more lyophilizers or when consideringfuture expansion. A financial and process flow analysis mustbe conducted and the facility expansion potential consideredwhen deciding which system to use when three lyophilizersare involved.

A fully automated, flexible loading system, including load-ing accumulation table, transfer cart, and unloading accumu-lation table, all with dedicated laminar flow, plus guide rails,power rail, control systems, and operator interfaces canrequire an investment of roughly $3 million. However, thevalue of one or two lyophilizer loads of vials having to bedestroyed due to potential contamination by an operatorinvolved in manual loading could far exceed this investment,particularly in the case of high priced biotech products. As anexample, a 270 square foot (25 square meter) shelf area freezedryer holds greater than 77,000 5ml/18mm diameter vials. Ata selling price of $40 per vial, the value of a fully loadedlyophilizer would be $3 million.

While most flexible loading systems employ floor-mountedrails to precisely guide the transfer cart through the lyo-philizer corridor - Figure 1, at least one vendor has opted touse laser technology for this purpose. Laser is emitted fromthe transfer cart and returned by wall-mounted reflectors,providing continuous feedback as to the cart’s precise posi-tion within the corridor - Figure 4. Such a system obviates theneed for guide rails and makes floor cleaning easier. Thisallows greater flexibility for the addition of new lyophilizers(requiring only a PLC program change) and permits a greaterdegree of latitude in transfer cart motions. Carts then can beturned through multiple angles and negotiate tight corners,which permits greater choice in room layout. There is lessindustry history on the performance of these laser-guidedcarts, and operator traffic in the area must be consideredwhen implementing the system to avoid multiple stops-and-starts of the transfer cart which can lead to greater downtime

Figure 2. Vial loading through slot door in lyophilizer.

Figure 3. Unloading Accumulation Table (UAT).




Figure 5. Fixed loading systems with row loading of vials using a push bar.

for equipment maintenance.Power for transfer carts is supplied either through a power

rail running parallel to the guide rails and mounted either onthe floor or on the wall below the lyophilizer, or from an on-board battery that gets recharged when the T-cart is dockedto the loading table, unloading table, or placed in a mainte-nance position further down the corridor.

Vials can be loaded an entire shelf at a time or one or morerows at a time. A third approach, loading multiple shelves atonce, was employed on early installations, but requires a fullyopen lyophilizer door and long atmospheric exposure time ofvial contents while vials are being accumulated, and thus hasfallen out of favor. Flexible systems generally load a shelf ofvials at once, whereas fixed systems can load row-by-row (ora few rows at a time) or less commonly, shelf-by-shelf. Rowloading is best suited for barrier isolation as it presents amuch smaller footprint to be enclosed - Figure 5. Another

benefit of fixed row-by-row systems is that the transfer timebetween filler and pre-cooled shelves is minimized.

Row loading of vials means that the slot door on thelyophilizer must remain open for extended periods of time. Tominimize frost formation on pre-cooled shelves, which cancomplicate vial loading and inhibit heat transfer, some endusers have opted to maintain the relative humidity of theloading area as low as 5% RH, and reduce the local tempera-ture to 10 to 12°C versus the normal 15 to 20°C.1

Roughly 50% of the installed systems are of flexible de-sign, 40% fixed, and 10% hybrid. Of the fixed loading systems,80% or more load by rows rather than by full shelves.

The decision to go with a pass-through lyophilizer (i.e.,front load/rear unload rather than front load and unload)necessitates that the condenser be positioned either to theside of or below the freeze dryer chamber, thereby creatingsome space restrictions for adjacent process equipment andthe potential need for a classified environment on the unload-ing side.

Most ALUS systems in the field are outfitted with local-ized, dedicated laminar flow units. Fewer than 20 systemshave been built with true barrier isolators - Figure 6 and a fewwith RABS. Vendors have seen increasing demand recentlyfor barrier isolated systems, but these still remain a minorityof requests.

In addition to an integral laminar flow hood, a RABSincludes rigid barriers and often gloveports like an isolator.However, unlike an isolator, some end users permit the doorsof a RABS to be opened to clear jams during processing ormake other minor interventions with a required subsequentline clearance and manual sanitization step before resumingoperation. A secondary laminar flow unit or facility-basedlaminar flow may be provided on the perimeter of the RABSto compensate for these occasional door openings, includingthose made during cleaning and set-up between lots.

Figure 4. Laser-guided transfer cart.




Figure 6. Transfer cart with barrier isolator.

The doors of an isolator are intended never to be openedduring processing. The enclosure is air tight and maintains adesired overpressure, and the interior is automatically ster-ilized or sanitized via Vaporized Hydrogen Peroxide (VHP) toa desired bio-burden reduction level. In contrast, a RABS isnot usually designed to maintain an overpressure and itsinterior is typically manually sanitized using liquid hydrogenperoxide or other agents.

ChallengesA common issue encountered during the start-up and con-tinuing operation of an ALUS is the alignment of the loadingsystem with the shelves of the lyophilizer. In the case of aflexible ALUS, the transfer cart must align with multiplelyophilizers, each with slightly different chamber and shelfpositioning control. Chambers must be carefully positionedupon initial installation, or occasionally repositioned in thecase of existing machines being retrofitted with an ALUS,and docking locations must be precisely programmed into thecart’s control system.

Features such as linear encoders have been added to thehydraulic cylinder on the shelf control system to moreprecisely position the shelves relative to the vial transferplate and loading system. The signal output also may besent to the transfer cart that may have its own automaticheight adjustment system. Clamps may be incorporated onthe transfer cart to match up with blocks on the slot door,alignment fixtures may be installed directly on the shelves,and interlocking pins or fingers also have been successfullyemployed. To compensate for the shelves not being flat orslight misalignments, some transfer plates have been fabri-cated of thin, flexible 301SS. Through these efforts, theplates can be aligned to within 0.5mm or better, therebyavoiding vial tipping or jamming during the loading orunloading process.

For standard installations not including an ALUS, prod-uct temperature is normally monitored over the entire freezedrying cycle by placing thermocouples in vials on differentshelves and different areas of each shelf, as the vials are beingloaded tray-by-tray into the lyophilizer. Using this informa-tion, operators can monitor progress of the cycle, therebyensuring that product is fully frozen throughout the chamberat the time the primary drying phase is automatically initi-ated, that product is maintained below its collapse tempera-ture during this drying phase, and that the sublimationprocess is complete for all vials at the time the secondarydrying step is triggered.

In the case of an ALUS-equipped lyophilizer, automaticloading generally precludes placement of thermocouples invials. Therefore, the monitoring of a production run is limitedto observing temperature data coming from the shelf inletand outlet RTDs and condenser, chamber pressure, and insome cases, automated chamber pressure rise measurementstaken at the end of a drying phase. This makes the originalfreeze drying cycle development/scale-up process and itsqualification critical, as well as the lyophilizer control system’sability to carry out consistent, reproducible cycles. Cycle

scale-up work is conducted with the aid of vial thermocouplesinstalled in the lyophilizer solely for this purpose, but notused during production.

A common problem encountered during unloading is thepresence of stoppers (and their connected vials) stuck to theunderside of the shelf above, originating from the stopperingprocess at the end of secondary drying when the shelves arecompressed with high force onto the stoppers. Factors affect-ing this adhesion are thought to include the rubber formula-tion and surface coating, shelf surface geometry, stopperingwith partial vacuum in the chamber (leading to a suctioneffect between stopper and shelf) followed by chamber vent-ing to atmosphere, residual silicone on the shelf and/orstopper, and shelf temperature.

With manual unloading, operators often shake the four-sided stainless steel tray ring back-and-forth until adjacentvials cause the hanging vial to drop down onto the shelf andthen unload. With an ALUS, such an action is much moredifficult to achieve, thereby resulting in a higher incidence ofsuch vials falling and knocking over adjacent vials whileunloading. This has to be addressed by incorporating fallenvial detection/rejection systems at unloading.

Another aspect often overlooked during the detailed de-sign phase is accessibility for cleaning and maintenance ofthe transfer cart. The cart rails may be positioned close to thewall of the lyophilizer aisle, thereby making access to one sideof the cart difficult. Some owners have had the foresight todedicate a separate section of the corridor for cart mainte-nance, whereas others have had to install a removable panelin the wall as an afterthought.

One of the justifications used in the past for the costlyinvestment in automating loading and unloading is thecomplete elimination of operators from the lyophilizer corri-dor, but this seldom has been realized. In many facilities, anoperator still keeps a watchful eye on the process and isavailable to quickly clear jams and other faults. Opticalsafety guards are mounted on the sides of transfer carts toprevent collisions with operators and other objects. As thealready reliable ALUS systems improve even further, the




Figure 7. ALUS installations by year.

hope remains that human interventions in the process can bekept to an absolute minimum.

Industry GrowthThe overall growth in pharmaceutical company adoption ofALUS systems can be seen in Figure 7. The figure shows thetotal number of annual installations since 1989 by two of thekey industry suppliers. The growth in ALUS lines follows asimilar pattern to that of isolators, as published recently byJ. Lysfjord and M. Porter in their biannual update,2 acceler-ating in the second half of the 1990s.

Technology TrendsIn the future, a variety of ALUS designs will continue to beinstalled, but at least one equipment supplier is anticipatinga reduced demand for shelf-by-shelf loading systems in favorof row-by-row systems.

New features added in recent years to ALUS have includedon-board particle counters within transfer carts and wirelessEthernet data transmission, as well as VHP-sterilizable,isolator-enclosed loading carts, and conveyors.

Buffer systems, in which multiple shelf loads of recentlyunloaded, stoppered vials are stored prior to capping, shouldfind more use as regulatory agency focus grows on themaintenance of sterility of the top of the stopper beforeenclosure by the aluminum crimp seal.

End users are expected to more clearly designate anddefine procedures and environmental controls for a technicalarea for ALUS cleaning and maintenance adjacent to, or partof, the lyophilizer corridor.

Lastly, greater focus will be given to the areas of errorprevention, detection, and recovery throughout the process,including provisions for manual backup where warranted.

References1. “HollisterStier Contract Manufacturing Invests $5.7 mil-

lion in Customized Lyophilizer and Facility Upgrade,”Pharmaceutical Online, 24 May 2004. http://www.pharmaceuticalonline.com/content/news/article.asp?docid={72f96228-b5bf-4ebe-b4c4-babc57b3005c}.

2. Lysfjord, J., and M. Porter, “Barrier Isolator Filling Line– Deliveries by Year,” presented at the June 2004 ISPEBarrier Isolator Conference, Arlington, VA.

About the AuthorPeter Heyman is a Senior Manager of Phar-maceutical Manufacturing Technology atCH2M Hill. He is responsible for providingengineering services to pharmaceutical cli-ents in process flow design and facility lay-out, as well as the specification, procure-ment, and testing of aseptic fill and finishmanufacturing equipment. He joined CH2M

Hill in 2006 after 21 years on the client side, working forGenentech, Merck, and Becton Dickinson. He holds an MS inbiomedical engineering from the University of Utah and a BSin the same field from the University of Pennsylvania. He canbe contacted by phone at: +1-732-868-2253 or by e-mail [email protected].

CH2M Hill Lockwood Greene, 270 Davidson Ave.,Somerset, New Jersey 08873.


Biofilm Formation


Figure 1. The threephases of biofilm growth(after Flemming6).

This articlepresents anoverview ofhow bacteriacan grow andmultiply incompendialwater systems.

Biofilms – Survival and Growth ofBacteria in Compendial High PurityWater Systems

by Frank Riedewald and Aidan W. Sexton

Introduction

The pharmaceutical industry goes togreat lengths and expense to controlbacterial growth in compendial watersystems. Despite all efforts to make it

difficult for bacteria to survive in compendialwater systems, bacterial survival and growthis observed all too often.

How do bacteria manage to survive and

even multiply in compendial water systemsthat have been artificially created to be hostileto bacterial growth? Microbiologists havestruggled to give a satisfactory answer to thisquestion for quite some time. However, overthe last 15 years, it has become more and moreevident that bacteria form so called biofilms inhigh purity water systems. Once bacteria haveformed a sessile biofilm on any available sur-

face, their chances ofsurvival and growth in-crease significantly. Infact, in nature morethan 99% of all micro-bial activity occurs inbiofilms, and not as tra-ditionally thought, asfree-floating (plank-tonic) individual bacte-ria.1,2 Kolter has re-cently commented that“biofilm formation ismost likely a universalfeature of microbes.”3

The success story ofbiofilms stretches fromthe very beginning oflife three billion yearsago, to today’s tenaciousbiofilm infections,which constitute about60% of infections treat-ed by physicians in thefirst world.4,5 This ar-ticle will first outlinehow biofilms form incompendial water sys-tems, and then discusshow biofilms will ac-

Reprinted from





Biofilm Formation


Figure 2. Mechanism and timeframe for formation of a biofilm. (Image adopted after J. W. Costerton and Peg Dirckx, Center for BiofilmEngineering at Montana State University, Bozeman, USA - see Figure 3.)

tively contaminate compendial waters and show that theyare the root cause of ongoing bacterial contamination of thewater system.

What is a Biofilm?A biofilm is defined as bacterial cells adherent to each otherand/or to surfaces or interfaces and are covered by a slimysubstance, which acts as a shield, protecting the biofilm fromphysical and chemical attack. Planktonic cells on the otherhand are free single motile cells in the bulk phase of thewater.2 The nature of the growth of sessile (non-mobile)bacteria films, while not fully understood, is almost certainlysignificantly different to planktonic growth. A typical growthcurve for a biofilm is shown in Figure 1.

Formation of a BiofilmFigure 2 illustrates the formation of a biofilm starting froma clean surface, from initial contamination by free floatingplanktonic cells, to the formation of a mature biofilm, and theapproximate time scale involved.

In an aquatic environment, an invading planktonic bacte-rial cell is attracted to a surface, where it may attach withina short timeframe.6,7 Indeed speed of attachment is thoughtto be a survival strategy of bacteria.7 Bacteria that attachfirst get the best location for survival.

Initial surface attachment is thought to occur through amechanism whereby the bacteria accumulate at the liquid/pipe (or vessel) interface due to the increased ionic nutrientconcentrations that occur at the water/steel interface in highpurity water systems.2 Since growth conditions are signifi-cantly enhanced in this interface zone, biofilm growth isinherently favored over planktonic growth at this location.

The time scale for a biofilm to form depends on a numberof factors, such as nutrient availability, hydrodynamics and

substratum characteristics, to name but a few. Since everywater system will be different, exact predictions of how longit takes for a biofilm to form is difficult. In a laboratory watersystem, for instance, an extensive biofilm was found afterthree months of operation. However, the bulk water qualitywas always within microbial quality specification.8 Thus, amature biofilm can develop remarkably fast, especially inenvironments where all requirements for biofilm growthexist, namely:

• an available surface• a source of water• initial contamination by an invading planktonic bacte-

rium• adequate, albeit very low concentrations of nutrients

Once the biofilm has grown to its mature state (Figure 3), acomplex collection of micro-colonies, which are protected by apolysaccharide coating, can be observed. The biofilm is inter-spersed with water channels through which the bulk waterflows, transporting nutrients to and waste from the biofilm.Some micro-colonies may be conical in shape, while othersmay be mushroom shaped.2 A further advantage of biofilmformation is that it may allow for enhanced exchange oftransmissible genetic information.7

The structure of a mature biofilm is not random.2 A levelof organization can develop in which the cells specialize toachieve a structurally and physiologically complex biofilm.2

Bacteria must communicate and coordinate their behavior toachieve a complex biofilm structure as shown in Figure 3,7,9

behaving more like an organism, rather than just singlebacterial cells. Bacteria are able to sense their environment,process information, and react appropriately.7,10 Their abilityto sense cell density of other species as well as their own,


Biofilm Formation


communicate with each other, and to behave as a communityis called quorum sensing.11-13 This ability of bacteria, onlyrecently discovered, significantly increases their chances ofsurvival.10,13 Planktonic cells on the other hand are not able tocommunicate with each other, as the chemical signals in-volved are more likely to be carried away by the bulk waterphase without reaching another cell.7

Furthermore, the cells composing a biofilm are not simplyplanktonic cells attached to a surface, as could be deduced fromthe above definition of a biofilm. Cells, once attached, aredemonstrably and profoundly different to their planktoniccounterparts.2 Their morphology and their phenotype havechanged. For instance, a bacterium can only become a produc-tive member of a biofilm community if it refrains from produc-ing a flagellum that might destabilize a biofilm.7 These alteredcells secrete a polysaccharide substance, which acts both as aprotective layer and a nutrient trapping device encasing thebiofilm. Different bacterial growth patterns occur, graduallydeveloping a structurally complex mature biofilm.2,7

Within the biofilm, gradients of nutrient, pH, and oxygenexist. In single species biofilms, bacteria alter their geneexpression to maximize their survival in their particularmicroenvironment, while in multi-species biofilms the vari-ous bacteria are distributed according to the best require-ments of each one.7,9 In nature, bacteria usually live incomplex mixed species biofilms.2,13 However, industrial

biofilms can be single species biofilms.14

The image of a biofilm shown in Figure 3 is somewhatidealized. Figures 4 and 5 show a real biofilm in an industrialwater system. Hydrodynamics also have an influence on thestructure of biofilms. Under the influence of shear stress athigher fluid velocities, biofilms tend to be denser and theirsurfaces tend to be smoother.10 In addition, the structures ofthe biofilm as shown schematically in Figure 3 would becomemore elongated in the downstream direction10 influenced byhigh fluid velocities near the surface.

Nutrient Levels in Purified Water SystemsAlthough the nutrient concentration in purified water sys-tems is artificially low, bacteria still manage to grow andmultiply in these systems. Bacteria can grow and multiply inpurified water systems because the nutrient concentration isstill not low enough to prevent bacterial growth. In fact, the

Environment Nutrient Concentration Growth Mode of(mg/L) Bacteria

Deep Ocean 0.5 – 0.8 Grow in Biofilm only

Distilled Water Up to 0.5 Grow in Biofilm only

Petri Dish > 2,000 Grow in Biofilm or asPlanktonic cell

Table A. Nutrient levels in different water types.2,6,15

Figure 3. Schematic of mature biofilm. (Image reproduced with permission from J. W. Costerton and Peg Dirckx, Center for BiofilmEngineering at Montana State University, Bozeman, USA.)


