European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326

Contents lists available at SciVerse ScienceDirect

European Journal of Pharmaceutics and Biopharmaceutics

journal homepage: www.elsevier .com/locate /e jpb

Research paper

Investigating factors leading to fogging of glass vials in lyophilized drugproducts

0939-6411/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.ejpb.2013.06.007

⇑ Corresponding author. Tel.: +41 61 68 79487.E-mail address: holger.roehl@roche.com (H. Roehl).

Ahmad M. Abdul-Fattah a, Richard Oeschger b, Holger Roehl a,⇑, Isabelle Bauer Dauphin a, Martin Worgull a,Georg Kallmeyer a, Hanns-Christian Mahler a

a Pharmaceutical Development & Supplies, Pharma Technical Development Biologics EU, F. Hoffmann-La Roche Ltd, Basel, Switzerlandb Pharmaceutical Bulk Operations Parenterals, F. Hoffmann-La Roche Ltd, Basel, Switzerland

a r t i c l e i n f o a b s t r a c t

Article history:Available online 19 June 2013

Keywords:LyophilizationFogging of glass vialsMonoclonal antibodiesSolid state characterizationVial washingVial depyrogenationMarangoni effectVisual inspectionCosmetic defect

Vial ‘‘Fogging’’ is a phenomenon observed after lyophilization due to drug product creeping upwardsalong the inner vial surface. After the freeze-drying process, a haze of dried powder is visible insidethe drug product vial, making it barely acceptable for commercial distribution from a cosmetic pointof view. Development studies were performed to identify the root cause for fogging during manufactur-ing of a lyophilized monoclonal antibody drug product. The results of the studies indicate that drug prod-uct creeping occurs during the filling process, leading to vial fogging after lyophilization. Glass quality/inner surface, glass conversion/vial processing (vial ‘‘history’’) and formulation excipients, e.g., surfac-tants (three different surfactants were tested), all affect glass fogging to a certain degree. Results showedthat the main factor to control fogging is primarily the inner vial surface hydrophilicity/hydrophobicity.While Duran vials were not capable of reliably improving the level of fogging, hydrophobic containersprovided reliable means to improve the cosmetic appearance due to reduction in fogging. Varying vialdepyrogenation treatment conditions did not lead to satisfying results in removal of the fogging effect.Processing conditions of the vial after filling with drug product had a strong impact on reducing butnot eliminating fogging.

� 2013 Elsevier B.V. All rights reserved.

1. Introduction

Cosmetic defects in lyophilizates have recently gained increasedattention. A phenomenon that has been referred to as ‘‘fogging’’ ofglass vials [1] can be described as a white haze or cloud of differentpatterns and forms after freeze drying, e.g., in the form of finger-like protrusions, branching, or uniform haze. Fogging of glass vialshas been observed for some time now in the pharmaceutical indus-try, but it had not gained much scrutiny because it has beenviewed as a non-critical cosmetic defect. However, given that fog-ging may also be observed in the vial neck region, there may beconsiderations on whether this may have an impact on containerclosure integrity (CCI), though there is no direct data or evidenceof such concern. High incidents of glass vial fogging may lead tosignificant reject rates for lyophilized drug product (DP) duringinspection (manual, semi-automatic, or automatic), by virtue ofjust being a cosmetic defect and/or fogging reaching all the wayup to the vial shoulder. Furthermore, appearance can be of specificinterest and focus for specific markets and may become costly to

the company – as is in the case of fogging – if the problem cannotbe solved or controlled.

Root causes of fogging are complex and not well researched. It isbelieved that it starts with solution wetting of glass vial walls andadsorption of solution components onto the glass inner surface,followed by solution creeping up vial inner walls due to gradientsin surface tension driven by thermal and/or compositional factors.The solution remains on the inner walls of the vials until loadedinto the freeze dryer and is dried in such state. Interestingly, thephenomenon of solution creeping up an inner container surfaceduring filling can also be observed in daily life: when filling coffeeinto a clean coffee mug, the quick rise of a film of coffee on con-tainer surface can be observed. The rise of wine along the wallsof a glass cup is another example that has been brought forwardfor creeping (Tears of Wine). When the wine is placed in a glasscup, it climbs along the walls to wet the walls in the same mannercapillary rise does [2], although the ‘‘Schlieren’’ phenomena inwine can also be attributed to alcohol content or other parameters.

