The Application of Noninvasive Headspace Analysis to Media

10.5731/pdajpst.2015.006312Access the most recent version at doi: 230-24770, 2016 PDA J Pharm Sci and Tech

Derek Duncan, Tony Cundell, Lauren Levac, et al. Fill InspectionThe Application of Noninvasive Headspace Analysis to Media

on June 1, 2016journal.pda.orgDownloaded from on June 1, 2016journal.pda.orgDownloaded from

http://journal.pda.org/lookup/doi/10.5731/pdajpst.2015.006312

http://journal.pda.org/


TECHNOLOGY/APPLICATION

The Application of Noninvasive Headspace Analysis toMedia Fill InspectionDEREK DUNCAN1,*, TONY CUNDELL3, LAUREN LEVAC2, JAMES VEALE2, SUZANNE KUIPER1, andRAVI RAO1

1Lighthouse Instruments BV, Amsterdam, The Netherlands; 2Lighthouse Instruments LLC, Charlottesville, VA, USA;and 3Microbiological Consulting, LLC, Scarsdale, NY, USA ©PDA, Inc. 2016

ABSTRACT: The results of a proof-of-principle study demonstrating a new analytical technique for detectingmicrobial growth directly in pharmaceutical containers are described. This analytical technique, laser-based head-space analysis, uses tunable diode laser absorption spectroscopy to nondestructively determine gas concentrations inthe headspace of a media-filled pharmaceutical container. For detecting microbial growth, the levels of headspaceoxygen and carbon dioxide are measured. Once aerobic microorganisms begin to divide after the lag phase and enterthe exponential growth phase, there will be significant consumption of oxygen and concomitant production of carbondioxide in the sealed container. Laser-based headspace analysis can accurately measure these changes in theheadspace gas composition. The carbon dioxide and oxygen measurement data for the representative microorganismsStaphylococcus aureus, Bacillus subtilis, Candida albicans, and Aspergillus brasiliensis were modeled using theBaranyi-Roberts equation. The mathematical modeling allowed quantitative comparisons to be made between the datafrom the different microorganisms as well as to the known growth curves based on microbial count. Becauselaser-based headspace analysis is noninvasive and can be automated to analyze the headspace of pharmaceuticalcontainers at inspection speeds of several hundred containers per minute on-line, some potential new applications areenabled. These include replacing the current manual human visual inspection with an automated analytical inspectionmachine to determine microbial contamination of media fill and pharmaceutical drug product vials.

KEYWORDS: Microbial detection, Microbiological test method, Media fill inspection, Headspace analysis, Head-space oxygen, Headspace carbon dioxide, Microbial growth, Frequency modulation spectroscopy.

LAY ABSTRACT: A novel analytical technique has been demonstrated for detecting microbial growth in media-filledpharmaceutical containers. This analytical technique, laser-based headspace analysis, uses tunable diode laserabsorption spectroscopy to determine gas concentrations in the headspace of a pharmaceutical container. For detectingmicrobial growth, the levels of headspace oxygen and carbon dioxide are measured. The study shows that onceaerobic microorganisms begin to grow after the lag phase and enter the exponential growth phase there will be asignificant consumption of oxygen in the sealed container as well as a corresponding production of carbon dioxide.Headspace analysis can accurately measure and monitor these changes in the headspace gas composition and couldtherefore be used to detect contaminated pharmaceutical containers. Because the technique can be automated toanalyze hundreds of containers a minute on-line, there are opportunities for implementing a headspace inspectionmachine to perform automated inspection of media fills used to validate aseptic filling operations.

Introduction

It was recently demonstrated that microbial growth inmedia-filled vials could be detected by monitoring the

consumption of oxygen in the vial headspace (1). Thiswas accomplished by using a tunable diode laser ab-sorption spectroscopy (TDLAS) method. The diodelaser spectroscopy technique, frequency modulationspectroscopy (FMS), is a powerful analytical tech-nique that can achieve a high signal-to-noise ratio

* Corresponding Author: Derek Duncan, LighthouseInstruments BV, Science Park 408, NL-1098 XH Am-sterdam, The Netherlands; Telephone: �31 6 20176502; e-mail: [email protected]

doi: 10.5731/pdajpst.2015.006312

230 PDA Journal of Pharmaceutical Science and Technology

on June 1, 2016journal.pda.orgDownloaded from


enabling the measurement of trace amounts of gases(2). Over the past decade this analytical method hasbeen implemented in the pharmaceutical industry fornoninvasive measurements of headspace gas compo-sition and headspace pressure/vacuum levels in phar-maceutical drug containers. Previous publicationshave described the use of FMS to characterize theheadspace in sterile pharmaceutical product vials forthe confirmation of vacuum or inert gases, container-closure integrity, and moisture content of lyophilizedproducts (3– 8). One of the applications has been tointegrate the laser-based headspace sensors into auto-mated inspection machine platforms for the on-lineinspection of finished pharmaceutical product (9). Themethod is also described as a deterministic containerclosure integrity test method in the revised USP�1207� Sterile Product Packaging—Integrity Eval-uation recently released for public comment (10). Theresults described in this report show that nondestruc-tive, laser-based headspace analysis can detect micro-bial growth in stoppered and sealed pharmaceuticalvials filled with microbiological growth media bymonitoring headspace oxygen consumption and corre-sponding carbon dioxide production. Using this ap-proach, the determination of microbial growth can bemeasured analytically directly in pharmaceutical con-tainers in a nondestructive manner. Any contamina-tion, whether by aerobic microorganisms, anaerobicmicroorganisms as well as mycoplasamas, that causesdetectable changes in the headspace gas compositionof the sealed container will be detected. The primaryproof-of-principle study described demonstrates usingheadspace analysis to detect media vials contaminatedwith five representative compendial aerobic microor-ganisms. Additional secondary studies also describedin this paper demonstrate the ability of headspaceanalysis to detect media vials contaminated with aer-obic or anaerobic microorganisms and mycoplasmas.Because headspace analysis can be automated at in-spection speeds of several hundred vials per minuteon-line, a potential application is pharmaceutical aer-obic media fill inspections as part of aseptic fillingprocess validation. This represents an opportunity forreplacing the labor intensive human visual inspectionprocess (a subjective method) with an automated an-alytical inspection for media fills. This change wouldimprove the reliability of media fill inspection, reducethe media fill inspection time and the required humanlabor, and improve vial reconciliation and data integ-rity. In addition, headspace analysis could potentiallybe used to determine the lag time and growth rate ofmicroorganisms in microbial challenges for beyond-

use dating studies and for the nondestructive verifica-tion of presumptive sterility test failures. A presump-tive sterility test failure could be confirmed byperforming headspace analysis of the sterility test vi-als or canisters to check for headspace oxygen con-sumption and/or carbon dioxide production in the can-ister due to microbial growth prior to subculturing thesterility test media. Finally, the measured headspaceoxygen and carbon dioxide data curves can be ana-lyzed with mathematical modeling. The results of theprimary study were used to demonstrate that thismathematical approach enables the differences in thegaseous exchange characteristics from microorganismto microorganism to be quantified as well as to inves-tigate if the headspace gas curves mimic bacterialgrowth curves.