Biofilm Formation


Figure 4. Scanning electron micrograph of a biofilm (Bar 1 μm).(Image reproduced with permission from Rodney Donlan and DonaldGibbon, Center for Disease Control and Prevention, Atlanta, USA.)

Sidebar 1 - The Difficulties ofExamining Biofilms

Why have we missed biofilms for so long? This isbecause they are difficult to study, whereas laboratorywork on planktonic cell cultures is easier and moreconvenient.

At the beginning of the 1980s scientists realized theimportance of biofilms in natural and industrial watersystems. At this time, direct recovery methods showedunequivocally that more than 99.9% of bacteria in anaquatic system grow in biofilms on surfaces. It tookanother few years before the complex open structureof biofilms (Figures 3 to 5) came to light and was fullyappreciated. Originally, biofilms were thought to besimple mats of bacteria stuck to surfaces. Using anumber of techniques in the early 1980s such as theScanning Electron Microscopy (SEM - Sidebar 2) andthe Confocal Laser Scanning Microscope (CLSM -Sidebar 3), the complexity of biofilms was finallyrevealed.

It is the emphasis placed on laboratory work that hasresulted in our missing the significance of biofilms forso long. Only recently have microbiologists realizedthat in nature, bacteria grow in biofilms and not assingle cell culture as studied in the laboratory.2 “Micro-biologists have been barking up the wrong tree sincethe time of Pasteur,” says Costerton.1

nutrient concentration of Deep Ocean waters and distilledwaters (Table A) is similar. In both cases, the nutrients areavailable in infinite quantities, albeit in very low concentra-tions. The main source of nutrients in compendial watersystems is generally the feed water.6,24 One should keep inmind that compendial water systems are open systems,meaning that new feed water ensures a constant supply of

nutrients, albeit at a very low concentration.Since the nutrient concentrations found in deep ocean

water and distilled water are not significantly different,bacterial growth in distilled waters should be expected,albeit only in a biofilm. Direct microscopic observationshave shown that bacteria cannot grow as free floatingplanktonic cells at nutrient levels found in distilled watersystems.16 However, by forming a biofilm, not alone willthese bacteria survive, but they will actually grow andreproduce.6,16 Therefore, biofilm bacteria succeed in growingand multiplying in low nutrient environments, which arehostile to the growth of planktonic cells. This is because thefood is brought to the sessile bacteria cells in the watersystem, instead of individual cells wasting energy swim-ming toward the food.

Bacteria are used to starving conditions. In nature, thebulk of bacterial existence must be spent in starving or slowgrowing conditions, because no natural environment con-tains enough nutrients to sustain the voracity of rapid bacte-rial growth for very long.17 Paradoxically, these low nutrientconcentrations, as found in nature, are actually created bythe bacteria themselves.15 Whenever there is a high nutrientconcentration available for bacteria, bacteria move in, multi-ply, and consume the food source very quickly.

Bacteria can only multiply as planktonic cells if the nutri-ent concentration is very high. Examples of planktonic cellsgrowing and reproducing can be found in laboratory nutrientrich Petri dishes, bottles of growth media, or inside a host(e.g., human, animal, etc.).1,11

Highly structured biofilms cannot develop if the nutrientconcentration is very low. For instance, in Ultrapure Watersystems (nutrient concentration < 30 ppb TOC), a biofilm canonly develop to a thickness of a couple of cell layers.18

Why Do Bacteria Grow in Biofilms?There is insurmountable scientific evidence that bacteriaprefer to live in biofilms rather than existing as free-floatingplanktonic bacteria. Extensive investigations on aquatic sys-tems have shown that in all natural, industrial, and medicalecosystems, biofilm bacteria hugely outweigh their plank-tonic counterparts in biomass. In nature more than 99% of allmicrobial activity occurs in biofilms.1,2 Since attachment ofbacteria to surfaces is so prevalent, it is seen as the mostimportant survival strategy of bacteria in low nutrient wa-ters.9,13,19

Considering all of the advantages that bacteria find in theformation of biofilms, microbiologists now conclude that thepreferred mode of bacterial growth is in a biofilm, while theplanktonic state is a mechanism for biofilm dissemination ordispersal.2

There are a couple of reasons why bacteria may prefer tolive in a biofilm, rather than in the planktonic state. Thebiofilm is a favorable habitat as it allows access to scarcenutrients. Biofilms provide a defense mechanism against awhole range of stresses, not to understate it; there is strengthin numbers.


Biofilm Formation


• Access to resources and niches that cannot be utilized byisolated cells.9 For example, bacteria have been foundflourishing inside UV disinfectant cells. In nature, manybacteria break down complex organic polymers, requiringthe concerted action of many cells.9 On the other hand, inlaboratory cultures, bacteria are supplied with simplenutrients, which are readily utilized by individual cells.

• Collectively protect against antagonists that eliminateisolated cells9 (i.e., protection against biocides, phages,etc.).

• Optimize population survival by differentiation into dis-tinct cell types9 (i.e., formation of dormant cells or sub-populations). The resulting genetic diversity, even in singlespecies biofilms, greatly enhances the chances of thebiofilm community to survive attack. Diversity will ensurethe survival of some bacteria; much in the same way thatmixed woodlands survive droughts or insect attacks betterthan mono-species forests.22

• DNA exchange between bacterial cells – even with differ-ent species – is possible, further increasing the survivalabilities of the community.7

Contamination of the Bulk Water Phase by Biofilm BacteriaIn high purity water systems, the presence of a biofilm wouldnot be a problem if the bacteria remained only in the biofilm.Unfortunately, biofilms actively contaminate the bulk phaseof purified water systems with planktonic bacteria and biofilmcell aggregates. It is these planktonic bacteria and aggre-gates released from the biofilm, which are detected by typicalwater quality assessments. In other words, bacterial con-tamination of the bulk water phase is directly linked to theactivity of a pre-existing biofilm. The quality of the bulkwater can only be improved by suppressing the biofilmactivity below a certain acceptable threshold.6

Bacteria living in the sessile community of a biofilm musthave a mechanism to colonize new areas, or leave the micro-

Favorable HabitatAs already stated, in low nutrient environments, bacteria canonly grow while residing in a biofilm. One of the reasons whybiofilms are so successful is that they form right where thenutrient concentration is at its highest since nutrients tendto accumulate at surfaces. It is energetically more favorablefor organic molecules - nutrients - to be at an interface thanremain in solution.2 In other words, in low nutrient waters,such as compendial water systems, there is a natural flow oforganic materials (i.e., nutrients) from the bulk water phaseto the surfaces. By forming a biofilm, bacteria can utilizethese surface nutrients.

But the biofilm mode of growth makes the habitat evenmore favorable. The polysaccharide film encasing the biofilmserves as a nutrient trapping device.2 This ability of biofilmsto “actively” extract nutrients from the bulk water explainswhy biofilms can grow in high purity water systems, whentheir planktonic counterparts starve.2 A further advantagefor cells in a biofilm is that attached cells, which have died,can be used as nutrient sources by neighboring cells.

DefenseThe polysaccharide coating encasing the biofilm also offers aprotective shield against physical and chemical attack. Gener-ally, bacteria within biofilms can survive much greater concen-trations of biocides (chlorine, penicillin, etc.) than their motilebacterial counterparts in the bulk phase of the water.2,19

Very high concentration and very long exposure timesexceeding the recommended concentrations and exposuretimes of the manufacturers may be required to eradicatebiofilms with typical disinfectants as for instance used in thefood-processing industry (namely, chlorine, ammonium com-pounds, iodophores, etc.).20 This is because recommendedbiocide concentrations and exposure times have been deter-mined by assessing planktonic cell deaths, while ignoring theeffect, if any, on the presence of a biofilm.2

One mechanism of biofilm resistance to antimicrobialagents is related to the inability of an antimicrobial agent topenetrate the full depth of a biofilm.21 A second theory is thatat least some of the cells in a biofilm experience nutrientlimitation, and therefore exist in a slow growing or starvedstate. It is known that starved cells are not very susceptibleto many antimicrobial agents.19,21

Protection from heat and drying is also provided by theencasing slime layer, as it is composed of fibrous polysaccha-rides and contains about 99% water, which protects thebiofilm remarkably well from drying,2 an important survivalstrategy as water levels in nature can often fluctuate widely.

Strength in NumbersAs is so often the case, there is strength in numbers. With theadvantages of its communal structure, biofilms constitute aprotective mode of growth, that allows survival and growth inhostile environments.9 Research has shown that bacteriahave communication and decision making capabilities, whichenable them to coordinate growth, movement, and biochemi-cal activity to achieve:

Figure 5. Scanning confocal micrograph of a biofilm in anindustrial water system showing the mushroom architecture.(Image reproduced with permission from Paul Stoodley, Center forBiofilm Engineering at Montana State University, Bozeman, USA.)


Biofilm Formation


with bacteria.10,19,23 For rippling, detaching, and rolling dis-persal (Figure 6), external fluid forces are required. However,no external fluid forces are required for seeding dispersal.The latter mechanism of biofilm cell release to spread to newareas is more sophisticated as the biofilm actively releasessingle motile planktonic cells. This has been shown to be aprogrammed set of events leading to a localized hydrolysis ofthe extra-cellular polysaccharide coating and conversion of asub-population of cells into motile planktonic cells, whichleave the biofilm19,23 as indicated in Figure 6.

ConclusionBiofilms, to date, have been underestimated, both in theirextensive occurrence in water systems and their contribu-tion to the overall bioburden in all compendial water sys-tems.2,10,14,24 Biofilms in high purity water systems should beseen as normal since currently we neither have the meansto prevent a biofilm from forming nor to totally removebiofilms from the surfaces of water distribution systems.6

The method of detection of biofilms or their managementin high purity water systems is beyond the scope of thisarticle. Overviews are readily available regarding biofilmmanagement strategies for high purity water systems.6,25 Thefollowing recommendations can be drawn from the literatureto minimize biofilm growth:25

• minimization of nutrient concentration in the bulk waterphase and thereby at surfaces

• use of disinfectants in high concentrations to kill biofilmcells

• use of mechanical operations to detach biofilms• selection of materials of the correct smoothness/roughness

to make cleaning easier26

• cleaning or removal of the biofilm at short intervals• drying the system completely whenever possible• checking the efficiency of countermeasures on surfaces• use of high water linear velocities throughout the distribu-

tion piping

DefinitionsBacteria - Traditionally regarded as single cell entities,bacteria are a very heterogeneous group. The only propertythey have in common is that they are all prokaryotes. Simplystated, prokaryotes are molecules surrounded by a mem-brane and cell wall. Prokaryotic cells lack characteristiceukaryotic sub-cellular membrane enclosed “organelles,” butmay contain membrane systems inside a cell wall.

Typical bacteria: (Eubacteria): Rigid cell walls; unicellu-lar; multiply by binary fission; if motile move using flagella.

Cell - The basic structural unit of living things whether plant,fungus, or animal.

Dissemination - Scattering, distributing over a consider-able area.

Metabolism - The sum of all the chemical reactions in a cell.

bial community if conditions become unfavorable (i.e., if thearea is getting too crowded or nutrients are becoming toolimited).7 More than one mechanism has been identified bywhich biofilms spread to other areas or (from our point ofview) contaminate the bulk phase of a purified water system

Sidebar 2 - The Principle of theScanning Electron Microscope

Figure 4 was recorded with a Scanning Electron Micro-scope (SEM). The SEM is capable of producing highresolution (< 1nm) images. It allows a greater depth offocus than the optical microscope so it can produce arepresentation of a three-dimensional sample. Sincethe SEM operates under vacuum, a biofilm sample mustfirst be dehydrated before imaging. Secondly, as abiofilm is not conductive, it also must be gold plated.Of course, these two processes somewhat alter thebiofilm sample.

In SEM, elec-trons are emittedfrom a cathode andaccelerated towardan anode. The elec-tron beam passesthrough pairs ofscanning coils in theobjective lens,which deflect thebeam in a rasterfashion over a rect-angular area of the sample surface. The beam isfocused by one or two condenser lenses into a beamwith a very fine focal spot sized one to five nm2 (seeimage).

Images can be generated from two different electronbeams. The first method to generate an image with theSEM is by detecting the backscattered electrons. Whenthe electron beam strikes the sample, some of theelectrons will interact with the nuclei of the sampleatoms and will bounce back out of the sample withoutslowing down. This effect is called backscattering.When these backscattered electrons hit the detector, asignal is generated which is used to generate an image.The second method depends on the detection ofsecondary electrons to generate an image. In this case,the electrons of the atoms of the specimen are pushedby the high velocity electron beam out of the atom andexit the sample surface. These electrons move veryslowly, in comparison to backscattered electrons, andsince they are negatively charged, they can be at-tracted to a positively charged detector. This attractionforce pulls in electrons from a relatively large area,which is what gives secondary electron images a three-dimensional look, similar to that seen in Figure 4 in themain text.


Biofilm Formation


Figure 6. Bacterial dispersal from a biofilm. (Image reproducedwith permission from Peg Dirckx, Center for Biofilm Engineering,Montana State University, USA.)

Morphology - The form or shape of a particular organism orstructure.

Motile - Having spontaneous, but not conscious or volitionalmovement.

Polysaccharide - Long chains of monosaccharide (sugar)subunits.

Sessile - Permanently attached or fixed; not normally free-moving.

References1. Potera, C., “Biofilms Invade Microbiology,” Science, 27

September 1999, Vol. 273, p. 1795-1797.2. Costerton, J.W, Lewandowski, Z., Caldwell, D.E., Korber,

D.E., and Lappin-Scott, H.M., “Microbial Biofilms,” An-nual Review of Microbiology, 1995, Vol. 49, p. 711-745.

3. Kolter, R., “Surfacing Views of Biofilm Biology(editorial),”Trends in Microbiology, Vol. 13, No. 14, 2005,p. 1-2.

4. Nisbet, E.G. and Sleep, N.H., “The Habitat and Nature ofEarly Life,” Nature, Vol. 409, 22. February 2001, p. 1083-1091.

5. Fux, C.A., Costerton, J.W., Stewart, P.S., and Stoodley,P., “Survival Strategies of Infectious Biofilms,” Trends inMicrobiology, Vol. 13, No. 14, 2005, p. 34-40.

6. Flemming, H.C., “Biofouling bei Membranprozessen,”Pharma-Technologie Journal, CONCEPT Heidelberg,1992.

7. Watnick, P. and Kolter, R., “Biofilm, City of Microbes,”Journal of Bacteriology, Vol. 182, No. 10, 2000, p. 2675-2679.

8. McFeters, G.A., Broadaway, S.C., Pyle, B.H., and Egozy,Y., “Distribution of Bacteria Within Operating Labora-

tory Water Purification Systems,” Applied and Environ-mental Microbiology, Vol. 59, (1993), p. 1410-1415.

9. Shapiro, J. A., “Thinking about Bacterial Populations asMulticellular Organisms,” Annual Review of Microbiol-ogy, Vol. 52, 1998, p. 81-104.

10. Stoodley, P., Sauer K., Davies D. G., and Costerton J. W.,“Biofilms as Complex Differentiated Communities,” An-nual Review of Microbiology, 2002, Vol. 56, p. 187-209.

11. Sauter, K., Camper, A.K., Ehrlich, G.D., Costerton, J.W.,and Davies, D.G., “Pseudomonas aeruginosa DisplaysMultiple Phenotypes during Development as a Biofilm,”Journal of Bacteriology, Vol. 184, No. 4, 2002, p. 1140–1154.

12. Strauss, E., “A Symphony of Bacterial Voices,” Science, 21May 1999, Vol. 284, No. 5418, p. 1302-1304.

13. Jefferson, K.K., “What Drives Bacteria to Produce aBiofilm?,” FEMS Microbiology Letters, 236, (2004), p.163-173.

14. Mittelman, M. W., “Biofilm Development in PurifiedWater Systems,” in Microbial Biofilms, Lappin-ScottH.M., Costerton J.W., Eds., Cambridge University Press,Cambridge, ISBN 0521 454123, 1995, p 133-147.

15. Morita, R.Y., “Bioavailability of Energy and its Relation-ship to Growth and Starvation Survival in Nature,” Cana-

Sidebar 3 - The Principle of ConfocalLaser Scanning microscope

Figure 5 was obtained with a Confocal Laser ScanningMicroscope (CLSM). With CLSM, 3D images can berecorded without the need to dehydrate or plate sampleswith a metal. Apart from staining the cells to ensure thecells will fluoresce when hit by the laser beam, the cellscan be observed in their natural environment.

The simplestCLSM uses a laserbeam and imagesare taken point-by-point and recon-structed using acomputer. The la-ser beam is focusedinto a small volumewithin a fluorescentspecimen. Typi-cally, the specimenmust be stained inorder for it to fluo-resce. Only the fluo-rescent light will beallowed to pass intothe detector.

CLSM offers several advantages over conventionaloptical microscopy, including controllable depth offield, the elimination of image degrading out-of-focusinformation, and the ability to collect serial opticalsections from thick specimens.


Biofilm Formation


dian Journal of Microbiology, Vol. 34, (1988), p. 436-441.16. Warburton, D.W., Austin, J.W., Harrison, B.H., and Sand-

ers, G., “Survival and Recovery of Escherichia coli O157:H7in Inoculated Bottled Water,” Journal of Food Protection,Vol. 61, No. 8, 1998, p. 948-952.

17. Chesbro, W., “The Domains of Slow Bacterial Growth,”Canadian Journal of Microbiology, Vol. 34, 1988, p. 427-435.

18. Gillis, R.J., “A Comparative Study of Bacterial Attach-ment to High-Purity Water System Surfaces,” UltrapureWater, September 1996, p. 27-32.

19. Costerton, J.W, Stewart, P.S., and Greenberg E.P., “Bac-terial Biofilms: A Common Cause of Persistent Infec-tions,” Science, 21. May 1999, Vol. 284, p. 1318-1322.