Studies are available in the literature that investigated themechanism of creeping and film transfer through observing creep-ing behavior of charged nanoparticles in aqueous solutions to theinterior glass surface of vials or containers [3–5]. It was suggestedthat Marangoni flow [2–6] was a possible mechanism for film

A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326 315

transfer to a glass solid surface. In aqueous formulations, especiallywhen containing a surfactant, and possibly even with a hydrationfilm alongside the walls of a vial (i.e., pre-wetting of the interiorsurface, e.g., by a thin layer of condensed water vapor from the sur-rounding) [3], a difference in the surface tension (dc) between twopoints on the solution surface (i.e., surface tension gradient) trig-gers a driving force for fluid flow toward the region of high c value[2,3,7]. As described by Levich [7], surfactant molecules adsorbalong fluid interfaces, where they lower the interfacial tension.Convection in solution tends to increase (or decrease) the surfaceconcentration of adsorbed surfactant near zones where the flowconverges (or diverges). However, both adsorptive/desorptive andbulk diffusive fluxes tend to reduce gradients in surface concentra-tion. If either of these fluxes is slow, a non-uniform distribution ofadsorbed surfactant is established, causing a gradient in the inter-facial tension [7]. At the end, the driving force for fluid to flow to-ward the region of high c must be strong enough to overcome theresistance of the fluid to flow (viscosity, g) and fast enough to avoidequilibration of all gradients by diffusion [8]. The resulting transferof adsorbed surfactant molecules from the regions of lower surfacetension toward the regions of higher surface tension constitutesthe Marangoni effect or Marangoni convection (Fig. 1). Due to massconservation, the fluid recirculates in the bulk, which creates thetypical pattern for Marangoni convection. The difference dc canbe due to temperature gradients at the interface (the thermocapil-lary effect) or concentration gradients (the destillocapillary effect)[9]. Marangoni convection can manifest as macro-convection,where convection originates from concentration or temperaturedifferences due to an asymmetry in the system, or micro-convec-tion, where the convection is initiated by small (random) temper-ature or concentration disturbances that grow with time [8–11].

Previous studies suggest that Marangoni flow/convection isinfluenced by the interaction between formulation and primarypackaging container. The extent of this interaction will influencesolution creeping behavior and ultimately influence the degree offogging after lyophilization. The degree of interaction between for-mulation and primary packaging container is a function of formu-lation composition (e.g., pH, composition, ionic strength, viscosity),as well as surface properties of glass. The current study was donein the context of researching glass fogging of a commercial product

Fig. 1. A simplistic illustration of the interaction between surface active solutionsand hydrophilic glass vials. (For interpretation of the references to color in thisfigure legend, the reader is referred to the web version of this article.)

(labeled as mAb1 in this study). During the investigation to solvethe problem, different variables were systematically evaluated:

1. Formulation composition: Surface activity is known to impactMarangoni effect and subsequent potentially creeping, but itis not clear to what extent it will impact fogging, especially withingredients known to adsorb onto glass. To our knowledge,there is no study that systematically correlates fogging with for-mulation properties/composition. We investigated what formu-lation ingredients contribute to fogging (while keeping pH andionic strength in all test solutions relatively the same) usingcommonly used vials for lyophilization.

2. Surface properties of glass: Glass surface properties are known todiffer between different glass types/vendors, as well as withinone single lot of glass vials [12]. The inner surface propertiesof glass vials are influenced by the glass composition, the glassforming process [12,13] as well as storage conditions of the pri-mary packaging containers. The resulting differences in surfaceproperties between different glass vials are expected to result invarying degree of interaction with formulation ingredients andhence influence creeping and ultimately fogging. To our knowl-edge, there is no study that investigates fogging as a function ofglass surface. In this study, we investigated the impact of vialglass quality/inside coating on fogging. Vials with inner surfacecoated with baked-on silicone versus ‘‘unsiliconized’’ surfaceswere used.