Media Fill Validation for Aseptic Filling Processes

To validate an aseptic filling operation, it is pharma-ceutical industry practice to fill, stopper, and sealcontainers used for product with tryptic soy broth(TSB), incorporating likely interventions within themedia fill to mimic the standard filling process. Thenumber of media-filled containers could be in excessof 10,000 depending on the typical product batch size.Media fill incubation conditions can vary but a com-mon practice is to incubate the containers at 20 –25 °Cfor 7 days, followed by a further 7 days at 30 –35 °C.The containers are then visually inspected under thesupervision of a trained microbiologist. Filled unitslacking container-closure integrity may be rejectedprior to incubation. A 100% inventory control of themedia fill containers is expected by the regulators.After incubation, the vials are visually inspected formicrobial contamination in the form of turbidity, pel-licle formation, floccular-suspended material, and/orprecipitation. In general, a single turbid, filled con-tainer after incubation, irrespective of the batch size,will trigger an investigation, while two or more turbid,filled containers require a repeat of the media fill. Thecontents of the so-called turbid vials are microscopicallyexamined and subcultured to confirm the presence ofmicroorganisms and the microbial isolates identified tothe species level to aid investigation (11, 12).

Microbial Metabolism as Related to Media Fills:Microorganisms can be conveniently classified intoaerobic, facultative anaerobic and anaerobic organ-isms. For example, obligate aerobes cannot grow inthe absence of oxygen or other electron acceptors suchas nitrate and lack a functional fermentative metabo-

231Vol. 70, No. 3, May–June 2016



lism. These organisms convert glucose to carbon di-oxide using the Embeden-Meyerhof-Parnas (EMP)pathway or glycolysis, the Krebs or citric acid cycleand the electron transport chain to obtain energy in theform of ATP with oxygen being the terminal electronacceptor (13). The overall reaction can be representedas C6H12O6 � 6O2 3 6CO2 � 6H2O � 38ATP. Withaerobic metabolism one would expect to find in theheadspace of a media fill vial evidence of a stoichio-metric oxygen consumption and carbon dioxide gen-eration.

In contrast, facultative and obligate anaerobes grow ineither the presence or absence of oxygen, respectively,(in the absence of oxygen they use fermentative path-ways or anaerobic respiration) so their patterns ofoxygen consumption and carbon dioxide production ina media fill vial will be more complex. They containtwo groups, either cytochrome independent, (e.g. lac-tic acid bacteria), or cytochrome dependent (e.g.,member of the family Enterobacteriaea). The variousfermentative pathways of microorganisms in these twogroups will produce end products that include lacticacid, alcohol, a mixture of organic acids (acetic, lactic,and formic acid), carbon dioxide, ethanol, and butane-diol (13). It is therefore more difficult to predict ex-actly what the headspace gas dynamics will be as thevarious microorganisms begin to grow. As a media fillvial is a closed system, oxygen will be limited. TheTSB, if saturated with oxygen, will contain less than 8mg of oxygen per liter and the headspace will containaround 270 mg/L. It is possible that soon into theexponential growth phase the oxygen will be depletedwith the resulting pattern of metabolism shifting fromaerobic to anaerobic.

Oxygen Consumption

A recent publication reported the lag phase oxygenconsumption per colony-forming unit (CFU) (Qo2) forEscherichia coli under nutritional-limited growth as2.4 � 10–7 mol O2 CFU–1 day–1 or 1 � 10– 8 mol O2

CFU–1 h–1 (14). The oxygen depletion in a pharma-ceutical vial due to exponential microbial growth willbe detected by headspace analysis. It is assumed thatthis type of oxygen depletion directly links to themicrobial count for aerobic growth. Growth of facul-tative anaerobes would also be detected through oxy-gen depletion in the sealed vial, but the correlation tomicrobial count could be more complicated. In bothcases, exponential microbial growth will also be iden-tified by carbon dioxide production in the headspace.

Media Fill Broth

(a) TSB (Soybean-Casein Digest Broth): The com-position of the compendial TSB is given in Table I.

What may be lacking for the broadest aerobic micro-bial growth in TSB includes essential minerals, somevitamins, and fatty acids. Furthermore, an importantpoint for this study is that traditional microbiologicalgrowth media are optimized for microbial growthidentified by turbidity of the media and not by head-space analysis. It might be that microbiologicalgrowth media, as well as incubation conditions, couldbe optimized for maximum change in the headspaceoxygen and/or carbon dioxide levels to identify mi-crobial growth.

In a static, batch culture like a media fill vial, a typicalbacterial growth curve will consist of five distinct

TABLE IThe Ingredients, Composition, and Function of the Components of TSB

Ingredient Composition (g/L) Metabolic Role

Pancreatic Digest of Casein 17 Enzymatic digests of casein and soybeanmeal provide for amino acids andother complex nitrogenous substances.Carbon, nitrogen and energy source.

Papaic Digest of Soybean Meal 3 As above. Contains high vitamin andcarbohydrate content from soybeans.

Sodium Chloride 5 Maintains osmotic equilibrium

Dipotassium Phosphate 2.5 Buffer to control pH

Dextrose 2.5 Carbon and energy source




phases, namely, the lag, exponential growth, station-ary, death, and long-term stationary phases. It hasbeen assumed that the lag phase allows for the adap-tation of the bacterial cells to the new culture condi-tions and may include repair to cellular damage andsynthesis of enzymes needed to enter the exponentialgrowth phase. The most rapid oxygen consumptionand carbon dioxide production will be delayed untilthe microorganism enters the mid-exponential growthphase where there will be a sufficient number ofrespiring microbial cells to result in significant changeto the headspace composition.

(b) Fluid Thioglycollate Medium (FTM): It is stan-dard practice for pharmaceutical media fills to useTSB. However, to enhance the growth of anaerobeswhen they have been implicated in sterility test fail-ures or for sterile products with a nitrogen blanket intheir headspace, FTM is usually used. The composi-tion of the compendial FTM is given in Table II.

The top one-third of the medium will be oxygenatedwith the redox indicator resazurin appearing red, andwill support the growth of facultative anaerobes andstrict aerobes, while the remaining two-thirds willsupport facultative aerobes and strict anaerobes. Un-der one of the experimental conditions used in thisstudy, where the medium-filled vial is purged withnitrogen, the entire medium will be anaerobic. Ingre-dients that may be lacking in FTM for full-rangeanaerobic growth are hemin and vitamin K.

Although FTM is suitable for sterility testing as itcontains neutralizers for mercuric preservative andsupports the growth of aerobic as well as facultative

and obligate anaerobes, it is not used to isolate anaer-obes in a clinical setting.

Growth of Mycoplasma in Media Fills

Members of the order Mycoplasmatales are flexible,pleomorphoric organisms as small as 0.2 micron thatlack a cell wall and are commonly known as contam-inants of mammalian cell cultures. As they have asmall genome (600 –1400 kB) compared to the 4.64mB of E. coli K-12, they have a reduced syntheticcapability and rely on host or cell-culture metabolicproducts such as fatty acids, sterols, and other com-plex lipids for growth. Growth media containing fetalcalf serum, cholesterol, and yeast extract are widelyused to isolate mycoplasma from cell culture, vac-cines, and other biologics. However, mycoplasma area widely diverse group of facultative anaerobic micro-organisms with a broad host range including humans,other animals, insects and plants, as well as differentmetabolic patterns so that the headspace dynamics aredifficult to predict.