20. Costerton, J.W., P.S. Stewart, “Battling Biofilms,” Scien-tific American, July 2001, p. 74-81.

21. Lewis, K., “Riddle of Biofilm Resistance,” AntimicrobialAgents and Chemotherapy, April 2001, p. 99-1007.

22. Boles, B.R., Thoedel, M, and Singh, P.K. “Self-GeneratedDiversity Produces ‘Insurance Effects’ in Biofilm Com-munities” Proceedings National Academy of Science, No-vember 2004, Vol. 101, No. 47, p. 16630-16635.

23. Hall-Stoodley, L., Stoodley, P., “Biofilm Formation andDispersal and the Transmission of Human Pathogens,”Trends in Microbiology, Vol.13, No.1, 2005, p. 7-10.

24. Flemming, H.C., “Biofouling in Water Systems -Cases,Causes, and Countermeasures,” Applied Microbiology &Biotechnology, 2002, 59, p. 629-640.

25. Riedewald, F., “Biofilms in Pharmaceutical Waters,”Phar-maceutical Engineering, Nov/Dec 1997.

26. Riedewald, F., “Bacterial Adhesion to Surfaces: The In-fluence of Surface Roughness,” PDA Journal of Scienceand Technology, Vol. 60, No. 3, May-June 2006, p. 164-171.

AcknowledgementWe would like to thank Martin Doody for drawing the imagesof the SEM and the CLSM.

About the AuthorsFrank Riedewald is Process EngineeringManager of CEL International Ltd. in Cork,Ireland. CEL is a multi-disciplinary engi-neering company engaged in the pharmaceu-tical, biotechnology, and fine chemicals in-dustries. CEL has offices throughout the UKwith its head office in Coventry. In addition,CEL has offices in Cork, Ireland, and Shang-

hai, China. Riedewald holds a BE in mechanical engineeringfrom the Technische Universität Clausthal and a ME inchemical engineering from the Technische Universität Ber-lin. He is a chartered chemical engineer and a member ofISPE and PDA. He is European Chairman of the ASME BPE“Design For Sterility And Cleanability” Subcommittee. Hecan be contacted by telephone at +353-(0)21-435 7705 or e-mail: [email protected]

Aidan Sexton is Validation Manager ofMott MacDonald Pettit in Cork, Ireland.Sexton holds a BSc in industrial chemistryand a PhD in applied chemistry from theUniversity of Limerick. He is a qualifiedtrainer and has organized, developed, anddelivered a range of CPD training courses inthe pharmaceutical field. He is a member of

ISPE and is an Associate Member of the Institution ofChemical Engineers (IChemE). He is a member of theorganizing committee of the Irish Branch of the IChemE. Hecan be contacted by phone at +353-(0)21-4809800 or by e-mail at: [email protected].


Virtual Infrastructures


This articlepresents theability toconsolidatevariousregulated andnon-regulatedapplicationsonto the samephysical serverhardware, anddiscusses howsystemvirtualizationworks, benefitsand limitations,and how toqualify such aninfrastructure.

The Use of Virtual Infrastructures inPharmaceutical Manufacturing

by Adrian Wildangier and Jon Jensen

Introduction

Hardware virtualization is one of themost revolutionary achievements incomputer architecture of the past de-cade. Virtualization provides Informa-

tion Technology (IT) managers the ability toreliably consolidate multiple servers onto onephysical server. This offers significant costsavings since applications can now reside ontheir own virtual server completely isolatedfrom other applications, while sharing the un-derlying physical hardware. Additionally,virtualization also provides excellent function-ality for disaster recoverability, high availabil-ity, and system manageability. The overallpicture is a high return on investment and lowtotal cost of ownership of the server infrastruc-

ture. Hardware virtualization is changing theway people look at systems.

The question remains on how to implementthis technology in highly regulated industries,where, from a regulatory standpoint, such aparadigm shift can often be met with skepti-cism and resistance. The purpose of this articleis to provide an introduction to virtualizationtechnology, its benefits and uses, and to offersome context to address qualification in thepharmaceutical industry.

Overview of System VirtualizationIn the virtual world, physical server compo-nents are virtualized and presented to the guestoperating system (e.g., Microsoft Windows 2003or Redhat Linux) as hardware components.

Figure 1. Virtualizationsoftware vs. physicalserver.

Reprinted from







Some of these components include theprocessor, storage media, memory,video card, and network adapter. Theguest operating system seems to becommunicating directly with the physi-cal hardware, but in fact, it is commu-nicating with virtualized hardware, thevirtual machine.

The virtualization software provideseach virtual machine access to thephysical hardware resources. It thenalso is responsible for scheduling thoseresources for each virtual machine.Since the virtual machine does nothave direct access to the physical hard-ware, the virtualization software actsas the bridge between the physical andvirtual hardware.

As an example, think of two coun-tries separated by a bridge. The coun-tries regularly trade goods with eachother, but neither is allowed to walkover the bridge. For the goods to bepassed back and forth, there is a bridgekeeper that handles this task. As atrader of the resources, it does notmatter who or what is on either side ofthe bridge, and as a matter of fact, nordoes it matter which country the bridgeis attached to. The only important as-pect is that the bridge keeper has theresources required when they are re-quested by the trader. In this example,one side of the bridge is the virtualmachine asking for the resources, thebridge keeper is the virtualization soft-ware acting as the mediator and on theother side of the bridge is the physicalhardware providing the resources.

There can be multiple virtual ma-chines that reside on the same physicalserver without directly impacting orinteracting with each other. To under-stand this scenario, let us expand onthe bridge example. Instead of thebridge being a stationary object, thinkof it as being a rotating bridge. Thebridge is not bound to two countries,but instead it can rotate to link todifferent countries. Every few days thebridge repositions itself so the newcountry can request from the bridgekeeper the resources it needs. As well,each country asking for resources hascement walls, barb wire, and electricfences that prevent any interaction witheach other. All interaction has to be

done through the bridge keeper. This isa simplified example, but the rotatingbridge manages all the resources sched-uling for each country, and the coun-tries’ walls represent the isolation be-tween each virtual machine. In reality,the resource scheduling happens veryquickly, and depending on the proces-sor, it can happen up to a billion timesa second, which gives the impression ofa multi-tasking system.

The virtualization software alsoadds additional complexity when com-paring it to a single guest operatingsystem and server. Extra layers createoverhead that typically reduce appli-cation and guest operating system per-formance. The trade-off is that withthe additional complexity comes en-hanced resource management, and thebenefits will out-weigh the minimalperformance loss.

System Benefits and UsesWith virtualization software, the ap-plications and guest operating systemare encapsulated inside a virtual ma-chine which gives an organization ad-ditional system benefits. The ability tocreate virtual machines opens the doorto high disaster recoverability, avail-ability, and manageability.

Disaster recovery plans are designedto help IT managers provide businesscontinuity in the event of an organiza-tional disaster. The plan can includeinformation on what restoration pro-cesses are required in the loss of anentire datacenter to a single file. Whena disaster occurs, the expense involvedin providing business continuity canbe high. Generally, compatible equip-ment should be maintained at a standbysite for quick recovery. Withvirtualization, disaster recovery of avirtual machine is simplified.

When restoring a guest operatingsystem in a virtual environment, thereare two components that are required.The virtualization software needs to beinstalled on compatible (i.e., virtualiza-tion vendor approved) physical hard-ware and the virtual machine systemfiles (i.e., configuration and disk files)need to be restored to a location acces-sible by the virtualization software. Oncethis is done, the virtual machine can be

turned on without having to reconfigurethe guest operating system to properlyuse the new environment, as it is un-aware of any physical changes.

The requirement to install a backupagent on the virtual machine shouldnot be overlooked, but can be mitigatedin some instances. Guest operatingsystem backup agents play an impor-tant role when specific files need to berecovered from a virtual machine. Back-ing up the entire virtual machine sys-tem files will restore the entire systemstate of a guest operating system at aspecific point in time. While this isuseful for disaster recovery purposes,it can be a time consuming process ifindividual files need to be restoredfrom a guest operating system. Thebackup agent provides the ability torestore individual files, while a virtualmachine is running, mitigating alengthy restore process. Since backupagents could be an organizational ex-pense, it might be useful to have a cost/benefit discussion to understand whichbackup strategy would work for eachvirtual machine. For example, if a vir-tual machine has a large user commu-nity that frequently requests restores,then a backup agent might be useful inthis situation. On the other hand, if avirtual machine has a negligible im-pact on the user community or manu-facturing process, then this systemcould be a good candidate for a virtualmachine system file backup.

System restores can be performedon a variety of hardware without im-pacting the guest operating systemsqualification or application validationstate. Table A presents organizationalbenefits of using virtualization soft-ware.

Depending on the criteria, the abil-ity to keep a system running duringbusiness hours is an important task,especially when high availability isrequired. In a traditional non-clusteredserver environment, a system may needto be taken offline for a variety of rea-sons, some of which include:

• firmware or hardware upgrades• troubleshooting of physical compo-

nents• replacement of system hardware




Even in a high availability cluster en-vironment, there may still be the needto take a system offline due to certainsoftware/hardware compatibility is-sues. In the VMware ESX virtualizationinfrastructure, there is the ability tomigrate virtual machines betweenphysical servers while a guest operat-ing system is running. The affectedhardware can then be taken offlineduring business hours. This technol-ogy can eliminate time consuming andexpensive server outages. As well, itcan reduce the requirement to pur-chase expensive hardware and licensefor a high availability environment.

If hardware needs to be upgraded asa result of a system running low onresources, then in a traditional serverenvironment, this could mean a backupand restore, and additional documen-tation and re-qualification of the newenvironment. In a qualified virtualenvironment, there is the ability tomigrate virtual machines betweenphysical components, without the com-plex backup and restoration procedure.Also, the GxP applications and guestoperating systems are not impacted.This can mitigate the work required ina traditional server scenario. The onlyrequired process is the shut down ofthe virtual machine, the click of a mi-gration and power-on button. This re-duces the time and cost of bringing anew server environment online.

The majority of the virtual infra-structure also can be managed throughone central console. The console is flex-

ible enough so that there is the abilityto reconfigure or upgrade the proces-sor, memory, network, and disk set-tings for each of the virtual machines.Since the console only requires net-work connectivity to the virtualizationsoftware, there is the ability to up-grade virtual machines resources atremote locations hence minimizingtravel and labor expenses.

The key use of virtualization in thepharmaceutical industry is in maxi-mizing server utilization. This can beparticularly applicable for pharmacycompanies that do not have the volumeof work to justify installing regulatedapplications on dedicated servers. Forexample, a QC lab may require a serverfor a chromatographic data manage-ment, laboratory information manage-ment, and calibration management sys-tem. In a physical environment, eachof these applications would require adedicated server, whereas in a virtualenvironment, all three applicationscould be hosted in complete isolationon one physical server. Since there aremultiple virtual machines running onone physical host, unwanted costs andtasks can be eliminated. For example,there is a reduction in costs for power,cabling, and management switches.

Virtual LimitationsA business decision must be made whenperforming server consolidation andcapacity planning. Not every server inthe datacenter may be a candidate for

virtualization. A virtual infrastructuremust be properly sized and configuredto obtain optimal system performance.Virtualization overhead needs to beaccounted for when installing or mi-grating servers into a virtual infra-structure. However, existing processormanufacturers are in discussion to in-clude virtualization support for pro-cessors, and therefore, virtualizationoverhead may no longer be an issue bythe time of this publication.

When considering the use of a vir-tual machine, several criteria need tobe considered. For each guest operat-ing system required, an organizationneeds to understand hardware utiliza-tion (i.e., disk and processing), amountof concurrent connections (i.e., memoryusage per each connection), and anyspecial hardware requirements (i.e.,RS232 connections). In this event, acandidate matrix may assist in identi-fying the most and least suitable sys-tems for virtual infrastructure use oran experienced consultant or vendor.

Since virtualization is about re-sources consolidation, storage and net-work strategies need to be addressed.A storage area network or shared highcapacity storage solutions are recom-mended due to their efficiency. Thenetwork environment also should beable to handle gigabit Ethernet speedswith potential for link aggregation, suchas MAC Address or IP Address loadbalancing (via the EtherChannel orLACP protocol).

Table A. Organizational benefits of using virtualization software for disaster recovery.

Organizational Benefit Explanation

Reduced capital and • Less equipment needs to be stored offsite, as multiple virtual machines can be restored on the same physical hardware.operational expenses • Since virtual machines are not dependent on physical hardware, like-for-like processors (i.e., AMD vs. Intel) or storage controllers

(i.e., SCSI vs. HBA) are not required.

Simplified disaster recovery • Since the virtual machine hardware environment requires no modification, the qualified state of the guest operating system andprocesses validated state of the GxP application does not change.

• Depending on the style of backup done, the restore process can be quick and simple to get the entire virtual infrastructure fullyfunctional (e.g., installation of virtualization software and restore of virtual machine configuration and disk files).

Faster business continuity • The virtualization software can be installed on a wide range of compatible hardware.after a disaster • The virtual infrastructure can be pre-installed at a remote disaster recovery site ready for virtual machine restores, as the

virtualization software does not have guest operating system or application dependencies, so there is no reconfiguration requiredto host additional virtual machines.

Less documentation and • The installation of additional virtual machines hosting GxP applications does not constitute the re-qualification of the entire virtualvalidation infrastructure.

• Since the hardware isolation of the virtual machines has been tested and qualified, restored GxP applications only require minimaltesting to keep their qualified state (e.g., Confirm the hardware resources are not below a specified threshold). Installation ofadditional virtualization software and hardware to the virtual infrastructure is done through a subset of qualification tests.




Figure 2. Candidate matrix.

Reduction in CostsThere are immediate cost benefits whenrunning in a virtual environment. Firstis the ability to consolidate multiplephysical server resources. For example,the VMware ESX software has the abil-ity to consolidate 64 physical serversinto one virtual environment. With aneight-way server, this may be possible,but it would not be best business prac-tice, since hardware failure on thephysical hardware can cause down timefor a greater number of virtual ma-chines. Typically, one should stick tothe rule of four to six virtual machinesper one physical processor (this valuemay vary depending on the criterialisted in the virtual limitations sectionand the candidate matrix). A typicalmedium to enterprise server can costanywhere from $10,000 to $40,000 (allprices shown are for example purposesand will vary with hardware, vendor,and region). If five phys-ical serversare purchased at an average cost of$25,000 per server, a capital expense of$125,000 is incurred.

If the same logic is applied to thevirtual infrastructure, the only incurredcost is for the physical hardware andthe virtualization software license.Each additional virtual machine in-stallation is essentially free (guest op-erating system and software licenseand GxP validation costs are not in-cluded).3

Virtualization can have an impact

on other externalities that affect theoperation and maintenance of adatacenter. Resource consolidation hasa domino effect that trickles down toitems such as price per network port,KVA UPS power, and BTU HVAC con-sumption, to even rack real estate, andseismic costs. The overall picture in-cludes significant savings, stream-linedoperations, and efficient datacentermanagement. Virtualiza-tion can bean appealing process for all organiza-tions, especially ones that operate un-der tight budgets.

Overview of QualificationThe need to qualify the virtual envi-ronment stems from the intention thatthe system will be used to host regu-lated applications. Virtualization soft-ware runs directly on the system hard-ware and exists as a layer between theserver hardware and the server oper-ating systems. GAMP 4 categorizesoperating systems as a Category 1, andtherefore, the functions called uponwill be challenged indirectly by thefunctional testing of the application.However, since virtualization takesplace between these two layers, andvirtualization software has some dis-tinct functionality of its own, a hybridapproach to qualification is called for.It is worth noting that end users ofapplications running on virtualizedhardware will effectively have no in-teraction with the virtualization layer.

The virtualizationsoftware does notprovide any viableservice to the enduser on its own.Once thevirtualization soft-ware is qualified, avirtual machine canbe configured to hosta validated applica-tion without regula-tory impact.

Due to the tech-nical nature ofvirtualization, thevirtual environ-ments are most suit-ably managed by theIT group and are

qualified as part of the corporate infra-structure. Therefore, User Require-ment Specifications (URS) need to bedefined by the appropriate networkpersonnel that will be utilizing thefunctionality. Test cases can then bewritten based on the URS to test thesystem.

The first aspect to consider is thehardware itself and the design of aserver farm (i.e., cluster). How manyservers will be consolidated into a singlefarm? Which applications and howmany users will share the resources ofvirtualized hardware? These questionswill be addressed in the design phase ofthe project, which will ensure that ad-equate resources are allocated and allassociated system components are com-patible. The server configuration willbe documented as part of the Installa-tion Qualification (IQ). Each server inthe farm will be built with a standard-ized IQ protocol to ensure that thesoftware is installed properly and func-tions according to manufacturers’ in-structions.

Operational Qualification of thevirtualization software can now takeplace. This will include testing of theadditional functionality that was de-fined in the URS. Performing a riskanalysis on each URS will clearly illus-trate where testing should be focused.For example, one key risk area is theisolation of virtual machines; a testshould be performed to verify that thevirtual machines created are truly iso-lated and do not interact with eachother. This test may be achieved by“crashing” (i.e., blue screen a Windowsserver or Kernel panic a Linux server)one or more virtual machines runningon the same physical hardware, andverifying that there is no impact to thestate of the other virtual machines.

The point at which the end user getsinvolved in this process is when theregulated application is installed on avirtual system. This, of course, is per-formed in conjunction with IT person-nel that will need to make sure that thespecific application is compatiblewithin a virtual environment. Oncecompatibility has been established, theusual sizing exercises need to be car-ried out so that appropriate resources




can be allocated to the virtual machinebuild. This can be documented as partof the IQ associated with the applica-tion, and once completed, will providethe end user with a qualified environ-ment on which they can perform theOperational and Performance Qualifi-cation as part of the overall applicationvalidation. As per GAMP 4, the valida-tion will inherently test the underlyingoperating system, and therefore, anyfunctionality called upon from thevirtualization layer as well.

Maintenance of the qualified sys-tem must be addressed to ensure thatthe validated states of the regulatedapplications are not impacted. Anyanticipated changes should be definedin the documentation and guidanceshould be given on how the impact canaffect those changes. A few examples ofchanges could be the addition or re-moval of hardware or virtual machinesto a physical server. This could eveninclude the allocation of resources to aparticular application. These and otherchanges can be defined within the op-

Table B. The average cost of physical server hardware (all pricesshown are for example purposes and will vary with hardware,vendor, and region).