3. Process conditions: The extent of formulation-container closureinteraction can be influenced, in theory, by altering process con-ditions. Processing conditions that ‘‘favor’’ this interaction areexpected to worsen the fogging problem post-lyophilization,and vice versa. For example, it was shown that changes in topo-logical structure and chemical composition of the inner surfaceof unsiliconized glass vials occur after washing and depyrogen-ation, depending on the process and related controls [12].Because glass vials are to be used for parenterals, they needto be washed, depyrogenated, and sterilized according to theprescribed methods (e.g., EU and US Good Manufacturing Prac-tice) before the pharmaceutical solutions are filled [1]. Theimpact of washing, depyrogenation, and sterilization on glassvial inner surface is a subject of research. To our knowledge,there is no study that investigates fogging as a function of pro-cess or cold temperature exposure. In our studies, we investi-gated if there is an impact of the vial washing anddepyrogenation step on fogging post-lyophilization (using unsi-liconized and siliconized glass vials). Furthermore, we lookedinto the impact of modifying pre-lyophilization conditions(exposing vials containing solution of drug product to cold/refrigerated temperatures prior to freeze drying) on fogging.

Through our studies, we will discuss possible process improve-ments and solutions to control the fogging problem in develop-ment and at production scale.

2. Materials and methods

2.1. Materials

2.1.1. ChemicalsPharmaceutical quality recombinant humanized monoclonal

antibody 1 (mAb1) was produced and purified (>99%) at Roche,Penzberg. The antibody formulation used was 0.01% w/v Polysor-bate 20, 60 mM Trehalose, 5 mM Histidine/Histidine HCl all at pH6.0, at 25 mg/mL of mAb1.

Similarly, a pharmaceutical quality recombinant humanizedmonoclonal antibody 2 (mAb2) was produced and purified

Table 1Washing and depyrogenation treatments applied to unsiliconized Duran vials(Vendor 2) and siliconized Duran vials (Vendor 1).

Glass type Washing Pre-drying/pre-tunneltemperature(�C)

Depyrogenationtemperature(�C)

Batchlabel

15 mLunsiliconizedDuran vials(US)

No – – UntreatedUS

Yes 25 280 US25280Yes 60 280 US60280Yes 60 300 US60300Yes 25 320 US25320Yes 60 320 US60320

15 mL siliconizedvials (S)

No – – UntreatedS

Yes 25 280 S25280Yes 60 280 S60280Yes 60 300 S60300Yes 25 320 S25320Yes 60 320 S60320

316 A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326

(>95%) at Roche, Penzberg. Formulations used for mAb2 were thefollowing:

1. 20 mM Histidine/Histidine HCl all at pH 5.5, at 25 mg/mL ofmAb2.

2. 0.01% w/v Polysorbate 20, 5 mM Histidine/Histidine HCl all atpH 5.5, at 25 mg/mL of mAb2.

3. 0.01% w/v Polysorbate 20, 250 mM Trehalose, 20 mM Histidine/Histidine HCl all at pH 5.5, at 25 mg/mL of mAb2.

Neither the impact of pH nor ionic strength variations on eachprotein were tested. Besides mAb solutions mentioned above,some placebo solutions were prepared for testing at developmentscale. The compositions of most placebo solutions prepared werebased on the composition to the mAb1 DP formulation. The follow-ing placebo/formulations were prepared for this study:

1. Placebo 1: 250 mM Trehalose, 20 mM Histidine/Histidine HClall at pH 5.5.

2. Placebo 2: 0.01% w/v Polysorbate 20, 60 mM Trehalose, 5 mMHistidine/Histidine HCl all at pH 6.0.

3. Placebo 3: 0.03% w/v Polysorbate 20, 60 mM Trehalose, 5 mMHistidine/Histidine HCl all at pH 6.0.

4. Placebo 4: 0.05% w/v Polysorbate 20, 60 mM Trehalose, 5 mMHistidine/Histidine HCl all at pH 6.0.

5. Placebo 5: 0.01% w/v Polysorbate 80, 60 mM Trehalose, 5 mMHistidine/Histidine HCl all at pH 6.0.

6. Placebo 6: 0.05% w/v Polysorbate 80, 60 mM Trehalose, 5 mMHistidine/Histidine HCl all at pH 6.0.

7. Placebo 7: 0.2% w/v Polaxamer 188, 60 mM Trehalose, 5 mMHistidine/Histidine HCl all at pH 6.0.

8. Placebo 8: 0.5% w/v Polaxamer 188, 60 mM Trehalose, 5 mMHistidine/Histidine HCl all at pH 6.0.

9. Placebo 9: 25 mg/ml Bovine Serum Albumin, 0.01% w/v Polysor-bate 20, 60 mM Trehalose, 5 mM Histidine/Histidine HCl all atpH 6.0.

Polysorbate 20 and Polysorbate 80 were purchased from Croda,and Polaxamer 188 (Pluronic F68) was purchased from BASF. Bo-vine Serum Albumin (BSA) was purchased from Sigma. Trehalosewas purchased from Ferro Pfanstiehl Laboratories.

2.1.2. Vials and closure systemsAll vials used for this study were 15 mL tubing (tubular glass)

vials. Duran vials (33 Expansion Glass) were purchased from 2 dif-ferent vendors labeled as Vendor 1 and Vendor 2. A siliconized ver-sion of the vials (51 Expansion Glass) with silicone baked-oninternal vial walls (Silicone emulsion sprayed/fixed by heat) waspurchased from Vendor 1. 20 mm Lyo stoppers were obtained fromDaikyo.

2.2. Methods

2.2.1. Surface tension measurementsSurface tension of some solutions was measured using a Krüss

easy drop (Hamburg, Germany), DSA Version 1.92.1.1 (workingprinciple of instrument relies on pendant drop method). 0.8 mLof the sample solution was drawn into a 1 mL syringe with a1.8 mm diameter cannula and the syringe clamped onto the devicefollowed by drop application and surface tension measurements.Surface tension measurements were done in triplicate measure-ments. Data were analyzed using DSA1-software.

2.2.2. Vial washing and depyrogenation/sterilizationVials were handled either untreated, washed followed by steam

sterilization at P121 �C (Lytzen Dry Heat Sterilizer, Herlev, Den-

mark), or washed followed by sterilization in a depyrogenationtunnel. A summary of processing conditions in the tunnel on differ-ent vial types is provided in Table 1.

Unless otherwise specified, vials that underwent tunnel treat-ment were washed in a Bausch & Ströbel FAU6000 vial washer con-nected to a Bausch & Ströbel, DHT 3670 drying tunnel throughwhich the vials were passed for depyrogenation. Washing wasdone at 60 �C using distilled water. Depyrogenation conditions inthe tunnel are described in Table 1. The drying tunnel is essentiallymade of three zones: a heating/pre-warming zone 740 mm long (todry wet vials), a depyrogenation zone 1380 mm long (wheresterilization is performed), and a cooling/destabilizing zone1626.5 mm long (where vials are cooled and brought down toroom temperature before filling). The belt speed of this model isadjusted at a constant value of 97 mm/min. This means each vialspends an average of �35 min in the entire cycle, with a residencetime of �14 min in the depyrogenation zone. The machine settingsdo not allow for varying belt speed; hence, we were not able tovary tunnel drying time in our studies. Batch size was roughly500 vials per run. During the pre-heating/pre-drying and depyro-genation runs, the actual temperature of the tunnel was monitoredto observe differences from the set temperature. Additionally, tem-perature of select vials at different locations was monitored duringthe sterilization run using Kaye Validator 2000 thermocouplesconnected to a Kaye Validator control unit (GE Kaye Instruments,Billerica, MA) to measure variability in vial temperature duringvial depyrogenation. Regardless of location, vial temperatureprofiles were similar (results not shown). Additionally, no differ-ences were observed between vial temperatures of siliconizedversus unsiliconized vials under the same depyrogenation condi-tions (results not shown).