Mycoplasma became, in addition to a cell culturecontaminant, a compliance concern when a major bio-technology company observed multiple media fill val-idation failures due to Acholeplasma laidlawii con-tamination. Turbid media, when subcultured andsubjected to microscopic examination, initially failedto implicate a microrganism. The contamination waseventually identified as due to A. Laidlawi using 16Sribosomal RNA (rRNA) base sequencing. Further in-vestigation showed that the batches of dehydrated TSBused were contaminated with the mycoplasma, whichwere not removed from the media fill by sterile filtra-

TABLE IIThe Ingredients, Composition, and Function of the Components of FTM

Ingredient Composition (g/L) Function

Pancreatic digest of casein 15 Enzymatic digests of casein provide for amino acidsand other complex nitrogenous substances.Carbon, nitrogen and energy source.

Yeast extract 5 Vitamin source

Dextrose 5.5 Carbon and energy source

Sodium chloride 2.5 Maintains osmotic equilibrium

L-cystine 0.5 Reducing agent

Sodium thioglycollate 0.5 Reducing agent. Neutralization of mercerical agents

Agar 0.75 Medium stabilizing agent to maintain anaerobosis

Resazurin .001 Redox indicator

233Vol. 70, No. 3, May–June 2016



tion using a 0.2 micron sterilizing-grade filter. Thiswas mitigated either by using a 0.1 micron sterilizing-grade filter, purchasing irradiated media, or movingfrom animal- to plant-based media. The latter wasultimately unsuccessful as the plant-based materialwas also contaminated by mycoplasma (15).

Baranyi-Roberts Equation Application to the CO2

and O2 Measurements: The microbial headspace ox-ygen and carbon dioxide measurement curves overtime were modeled using the Baranyi–Robert equationshown below (16). While this equation is normallyused to model microbial growth measured by platecount or turbidity, the sigmoidal shape of the micro-bial CO2 production curves and adjusted O2 consump-tion curves suggested it would be a candidate formodeling the headspace gas curves.

y � y0 � �max � A � ln�1 �e�maxA � 1

e� ymax�y0 � (1)

In its standard use, the variable y represents the mi-crobial count in log10 CFU mL–1, with y0 representingthe initial bacterial population count, and ymax repre-senting the final bacterial population count. For thepurposes of this study, the variable y was adapted torepresent the log10 gas concentrations, with y0 andymax signifying the initial and final log10 headspace gasconcentrations. The two remaining parameters of theequation are �max, the maximum specific growth rate(h–1), and h0, the bacterial adaptability factor (relatingto the initial physiological state of the microorganismbeing examined), with A being defined as

A � t �1

�max� ln(e��max � t) � e�h0 � e (2)

The lag time is related to both h0 and �max by

tlag �h0

�max(3)

The adjusted O2 consumption curves were generatedby using the magnitude of the difference in O2 levelsfrom point to point to construct a sigmoidal curve thatincreases according to the magnitude of O2 consump-tion in the original curve. This changes the orientationof the curve to be similar to the CO2 production curvethereby enabling application of the Baranyi-Robertequation for modeling the O2 results.

Materials and Methods

Media Fill and Inoculated Vial Preparation

In the primary experiment, the 10 mL borosilicateglass vials were sterilized, aseptically filled with ap-proximately 5 mL of TSB, (Becton, Dickinson andCompany, Franklin Lakes, NJ), stoppered, and cappedin a clean room located at Afton Scientific (Charlot-tesville, VA). After a 2 week incubation period toconfirm sterility, the vials were inoculated with�100 CFU of five representative challenge micro-organisms using commercially prepared cultures asper the instructions of the manufacturer (EZ-CFU,Microbiologics, Inc., St Cloud, MN). The five chal-lenge microorganisms represent compendia-speci-fied microorganisms (USP Chapters �61�, �62�,and �71�).

The flip-tops of the caps were removed and the vialswere injected with the different inocula (Table III) viasterile insulin syringes. Self-sealing rubber stopperswere used to maintain container-closure integrity after

TABLE IIIThe Seven Sets Each with Different Inocula Prepared in This Study

Set Inocula (100 �L/vial)ATCC

NumberMicrobiologics Inc

Catalog #Spores or

cellsCFU per100 �L

Baseline Control None N/A N/A N/A N/A

Negative Control Air N/A N/A N/A N/A

Aspergillus brasiliensis Aspergillus brasiliensis 16404 0392C Spores �100

Bacillus subtilis Bacillus subtilis 6633 0486C Spores �100

Candida albicans Candida albicans 10231 0443C Cells �100

Pseudomonasaeruginosa


9027 0484Z Cells �100

Staphylococcus aureus Staphylococcus aureus 6538 0485C Cells �100




injection. Seven sample sets were prepared including abaseline control, a negative control inoculated withsterile air, and five different challenge microorgan-isms. Each set consisted of twenty 10 mL vials.

Incubation Conditions

The media fill vials were stored in an environmentalchamber for the course of the 14 day study. Thesamples were incubated at 20 –25 °C for the first 7days and 30 –35 °C for the following 7 days. Thesamples were removed from the environmental cham-ber shortly before visual inspection and headspacemeasurement sessions to come to ambient tempera-ture. Table IV lists the general description, typicaloptimum growth temperature and the appearance ofexponential growth in TSB for each of the challengeorganisms.

Visual Inspection

The vials were visually inspected and assessed for thepresence or absence of microbial growth daily byqualified laboratory technicians. The visual conditionsof the samples were then recorded and later comparedwith the headspace analysis measurements. It shouldbe noted that without continuous monitoring of thevials it is not possible to determine exactly whenturbidity occurred in the vials in relationship to theheadspace changes.

Oxygen Headspace Analysis

Headspace oxygen measurements were performed us-ing a nondestructive headspace oxygen analyzer

(Model FMS-760, Lighthouse Instruments, Charlottes-ville, VA). Analyzer calibration was performed usingcertified 20% and 0% oxygen standards. National In-stitute of Standards and Technology (NIST)-traceablecertified reference standards were used to evaluatesystem performance (accuracy, precision, linearity,and limit of detection) of the FMS-760 and to verifysystem calibration.