Type of Expense Equipment Cost

Capital Costs 1x Physical Server $10,000 - $40,000

Operational Costs 1x Hardware Maintenance $1,000 - $2,000 (recurring) (ea. 2 Yrs)

Maximum Cost Average for 1 Server $25,000 (1 Time) Incurred + $1,500 (ea. 2 Yrs)

Average of 5 Servers $125,000 (1 Time) + $7,500 (ea. 2 Yrs)

Table C. The average cost of virtualization software with physicalserver hardware utilizing multiple virtual machines (all pricesshown are for example purposes and will vary with hardware,vendor, and region).

Type of Expense Equipment Cost

Capital Costs 1x Physical Server $10,000 to $40,000

1x Virtualization Software $8,000License

Additional virtual machines $0(Servers)

Operational Costs 1x Hardware Maintenance $1,000 to $2,000(recurring) (ea. 2 Yrs)

1x Virtualization Software $3,000 (ea. 2 Yrs)Maintenance

Maximum Cost Average for 5 Servers $33,000 (1 Time) Incurred + 4,500 (ea. 2 Yrs)

erational SOP forthe virtual infra-structure. By pre-defining variouschanges and per-forming some pre-liminary impact as-sessments, clearguidance on how toimplement changescan be given rang-ing from formalchange control tologbook entries.This will allow forefficient operationand will lessen thechances of costlymistakes. Anotherkey area to addressis component com-patibility. The op-erational SOPshould contain a listof compatible hard-ware and supportedoperating systemsto ensure compli-ance. Setting upmonitoring param-

Sixth Edition, 2003, p.55 - 90.2. Tanenbaum, A., Structured Com-

puter Organization, Fourth Edition,1999, p. 403 - 477.

3. Oblesby, R. and S. Herold, VMwareESX Server Advanced TechnicalDesign Guide, 2005, p. 18 - 263.

4. GAMP® 4, Good Automated Manu-facturing Practice (GAMP®) Guidefor Validation of Automated Sys-tems, International Society for Phar-maceutical Engineering (ISPE),Fourth Edition, December 2001,www.ispe.org.

About the AuthorsAdrian Wildangieris a lead technical con-sultant with CFRTechnologies, Inc. Heholds a BTech. in com-puter systems, and hasworked in the informa-tion technology indus-

try for 10 years. He specializes in de-ploying enterprise systems for the regu-lated industry. He has implementedvarious networks and server architec-tures, and has been involved in thecommissioning and decommissioningof various local and remote datacentersites. He can be contacted by telephoneat: +1-604-628-2169 or by e-mail at:[email protected].

Jon Jensen is a prin-cipal consultant withCFR Technologies, Inc.He holds a BSc inchemistry and a di-ploma of technology inchemical engineering.He has more than 10

years of experience in the pharmaceu-tical industry specializing in computersystem validation and IT compliance,including implementation and valida-tion of various systems such as: LIMS,EDMS, ERP, SCADA, CDMS, and Vir-tual Infrastructures. He can be con-tacted by telephone at: +1-604-628-2169 or by e-mail at: jjenson@cfrtechnologies. biz.

CFR Technologies, Inc., 11-7460Moffatt Rd., Richmond, B.C., V6Y 3S1Canada.

eters for each virtual farm/machinealso will allow administrators toproactively stay on top of any perfor-mance issues.

Once the Virtual Infrastructure isqualified, it will provide a secure, reli-able, and scalable platform that can beefficiently managed to meet not onlybusiness demands, but also the increas-ing regulatory requirements facing theIT industry. There are substantial ben-efits when configuring applications andguest operating systems in the quali-fied environment. Some of these ben-efits provide the ability to consolidatephysical server hardware, while in-creasing resource manageability. Thisalso means a significant savings to thecapital and operational budget for theIT department. The initial investmentmade to set up and qualify a virtualinfrastructure will continue to provideexcellent long term benefits.

References1. Silberschatz, A., P. Galvin, and G.

Gagne, Operating System Concepts,


Industry Interview


Dr. Gold talksabout thechallenges of asmall companybringing a drugto clinical trialsandincorporatingadvancedmanufacturingmethods thatinvolvecooperationbetweenmanufacturerand equipmentvendors.

PHARMACEUTICAL ENGINEERING InterviewsLynn C. Gold, PhD, Vice President,Research and Process Development,Sonus Pharmaceuticals Inc.

Dr. Lynn Goldjoined Sonus in Janu-ary 2004 as VicePresident of Researchand Process Develop-ment. She has exten-sive experience withpharmaceutical de-velopment, specifi-cally with the appli-cation of emulsion for-

mulations, a major component of Sonus’ coretechnology platform. She oversees discoveryresearch, formulation development, analyticalchemistry, manufacturing, and quality controlat Sonus. Dr. Gold’s previous experience in-cludes 13 years at Fresenius Kabi (formerlyPharmacia and Upjohn) where she had operat-ing responsibility for product and businessdevelopment activities, including serving asVice President of Research and Development.She received a BS in chemistry from StateUniversity of New York at Buffalo and a PhDfrom the University of Rochester, New York.

Q Sonus Pharmaceuticals is a smallpharmaceutical company that has just

reached a milestone in its development as acompany with two oncology products in theclinic. Can you give us some background on thecompany?

A Sonus Pharmaceuticals is a West Coastpharmaceutical company that initiated

its first oncology drug candidate developmentprogram as it redefined the company in 2000.The company focus is to develop novel drugproducts that provide better therapeutic alter-

natives for cancer patients and their caregivers.Sonus does this through the people of Sonuswho bring this purpose to life by living ourvalues.

Q How does a small company bring a drugto clinical trials?

A Our dedicated group of scientists with acommitment to discovery and quality were

the primary ingredient required to get the firstidea into the clinic. In addition, a strategicdecision outsourcing various aspects of the pro-gram that were not feasible for a 50 man com-pany to manage was implemented. TheTOCOSOL® Paclitaxel development programhas now progressed to a Phase III study. Thenovel idea of using a vitamin E-based paclitaxelemulsion to reduce problems with a well-knownchemotherapeutic, Taxol, is being studied in ametastatic breast cancer trial with more than800 patients enrolled.

A second oncology drug candidate enteredPhase 1 trials in September 2006. This drugcandidate is a vitamin E based emulsion for-mulation (TOCOSOL) of a camptothecin de-rivative. The goal of the Sonus developmentprograms is to provide safe and easy to useoncology products that add value to the cancerpatient.

Q What is unique about TOCOSOL technol-ogy?

A One of the many problems encounteredin drug development is finding a tech-

nique to deliver non-water soluble drug candi-dates. Vitamin E offers a biocompatible oil thatcan support incorporation of various drug can-

by Cathy Middleberg, Wyeth Vaccines, ISPE EditorialCommittee

Reprinted from





Industry Interview


didates. Typically, these candidatesare insoluble in water. Sonus has iden-tified a variety of methods to emulsifyvitamin E with various drug candi-dates for intravenous delivery.

The TOCOSOL emulsion can beaseptically filtered prior to filling, thisaffords the benefit of not subjecting anoncolytic to terminal heat sterilization.The ability to filter sterilize this emul-sion is a result of the high energy pro-cess that produces a droplet size distri-bution of less than 200 nm.

Q Why is this process able to pro-duce a filter sterilizable emul-

sion?

A One aspect of the system is ofcourse the components of the for-

mulation with the primary excipientbeing the vitamin E. Some of the physi-cal properties of vitamin E contributeto the overall product quality. In addi-tion, the high energy homogenizationprocess incorporating the proprietarytechnology from the equipment vendoralong with optimization of the processparameters during scale-up contributeto the quality of the product producedand the reproducibility.

Q The FDA is very interested in theadvancement of pharmaceutical

manufacturing methods. What collabo-rative processes did Sonus embark onto develop your process?

A The first batches of TOCOSOLPaclitaxel were manufactured at

the 2L scale. The homogenization pro-

cess was successfully transferred andscaled up at a Contract ManufacturingOrganization (CMO) to 50L. This scaleup required cooperation between mul-tiple parties to reduce this to practicefor the first time and has been repro-duced multiple times. Sonus has dem-onstrated that this emulsion can bemanufactured under cGMPs at com-mercial scale.

Q How has working closely withthe equipment manufacturer

made an impact on the process?

A The process is a system built ofequipment supplied by multiple

vendors. Sonus was able to work withthe vendors of the equipment to makesystematic modifications as needed andthen study the impact of these changesto identify which changes were critical toproduct quality and which were not. Thiswould have been a difficult task withoutthe cooperation of all the vendors. Themodifications ranged from making the

system more pharmaceutically compat-ible and more end user friendly to mak-ing the process more efficient.

Q How will Sonus measure the suc-cess of the manufacturing pro-

gram?

A Quality and regulatory releaseof the validation batches that will

be manufactured to support the com-mercial scale with a cost effective andreproducible process will mean thatSonus has succeeded with theTOCOSOL Paclitaxel Injectable emul-sion manufacturing portion of the CMC.

Q Where is Sonus with themanufacturing program for the

TOCOSOL Camptothecin drug prod-uct, which is in Phase 1?

A The manufacturing process is inthe early development stages, but

has the benefit of building on the knowl-edge base of TOCOSOL Paclitaxel In-jectable emulsion. There will be changesto the program as we move forward, butthere is a higher level of confidence thatthis process will be robust andcommercializable based on the develop-ment history in the company.

Q Does Sonus plan to pursue othertherapeutic areas with the

TOCOSOL technology?

A Research and development atSonus is still focused on applica-

tions of TOCOSOL technology in theoncology area. Sonus continues to ex-pand its expertise in this technologyand will evaluate other therapeuticareas as time and resources permit.

The Sonus Team.

About Sonus Pharmaceuticals, Inc.Headquartered near Seattle, Sonus Pharmaceuticals is focused on the development ofdrugs that provide better therapeutic alternatives for cancer patients, includingimproved efficacy, safety and tolerability, and are more convenient to use. TheCompany’s lead product candidate is TOCOSOL Paclitaxel, a ready-to-use, injectableformulation of the widely prescribed anti-cancer drug, paclitaxel. In Phase 2 clinicaltrials, TOCOSOL Paclitaxel has demonstrated encouraging safety and anti-tumoractivity in patients with breast, ovarian, lung and bladder cancers. TOCOSOLPaclitaxel is currently in a Phase 3 pivotal trial in patients with metastatic breastcancer. Patient enrollment in the trial was completed in November 2006. Submissionof the TOCOSOL Paclitaxel New Drug Application is targeted for the end of 2007. InOctober 2005, the Company announced the signing of a development and commer-cialization agreement for TOCOSOL Paclitaxel with Schering AG. Sonus moved itssecond oncology product candidate, TOCOSOL Camptothecin (SN2310) InjectableEmulsion, into clinical development at the end of the third quarter of 2006. SN2310Injectable Emulsion is a novel camptothecin compound formulated with the Company’sproprietary TOCOSOL technology.


Filtration Technologies


This articlediscusses howthe use of thin-cake verticalcandle filtersand horizontalpressure platefilters canefficientlyremoveactivatedcarbon, metalcatalysts, andtrace insolublesfrom ActivePharmaceuticalIngredient (API)slurries.

Thin-Cake Filtration Technologies forRemoving Activated Carbon, Catalysts,and other Trace Solids from ActivePharmaceutical Ingredient (API) Slurries

by Barry A. Perlmutter

Introduction

As the pharmaceutical industry haschanged and grown since the mid-1980s, there are increasing concernsabout the safe handling of Active Phar-

maceutical Ingredients (APIs). To meet lowerexposure limits of unknown compounds and tohave batch-to-batch integrity with less opera-tor interaction, the industry’s need for newtechnologies has expanded.

In the chemical synthesis operations of apharmaceutical plant, the components include

reactors, separation equipment, drying equip-ment, containment systems, and other processsystems and utilities. The reaction step gener-ally includes hydrogenation catalysts, activatedcarbon, and other accelerators for the precipi-tation reaction. After this step, separation isrequired to remove these substances along withunreacted materials, process impurities andreaction by-products, and API residuals.

This article focuses on one area of impor-tance, which is the efficient removal of acti-vated carbon, metal catalysts, and trace insol-

ubles, such as diatomaceousearth, from API slurries. Cur-rently, most API slurries areclarified with the use ofmanual plate filters, filterpresses, bag filters, cartridgefilters, and other conventionalfilter equipment. All of theseunits require manual opera-tions for cake discharge andcleaning between batches orcampaigns, as well as sufferfrom high labor and mainte-nance costs, high disposalcosts, and the exposure of theoperators and the environmentto toxic and hazardous sol-vents and solids in addition toused and contaminated filtercloth, bag filters, and filter car-tridges.

Spinning disk filters alsoare commonly used for clarifi-cation. While these units over-

Figure 1. BHS candleshowing gas flow toexpand the filter mediasock for cake discharge.

Reprinted from







higher alloys. Within the vessel are horizontal manifoldscalled candle registers. Each candle is connected to a registerwith a positive seal to prevent bypass. Each register maycontain from one to 20 candles depending upon the filter size.The registers convey the liquid filtrate in the forward direc-tion as well as the pressure gas in the reverse direction forfilter media sock expansion. Each register is controlled withautomated valves to ensure optimum flow in both directions.Figure 2 illustrates the candle filter vessel.

Automatic Process CyclesFilling: The slurry feed enters the bottom of the filter vessel.

Filtration: The slurry is either pumped or pressurized fromthe reactor into the vessel. Cake will deposit on the outside ofthe candle; the separated filtrate will flow through the fil-trate pipe and the registers. This process continues until oneof the following conditions is achieved: maximum pressuredrop, maximum cake thickness, minimum flow, or time.

Washing: Displacement washing or recirculation washing.

Drying: Blowing gas, steam, or “shock” drying.

Heel (Falling-Film) Filtration: The liquid remaining inthe vessel cone after filtration or washing is completelyfiltered.

Cake Discharge: Gas flows sequentially through each of thecandle registers, down each of the filtrate pipes, and then isdistributed by the perforated core. The filter media sockgently expands by the gas flow and pressure allowing for cakedischarge - Figure 1. Alternatively, the cake can be dis-charged as a slurry.

Figure 2. BHS candle filter.

come some of the special handling requirements of manualfilters, they add mechanical complexity to the process as wellas special cleaning requirements. Spinning disk filters re-quire drive motors, gear box assemblies (including gear boxhousing, gear reducers, bearings, shaft bearing arrange-ment, and bushings, torque loadings, etc.), mechanical seals(either single or double with special maintenance and clean-ing), and unique installation concerns, such as center ofgravity due to the spinning plates, static and dynamic balanc-ing, and bearing lubrication and design to ensure no exposedthreads and cleanability issues and overall maintenance ofthese components.

This article discusses the use of thin-cake vertical candlefilters and horizontal pressure plate filters as alternatives tospining disk, manual, and conventional filter equipment.These new technologies are currently installed in many APIapplications and processes. The selection process as well asthe technolgies are described in the article. The articleincludes test data and case histories and concludes with adiscussion of clean-in-place operations and current GoodManufacturing Practices (cGMPs) guidelines. ANSI/ISA S88(and IEC 61512-1 in the international arena) batch processcontrol system standards also are examined for validation.Finally, factory and site acceptance testing is described.

Clarification of Slurries and Recovery of SolidsCandle filters and pressure plate filters are installed forclarification and recovery applications from liquids with lowsolids content. The candle filters are vertical candles, whilethe pressure plate filters are horizontal plates. The cakestructure as well as the process parameters determine theoptimum thin-cake technology.

Description and Operation of the Candle FilterCandle filters provide for thin-cake pressure filtration, cakewashing, drying, reslurry, and automatic discharge, as wellas heel filtration in an enclosed, pressure vessel. Units areavailable from 0.17 m2 up to 100 m2 of filter area per vessel.

Filter Candles and MediaThe filter candles (Figure 1) consist of three components:single-piece dip pipe for filtrates and gas, perforated corewith outer support tie rods, and filter sock media. The filtratepipe is the full length of the candle and ensures high liquidflow, as well as maximum distribution of the gas during cakedischarge. The perforated core can be a synthetic material,stainless steel, or higher alloys and is designed for the fullpressure of the vessel. The outer support tie rods provide foran annular space between the media and the core for a lowpressure drop operation and efficient gas expansion of thefilter media sock for cake discharge. Finally, the filter mediais a synthetic type with a clean removal efficiency to less thanone to three microns. As the cake builds up, removal efficien-cies improve to less than one micron.

Filter Vessel and Candle RegistersThe candle filter vessel is constructed of stainless steel or




Description and Operation of the Pressure PlateFilterThe pressure plate filter has similar operating characteris-tics to the candle filter. The filter design is shown in Figure3.

Automatic Process CyclesFilling: The slurry feed enters the bottom of the filter vessel.

Filtration: The slurry is pumped under pressure into thevessel or via gas pressure through the reactor. Cake willdeposit on the top of the plates. The separated filtrate willflow through the plates to the center main filtrate outlet. Thisprocess continues until one of the following conditions isachieved: maximum pressure drop, maximum cake thick-ness, minimum flow, or time.

Washing: Displacement washing or recirculation washing.

Drying: Blowing gas, steam, or “shock” drying.

Heel Filtration: The liquid remaining in the vessel coneafter filtration or washing is completely filtered.

Cake Discharge: As shown in Figure 5 on page 72, the motorson the top of the filter operate at different frequencies and theplates gently vibrate for cake discharge. The plates vibrate inthe vertical and horizontal planes and the solids are conveyedin an elliptical pattern to the outside of the vessel. Gas assisthelps in the discharge process. There are no rotating plates,gears, or bushings and mechanical seals are not required.

Figure 3. BHS pressure plate filter.

Figure 4. Filter plate.

Selection of Candle versus Pressure PlateFilter Technologies: Cake Structure and

Process ParametersThe major difference between the two technologies dependson the cake structure that is formed. Some cakes are betterhandled in the horizontal and some in the vertical.