2.2.3. Vial characterization studiesThe inner surfaces of vials used for lyophilization runs in the

current study were analyzed via surface energy measurementsand Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS).Note that none of these methods is suggested as a quality controlmethod for primary packaging and rather intended forcharacterization.

Surface energy measurements of different glass surfaces/vialswere performed using test inks with specified surface energies(Arcotest, Mönsheim, Germany). A definite volume of the test inkswas applied on the interior surface of the vials and surface energyestimated. Because no ink for each surface energy value is

A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326 317

commercially available, samples were evaluated against a refer-ence value of 72 mN/m (a characteristic value for clean hydrophilicglass surfaces) and 22 mN/m (a characteristic value for cleanhydrophobic glass surfaces). Three cases could occur with thetested vials;

Droplet formation: surface energy of the sample is lower thanthat of ink used (rsurface < rink).Ink drop spreads: surface energy of the sample is higher thanthat of ink used (rsurface > rink).No spreading of ink droplet: surface energy of the sample and inkare roughly equal (rsurface = rink).

Please note that the method to measure surface energy is only arigorous method for estimation. It is not suggested as a method formaking conclusions about subtle differences between differentglass vials. The determination of surface energy by inks with spe-cific surface energies gives only rough estimates of the exact value,especially for hydrophilic glass surfaces. Commercially availableinks have larger steps in surface energy. Contact angle measure-ments on curved vial surfaces exhibit large errors, especially forhydrophilic surfaces and contact angles below 10� cannot be mea-sured reliably.

Time-of-Flight Secondary Ion Mass Spectrometry (ToF-SIMS)spectra of different glass vial surfaces (static mode) were acquiredusing a ToF-SIMS IV-100 instrument (ION-TOF, Münster, Germany).Vials were cut and analyzed without any further preparation. Spec-tra were acquired using a 15 keV 69Ga+ primary ion beam. In order toachieve highest mass resolution (m/Dm � 10,000, where Dm is thewidth of the peak measured at 50% of the total peak height = fullwidth at half maximum (FWHM)), the spectra were acquired inthe so-called ‘‘Bunched mode’’ using a typical target current of theprimary Ga+ ions of about 2 pA. For depth profiling, the instrumentis equipped with a low energy (0.25–2 keV) dual ion source column(DSC) including a Cs+-liquid metal- and an Oþ2 electron impact-gun.The depth profile acquisitions were done in the ‘‘dual beam mode.’’In this mode, the parameters of analysis gun can be selected inde-pendently from the sputter gun parameters. In order to suppresssecondary ion signals originated nearby the crater edges, the analy-sis beam was rastered over an area significantly smaller than thesputter crater size (‘‘gating’’). In all cases, charge compensationwas achieved by applying low-energy electrons (20 eV) from apulsed electron flood gun.

2.2.4. Freeze-drying experimentsFreeze drying in development was performed using a LyoStar II

freeze dryer (FTS Systems, Stone Ridge, NY) associated with Lyo-manager II, FTS Systems software. The freeze dryer has 2 shelves,with 16 thermocouple plugs to monitor temperature. Total shelfsurface area is 0.28 m2. The chamber pressure was measured andrecorded by a capacitance manometer, and the apparent chamberand condenser pressures were measured and recorded by Piranigauges.

All formulations were filtered through a Millipore 0.2 lm filterunder a laminar flow hood (LAF) prior to filling the vials. Solutionswere filled into different vial types/treatments (as described under‘‘vials and closure systems’’) using a repeater pipette (EppendorfXStream) and 50 mL sterile Combitips with a fill volume of6.2 mL. Sterile stoppers were placed in a semi-stopper positiononto vials using sterile forceps to allow for efficient sublimationand desorption of water molecules during the freeze-drying cycles.After labeling, filling and semi-stoppering, vials were loaded ontothe lyophilizer for freeze drying. For freeze-drying experiments,50–53 vials per vial type/treatment were filled for mAb solutions,while 25–26 vials per vial type/treatment were filled for non-mAb solutions (placebos). For each lyophilization cycle, one shelf

was loaded with approximately 150 vials. Product temperaturewas monitored in 1–2 vials per formulation per run using a gaugethermocouple positioned in solution center slightly above the bot-tom of the vial. Before performing the lyophilization runs, vialswere placed in trays in a hexagonal array to ensure uniform heatdistribution during the runs. To reduce the impact of radiative heattransfer during the lyophilization runs, in each tray the 3 frontrows of vials and all vials at the periphery were empty. In addition,aluminum foil was placed on the Plexiglas front door.