The certified standards were made by backfilling 10mL vials with certified NIST-traceable oxygen mix-tures and then flame sealing them. The standardswere made with glass vials identical to those used inthe media fill. Table V shows the results of repeatedmeasurements of the known oxygen standards overthe course of the study and demonstrates the accu-racy and precision of the headspace oxygen mea-surements for the 10 mL vial configuration. Before

TABLE IVDescription of Primary Challenge Organisms

Microorganism Description

OptimumGrowth

Temperature(°C) Appearance in TSB


Gram-negative, rod-shaped,obligate aerobe

37 Turbid growth

Staphylococcus aureus Gram-positive cocci,facultative anaerobe

35–37 Turbid growth

Bacillus subtilis subsp.spizizenii

Gram-positive, spore-formingrod, obligate aerobe

30–37 Flocculent/surface growth

Candida albicans Yeast, facultative anaerobe 25–30 Flocculent/surface growth

Aspergillusbrasiliensis

Fungus, obligate aerobe 24–30 Flocculent/surface growth

TABLE VResults of Measurements on Certified Standardsof Known Oxygen Concentration throughout the14 Day Study

Oxygen StandardValue (% atm)

Measured OxygenConcentration (% atm)

Mean SD

0.00 0.06 0.07

1.00 1.15 0.14

2.00 2.14 0.17

4.00 4.19 0.17

8.00 8.14 0.14

20.00 20.14 0.20

235Vol. 70, No. 3, May–June 2016



every measurement session, one measurement wastaken of each of the six known oxygen standards toverify the performance of the system. The head-space oxygen concentration in each sample vial wasthen measured and recorded. Each sample vial wasmeasured once. The first vial in each set was mea-sured five consecutive times each day to determinemeasurement accuracy on the sample vials.

Carbon Dioxide Headspace Analysis

Headspace carbon dioxide measurements were per-formed using a nondestructive headspace carbon di-oxide analyzer (Model FMS-CO2, LighthouseInstruments, Charlottesville, VA). Calibration wasperformed using a certified carbon dioxide standard.Certified reference standards were used to evaluatesystem performance (accuracy, precision, linearity,and limit of detection) of the FMS-CO2 and verifysystem calibration.

The certified standards were made by backfilling 10mL vials with certified NIST-traceable carbon dioxidegas and then flame sealing them. The standards weremade with glass vials identical to those used in themedia fill. Table VI shows the results of repeatedmeasurements of the known carbon dioxide standardsover the course of the study that demonstrates theaccuracy and precision of the headspace carbon diox-ide measurements for the 10 mL vial configuration.

Carbon dioxide measurements are reported in units oftorr. However, because media fill vials are at near-atmospheric pressure, it is possible to convert thosemeasurements to units of % atm. This allows for easy

comparison of the carbon dioxide and oxygen mea-surements of the media fill vials.

One measurement was taken of each of the four knowncarbon dioxide standards to verify the performance ofthe system. The headspace carbon dioxide in eachsample vial was then measured and recorded. Eachsample vial was measured once. The first vial in eachset was measured five consecutive times each day todetermine measurement accuracy, that is, system suit-ability on the sample vials.

Chemical Oxidation of the TSB

Prior to any microbial growth, a difference was ob-served between the expected initial headspace oxygen(20.9% atm) and measured oxygen concentrations inthe media fill vials. This was due to two factors:spectroscopic broadening of the oxygen signal due tomoisture in the headspace and chemical oxidation ofthe broth. To quantify the signal broadening, five vialswere filled with water and five with fresh TSB with allvials having an atmospheric air headspace. The head-space oxygen and carbon dioxide levels in every sam-ple were measured and montiored over the course ofthe 2 week incubation period. In addition, the head-space oxygen levels of the original baseline and neg-ative control samples were monitored over the courseof 3 months.

Secondary Studies

As described above, the primary proof-of-principleexperiment focused on detecting the growth of aer-obic microorganisms inoculated into vials filledwith TSB. Secondary measurements were also doneusing headspace analysis to demonstrate the growthof anaerobes and mycoplasmas. For anaerobes, FTMwas used and the media vials were purged withnitrogen so that facultative and strict anaerobes—but not aerobes—would grow in the FTM. A total of2 mL of FTM medium was added to a standardVCDIN2R vial. The vials were then inoculated withthe appropriate microorganism to a final concentra-tion of 10 CFU/mL. To ensure initial headspaceoxygen levels were below 1% atm, the vials wereagain purged for 30 seconds with nitrogen directlyafter inoculation. Triplicate samples were inocu-lated with the corresponding microorganism andwere incubated at a temperature of 30 –35 °C. Thegeneration of headspace carbon dioxide was thenfollowed over a period of more than 8 days.

TABLE VIResults of Measurements on Certified Standardsof Known Carbon Dioxide Gas Pressure in Unitsof torr throughout the 14 Day Study

Carbon DioxideStandard Value (torr)

MeasuredCarbon DioxidePressure (torr)

AverageStandard

Deviation (SD)

49.3 50.1 0.2

99.3 100.2 0.4

149.2 149.1 0.6

544.4 542.6 2.1




Table VII lists the general description, typical opti-mum-growth temperature, and the appearance of ex-ponential growth in FTM for each of the challengeorganisms.

For the detection of mycoplasmas, media-filled vialswere prepared with basic broth or Z� broth (MicrosafeInc., Leiden, The Netherlands) depending on the mi-croorganism being tested. Similarly, a total of 2 mL ofgrowth media was added to a VCDIN2R vial. Tripli-cate samples were inoculated with the mycoplasmaorganisms and incubated at a temperature of 35–37 °C.The broth was proprietary but would be based onpleuropneumonia-like organism (PPLO/mycoplasma)broth supplemented with yeast extract and horse se-rum. The headspace oxygen consumption and head-space carbon dioxide production curves were thenmonitored over a 12 day incubation time.

Finally, a proof-of-principle experiment was per-formed to investigate if contamination of actual phar-maceutical product could be measured with headspaceanalysis. Vials of commercial pharmaceutical humanalbumin, 5%, were sourced and then inoculated witheight standard test microorganisms. Headspace oxy-gen and carbon dioxide levels were monitored duringa 14 day incubation (20 –25 °C for 7 days, followed bya further 7 days at 30 –35 °C).

Results and Discussion

Experimental Data

Time plots of the measured headspace oxygen and car-bon dioxide levels in the media vials and the time pointwhen visual microbial growth was first observed duringthe incubation period are shown in Figures 1 to 7.

Figures 1 to 5 show the results from the five standardtest organisms. Figures 6 and 7 show the results fromthe negative control set and the baseline control set,respectively. The graphs show the oxygen consump-tion and carbon dioxide production in the headspace ofthree samples from each of the test organisms from thetime of inoculation of the media vial through the 14day media fill incubation period. For each microor-ganism, three samples from the set of 20 were chosento be plotted and represent the extreme high sample,extreme low sample, and a middle sample in terms ofheadspace oxygen and carbon dioxide levels. Thegraphs in Figures 1 to 5, therefore, illustrate howconsistent the headspace oxygen consumption and car-bon dioxide production were across the sample set of20 media vials for each microorganism. The error barsfor the measured data points can be deduced from thedocumented headspace analyzer performance in TableVII and Table VIII, and are smaller than the datapoints themselves in the graphs in Figures 1 to 7.