Cake Thickness and Filtration: The candle filter is lim-ited to cake structures that can be formed to about five-20mm. The pressure plate filter can handle cakes up to 75 mm.Both units can conduct filtration up to 150 psig.

Filter Media: The candle filter uses only synthetic mediawith a clean removal efficiency from one to three micronrange and finer down to 0.5 microns. The pressure plate filteralso can use metal media. For the pressure plate filter, theclean micron range removal efficiency is also one to threemicrons and finer.

Cake Washing: If the process requires washing to removethe API from the solids, then generally the pressure platefilter is a better alternative. If washing is not as critical, thenthe candle filter may be the optimum technology for clarifica-tion and recovery.

Heel Filtration: The remaining liquid in the vessel (liquidheel) after filtration or washing can be removed from thecandle filter or pressure plate filter by circulation, heel filterin the cone of the vessel, or additional heel filter plates in thepressure plate filter.

Cake Drying: The candle filter can produce cakes withapproximately 10% moisture. This moisture level dependsupon the specific cake, but the moisture lower limit is thatmoisture just above the cake cracking point. The pressureplate filter can produce bone dry cakes.

Cake Discharge: Both designs can easily discharge mostcakes equally with no residual heel.




Clean-In-Place (CIP)/Steam-In-Place (SIP): Both unitsconduct CIP/SIP operations in identical manners by fillingand circulating cleaning fluids, while blowing gas in thereverse direction to the filtration direction, which creates aturbulent mixture or a quasi-ultrasonic cleaning effect. Thepressure plate filter further enhances this operation withplate vibration.

Typical Testing to Determine the OptimumFiltration Technology of Vertical Candle

Filter or Horizontal Plate FilterOverview of Bench Top TestingThe bench top testing is conducted using a Pocket Leaf Filter(PLF) - Figure 6. The testing will analyze cake depths,operating pressures, filter media, washing and drying effi-ciencies, and cake discharge (qualitatively, based upon expe-rience of the vendor). The PLF is used to gather the basicfiltration, washing, and drying data.

Filtration: The first optimization concerns the cake depthversus the filtration rate. Other parameters that are variedsequentially include cake depth, filtration pressure, andfilter media. Cake depths can range between six to 75 mm.

Washing: Displacement washing tests also are performed inthe PLF. Washing pressure, time, and wash ratios are opti-mized to meet final quality specifications.

Drying: Product drying in the PLF is tested by blowingambient-temperature or hot gas through the cake. The pres-sure is kept constant and gas throughput is measured versustime.

Example of a Typical Bench Top Testing ProgramBench top tests are conducted on an API/Solids Slurry usinga PLF. The tests are conducted to demonstrate that thecatalyst and impurities can be removed from the API byfiltration. The tests demonstrate that the API can be easilyfiltered and that thin-cake pressure filtration using a candlefilter or pressure plate filter would provide excellent resultsfor this application.

The following filtration options are evaluated and aresuitable for this application: candle filter and pressure platefilter.

Test PurposesThe purposes of this test were to:

• Determine if the catalyst and impurities can be separatedfrom an API. The current filtration process is with dispos-able cartridge and bag filters:- A 12,000 liters batch of slurry is filtered in approxi-

mately six hours.- The slurry contains the catalyst and impurities and is

a dark color. The filtrate is the product and should be aclear liquid.

- The cake is washed and then discharged as a slurry fordisposal.

• Determine which type of thin-cake filtration is suitable forthis process: candle filter or pressure plate filter.

• Determine the required size for the production equipment.

Test MethodsThe Pocket Leaf Filter was used to gather data and makeobservations on this product. The following information wasgathered during this test:

Figure 5. Cake discharge in the pressure plate filter.

Figure 6. Pocket leaf filter.




Figure 7. Filtration time versus feed volume.

• Filtrate Quality vs. Filter Media• Filtration Time vs. Feed Volume (Cake Height) and Filtra-

tion Pressure• Cake Height vs. Feed Volume• Media Blinding vs. Number of Runs

Test FacilitiesTests were performed in a suitable laboratory with originalslurry produced by the customer.

Test device: 400 ml Stainless Steel Pocket Leaf Filterwith 12 cm² filter area

Filter cloths: FDA-Approved Polypropylene and Tefloncloths

Temperature: - Ambient Slurry- Ambient Pocket Leaf Filter- Ambient Nitrogen for pressurizing the fil-

ter and drying

Test Data – Confidential

Test Results• Filtrate Quality - The feed slurry is a dark color and the

filtrate should be a clear liquid. An FDA-Approved Teflonmedia produced clear filtrate that contained no visiblesolids.

• Filtration Time

Filtration Time vs. Feed VolumeThe filtration time increased with the square of the feedvolume (or the cake height) as expected - Figure 7. The data

clearly demonstrate that the filtration time for smaller feedvolumes (i.e., thinner cakes) is the preferred filtration method.

The following equation can be used to predict the filtrationtime based on the data:

Equation 1: tF = a + b (V/AF)2, wheretF = the filtration time in minutesa = a constant in minutesb = a constant in minutes * m4/m6

V = the feed volume in m3

AF = the area of the filter in m2

A least squares regression of the data yields the followingconstants for Runs 1, 2, 3, and 4:

a b R2Run 1, 15 psi 0.32 minutes 27.16 minutes * m4/m6 0.982Run 2, 15 psi 0.89 minutes 69.53 minutes * m4/m6 0.975Run 3, 30 psi 1.34 minutes 15.75 minutes * m4/m6 0.983Run 4, 45 psi 1.72 minutes 9.67 minutes * m4/m6 0.992

Filtration Time vs. PressureThe filtration time varied with the inverse of the filtrationpressure, indicating that the cake is non-compressible -Figure 8. This means that pressure filtration will result in thehighest filtration rates and the smallest filter area. Thefollowing equation can be used to predict the filtration timefor a given amount of feed vs. the filtration pressure:

Equation 2: tF = a’ + b’/P, wheretF = the filtration time in minutesa’ = a constant in minutesb’ = a constant in minutes*psiP = the filtration pressure in psi

Figure 8. Filtration time versus 1/pressure.




A least squares regression of the data yields the followingconstants for the filtration time for 350 ml of feed slurry:

a’ b’ R2Runs 2, 3, 4 -0.0315 minutes 96.092 minutes * psi 0.992

Cloth BlindingThe same filter cloth was used for each trial and the timerequired for 400 ml of water to flow through the filter at 15 psiwas recorded before the trials and after each run. The flowrate through the media slowed down after the first run, andthen reached a steady rate. This indicates that a small andconsistent amount of solids remains on the cloth after eachrun. This result is normal for a cloth media (unlike a car-tridge, bag, or paper filter that tends to blind over time due tothe particles being trapped in the depth of the filter) anddemonstrates that the cloth did not blind during these trials.

Cake DischargeThe cake was discharged dry prior to drying on test Runs 1 to5, after one minute of drying on Runs six to 10, and as a slurryon Runs 13 to 20. The cake discharge was excellent in eachcase. This allows the customer to choose the discharge methodthat is best for their particular situation.

Selection of Production Technologies and Scale-UpThe data indicates that thin-cake pressure filtration is thecorrect separation method for this product. The result of thetests indicate that either a candle filter or a pressure platefilter are suitable for this application. These filters are batchdevices and the cycle time for a batch is the sum of the timesfor each step in the process.

The actual cycle time for a batch filter is the sum of:tFill = Filling Time = 5 minutestTurbid Filtration = Turbid Filtration Time = NAtF = Filtration Time = See Each Case BelowtW = Washing Time(s) = 10 minutestDrain = Draining Time = NAtDrying = Drying Time = NAtDis = Discharge Time = 5 minutes_____________________________________________________

tTotal = Total Cycle Time = Filtration Time+ 20 minutes

Option 1: Process the Entire Reactor Batch in OneDrop with a Candle FilterThe cycle time should not exceed six hours (240 minutes) sothe allowable filtration time if the entire batch is processed inone batch is 240 minutes - 20 minutes = 220 minutes. Thecandle filter operates with a typical cake thickness of 10 mm,and Equation 3 can be rearranged to determine the requiredfiltration area to process a 12,000 liter batch.

AF = b” * V/(h - a”) = 12.846 mm *m2/m3 * 12 m3/(10mm - 0.2436mm) = 15.8 m2

A Candle Filter with an area of 18.8 m2 is the inital choice forthis option. We now use Equation 1 and the data from thespecific run to confirm that the filtration time in this filter isacceptable:

tF = a + b (V/AF)2 = 1.72 minutes + 9.67 minutes * m4/m6 * (12m3/18.8 m2)2 = 5.7 minutes.

This filtration time of 5.7 minutes is much less than the 220minutes allowed, and this confirms that the candle filter islarge enough for Option 1. A smaller filter with multipledrops is also possible, but the cGMP requirement is for onecomplete batch for batch-to-batch integrity.

Option 2: Process the Entire Reactor Batch in OneDrop with a BHS Pressure Plate FilterFor this application, the pressure plate filter can operate witha cake thickness of 55 mm (maximum cake thickness = 75mm) and Equation 3 can be rearranged to determine therequired filtration area to process a 12,000 liter batch.

AF = b” * V/(h - a”) = 12.846 mm *m2/m3 * 12 m3/(55 mm - 0.2436mm) = 2.82 m2

A pressure plate filter with an area of 2.9 m2 is the initalchoice for this option. We now use Equation 1 and the datafrom the specific Run to confirm that the filtration time in thisfilter is acceptable:

tF = a + b (V/AF )2 = 1.72 minutes + 9.67 minutes * m4/m6 * (12m3/2.9 m2)2 = 167 minutes.

This filtration time of 167 minutes is less than the 220minutes allowed, and this confirms that the pressure platefilter is the correct size for Option 2.

SummaryThe testing demonstrated that thin-cake pressure filtrationtechnologies of candle or pressure plate filters with auto-matic operations can replace the bag and cartridge filtersthat are currently being used. The benefits include fullyautomatic operation, complete containment, no operator in-volvement, and low maintenance and operating costs. Thecandle filter requires 19 m2 of filter area while the pressureplate filter requires 3 m2 of filter area. Further discussion isrequired of the other process parameters to determine theappropriate choice of technology.

Typical Case HistoriesThe following are installation process details from candle andpressure plate API applications.

Application 1: Candle Filter with 6 m2 of Filter AreaIn a recent API installation, the customer installed twocandle filters for removing activated carbon and diatoma-ceous earth from a 3000 kg API slurry. The details are asfollows:




• Installation: Duplex candle filters, each with 6 m2 of filterarea

• Slurry has 25 kg of activated carbon + 15 kg of diatoma-ceous earth

• Batch size is 3000 kg of an API• Filtration Pressure = 5 bar• Drying with nitrogen to 5% final moisture• Cycle Times:

- 20 minutes of recirculation of the initial turbid filtratecontaining solids. This recirculation was necessary forproduct clarity

- 120 minutes of filtration for final production- 20 minutes for draining, drying, and cake discharge

Application 2: Candle Filter with 5 m2 of Filter Area• Installation: One candle filter with 5 m2 of filter area for

de-colorization of an API• Slurry has 8 Kg activated carbon + 7 Kg of diatomaceous

earth.• Batch size is 4,000 liters of slurry.• Filtration pressure is 2-6 bar.• Cycle Times:

- Filtration = 30 minutes with a cake depth of 20 millime-ters

- Drying = 5 minutes- Discharge = 10 minutes

• Process Details:- Production results are identical to pocket leaf filter

tests of 1200 liters/m2/hour- Cake discharge is 100%; no residual heel.

Application 3: Pressure Plate FilterIn this next application, a pressure plate filter was chosenrather than a candle filter. The API is bound to the activatedcarbon so intense washing is required. The benefit of thevibrating plates allowed the solvent and carbon to mix in thevessel and then reset the bed by filtration. The horizontalplates provided for a well-defined cake structure for cakewashing and then bone-dry cake discharge. The details are asfollows:

• Installation: Pressure plate filter with 10 m2 of filter area• Slurry has 100 kg of activated carbon.• Batch size is 6500 kg of an API.• Filtration Pressure = 3-5 bar.• Drying with nitrogen to less than 10% final moisture.• Cycle Times:

- 4 minutes of recirculation of the initial turbid filtratecontaining solids. This recirculation was necessary forproduct clarity

- 120 minutes of filtration for final production- 20 minutes for draining, drying, and cake discharge

Typical Installation Drawings and P and IDsCandle FilterThe question for API installations generally focuses on con-tainment. Both the candle and pressure plate filters can be

installed in a full glove box or only with a glove box at the cakedischarge flange. Figure 9 illustrates a typical full glove boxinstallation for a candle filter.

Pressure Plate FilterFigure 10 on page 78 illustrates a typical installation wherecontainment is not critical and the cake be discharged intoopen totes. The polishing filters produce “absolute-rated”final product quality for the downstream operations.

Process Controls and TestingWhen evaluating an API application, bench top testing is thefirst step. After this step, the project continues throughvarious stages until completion. The final step before the newequipment and systems are sent to their final destination isthe Factory Acceptance Test (FAT). The FAT may includePLC testing, CIP tests using riboflavin testing, swab tests,and other client specified tests.

Candle filters and pressure plate filters must be con-trolled either through a local Programmable Logic Control-ler (PLC) or a Distributed Control System (DCS). For PLCsystems, the Batch-S88 standards allows for modular con-trol system operation. It defines the process operations,including the cleaning operations in discrete and individualmodules or steps so that operators can perform certain tasksreliably and without variation to ensure a unit that isoperated correctly and produces reproducible batches aswell as is defined “as clean.” A typical PLC sequence wouldbe as follows:

Figure 9. Candle filter with cGMP candles installed in a glove box(not all nozzles are shown; glove box designs require fullcustomization).




• Idle - in semi-automatic/automatic mode the sequence isnot running

• Running - main operation is running without any activesequences

• Typical sequences of operation:- Nitrogen purging- setup- filling- wash 1- wash 2- blowing - drying- discharging- clean-in-place

• Typical batch outcomes:- cycle completed and the sequences begin again- aborted and corrective actions are required- holding in a sequence and waiting for an action- restarting a sequence from a Holding position- grounding to put the filter in its fail-safe position at the

end of the reactor batch

After successful completion of the FAT, the unit can beshipped. Site acceptance tests as well as the IQ/OQ tests arethen conducted using the S-88 standards and formats andrepeating the procedures, as necessary.

Figure 10. BHS pressure plate filter for cGMP production.

SummaryThin-cake filtration operations provide many benefits to theproduction/clarification process. By selecting the optimumthin-cake technology of candle filter or pressure plate filter,engineers can realize a more efficient process approach,including solids handling and cleaning of equipment withminimal operator involvement for improved safety and envi-ronmental concerns.

References1. Bryan, Sonja, “Cut the FAT: Plan Early to Streamline

Your Next Factory Acceptance Test,” Pharma Manufac-turing, (2003): 18-28.

2. Forsyth, Richard J., “Cleaning Validation of ActivePharmceutical Ingredients,” American PharmaceuticalOutsourcing, (2006): 16-23.

3. Perlmutter, Barry A., “Test It Right,” Chemical Process-ing, (2003): 29-31.

4. Perlmutter, Barry A., “Clean-In-Place Operations for Thin-Cake Filtration Technologies,” PharmaChem, (2005): 6-9.

About the AuthorBarry A. Perlmutter is currently Presidentand Managing Director of BHS-FiltrationInc., a subsidiary of BHS-Sonthofen GmbH.BHS is a manufacturer of thin-cake filtra-tion, washing, and drying technologies.Perlmutter has more than 24 years of techni-cal engineering and business marketing ex-perience in the field of solid-liquid separa-

tion, including filtration and centrifugation and process dry-ing. He has published and lectured extensively worldwide onthe theory and applications for the chemical, pharmaceuti-cal, and energy/environmental industries, and has been re-sponsible for introducing many European companies andtechnologies into the marketplace. Perlmutter began hiscareer with the US Environmental Protection Agency. Hereceived a BS in chemistry from Albany State (NY) Univer-sity, MS from the School of Engineering at WashingtonUniversity, St. Louis, and an MBA from the University ofIllinois. He serves on the Board of Directors of the AmericanFiltration and Separations Society (AFS) and is a member ofISPE. He can be reached by e-mail at: [email protected].

BHS-Filtration Inc., 9123-115 Monroe Rd., Charlotte,North Carolina 28270.


Global Regulatory News


InternationalThe International Conference onHarmonisation (ICH)1 Steering Com-mittee plus the expert working groupsmet in Chicago, Illinois from 21 to 26October 2006. Quality experts from thechemical and biotech areas held Qual-ity Strategy sessions to identify areasto be addressed.

PIC/SThe Pharmaceutical InspectionConvention and the Pharmaceuti-cal Inspection Cooperation Scheme(PIC/S)2 issued a revision of the PIC/SGMP Guide, which came in to force inAugust 2006. The revision was made inparallel with the EU GMP guide toinclude reference to counterfeiting.

ArgentinaIt has been reported3 that Argentina’sregulatory agency, Anmat, has pub-lished a new regulation dealing withgood bioavailability/bioequivalencestudy practices.

AustraliaThe following items were added to theTherapeutic Goods Administration(TGA) Web site4 in October/November2006:

• TGA news issue 51 November 2006,including an update on the consul-tation concerning the Australia NewZealand Therapeutic Products Au-thority (ANZTPA) and details of therelease of the next phase of consul-tation documents on 18 October 2006with submission to be made by 6December 2006. Phase III of theConsultation will start in March2007. Also reported were thechanges to the fees and payments, areminder that medical devices mustbe transitioned to the new medicaldevice regulatory system prior to 4October 2007, information that 11new EU guidelines have beenadopted by the TGA since the lastissue.

• a listing of Australian Manufactur-ers licensed to manufacture thera-peutic goods

• guidelines on the GMP clearance ofoverseas medicine manufacturers

• the release of the second phase ofconsultation on the proposed trans-Tasman regulatory scheme fortherapeutic products

• reminder of 2006/2007 annualcharges and the installment datesfor quarterly payments

• EU guidelines adopted, includingGuidelines on the PharmaceuticalQuality of Inhalation and NasalProducts (EMEA/CHMP/QWP/49313/2005)

CanadaAdded to the Health Canada Drugsand Therapeutic Products Direc-torate (TPD) Web site5 was the News-letter giving details of a memorandumof understanding that is being devel-oped between the TPD and Australia’sTGA to allow for mutual recognition ofquality management system certifica-tions for medical device manufactur-ers in Australia, New Zealand, andCanada to promote regulatory coop-eration. Also included were details of aClinical Trials Manual issued to pro-vide guidance for the filing of a ClinicalTrial Application (CTA) in Canada.

IsraelIt has been reported6 that the Pharma-ceutical Administration in Israel hasissued a circular in August 2006 relat-ing to accelerated approvals for me-dicinal products that have already beenapproved by the US FDA or recom-mended for approval by the EuropeanMedicines Agency (EMEA). This ap-plies to medicinal products containingnew chemical entities, biological me-dicinal products, and additional indi-cations to products already registeredin Israel.

A guideline has been issued whichapplied to submissions filed from 1 Oc-tober 2006 detailing the requirementsand the procedure for the approval ofbrand names for medical products.7

PakistanA new autonomous regulatory author-ity is to be established in Pakistan toregulate medicines and medical de-vices.8 This is expected to streamlineregistration procedures and ensurehigh quality standards.

EuropeReported on the Web site for the Euro-pean Medicines Agency (EMEA)9

in October/November 2006 were:

• the Committee for Medicinal Prod-ucts for Human Use (CHMP)monthly report10 from the October2006 Plenary meeting held 16 to 18October, and the monthly reportfrom the September meeting held18 to 21 September 200611

In view of the anticipated enlargementof the EU with Bulgaria and Romaniaon 1 January 2007, Marketing Autho-rization Holders and Applicants areadvised that they should provide Mod-ules 1-2 of pending applications to thecontact points in the new MemberStates.

Documents prepared by theBiologics Working Party shown belowwere adopted at the September CHMPmeeting.

• guideline on validation of immu-noassay for the detection of anti-body to human immunodeficiencyvirus (Anti-HIV) in plasma pools(CHMP/BWP/298388/2005)

• overview of comments received ondraft guideline on validation of im-munoassay for the detection of anti-body to human immunodeficiencyvirus (Anti-HIV) in plasma pools(CHMP/BWP/94182/2006)

• guideline on validation of immu-noassay for the detection of hepati-tis B virus surface antigen (HBsAg)in Plasma Pools (CHMP/BWP/298390/2005)

• overview of comments received ondraft guideline on validation of im-munoassay for the detection of hepa-titis B virus surface antigen (HbsAg)in Plasma Pools (CHMP/BWP/94181/2006)

• Also released for six months reviewwas the Guideline on the Quality ofBiological Active substances pro-duced by stable Transgene Expres-sion in Higher Plants (CHMP/BWP/48316/2006).

The CHMP October Plenary meetingheld from 16 to 18 October 2006 adopted

Reprinted from







the following documents which werereleased for three months consulta-tion:

• Reflection Paper on Pharmacoge-nomic Samples, Testing and DataHandling (EMEA/20194/2006)

• Guideline on Excipients in the Dos-sier for application for MarketingAuthorization of a Medicinal Prod-uct (CHMP/QWP/396951/2006) (re-released)

The EMEA Web site12 also has beenupdated with:

• details of the Qualified Person Fo-rum being held on 30 November2006 in Prague and the ISPE/PDAworkshop on 6 to 7 December 2006on Challenges of Implementing ICHQ8 and Q9

• the CHMP QWP Draft Guideline onExcipients in the Dossier for Appli-cation for Marketing Authorizationof a Medicinal Product

• details of the Joint CHMP/CVMPQuality Working Party (QWP) WorkProgram for 2007 including the fi-nalization of the revision of the Notefor Guidance on the use of the NearInfrared Spectroscopy (CPMP/QWP/3309/01) to take account of advancesin this area

• Guideline on Parametric Release(Veterinary) adopted by CVMP on12 October 2006 (EMEA/CVMP/QWP/339588/2005)

The Heads of Agencies13 Web sitehas been updated with reports fromthe CMD(h) meetings held 16 to 18October 2006 and 18 to 19 September2006.

The CMD(h) also has agreed anupdated QRD template for MutualRecognition and Decentralized Proce-dures to address in the labeling – infor-mation in Braille.

The CMD(h) also has agreed that inexceptional cases it should be possibleto validate a decentralized procedureapplication, where an inspection of sitesoutside of the EU has not yet beencarried out. The manufacturing autho-rization must be available for the re-start of the procedure on Day 106.

A Question and Answer documenton the implementation of the new leg-islation has been provided which out-lines practical considerations concern-ing the phasing in of Directive 2004/27/EC amending Directive 2001/83/EC.

The European Commission DGEnterprise14 Web site has publishedan up to date list of substances consid-ered as not falling within the scope ofCouncil Regulation (EEC) No 2377/90.Also added was Notice to ApplicantsVolume 2C – Draft “Guideline on thereadability of the label and packageleaflet of medicinal products for hu-man use” update 2006. The updatesinclude specific recommendations forblind and partially sighted patientsand guidance concerning consultationwith target patient groups.

The European Directorate forthe Quality of Medicines (EDQM)15

Web site was updated with details of2007 Proficiency Testing Studies forOfficial Medicines Control Laborato-ries and a press release issued follow-ing the last Pharmacopoeial Discus-sion Group Meeting. Five new Ph Eurbiological reference preparations alsowere added. The Certification Proce-dure section has been updated withtwo new pages launched for TechnicalAdvice and Inspections. The 2007 tech-nical advice dates also have been added.

Czech RepublicThe State Institute for Drug Con-trol (SUKL) has published16 on itsWeb site: the SUKL Bulletins (9/2006and 10/2006), including guidelines validas of 1 November 2006, information forholders of authorization for manufac-ture of medicinal products and controllaboratories, list of medicinal productswhose marketing authorization willexpire in November and December2006, VYR-27 Version 1 concerningApplication for an authorization/change to authorization for manufac-ture of medicinal products, and guid-ance for provision of detailed informa-tion on manufacture, including newapplication forms resulting fromamended legislation. Also added wereupdated lists of distributors and manu-facturers.

DenmarkThe Danish Medicines Agency hasadded the following to its Web site17 inOctober/November 2006:

• Guideline on the Sunset Clause re-lating to the notification about ini-tiation or cessation of marketing ofmedicinal products. (According toarticle 23a of the Directive on me-dicinal products for human use(2001/83/EC) and article 27a of theDirective on veterinary medicinalproducts (2001/82/EC)). A new noti-fication form also is available.

LithuaniaThe State Medicines ControlAgency has included on its Web site18

information for Marketing Authoriza-tion Holders concerning the submis-sion of upgraded documentation. Ac-cording to the Accession Treaty, whenLithuania joined the EU, the transi-tional period for upgraded MarketingAuthorizations ended on 1 January2007.

MaltaAdded to the Web site of the Medi-cines Authority in Malta19 was up-dated information concerning fees forMarketing Authorization Applicationsand Variations which came in to force6 October 2006.

Also now available on this Web siteare Package Leaflets and Summary ofProduct Characteristics of authorizedproducts.

NetherlandsThe Web site of the Medicines Evalu-ation Board (MEB) has been up-dated20 to include details of an agree-ment signed between the MEB, theNetherlands Health Care Inspectorate(IGZ) and the US FDA for the exchangeof confidential information; the News-letter for Marketing AuthorizationHolders and a reminder that annualfees for 2007 are due for all productsregistered on 1 January 2007. TheMedicines Data Bank on the MEB Website now includes the possibility tosearch for recently updated packageleaflets and Summary of ProductsCharacteristics (SPCs).




SwedenThe Medical Products Agency Website has been updated21 on 1 Novemberto confirm that applications for mar-keting authorizations should be in com-pliance with current Directives andRegulations implemented in Swedishlegislation, including multiple appli-cations or duplicates.

References1. ICH - http://www.ich.org/cache/

compo/276-254-1.html2. PIC/S - http://www.picscheme.org/

index.php3. RAJ Pharma, Vol. 17, No. 10, Octo-

ber 2006, p. 660.4. TGA - http://www.tga.gov.au/me-

dia/index.htm5. TPD - http://www.hc-sc.gc.ca/dhp-

mps/prodpharma/index_e.html6. RAJ Pharma, Vol. 17, No. 9, Sep-

tember 2006, p. 608.7. RAJ Pharma, Vol. 17, No. 10, Octo-

ber 2006, p. 674.8. RAJ Pharma, Vol. 17, No. 11, No-

vember 2006, p. 756.9. EMEA - http://www.emea.eu.int/

whatsnewp.htm10. EMEA/CHMP/410526/2006 - 27

October 2006.11. EMEA/CHMP/325128/2006 - 29

September 2006.12. EMEA - http://www.emea.eu.int/

Press%20Office/presshome.htm#13. HOA - http://heads.medagencies.

org/14. EC - http://ec.europa.eu/enterprise/

pharmaceuticals/index_en.htm15. EDQM - http://www.pheur.org/16. SUKL (Czech) - http://www.sukl.cz/

enindex.htm17. DMA - http://www.dkma.dk/1024/

visUKLSArtikel.asp?artikelID=74618. SMCA - http://www.vvkt.lt/engl/19. MA - http://www.medicinesauthority.

gov.mt/news&events.htm20. MEB-CBG - http://www.cbg-meb.nl/

index.htm21. MPA - http://www.lakemedelsverket.

se/Tpl/StartPage____395.aspx

This information was provided by KarenFlatt, Pharmaceutical Research Asso-ciates (UK).


New Products and Literature


Safety Manager SystemHoneywell has announced the releaseof the newest version of its safety in-strumented system (SIS), HoneywellSafety Manager Release 110. SafetyManager is a component of theExperion® Process Knowledge System(PKS), Honeywell’s process control andautomation platform that ties togethercritical subsystems within a manufac-turing facility to give operators a bet-ter view of how processes are function-ing. The latest version includes newintegration capabilities with systemssuch as process simulation tools andfire and gas detectors, as well as a dataexchange and control communicationprotocol that will allow safety engi-neers to design and build large inte-grated and distributed plant-widesafety strategies.

Honeywell International Inc.,www.honeywell.com.

Fermentor Feed Pump

Watson-Marlow Bredel, a leadingmanufacturer of peristaltic pumps,announces its 520 Series dispensingpump for accurate metering, dosing,and transferring of fermentor nutri-ents in sanitary environments. Thepump is ideal for handling huge flowfluctuations where nutrient require-ments vary during the organisms’growth profile. The 520 Series acceptseight different tubing materials andsizes up to 9.6mm for flow rates rang-ing from 4 microliters/min up to 3.5liters/min and pressures up to 100 psi.

Watson-Marlow Bredel Inc.,www.watson-marlow.com.

Pure Water BrochureA new brochure from Veolia WaterSolutions & Technologies provides aconcise discussion of water purifica-tion in the pharmaceutical industry,and introduces a range of solutionsspecifically designed to meet, reliablyand cost effectively, that industry’s

uniquely demanding requirements. Il-lustrated throughout with photographsand diagrams, the brochure discussesthe production of purified water, highlypurified water, pyrogen free water andWFI as critical processes, and lists thekey standards with which water treat-ment plants for use in the pharmaceu-tical industry must comply.

Veolia Water Solutions & Technolo-gies, www.elgaprocesswater.com.

Process Container FilmMillipore Corp. has announced a newmultilayer single-use process containerfilm with exceptional clarity and ro-bustness. Disposable manufacturing isan industry trend that offers impor-tant advantages, including reducedcleaning and validation requirements,increased flexibility, and lower risk ofcross contamination. PureFlex film willenable customers to have better vis-ibility into their process container whilealso providing enhanced containmentof their product.

Millipore Corp., www.millipore.com.

Diagnostics SuiteEmerson Process Management hasannounced the availability of AdvancedDiagnostics on its industry leadingRosemount 3051S Series of Instrumen-tation for pressure, flow and level solu-tions using the HART communicationsprotocol. The ASP™ Diagnostics Suite,embedded in Emerson’s scalable, high-est performing, and most reliable pres-sure transmitter, provides users withnew tools for troubleshooting, detect-ing, and preventing abnormal situa-tions. Patented statistical process moni-toring technology, integral to the trans-mitter, provides users with an earlywarning of abnormal process or equip-ment conditions such as plugged im-pulse lines, changes in fluid composi-tion, or other events signaled by achange in the noise characteristics ofthe pressure signal.

Emerson Process Management,www.emersonprocess.com.

Condition ManagerInvensys has introduced a new intelli-gent real-time condition managementcomponent for the company’s InFusion

enterprise control system. TheInFusion Condition Manager buildsupon Invensys’ award-winningAvantis.CM technology to provide theability to collect real-time conditiondata from an even broader range ofplant data sources and to interoperatewith the full range of Invensys andthird-party applications supportedthrough the InFusion application en-vironment. The InFusion ConditionManager provides powerful tools foranalyzing and contextualizing data andapplies an expert rule set that firsttriggers and then manages the appro-priate operations, engineering, or main-tenance actions.

Invensys plc, www.invensys.com.

Tubing

NewAge Industries has announced theavailability of PTFE, FEP, and PFAfluoropolymer tubing (often referred toas Teflon® tubing) in six styles: straight,thin wall, coiled, corrugated, convo-luted, and stainless steel over-braided. A large inventory is stockedfor applications involving all typesof fluid and air transfer, includingchemical, pure water, food and bever-age, pharmaceuticals, adhesives, medi-cal and laboratory use, robotics, andothers. Fluoropolymer tubing is bestknown for its non-stick properties andheat resistance.

NewAge® Industries Inc., www.newageindustries.com.

To submit material for publicationin Pharmaceutical Engineering's

New Products and Literatureor Industry and Peopledepartments, e-mailpress releases with

photos to [email protected] consideration.

Reprinted from





Industry and People


Extract Technology NameRevived Under New

Ownership

The US-based private investment firm,Insight Equity, acquired the WalkerGroup, resulting in the UK-based manu-facturer of pharmaceutical equipmentto revert to its former name, ExtractTechnology Limited. David Pallister hasbeen chosen to lead the company.

Extract Technology, www.extract-technology.com.

GlaxoSmithKlineto Acquire Domantis

GlaxoSmithKline (GSK) plc has an-nounced that it has entered into an agree-ment to acquire Domantis Ltd, a leaderin developing the next generation of an-tibody therapies, for £230 million (US$454 million) in cash. Domantis, a pri-vately owned company, will become partof GSK’s Biopharmaceuticals Centre ofExcellence for Drug Discovery while con-tinuing to operate from laboratories inCambridge, UK. The acquisition agree-ment is subject to clearance under theHart-Scott-Rodino Antitrust Improve-ments Act and is expected to complete inJanuary.

GlaxoSmithKline plc, www.gsk.com.Domantis Ltd., www.domantis.com.

New Scholarship Programat Sartorius

Sartorius AG has announced its new“Sartorius International BiosciencesScholarship.” The scholarship will ini-tially offer ten positions to students ofupper semesters, those who have al-ready earned a degree in the naturalsciences and/or technology (such asbioprocess engineering, biochemistryor sensorics), and to those who wouldlike to gain experience working at aGerman technology group. Integrated

into global teams, the scholarship pro-gram participants are to work on se-lected projects in research and devel-opment as well as product marketingin the Biotechnology Division.

Sartorius AG, www.sartorius.com.

New Company to PromoteLondon’s Genetic

Research CapabilityLondon Genetics, a specialist agencycreated to facilitate partnerships be-tween industry and academic centresof excellence in genetics and genomics-based research across London, waslaunched 12 December at the Genesismeeting. London Genetics is a uniqueorganization that generates and man-ages partnerships between leading aca-demic and clinical researchers and thebiotech and pharmaceutical industries.

London Genetics, www.londongeneticslimited.co.uk.

Reprinted from





ISPE Update


ISPE Student Chapters – Paving the Way for the Society's Futureby Rochelle Runas and Phillip Pursifull

On a February evening in NewJersey in 1992, they broke breadand passed the pasta to celebrate.

ISPE’s North Jersey Chapter boardhad just handed 14 students from theNew Jersey Institute of Technology acharter certificate inaugurating themas Members of ISPE’s first studentchapter.

The historic moment marked thebeginning of what would later becomea Society program that today, boastsmore than 40 active Student Chaptersacross the globe with more than 1,000Student Members.

Recognizing that students are thefuture of the industry, one of ISPE’sgoals is to develop students into com-petent pharmaceutical professionalsand encourage them to pursue phar-maceutical careers. To highlight theefforts and initiatives that are part ofthis endeavor and to encourage Society

participation,Pharmaceuti-cal Engineeringwill profile ac-tive ISPE Stu-dent Chapters.The profileswill shine aspotlight on thepeople behindthe scenes whohave given andcontinue to givetheir time andenergy shapingthe next gen-eration of industry professionals.

No Such Thing as aPharmaceutical Engineer

When ISPE Members Joe Manfrediand Bob Lechich first approached theNew Jersey Institute of Technology

It’s More than Just a PosterIf you attended any of the past few Annual Meetings, it was hard not to missthe sight of students standing nervously in front of their huge poster displays.

The ISPE Student Poster Competition is a high-pressure event where eachstudent gives a five to 10 minute summary of their research, using a detailedposter to illustrate advanced and complex information for the judges. Thejudges add to the tension by quickly firing questions at the delegates to seehow well they know and can communicate their subject matter.

It is common to hear students say that the poster competition does morethan help them get real-world feedback from perusing Industry Members.

“I got my job when I was participating in the 2004 poster contest,” said IyadAl-Rabadi. “The vice president of my current company, Austin AECOM,stopped by my poster and asked me to give him a presentation about my poster

without realizing thathe was actually inter-viewing me. After thepresentation, he askedme if I would be inter-ested in a position as aprocess engineer. Onemonth after that, Iwent to AustinAECOM for an inter-view after which I gotmy current job!”

(NJIT) with the idea of forming anISPE Student Chapter, it drew a re-sponse that typified the pharmaceuti-cal industry’s place in academia at thattime. “They looked at me like I hadthree heads and asked ‘What is ISPE?’There was no such thing as a pharma-ceutical engineer,” said Manfredi.