TABLE VIIAnaerobes Used in the Secondary Study

Microorganisms DescriptionOptimum GrowthTemperature (°C) Appearance in FTM

Escherichia coli Gram-negative rod;facultativeanaerobe

37 Turbid growth

Staphylococcus epidermidis Gram-positivecoccus;facultativeanaerobe

35 Turbid growth

Candida albicans Yeast; facultativeanaerobe

37 Floccular/surface pellicle

Clostridium sporogenes Gram-positive,sporeformingrod; obligateanaerobe

35 Turbid growth

Bacteriodes fragilis Gram-positive rod;obligate anaerobe

35 Turbid growth

237Vol. 70, No. 3, May–June 2016



Although the visual observations were not continuous,it is notable that microbial growth was visually ob-served just prior to, or early in the period of rapidchange in headspace composition. Figures 8 and 9show the results from all twenty samples inoculatedwith the bacterium S. aureus which clearly demon-strate the consistency of the measured oxygen con-sumption and carbon dioxide production curves acrossthe set of 20 samples.

The results of both the negative control set and thebaseline control in Figures 6 and 7, respectively, show

a slight decrease in the headspace oxygen levels andno carbon dioxide production over the incubation pe-riod (336 h). The slight decrease in headspace oxygenlevels is due to TSB auto-oxidation. The TSB auto-oxidation was confirmed with a separate media study.Five vials were filled with a fresh batch of TSB andfive vials were filled with water. Initial headspaceoxygen measurements of both sets gave very similarreadings of 19.8% atm, implying that the aforemen-tioned water vapor broadening of the headspace-oxy-gen signal in a TSB vial (see section Chemical Oxi-dation of the TSB) is similar to that for water. In otherwords, at atmospheric levels, the initial headspaceoxygen content in a TSB media vial is underestimatedby 1.1% atm. The new broth vials proceeded to

Figure 1

Measured headspace oxygen and carbon dioxidelevels over the 14 day incubation period in mediavials innoculated with A. brasiliensis. The timepoint at which growth was detected visually is in-dicated by the vertical dotted line. The results ofthree samples are plotted to indicate the spreadbetween extreme high, medium, and low measure-ment results.

Figure 2

Similar plot of measured headspace oxygen andcarbon dioxide levels over the 14 day incubationperiod in media vials innoculated with B. spizizenii.

Figure 3

Similar plot of measured headspace oxygen andcarbon dioxide levels over the 14 day incubationperiod in media vials innoculated with C. albicans.

Figure 4

Similar plot of measured headspace oxygen andcarbon dioxide levels over the 14 day incubationperiod in media vials innoculated with P.aeruginosa.




chemically oxidize and the oxygen levels dropped to17.9% atm in 2 weeks, matching what was observedin the study control vials 2 weeks after fill. Thisexplains the initial headspace oxygen levels 17.9%in the study results, as the starting point of the inoc-ulation study was 2 weeks after filling of the vials withTSB. The oxygen levels in the water vials stayedconstant. The oxidation behavior of the TSB mediawas studied in more detail by monitoring headspaceoxygen as well as carbon dioxide levels in the studycontrol vials over a period of 100 days (2400 h). Theresults are plotted in Figure 10.

Experimental Data Secondary Studies

In addition to the results of the principal study justdescribed, secondary data was also generated to inves-

tigate the possibility of detecting the growth of anaer-obes and mycoplasmas with noninvasive laser-basedheadspace analysis. As described in the Materials andMethods section, standard VCDIN2R vials were filledwith 2 mL of FTM for the anaerobe experiment. Theheadspace of each vial was purged with nitrogen tocreate anaerobic conditions. The initial headspace ox-ygen levels were near zero and did not increase duringthe incubation. The generation of headspace carbondioxide was followed over an incubation period of 9days at a temperature of 30 –35 °C (Figure 11). Thestrict anaerobe C. sporogenes has the greatest measur-able CO2 production and B. fragilis the least. Thegrowth of B. fragilis is reported to be stimulated by thepresence of hemin and vitamin K which, as statedearlier, are not present in FTM. The reason for theabsence of growth indicated by zero CO2 productionfor E. coli and Staphylococcus epidermidis is notobvious to the authors but may be related to renderingthe FTM fully anaerobic with 100% nitrogen in theheadspace. It should be noted that the media vials inthis study inoculated with E. coli and S. epidermidisdid not display any turbidity and that these sampleswere not subcultured to evaluate the remaining micro-bial count after incubation. In general, the resultsdemonstrate the possibility of detecting the growth ofsome anaerobic organisms by monitoring the produc-tion of carbon dioxide in the headspace.

Figures 12 and 13 show the headspace oxygen con-sumption and headspace carbon dioxide productioncurves, respectively, over a 12 day incubation periodat a temperature of 35–37 °C for the mycoplasmastudy. In this study, a proprietary growth media wasused. Without a detailed knowledge of the medium

Figure 7

Similar plot of measured headspace oxygen andcarbon dioxide levels over the 14 day incubationperiod in the baseline control vials.

Figure 5

Similar plot of measured headspace oxygen andcarbon dioxide levels over the 14 day incubationperiod in media vials innoculated with S. aureus.

Figure 6

Similar plot of measured headspace oxygen andcarbon dioxide levels over the 14 day incubationperiod in the negative control media vials.

239Vol. 70, No. 3, May–June 2016



composition and metabolic patterns of the differentspecies of mycoplasma, the authors cannot commenton the headspace changes. However, from the resultsgraphed in Figures 12 and 13, it is demonstrated thatthe headspace oxygen and carbon dioxide measure-ments were able to detect growth of some mycoplas-mas. It should be noted that the media vials in thisstudy that did not show clear changes in the headspacegas content also did not display any discoloration ofthe proprietary growth media.

Finally, measurements were also made on vials ofcommercial pharmaceutical human albumin, which is

known to support microbial growth. The human albu-min product vials were inoculated with standard testmicroorganisms. Also in this case, growth of organ-isms was detected with laser-based headspace analysisduring a 14 day incubation (20 –25 °C for 7 days,followed by a further 7 days at 30 –35 °C) as shown inFigures 14 and 15. Unlike a microbiological culturemedium, 5% human-albumin solution is not optimizedfor microbial growth. It is noteable that the two strictaerobes P. aeruginosa and B. subtilis, which are richin proteolytic enzymes and not reliant on vitamins andgrowth factors, grew most abundantly while the strictanaerobe C. sporogenes grew least abundantly.

TABLE VIII.Mycoplasma Used in the Secondary Study

Species ClassificationCholesterol

Requirement Properties Host

A. laidlawii Genus Acholeplasma No Facultative anaerobe.Optimum growthtemperature30–35 °C.

Animals, plants, and insects.Cell culture contaminant.

M. fermentans Genus Mycoplasma Yes Facultative anaerobe. Cell culture contaminant.

M. hyrohinis Genus Mycoplasma Yes Strict aerobe. Respiratory disease inpigs.Cell culture contaminant.

M. pneumonia Genus Mycoplasma Yes Facultative anaerobe.Lacks aminoacidsynthesis.

Human respiratory disease.

M. gallispticum Genus Mycoplasma Yes Facultative anaerobeOptimum growthtemperature 37 °C.

Respiratory disease inTurkeys and chickens.

M. arginini Genus Mycoplasma Yes Facultative anaerobe. Animal respiratory infections.Cell culture contaminant.

Figure 8

Measured headspace oxygen levels over the 14 dayincubation period for all 20 media-filled vials inn-oculated with S. aureus.