Manfredi and Jon Tomson, who vol-unteered to form the second ISPE Stu-dent Chapter at Rutgers, The StateUniversity of New Jersey, were facedwith the challenge of planting a seedwhere there was no soil. Like manyuniversities at the time, departmentsof chemical engineering, mechanicalengineering, and industrial engineer-ing existed, but not pharmaceuticalengineering.

“The colleges did not have a compel-ling reason to be a part of ISPE,” saidISPE Founding Member Jim O’Brien,who has been involved in student de-velopment initiatives since their earlystages.

“After we began the early StudentChapters, we realized that many uni-versities did not offer courses specifi-cally targeted at the industry, so thenwe began to work with them to developcurricula, “ said ISPE Chairman JaneBrown, who helped start-up seven Stu-dent Chapters in the Carolina—SouthAtlantic Chapter (CASA).

The NJIT Student Chapter found

From left to right: Industry/ISPE Advisor Bob Lechich, Jim Ansbro, TraceyChodkwisz, Justin Mahalik, Chapter Advisor Demetri Petrides, Brian Huie,Rick Salsa, Larry Perfetto, James Carloni, Richard Lojek, and SteveWehner at the first student chapter charter ceremony at NJIT in 1992.

Concludes on page 2.

Reprinted from





ISPE Update


ISPE Student Chapters...Continued from page 1.

its home in the chemical engineeringdepartment in 1992. Not until a decadelater did NJIT establish its own phar-maceutical engineering program,Manfredi said.

An Eye OpenerThe formation of a Student Chapter isa grassroots effort between a univer-sity and local ISPE Affiliate or Chap-ter. Each Student Chapter is spon-sored by a local affiliate or chapter,which along with industry and facultyadvisors, provide instrumental devel-opment and support.

Currently, there are 41 active Stu-dent Chapters, 10 inactive chaptersawaiting new leadership, 11 chapterscurrently under formation, and twowith ongoing educational programs.The North American Chapters supportthe majority of the Student Chapters,with ISPE Affiliates sponsoring eightStudent Chapters in Germany, Ireland,Poland, Sweden, United Kingdom, andSingapore. The Singapore StudentChapter comprises students from fiveuniversities. The first student activityin France will take place this spring inParis.

Approximately 1,022 Student Mem-bers participate in local Student Chap-ter activities, as well as national con-ferences, poster competitions, studentleadership forums, and the ISPE An-nual Meeting.

By participating in a Student Chap-ter, students have access to network-ing opportunities, mentors (who canbecome their future employers), indus-try facility tours, professional develop-ment, interviewing practice, and otherbenefits that traditionally would notbe so readily available.

Students, as undergrads, only getwhat is available to them by way ofeducation, co-ops and internships, butthe paths to working in the pharmaceu-tical industry aren’t easily seen, accord-ing to Manfredi, who teaches in thepharmaceutical engineering Master’sprogram at NJIT. “Most students, inthe past, thought that with a degree in

chemical engineering, you could workin the petroleum and drug industries,”he said. “There was no differentiation.ISPE opened the eyes for a lot of stu-dents to careers in health care.”

“Joe, Jon and Jane Brown, and manyother visionary ISPE Members whofollowed, had as their objectives to de-fine pharmaceutical engineering, es-pecially as it related to career opportu-nities, and to develop a joint programbetween industry and academia for joband technology development,” saidConnie Muia, retired ISPE Director ofChapters and Affiliates who helpedexpand the Student Chapter program.

“There are many former StudentMembers who are employed in the in-dustry and serving in leadership rolesat ISPE,” she said.

Student MembershipFee: $15

Shaping the Future of theIndustry: Priceless

By the numbers, it’s obvious that Stu-dent Members aren’t exactly the mostinfluential constituency in the Societyor in the Industry. However, accordingto ISPE Members and staff, they willbe very soon, and so they are too impor-tant to ignore.

“They are the future of ISPE andour industry,” said Lynne Richards,ISPE Director of Affiliate and ChapterRelations. “The return on investmentis demonstrated in the multitude ofsuccess stories from past Student Mem-bers who credit ISPE for its role inopening doors to unmatched opportu-nities that helped define the profes-sionals they are today.”

“I became involved because I wantedto provide students with opportunitiesto make career-choice decisions thatweren’t available when I was in col-lege,” said Brown. “As I worked withthe students and helped them main-tain their Student Chapters, I alwayscame away feeling energized and ex-cited because of their eagerness to learnabout our industry and because of theirpositive attitudes.”

“I also realized how much I learnedfrom them each time I went to a Stu-dent Chapter meeting. It was a greatway to give something back to the So-ciety and made me feel like I was help-ing contribute to successful futures forthe students.”

Room to GrowOn average, ISPE establishes aboutone to two Student Chapters a year,and in the last year, chapter forma-tions increased 10 percent, accordingto Paul Lammers, Chairman, and BoCrouse-Feuerhelm, Past Chairman ofthe ISPE Student Development Com-mittee (SDC).

Together with the University Rela-tions Committee (URC), ISPE is look-ing at the possibility of implementingmore initiatives with elementary andhigh schools and developing an inter-national internship program, as wellas a global student leadership insti-tute.

Thanks to a handful of committedISPE volunteers in the early ‘90s, theStudent Chapter program is a world-wide success. The Society is seekingmore committed volunteers to continuecarrying the torch.

“I think what we do is very reward-ing, and entertaining, both personallyand professionally,” said Lammers.“But we don’t have enough Members inour committee. We’re looking for peoplewho have a passion for students andhave time to commit.”

Special thanks to Connie Muia for hercontribution to this article.

For more information on how tohelp shape the next generation of

industry professionals byjoining the SDC, visit

www.ispe.org/getinvolved.

If you’re a current or prospectiveStudent Member and would likemore information on what your

future could look like in theindustry with ISPE, visit

www.ispe.org/campusconnection.


ISPE Update


Dave DiProspero Honored with Max Seales YonkerMember of the Year Award

Dave DiProspero was awarded the Max Seales YonkerMember of the Year Award at the 2006 ISPE AnnualMeeting. This prestigious award honors the ISPE Mem-

ber who has made the most significant contribution to theSociety during the past year.

DiProspero has been an ISPE Member since 1994 and hasbeen heavily involved in all aspects of ISPE’s educationaldevelopment for several years, most recently as the Chair ofthe newly formed Training Advisory Committee. This pastyear he guided the initiative of the Certification TrainingTask Team to perform a gap analysis of programs availablein the ISPE library compared to the competencies requiredwithin ISPE’s new Professional Certification.

In this interview, he reviews his career, what is importantin life, and his years volunteering with ISPE.

How did you come to join ISPE?I joined ISPE in 1994 based on the recommendation of a goodfriend of mine, already involved in the Society. I had beenworking in the pharmaceutical industry for some time andknew of ISPE but was never a Member. I was told that ISPEwas a great place to expand your knowledge, through thevarious educational offerings, as well as the place to networkand meet others in the industry. With this in mind I signedup and took my first training course. I was quite satisfied.

Over the next few years I attended various meetings,events, and educational programs, without too much addi-tional involvement. In the late 1990s I was approached to be aspeaker at one of the conference programs. Reluctantly, Iaccepted and was introduced to a few of the volunteer leaders.I found these to include some sharp and dynamic people. The

session went very well and I was quite impressed with theinside working of the organization. This prompted me to reachout and see what other involvement options were available. Imade contact with Bob Chew who found a place for me on theNorth American Education Committee, where I served for acouple of years. It wasn’t long after that I got even moreinvolved with ISPE, via authoring technical articles and be-coming a conference leader for the Oral Solid Dosage (OSD)Tampa session. ISPE then gave me an opportunity to Chair aneducational committee and become part of the InternationalExecutive Education committee. I’ve been involved at thislevel for several years now and am honored to be part of suchan outstanding Society and dedicated group of individuals.

What is your title and what does your job entail?I am a Principal in Stantec’s Bio/Pharmaceuticals Group,Stantec Consulting Group, since 1993. Prior employers haveincluded IEDCO, Glatt Air Techniques, Matcon USA,Teledyne, and Lightnin Mixer. Stantec Consulting Group isa large full-service architectural and engineering firm, serv-ing a wide range of industries, including life sciences, whichis my particular area of responsibility. I am a senior leaderwithin the bio/pharmaceutical practice area and am typicallyinvolved in general consultation, front-end planning,strategizing and project set-up/implementation for our cli-ents. This includes big pharma, generics, contract manufac-turers, biotechs, and other firms.

My specific area of expertise is in Oral Solid Dosage Formsincluding the various aspects of OSD process, equipment, andfacility layout. Additionally, I have business developmentresponsibilities and work closely with our various regionalleaders to help grow the pharma side of our business. Overthe past couple of years, a rather large focus of my efforts hasbeen in Puerto Rico where Stantec has built a strong localpresence. Most recently I have been involved in developingand growing the business in the US Mid-Atlantic and South-east, as well as the West Coast, where Stantec has multipleoffices and strong engineering expertise. I am working for agreat firm that has talented people, strong executive man-agement, a solid growth plan and is well- positioned forfurther growth and longevity. Additionally, they fully sup-port my volunteer work with ISPE.

Where did you receive your education?I completed my undergraduate studies at the RochesterInstitute of Technology in upstate New York. As a part-timeevening student, I studied for about nine years in an interdis-ciplinary program and achieved a BS in business and electro-mechanical engineering. Over the course of my career I haveattended a wide range of continuing education sessions inprocess, engineering and business.From left to right: Jane Brown, ISPE Chairman, Dave DiProspero,

Gert Moelgaard, ISPE Immediate Past Chairman.Continued on page 4.


ISPE Update


“Collectively, all of us are making adifference in this world and our efforts arehelping people live longer, healthier, and

happier lives.”Dave DiProspero

Dave DiProsperoContinued from page 3.

How has ISPE helped your career?ISPE has helped my career in several ways. As a pharmaceu-tical consulting engineer, it is of vital importance that I (andmy firm) bring expertise and current knowledge on trends,technologies, regulations, and guidelines to our clients. TheSociety offers numerous avenues for staying current on theseissues relating to our industry and I have taken advantage ofthese. Additionally, getting to know my peers in the industry,through membership and various events has allowed me topersonally know who to turn to for advice and input on a givenexpertise. As a result, I have had reasonable success in mycareer and can credit ISPE for a good part of it.

What has been your best experience with ISPE?Probably my best experiences with ISPE have been related tothe people I have met and the friends I have made. Theindustry we work in is really a rather small community basedaround relationships. People frequently move from firm tofirm or from one side of the industry to the other. ISPE allowsyou to stay in touch and make use of a great network, watchingeach other move on and advance in career. Via volunteering, Ihave gotten to meet some great folks and have been able todevelop both personal and business relationships.

How did you feel when they announced yourname as the winner at the ISPE Annual Meeting?(President/CEO) Bob Best really caught me off guard on thisone. I had no idea it was coming. That particular day I hada red shirt on and was later told that my face tuned about asred as the shirt when my name was announced. Ironically,when I first saw Bob in Orlando, a few days before theMembership luncheon, he said to me that this was going to bemy best Annual Meeting. I just smiled and said, I hope so,without any idea of what was coming.

As he started to talk about the 2006 Member of the Year,during the award presentation and the various committeesthis person served on and then elements of the work done bymy Certification Training Task Team, I started to put thepieces together and thought, “No way, not me.” Then bang,my picture shows up on the big screen and my name isannounced. Wow! It was a great feeling. It was excellent to berecognized for the value of my volunteer contribution andequally important to the support my company Stantec gives,allowing me to volunteer.

Do you know any of the previous award winners?I actually know several of the previous award winners, all ofwhom I’ve met through ISPE. Last year’s winner, RandyPerez of Novartis, and I have served together on the Interna-tional Executive Education Committee. I also know pastwinners John Nichols, Gordon Farquharson, Jon Tomson,Jerry Roth, and Jim O’Brien.

After spending some time with Jon Tomson at this year’sAnnual Meeting, we learned that both of us grew up andspent a good part of our lives in Rochester, New York. We hada wonderful discussion about the city of Rochester and thevarious places we’d each go and things we used to do. Asstrangely as things work out, Jon and I currently live about50 miles apart from each other, here in New Jersey.

I would have to say that I am now in pretty good company.

Did you know Max Seales Yonker (for whomthe Award is named)?Unfortunately I did not know Max personally. However, I wasat the 2005 Annual Meeting Members Luncheon, when herhusband Tom gave an emotional and moving talk about herlife and struggle with her disease. I think everyone in atten-dance that day became a friend of Max and got to know her.

Tom’s talk also put a great perspective on the industry wework in and the contribution all of us make to the bettermentof human life. Tom’s recognition of the many pharmaceuticalcompanies and their products that helped extend Max’s lifewas enlightening. We often forget that when we are dealingwith our day to day efforts of providing a service or engineer-ing a system or getting wrapped in the details of a companyoperation or process. The Members of ISPE are working in anindustry that brings great benefit to many.

What is your greatest accomplishment?That is a very tough question. Hopefully the right answer isthat I have not yet made my greatest accomplishment. It isstill to come. However, one that does come to mind, which Iconsider significant, is maintaining a good balance betweenwork and family. This seems to be difficult these days. Toooften it ends up being one or the other. Rarely both. My wifeand I have raised two kids who have a good head on theirshoulders, a generally positive attitude, and a promisingfuture ahead of them. We’ve successfully maintained strongfamily ties, all while building a career and dealing withrelocations, business travel and other typical job-relatedissues. All in all I can say that I have done pretty well infamily and career.

What are your hobbies?Since a young age, music has always been and currently is avery big part of my life and is my major hobby and passion. Iam actually the lead singer and acoustic guitar player for aPhiladelphia-based classic rock cover band by the name of



ISPE Update


ISPE Annual Meeting Leads Call for Change inPharmaceutical Industry

The 2006 ISPE Annual Meeting was held 5-8 November inOrlando, Florida and its effects will be felt for years, as

Members continue to work with industry leaders, regulatoryagencies, and academia to lead the Society into innovativechanges.

More than 2,000 ISPE Members and non-Members at-tended the meeting, the Society’s premier opportunity andone of the industry’s foremost occasions to bring pharmaceu-tical professionals together for interactive workshops, dis-cussion forums, classroom seminars, and keynote sessions toexplore major trends.

“Change” was a recurrent theme throughout the keynotespeeches, and woven into more than 31 educational sessionswere topics focusing on ASTM, Quality by Design, RFIDtrends, biotechnology, drug shortages, risk management,and more, representing more than 250 speakers over thesuccessful four-day event.

Newly-elected Chairman of the Board, Jane R. Brown,addressed the changes by talking about the future includingnew ideas and new ways of working. “There will be challengesin the development of these new medicines, but many of thehurdles we now face will no longer exist because of thecollaborations we are forming today between industry, regu-latory agencies, and academia.”

Brown, manager of GMP Compliance for GlaxoSmithKline,was officially named Chairman at the meeting and, alongwith her fellow Board members, attended the Annual Meet-ing to meet delegates, participate in committee and educa-tional offerings, and help provide guidance for the industry’sfuture.

The 2006-07 International Board of Directors include ViceChair Bruce Davis, Treasurer Charles P. Hoiberg, SecretaryAlan MacNeice, and newly elected Directors Joan Gore,Tomiyasu Hirachi, and John Nichols. Returning Directorsinclude Gert Moelgaard (immediate Past Chair), Bob Chew,Jan Gustafsson, Linda McBride, Randy Perez, Andre Walker,Stephanie Wilkins, and Chris Wood.

Keynote MessagesChanges in the industry were outlined by keynote speakerMoheb Nasr, Director, ONDQA, CDER, of the US Food andDrug Administration (FDA), who gave the FDA’s perspectiveon the need for strategic innovation in the industry. NatRicciardi, President of Pfizer Global Manufacturing (PGM),provided a snapshot of the challenges facing the industrytoday and discussed how innovation will lead the way towardsuccessful adaptation to our new environment.

According to Ricciardi, some major industry challenges/pressures include new products; slower to market and rampup; hostile external environment (media and public percep-tions); decline of ‘trust’ in the industry; increased call for

generics; significant focus on cost (outsourcing); and govern-ment interventions.

How can we innovate? According to Ricciardi:• always seek sustained value• learn and apply best practices from non-pharmaceutical

industries• apply technology to mitigate external pressures on indus-

try• retain advantages of existing operating structures• foster a culture of innovation all across• bring effectiveness, stability, and predictability to a pro-

cess before seeking efficiency

In addition, Moheb Nasr gave insight into the challengesfacing the FDA. They include:• public concerns and political pressures• tight resources• adequacy of expertise for a rapidly changing science and

technology• inconsistency of regional and global regulatory systems

outside the US• Quality by Design, product/process design and develop-

ment, a robust quality system, and science- and risk-basedapproaches will help achieve the desired state of pharma-ceutical manufacturing

Ron Branning, Vice President of Commercial Quality atGenentech, also delivered a keynote presentation concerningISPE’s new strategic plan and outlined challenges the phar-maceutical industry is facing. Some of those identified chal-lenges include:

• generics are capturing a major share• small molecule focus being replaced by large molecules• lean manufacturing; shift to personalized medicine chang-

ing• manufacturing process and new technology• days of frequent blockbuster drugs are over• cost of drugs; pressures from consumers• R&D productivity declining• patent exclusivity diminishing

The positive news? ISPE is looking at those threats andchallenges and facing them head on.

“ISPE will embrace the future and is committed to playingan integral part of that future,” said Branning. “That meansplanning and leading and collaborating with key industryprofessionals and regulatory agencies.”

In addition to the education sessions and committee meet-ings, both ISPE and regulatory leaders were able to meet



ISPE Update


about critical issues, such as ICH Guidelines. They agreed tomeet again in June during the ISPE Washington Conferencein Arlington, Virginia, USA. ISPE Members can downloadvideo of the keynote session at www.ispe.org.