Figure 9

Measured headspace carbon dioxde levels over the14 day incubation period for all 20 media-filledvials innoculated with S. aureus.




Discussion of the Experimental Results

The results of the primary study demonstrated that anondestructive headspace gas measurement can iden-tify growth in TSB media vials after the vials wereinoculated with five representative challenge micror-ganisms. The headspace measurements could not onlyidentify the onset of exponential growth in the con-taminated media vials, but could robustly follow thegrowth evolution in the samples throughout the incu-bation period. During the writing of this paper, theauthors learned that these results have subsequentlybeen confirmed at other pharmaceutical microbiologylaboratories (17, 18). These recent preliminary studies

have also demonstrated that, in addition to the fiverepresentative organisms used in this study, headspaceanalysis can also detect the growth of typical houseisolates that have been found in sterile manufacturingfacilities (17). These results are promising indicationsthat headspace analysis could be used for an analyticalinspection of media fills. Undoubtedly, replacing thelabor intensive and subjective human visual inspectionprocess with an automated analytical headspace in-spection for media fills will require formal methoddevelopment and validation studies prior to implemen-tation. Formal method development and validationwould include a robust demonstration that headspaceanalysis identifies media fill vials that are contami-nated with a representative set of microorganisms

Figure 10

Measured headspace oxygen and carbon dioxide levels in TSB media vials over a period of 100 days (2400 h).

Figure 11

Measured headspace carbon dioxide levels over anincubation period of 9 days in nitrogen-purgedFTM medium-filled vials innoculated with different(facultative) anaerobic organisms. The plotted val-ues are the mean of measurements taken on threereplicate vials.

Figure 12

Measured headspace oxygen levels over an incuba-tion period of 12 days in media-filled vials innocu-lated with different mycoplasmas. The plotted val-ues are from a single sample vial.

241Vol. 70, No. 3, May–June 2016



including house isolates that are of interest for aspecific sterile manufacturing facility. For each micro-organism type and container configuration, it shouldbe demonstrated that a headspace carbon dioxideand/or oxygen measurement distinguishes a contami-nated vial from one that is sterile. The results frommethod development and validation studies shoulddefine headspace carbon dioxide and/or headspace ox-ygen limits that clearly distinguish all contaminatedvials from sterile vials after a defined incubation. Aheadspace inspection process can then be defined andqualified to robustly demonstrate that headspace mea-surements can be made with sufficient accuracy andprecision in the inspection process to detect all con-

taminated media vials. Future studies could investigatethe possibility of using headspace analysis to inspectmedia fills subjected to varying incubation conditionssuch as single temperature or shorter incubation times.

In addition to the primary application described in thispaper of using headspace analysis for media fill inspec-tion, there are other potentially interesting applicationsfor using headspace in microbiological activities. Theability to make headspace measurements directly in ster-ile pharmaceutical containers raises the additional oppor-tunity of using headspace analysis for sterility trouble-shooting in actual pharmaceutical product. For example,for product that is growth promoting (such as humanalbumin), a measurement of elevated carbon dioxidelevels could indicate a potentially contaminated con-tainer. Other potential applications envisioned by theauthors include using headspace analysis to determinethe lag time and growth rate of microorganisms (includ-ing anaerobes and mycoplasmas) in microbial challengesfor in-use dating studies, for the nondestructive verifica-tion of presumptive sterility test failures, and for growthmedia optimization studies. Each of these applicationscould be enabled by using nondestructive headspaceanalysis in containers or sterility test vials or canisters toquanitatively measure and monitor changing carbon di-oxide and/or oxygen levels resulting from microbialgrowth.

Mathematical Modeling of the O2 Consumption andCO2 Production Data

In the previous section, a general approach was de-scribed for method development and validation of

Figure 13

Measured headspace carbon dioxde levels over anincubation period of 12 days in media-filled vialsinnoculated with different mycoplasmas. The plot-ted values are from a single sample vial.

Figure 14

Measured headspace oxygen levels over an incuba-tion period of 14 days in pharmaceutical humanalbumin product vials innoculated with differentmicroorganisms.

Figure 15

Measured headspace carbon dioxide levels over anincubation period of 14 days in pharmaceuticalhuman albumin product vials innoculated with dif-ferent microorganisms.




headspace analysis to detect a contaminated media-filled unit in a media fill inspection process. It isassumed that the headspace gas dynamics in a sealedmedia vial correlate to actual microbial growth. It is,therefore, interesting to investigate if the measuredheadspace dynamics of oxygen consumption and car-bon dioxide production reflect microbial growthcurves based on microbial counts. A better under-standing of how headspace gas dynamics correlate tomicrobial growth can perhaps expand the possibleutility of headspace analysis to microbiological appli-cations such as the ones mentioned at the end of theprevious section.

In order to model the results of the primary proof-of-principle study described in this paper, the Baranyi-Roberts equation (16) was applied to fit a growth curveto both the carbon dioxide and modified oxygen datapoints for each microorganism being studied. Theresults from the mathematical modeling could then becompared to microbial growth parameters known frommicrobial count studies. In addition, the modelingenabled the quantification of differences in the gascurves from microorganism to microorganism. Mod-eling the results of the five representative organismsrequired frequent data points in the exponentialgrowth phase. All of the data collected in the studywas robust enough for the mathematical modelingexcept for the data collected for P. aeruginosa. Theextremely steep change of the headspace carbon diox-ide and oxygen curves for this microorganism in theexponential growth phase (see Figure 4) meant thattoo few data points were available for the mathemat-ical modeling. The model calculations determined themaximum growth rate, generation time, and apparentlag phase duration for each microorganism, which arethe three most meaningful growth parameters.

Discussion of the Carbon Dioxide Data Modeling

It was possible to obtain good quality fits to the carbondioxide production curves by varying the parametersof the Baranyi-Roberts equation using regression anal-ysis with the obtained fits to the gas curves of all fourmicroorganisms having average R2 values above 0.99.An example result is shown in Figure 16 which plotsthe carbon dioxide production curve for one of theS. aureus samples along with the corresponding fitusing the Baranyi-Robert equation.

Fitting the carbon dioxide gas curves measured in eachsample for each microorganism resulted in the calcu-

lated growth parameters summarized in Table IX.From a sensitivity analysis of the Baranyi-Robertsregression model used for the fitting, it could be con-cluded that the standard deviations (SDs) seen in thecalculated growth parameters from sample to sampleof a specific microorganism are probably due to nat-ural variation rather than any regression-related issues.

Discussion of the Oxygen Data Modeling

After providing the adjustments discussed previouslyto the original oxygen consumption data, the Baranyi-Roberts equation was successfully applied to fittingthe resulting modified oxygen data. With the high R2

value as an indication (approximately 0.999 on aver-age for all data sets), the quality of the fits proved tobe as good or better than for the carbon dioxide data.An example result is shown in Figure 17, which plotsthe modified oxygen gas curve for the same S. aureussample along with the corresponding fit using theBaranyi-Robert equation.

Fitting the oxygen gas curves measured in each samplefor each microorganism resulted in the calculatedgrowth parameters summarized in Table X. Again,from the sensitivity analysis of the Baranyi-Robertsregression model, the deviations seen in the calculatedgrowth parameters from sample to sample are mostlikely the result of natural variation rather than errorsin the regression process.