How ISPE will Help Lead the ChangeIn the new Strategic Plan, ISPE is firmly moving ahead as an“agent of change,” offering concrete solutions, including plansto:• promote industry innovation• work and collaborate with regulators, industry, suppliers,

and academia• work with a global mindset but with local actions• expand Membership into Process and Pharmaceutical

Development• play a major role in the pharmaceutical manufacturer of

the future• make a “stepwise” implementation• proceed with innovation• create a technology infrastructure to facilitate under-

standing and technological innovation• provide forums for industry, suppliers, academia, and

regulator collaboration

ISPE will continue to strive to lead the industry with thenewest and most necessary education courses, Webinars,technical documents, and scientific journals to provide edu-cation and professional development for its Members.

Be sure to visit www.ispe.org to view photos of awardwinners and Annual Meeting attendees, and learn more aboutthe events of the 2006 Annual Meeting. Next year, the AnnualMeeting will be held at Caesar’s Palace in Las Vegas, Nevada.

See You in Las Vegas forNext Year’s Annual Meeting

After the great success of the 2006 annual meeting, be sure tosign up for next year and make sure it’s in your budget.We have received an overwhelmingly positive response dur-ing and after the meeting. The meeting included up to theminute educational sessions, information on our new CPIPcertification, our new Journal of Pharmaceutical Innovation,and more than 280 vendors in the exhibition hall.

Your next chance will be 4-7 November 2007 at Caesar’sPalace in Las Vegas. Join pharmaceutical innovators fromaround the world for the premier opportunity of the year tonetwork with an international roster of industry profession-als. It features a host of important sessions, with a specialseminar reviewing the Baseline Guide for Oral Solid Dosages- new for 2007.

Watch www.ISPE.org for complete details and on-lineregistration. Event includes networking receptions, table topexhibits, and sponsorship opportunities. To register, pleasecall customer service at +1-813-960-2105.

ISPE Annual Meeting...Continued from page 5.

Dave DiProsperoContinued from page 4.

Right Turn at 40. There are five of us, all professionals withsuccessful careers by day. We do our best to get together andplay by nights or weekends, as much as family and jobresponsibilities allow. For me it is a perfect outlet from thestresses of work and life and is great fun. We are keeping alivethe classics from Lynyrd Skynyrd, Allman Brothers, Cre-dence Clearwater Revival, Eagles, Doors, ZZ Top, The Who,Eric Clapton, Bob Seger, and many others. Music the way itused to be. Beyond that, I like to travel and spend time withfamily and friends, enjoying life.

Who is in your family?I have a great supportive family that is clearly the mostimportant thing to me. I have to give them a lot of credit forputting up with me over the years. My wife Peggy and I havebeen married for 20 years and together for nearly 27. Wow!!Amanda, my daughter, will be 19 and is completing her firstyear of college, working toward a future in the health careindustry, likely in nursing or maybe even the pharmaceuticalindustry. Who knows? My son David, 14, is a freshman inhigh school and an avid athlete, something he certainlydoesn’t get from me. Soccer, basketball, baseball, football,and whatever else he can play. He keeps us running year-round. In addition to my immediate family, we have manyother family members and close friends with whom we try tospend as much time as possible.

Any other pertinent information that you feelMembers would like to know.I work very hard and often times go out of my way to maintainan upbeat and positive attitude on life. If the washing ma-chine breaks, why get upset about it? Just go out and buy anew one. My boss at Stantec has given me the tag, “the glasshalf-full guy.” I’m fine with this tag and prefer to live life thisway, than any other. I try not to sweat the small things andput my focus and energies on the good, not the bad. Mygeneral feeling is that “life is great” and we all need torecognize that, use it to our benefit and enjoy. Whenever theend comes, I hope to have no regrets. Smile and laugh as muchas you can, be nice to everybody and try to make a positivecontribution to Society. Isn’t that what it is all about?

I am very honored to have been chosen as the 2006 ISPEMax Seales Yonker Member of the Year. I would like to thankthe Members, staff, and leadership for this recognition. I amproud to be an ISPE Member and volunteer and look forwardto continuing contributions to the Society and great pharma-ceutical industry we all work in. Collectively, all of us aremaking a difference in this world and our efforts are helpingpeople live longer, healthier, and happier lives. It is the tragicdiseases, such as the one that cut Max’s life short, that shouldbe our drivers in the daily work we are doing.

Thanks to all my friends in the industry.


ISPE Update


April Paris Conference Highlights Nano and Micro Technology,and More

ISPE will hold its first European con-ference of 2007 in Paris, 16-19 April,

at the Hilton Hotel. One of the high-lights will be the Nano and Micro Tech-nology poster presentation (see articleon page 90) showing innovative con-cepts from the industry’s andacademia’s finest minds.

The ISPE Paris Conference programwill include eight seminars presentedby industry leaders on the followingkey topics:

16 APRILRevision of the GAMP® GoodPractice Guide: Validation ofProcess Control SystemsSeminar Leaders: Mark Cherry,AstraZeneca, UK; Martin Isherwood,Eli Lilly, UK

Since the GAMP Good Practice Guide:Validation of Process Control Systemswas first published, there have beendevelopments in a number of areas –risk assessment and Process AnalyticalTechnology (PAT) being two specificexamples. This seminar will discuss therevisions to the Guide which take intoaccount of such changes and, addition-ally, expand the consideration of globalregulatory requirements. The revisionswill also provide a clear understandingof the roles of the end users and suppli-ers, including suppliers of base applica-tions and systems integrators.

The seminar will be based on workcarried out by the Process Control Sys-tems Special Interest Group (SIG) ofthe GAMP Forum to improve under-standing of current regulatory trendsgoverning the use of automated sys-tems in the pharmaceutical industry.

16-17 APRILNano and Micro Technologyfor Pharmaceutical Productsand ProcessesSeminar Leaders: Chris Dowle, TheCentre for Process Innovation (cpi), UK:John Nichols, Foster Wheeler, UK: TomTaylor, The Centre for Process Innova-tion (cpi), UK

Technologists using the fundamentalcharacteristics of nano and micro tech-nologies are currently providing solu-tions to applications across a broadspectrum of industries; this is a rapidlydeveloping area. This seminar gives anup-to-date understanding of the un-derlying science; industrial manufac-turing technologies; the range of prod-uct applications; and environmental,safety, health and quality impacts. Itincludes nano particle generation, useof microreactors and nano diagnostics.

Process Analytical Technology(PAT)Seminar Leaders: Ian P. Flawn, FerringPharmaceutical A/S, Denmark; DavidSelby, Selby Hope International Ltd.,UK

PAT is a buzzword within the industryand yet it still holds many mysteriesfor us in everyday terms. There is nodoubt that PAT is pivotal to any com-pany achieving manufacturing excel-lence, it has an ongoing impact on Qual-ity Systems and modifies key activitiessuch as process development and themanagement of data. This seminar willlook at the key aspects of PAT and howeffective process control benefits manu-facturing. First-class speakers will ex-plain how the improved understand-ing of the process can lead to increasedyields and reduced waste, shortenedcycle times and ultimately deliversimprovements in product consistencyat reduced cost.

Pharmaceutical Water,Regulation and InnovationSeminar Leaders: Robert Walker, Rob-ert Walker GMP Consultancy Ltd., UK

Pharmaceutical water systems arecritical in virtually every aspect of APIand dosage form manufacture and regu-lators will subject them to scrutiny.This seminar will provide cutting-edgeinformation about ISPE’s new GoodPractice Guide and use case studiesand experts to highlight developments,

technology and materials to help pro-vide flexible and effective systems.

17 APRIL 2007Facility of the Year ExposéSeminar Leaders: Tony Felicia,AstraZeneca, USA: Andrew Signore,Integrated Project Services, USA

The ISPE Facility of the Year competi-tion has been running for the last threeyears and receives dozens of entrieseach year from major and emergingpharmaceutical companies. In the pastthe award was won by Novo Nordisk(2005) and Baxter (2006).

ISPE, INTERPHEX, and Pharma-ceutical Processing magazine spent2006 looking for the facility projectthat demonstrates global leadership;one that showcases cutting-edge engi-neering, innovative new technology oradvanced applications of existing tech-nology.

This awards program is about muchmore than just the science and technol-ogy of the facilities. It is about recog-nizing the shared commitment anddedication of individuals working fordifferent companies worldwide to in-novate and advance pharmaceuticalmanufacturing technology for the ben-efit of all global consumers.

18-19 APRILBiosafetySeminar Leaders: James V. Blackwell,BioProcess Technology ConsultantsInc., USA: David Harrison, Amgen,USA

This seminar will clarify the differentdesign requirements for each biosafetylevel. Speakers will highlight the vari-ous regulatory requirements associatedwith each level and describe whichelements define the appropriatebiosafety level for deployment. Del-egates will have an opportunity to hearabout case studies on design, valida-tion and implementation of systemscomplying with the different biosafetylevels.



ISPE Update


Design SpaceSeminar Leaders: Jan Gustafsson,Novo Nordisk A/S, Denmark; Pierre leMeur, SPEC Conseils, France

Introduced with ICH Q8, the new con-cept of Design Space is defined in theICH Q8 guideline glossary:

“Design space is the multidimen-sional combination and interaction ofinput variables (e.g. materials at-tributes) and process parameters thathave been demonstrated to provideassurance of quality. Working withinthe design space is not considered as achange. Movement out of the designspace is considered to be a change andwould normally initiate a regulatorypost approval change process. Designspace is proposed by the applicant andis subject to regulatory assessment andapproval.”

In this interactive discussion semi-nar speakers will explain the detailsbehind this definition. Delegates willexplore the reality of Design Space intoday’s pharmaceutical industry andwill examine what should be providedunder this concept in the applicationfile and during a regulatory inspection,as well as during any audit.

Seminar speakers will explain botha no change and a regulatory changescenario and they will point out thechallenges in the post-approval process.

Industry and regulators will elabo-rate on how design space can be imple-mented into daily practice within dif-ferent aspects of the pharmaceuticalindustry (sterile production, biotech-nology, APIs, etc.).

Project Management - FacingToday’s ChallengesSeminar Leaders: Andrew Ingleby,AstraZeneca, UK; Trish Melton, MIMESolutions Ltd., UK

Professional project management is arequirement in all industries, but pre-sents particular challenges in the phar-maceutical industry. So often manyprojects are poorly managed and do notdeliver their business success criteria.

April Paris Conference...Continued from page 7.

Nano and Micro Technology PostersWanted for Presentations

The ISPE Paris Conference will seethe launch of the Nano and Micro

Technology poster presentation fromthe industry’s and academia’s finestminds. Poster presentations will be ondisplay from 16-17 April 2007 at theNano and Micro Technology Seminar.

Topic Categories• Use of Microreactors for Pharma-

ceutical Processes• Nanotechnology in Pharmaceutical

Products• Nanodiagnostics• Relevant Metrology

Benefits of Presenting a Poster• Exposure to 500 pharmaceutical

industry professionals at the ISPEParis Conference

• Abstracts will be displayed at theISPE Paris Conference

• Forum for presenting work, com-munication topics, trends and ideas

• Less formal, interactive environ-ment

Eligibility• Professionals from the pharmaceu-

tical, biotechnology or medical de-vice industry

• University/college faculty• Students

TimelinesDeadline for Submissions:

26 February 2007Presenter Notification Date:

12 March 2007

Rules and Guidelines• Presenters will develop panel sec-

tions for the project abstracts, back-ground information, materials andmethods, graphics and charts if suit-able.

• Abstracts must demonstrate soundtechnical content.

• Abstracts must not exceed 400words.

• Number of submissions allowed perperson: 3

• The size of the posters should beapproximately 1 m × 1.5 m. Posters

will be displayed vertically.• Posters must be fully-assembled and

ready for viewing by 09.00 on Mon-day, 16 April 2007.

• No commercial advertisements arepermitted. One discrete companylogo is permitted.

• Collaborative poster presentationsare acceptable. All researchers mustbe acknowledged on the poster.

SubmissionTo submit an abstract, please completethe submission form available at thewebsite www.ispe.org/parisconferenceand submit it by e-mail to [email protected]. Early submission of an ab-stract is encouraged.

Review CriteriaSubmissions will be reviewed by theNano and Micro Technology seminarleaders and ISPE Education Advisorsfor technical relevance, merit and or-ganization. Those deemed most appro-priate/innovative will be chosen fordisplay at the ISPE Nano and MicroTechnology Seminar.

Poster Presentation Schedule• Poster Presentation Set Up

Sunday, 15 April, 17.00 - 19.00Monday, 16 April, 08.00 - 09.00

• Poster Presentations on DisplayMonday, 16 AprilTuesday, 17 April

• Poster Presentation Break DownTuesday, 17 April, 17.00 - 18.00

ContactAny questions about theposter presentation?Please contact the Nano and MicroTechnology seminar leaders:

John Nichols, Foster Wheeler, UKE-mail: [email protected]

Chris Dowle, The Centre for ProcessInnovation (cpi), UKE-mail: [email protected]

Tom Taylor, The Centre for ProcessInnovation (cpi), UKE-mail: [email protected]



ISPE Update


Find Solutions to Daily Work Problems at the ISPECopenhagen Classroom Training

Get the edge you need to succeed and register for the ISPE Copenhagen ClassroomTraining, 26 February - 1 March 2007, Copenhagen, Denmark. This course covers

topics such as computer systems validation; biopharmaceutical process development;pharmaceutical water systems; risk management; GMPs; and application of commis-sioning and qualification.

ISPE Classroom Training offers you a variety of courses to meet your needs forskills and practical knowledge and is a recognized opportunity for in-depth, topic-specific learning. Take this opportunity in order to enhance your career develop-ment and to find valuable solutions to daily work problems.

For more information and to register please visit www.ispe.org/CopenhagenTraining.

Tampa Conference 2007 Scheduled forFebruary

Seminars on Aseptic Processing, the new ISPE Packaging, Labeling andWarehousing Operations (PACLAW 2007) Baseline Guide, Oral Solid Dos-

ages, Sterile Manufacturing Facilities, and much more are scheduled for theISPE Tampa Conference, 12-15 February 2007, Hyatt Regency Tampa, Tampa,Florida, USA. For details and to register, visit www.ispe.org/tampaconference.

Sponsorship and table top exhibition opportunities are available for the ISPETampa Conference. For details, please contact Dave Hall, ISPE Director of Sales, bytel: +1-813-960-2105, Ext. 208, or e-mail: [email protected].

Mark Your Calendar with these ISPE Events

March 20071 San Francisco/Bay Area Chapter, Vendor Night,

South San Francisco Conference Center, SouthSan Francisco, California, USA

8 San Diego Chapter, Extended Education Class,La Jolla, California, USA

14 Great Lakes Chapter, Spring Meeting, Merillville,Indiana, USA

15 Greater Los Angeles Area Chapter, TechnicalTraining, California, USA

19 - 22 2007 San Francisco Classroom Training, TheFairmont, San Francisco, California, USA

21 Carolina-South Atlantic Chapter, 2007 AnnualTechnology Show - Quest for the Best!, Exhibits,Education Seminars, and Keynote Speaker,Durham, North Carolina, USA

Dates and Topics are subject to change

April Paris Conference...Continued from page 8.

This seminar will provide delegates withexamples of some of the key challengesfaced within pharmaceutical projectsand the tools to deliver the obligatoryquality, time, cost and regulatory com-pliance requirements of any project.

This seminar will focus on the ben-efits of well-structured and controlledpharmaceutical projects as highlightedby real life experience and case studiesand is supported by the Project Man-agement Community of Practice (PMCOP).

For more information and to register,please go to www.ispe.org/

ParisConference.


Classified Advertising


DID YOU KNOW...?More than 56% of Subscribers Forward

Pharmaceutical Engineering to One or More PeopleSource: 2005 Publications Survey

For advertising opportunities, call ISPE Directorof Sales, Dave Hall at +1-813-960-2105.

Architects, Engineers - Constructors

CRB Consulting Engineers, 7410 N.W.Tiffany Springs Pkwy., Suite 100, KansasCity, MO 64153. (816) 880-9800. See ourad in this issue.

IPS – Integrated Project Services, 2001Joshua Rd., Lafayette Hill, PA 19444.(610) 828-4090. See our ad in this issue.

Parsons, 150 Federal St., Boston, MA02110. (617)-946-9400. See our ad inthis issue.

Bioreactors/Fermenters

Cleanroom Products/Services

AES Clean Technology, 422 Stump Rd.,Montgomeryville, PA 18936. (215) 393-6810. See our ad in this issue.

Employment Search Firms

Jim Crumpley & Associates, 1200 E.Woodhurst Dr., Bldg. B-400, Springfield,MO 65804. (417) 882-7555. See our ad inthis issue.

Filtration Products

Siemens Water Technologies, 125Rattlesnake Hill Rd., Andover, MA01810. (978) 470-1179. See our ad in thisissue.

Label Removal Equipment

Hurst Corp., Box 737, Devon, PA 19333.(610) 687-2404. See our ad in this issue.

Passivation andContract Cleaning Services

Active Chemical Corp., 4520 Old LincolnHwy., Oakford, PA 19053. (215) 676-1111. See our ad in this issue.

Astro Pak Corp., 3187 Redhill Ave., Suite105, Costa Mesa, CA 92626. (800) 743-5444. See our ad in this issue.

Cal-Chem Corp., 2102 Merced Ave., SouthEl Monte, CA 91733. (800) 444-6786.See our ad in this issue.

Oakley Specialized Services, Inc., 50Hampton St., Metuchen, NJ 08840. (732)549-8757. See our ad in this issue.

Sterile Products Manufacturing

Used Machinery

Validation Services

Pharma-Bio Serv, POSITIONSAVAILABLE. USA: 168 W. Ridge Pike,Suite 131, Limerick, PA 19468. (215)997-3312. Puerto Rico: Rd. 693, MarginalCosta De oro, Sardinera Beach Blvd.,Suite 2, Dorado, PR 00646. (787) 278-2709. See our ad in this issue.

ProPharma Group, 10975 Benson Dr.,Suite 330, Overland Park, KS 66210;5235 Westview Dr., Suite 100, Frederick,MD 21703. (888) 242-0559. See our ad inthis issue.

Water Treatment

Christ Pharma & Life Science AG,Haupstrasse 192, 4147 Aesch,Switzerland. +41 617558111. See our adin this issue.

Reprinted from