Figure 16

Example of a measured CO2 growth curve, repre-sented by the log10 headspace carbon dioxide con-centrations, for TSB sample vial 134 innoculatedwith the bacterium S. aureus along with the fitobtained using the Baranyi-Robert equation.

243Vol. 70, No. 3, May–June 2016



Oxygen and Carbon Dioxide Modeling Comparison

The objective of the mathematical modeling was two-fold: (1) to gain insight into whether or not the mea-sured headspace gas dynamics reflect actual microbialgrowth, and (2) to quantify the measured gas curves sothat they could be compared with one another. Prog-ress toward meeting the first objective can be made bycomparing the calculated average lag times (repre-sented by � ) with the duration of the lag phasereported by other researchers from studies determiningmicrobial counts (19 –21). In general, the lag timescalculated from the headspace gas curves are longerthan those seen during in-use dating studies with ster-ile pharmaceutical products, that is, on the order of 6 h(22). Upon consideration, this is not surprising. Onecould theorize that microorganisms begin to grow byfirst consuming the available oxygen dissolved in theTSB media before accessing the oxygen in the head-space. In such a situation, headspace gas dynamicswould lag microbial count. It is also apparent from thegas curves in Figures 1–5 that there is not a one-to-oneconversion of oxygen molecules to carbon dioxidemolecules in the headspace as would be expected fromthe aforementioned respiration equation: C6H12O6 �6O2 3 6CO2 � 6H2O � 38 ATP. It is thereforeresonable to conclude that the headspace dynamicsare, in fact, a reflection of the specific metabolicpathways of each microorganism. Detection of micro-bial growth using headspace analysis is therefore ac-complished by measuring the results of the metabolicT

AB

LE

IXC

O2

Res

ults

Sum

mar

y

C.

albi

can

sA

.br

asil

ien

sis

S.

aure

us

B.

spiz

izen

ii

Ave

rage

SDM

inM

axA

vera

geSD

Min

Max

Ave

rage

SDM

inM

axA

vera

geSD

Min

Max

h017

.07

3.55

712

.05

25.7

24.

770.

567

4.10

6.28

20.1

91.

600

18.3

124

.53

5.32

0.29

14.

825.

96

mum

ax0.

3894

0.05

445

0.30

730.

5082

0.08

230.

0099

70.

0716

0.10

860.

4063

0.01

932

0.38

150.

4568

0.19

250.

0073

40.

1827

0.20

93

y00.

693

0.07

180.

575

0.85

50.

719

0.09

580.

555

0.84

90.

692

0.05

040.

563

0.80

40.

712

0.10

830.

524

0.89

3

ymax

4.65

10.

0544

4.52

84.

7581

4.28

60.

1752

3.92

84.

493

4.46

70.

0362

4.38

64.

547

4.74

60.

0364

4.65

64.

792

�[l

ag(h

)]43

.52.

9139

.250

.658

.01.

7853

.660

.749

.61.

9447

.153

.927

.70.

8926

.029

.1

Figure 17

Example of a modified O2 growth curve, repre-sented by the log10 headspace oxygen concentra-tions for sample 134 of the bacterium S. Aureus.




process rather than direct microbial count or evidenceof microbial growth as turbidity. A detailed discussionof the relationship between microbial metabolism andthe gaseous exchange patterns for a particular micror-ganism is outside the scope of this article. However, itis straightforward, as demonstrated here, to generateanalytical headspace gas data that correlates to micro-bial growth dynamics in a quantitative model. It issuggested by the authors that future work could inves-tigate this approach in more detail. The growth param-eters of the Baranyi-Roberts equation could be rede-fined for headspace analysis—that is, lag phase couldbe defined as gas-lag phase, maximum growth ratecould be defined as maximum gas consumption/pro-duction rate, and so on. In addition, studies could bedone to investigate the correlation between gas dy-namics and actual microbial count.

Finally, it can be concluded from the mathematicalmodeling that it is possible to make a thorough quan-titative analysis of the headspace gas dynamics mea-sured during microbial growth. This in turn enablesdetailed growth comparisons between different sam-ples of the same microorganism, between differentmicroorganisms, and between different conditions(growth media, incubation conditions). Using theBaranyi-Roberts equation, the gas curves can be de-scribed by quantitative parameters that include a gas-lag phase and a maximum oxygen consumption rate ora maximum carbon dioxide production rate. The abil-ity to straightforwardly collect analytical headspacegas data coupled with the ability to perform a thoroughquantitative analysis of the headspace gas curves mea-sured during microbial growth is a powerful combi-nation that could be applied in studies aimed at theoptimization of growth media and incubation condi-tions per microorganism.

Conclusions

Laser-based headspace analysis is a promising tech-nique for performing nondestructive, growth-basedmicrobial detection directly in pharmaceutical con-tainers. The nondestructive nature of the measurementand the fact that it can be automated for high-through-put measurements are unique advantages that enablepotentially interesting applications. Chief among theseis automated media fill inspection as the drasticchanges in the headspace composition of a contami-nated media vial during a standard incubation is ro-bustly detected by laser-based headspace analysis.Other potential applications include confirmation ofT

AB

LE

XO

2R

esul

tsSu

mm

ary

C.

albi

can

sA

.br

asil

ien

sis

S.

aure

us

B.

spiz

izen

ii

Ave

rage

SDM

inM

axA

vera

geSD

Min

Max

Ave

rage

SDM

inM

axA

vera

geSD

Min

Max

h04.

400.

734

3.31

6.44

4.29

0.62

43.

205.

515.

710.

988

4.51

8.37

4.96

0.46

34.

015.

74

mum

ax0.

0834

0.01

030.

0674

0.10

520.

0582

0.01

040.

0396

0.08

190.

1026

0.01

210.

0896

0.13

510.

1390

0.01

230.

1123

0.16

09

y01.

030

0.03

330.

974

1.08

51.

041

0.03

270.

932

1.08

41.

038

0.03

440.

983

1.09

90.

977

0.15

290.

538

1.09

8

ymax

3.08

80.

0583

2.94

53.

184

3.19

40.

1705

3.02

63.

717

2.77

60.

0361

2.68

92.

840

2.86

20.

1861

2.63

83.

412

�[l

ag(h

)]52

.53.

6347

.161

.274

.34.

3965

.781

.255

.32.

8850

.261

.935

.71.

4131

.637

.8

245Vol. 70, No. 3, May–June 2016



presumptive sterility test failures, and possible steril-ity troubleshooting of actual pharmaceutical productbatches. The headspace oxygen and carbon dioxidecurves measured to indicate microbial growth can beanalyzed with mathematical modeling. This approachenables the differences in the gaseous exchange char-acteristics from microorganism to microorganism tobe quantified and is potentially useful for calculatingrelative lag phase and maximum growth rate in termsof the headspace dynamics. The ease with which head-space gas curves can be measured nondestructively,coupled with robust quantitative analysis, enables anapproach for optimizing media and incubation timeand temperature conditions for microbial detectionusing headspace analysis.

Acknowledgements

The authors would like to thank staff at Afton Scien-tific for support in preparation and incubation of themedia vials for the primary study. For the studiesinvolving aenerobic organisms, mycoplasma, and hu-man albumin product, the authors would like to thankNigel Stapleton at Microsafe for fruitful discussion, aswell as the staff at Microsafe for support in prepara-tion and incubation of the media vials.

Conflict of Interest Declaration

The authors declare that they have no competing in-terests.

References

1. Kreyenbuehl, C.; Luemkemann, J. Determinationof Microbiological Growth in Media Filled Vialsby Tuneable Diode Laser Absorption Spectros-copy (TDLAS). Presented at Pharma Kongress,Duesseldorf, Germany, April 2012.

2. Bjorklund, G. C. Frequency-Modulation Spectros-copy: A New Method for Measuring Weak Ab-sorptions and Dispersions. Optics Letters 1980, 5(1), 15–17.

3. Templeton, A. C.; Han, Y.-H. R.; Mahajan, R.;Chern, R. T.; Reed, R. A. Rapid Headspace Ox-ygen Analysis for Pharmaceutical Packaging Ap-plications. Pharm. Technol. 2002, 26 (7), 44 – 61.

4. Lin, T. P.; Hsu, C. C.; Kabakoff, D. B.; Patapoff,T. W. Application of Frequency-Modulated Spec-

troscopy in Vacuum Seal Integrity of LyophilizedBiological Products. PDA J. Pharm. Sci. Technol.2004, 58 (2), 106 –115.

5. Mahajan, R.; Templeton, A. C.; Reed, R. A.;Chern, R. T. Frequency Modulation Spectrosco-py: A Novel Nondestructive Approach for Mea-suring Moisture Activity in Pharmaceutical Sam-ples. Pharm. Technol. 2005, 29 (10), 88 –104.

6. Cook, I.; Ward, K. Applications of HeadspaceMoisture Analysis for Investigating the Water Dy-namics within a Sealed Vial Containing Freeze-Dried Material. PDA J. Pharm. Sci. Technol.2011, 65 (1), 2–11.

7. Zuleger, B.; Werner, U.; Kort, A.; Glowienka, R.;Wehnes, E.; Duncan, D. Container/Closure Integ-rity Testing and the Identification of a SuitableVial/Stopper Combination for Low-TemperatureStorage at – 80 °C. PDA J. Pharm. Sci. Technol.2012, 66 (1), 453– 465.

8. Sudo, H.; O’Driscoll, M.; Nishiwaki, K.;Kawamoto, Y.; Gammel, P.; Schramm, G.;Wertli, T.; Prinz, H.; Mori, A.; Sako, K. Devel-opment of a Nondestructive Leak Testing MethodUtilizing the Head Space Analyzer for AmpouleProducts Containing Ethanol-Based Solutions.PDA J. Pharm. Sci. Technol. 2012, 66 (5), 434 –444.

9. Veale, J. R. New Inspection Developments. InPractical Aseptic Processing Fill and Finish, Lys-fjord, J., Ed.; Davis Healthcare International Pub-lishing/PDA: Bethesda, MD, 2009; pp 305–372.

10. Proposed Revisions to General Chapter Sterile Prod-uct Packaging—Integrity Evaluation �1207�.Pharmocopeial Forum, September 2014.

11. FDA Guidance for Industry. Sterile Products Pro-duced by Aseptic Processing—Current GoodManufacturing Practice. September 2004.

12. Cundell, A. Failure investigations in response toadverse aseptic filling simulation results. In Mi-crobial Risk and Investigations, McCullough,K. Z., Moldenhauer, J., Eds; PDA/Davis Health-care International Publishing: Bethesda, MD,2015; pp 541–560.




13. Doelle, H. W. Bacterial Metalbolism, 2nd ed.;Academic Press: New York, 1975; pp 738.

14. Riedel, T. E.; Berelson, W. M.; Nealson, K. H.;Finkel, S. E. Oxygen Consumption Rates of Bac-teria under Nutrient-Limited Conditions Appl. En-viron. Microbiol. 2013, 79 (16), 4921– 4931.

15. FDA Questions and Answers on Current GoodManufacturing Practices, Level 2 Guidance—Equipment. May, 2005.

16. Baranyl, J.; Roberts, T.A. A Dynamic Approachto Predicting Bacterial Growth in Food. Int. J.Food Microbiol. 1994, 23 (3-4), 277–294.

17. Roesti, D. Gas Headspace Analysis for Media FillInspection. Presented at the ECA Rapid Microbi-ological Methods Conference, Heidelberg, Ger-many, December 2014.

18. Brueckner, D. Determination of MicrobialGrowth by Laser Absorption Spectroscopy. Pre-sented at the Aseptikon Microbiology Conference,Mannheim, Germany, September 2015.

19. Monod, J. Growth of Bacterial Cultures. Ann. Rev.Microbiol. 1949, 3 (1), 371–394.

20. Novick, A. Growth of Bacteria. Ann. Rev. Micro-biol. 1955, 9 (1), 97–110.

21. Zwietering, M. H.; Jongenburger, I.; Rombouts,F. M.; Reit, K. T. Modeling of the BacterialGrowth Curve. Appl. Environ. Microbiol. 1990,56 (6), 1875–1811.

22. Lolas, A. G; Metcalfe, J. W. Evaluation of theMicrobial Growth Potential of PharmaceuticalDrug Products and Quality by Design. PDAJ. Pharm. Sci. Technol. 2011, 65 (1), 63–70.

247Vol. 70, No. 3, May–June 2016



Authorized User or for the use by or distribution to other Authorized Users·Make a reasonable number of photocopies of a printed article for the individual use of an·Print individual articles from the PDA Journal for the individual use of an Authorized User ·Assemble and distribute links that point to the PDA Journal·Download a single article for the individual use of an Authorized User·Search and view the content of the PDA Journal permitted to do the following:Technology (the PDA Journal) is a PDA Member in good standing. Authorized Users are An Authorized User of the electronic PDA Journal of Pharmaceutical Science and

copyright information or notice contained in the PDA Journal·Delete or remove in any form or format, including on a printed article or photocopy, anytext or graphics·Make any edits or derivative works with respect to any portion of the PDA Journal including any·Alter, modify, repackage or adapt any portion of the PDA Journaldistribution of materials in any form, or any substantially similar commercial purpose·Use or copy the PDA Journal for document delivery, fee-for-service use, or bulk reproduction orJournal or its content·Sell, re-sell, rent, lease, license, sublicense, assign or otherwise transfer the use of the PDAof the PDA Journal ·Use robots or intelligent agents to access, search and/or systematically download any portion·Create a searchable archive of any portion of the PDA JournalJournal·Transmit electronically, via e-mail or any other file transfer protocols, any portion of the PDAor in any form of online publications·Post articles from the PDA Journal on Web sites, either available on the Internet or an Intranet,than an Authorized User· Display or otherwise make any information from the PDA Journal available to anyone otherPDA Journal·Except as mentioned above, allow anyone other than an Authorized User to use or access the Authorized Users are not permitted to do the following:



Documents

The Application of Noninvasive Headspace Analysis to Media