Vials were loaded at a loading temperature of 5 �C. Equilibrationwas performed at 5 �C for 1–22 h prior to freezing (based on feed-back from production, 22 h was used to mimic the time the firstbatch of vials loaded onto the production scale freeze dryer remainat the cold loading temperature before initiation of the freeze-dry-ing cycle, as a ‘‘worst’’ case scenario). The lyophilization cycle fol-lowed equilibration at 5 �C according to a pre-defined program.

Commercial scale experiments with freeze drying were per-formed using a Klee (LYO 500) lyophilizer (OPTIMA pharma GmbH,Schwaebisch Hall, Germany) with a liquid nitrogen cooling systemand a total shelf surface area of 30 m2. The freeze dryer has 15shelves. Chamber pressure was measured and recorded by a capac-itance manometer, and the apparent chamber and condenser pres-sures were measured and recorded by Pirani gauges.mAb1 solutionwas filtered through a filter cartridge prior to filling the vials. Anew batch of Duran� vials were used for production runs, differentthan those used for development runs. A small number of vials(400) were filled manually with mAb1 solution using a Multipette�

filler. These vials were depyrogenated in the same tunnel used fordepyrogenating vials used for development. Pre-warming/depyro-genation conditions were the following: pre-warming at 25, 45,and 60 �C followed by depyrogenation at 320 �C; pre-warming at60 �C followed by depyrogenation at 300 �C; and finally pre-warm-ing at 25 and 60 �C followed by depyrogenation at 280 �C. Theremaining 46,400 vials underwent standard depyrogenation treat-ment and were filled with 1.5% w/v Mannitol solution. Before load-ing vials onto trays, the vial stoppers were placed onto vials in asemi-stopper position to allow for efficient sublimation anddesorption of water molecules during the cycle. Nominal fill vol-ume was set to 6.3 mL. Please note that filling of mAb1 solutionwas performed within 1 h of washing and depyrogenation of vials.After labeling, filling and semi-stoppering, vials containing mAb1were loaded onto the loading trolleys (at pre-defined locations).Loading of the lyophilizer for freeze drying was done automaticallyat a loading temperature of 10 �C. Product temperature was mon-itored in 16 vials using 16 Tempris probes (8 Tempris probesplaced 5 mm below the liquid level and 8 Tempris probes placed2 mm above vials bottom).

Vials containing mAb1 (including thermocouple vials) – for la-ter assessment on fogging – were placed on shelf 1 (first vials in.Vials would potentially be kept max. 24 h holding time at 10 �C be-fore freezing) and shelf 15 (last vials in. Vials may be kept for max.1 h holding time at 10 �C before freezing). Equilibration was per-formed at 5 �C for 1 h prior to freezing. The lyophilization cyclewas comparable to the development cycle. Preliminary studies alsoshowed that the extent of fogging did not differ according to lyoph-ilization cycle when the same type of product and vial was used(results not shown).

2.2.5. Visual assessment of fogging in lyophilized products100% of vials were visually assessed for fogging. Fogging in vials

was inspected using an APK-01 inspection machine (Eisai Machin-ery GmbH, Cologne, Germany) equipped with a VHX-600 DigitalMicroscope camera system (Keyence, Atlanta, GA, USA) and anadditional light source KL 2500 LCD (Schott, Mainz, Germany).

In general, visual inspection included criteria such as presenceor absence of fogging, extent of fogging around vial circumference,

318 A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326

vials where fogging has reached a potentially critical height on thevial wall but below shoulder (would be considered potential re-jects), the presence of fogging on shoulder (rejects), fogging pat-tern, and fogging density (light, medium, dense).

3. Results and discussion

3.1. Development studies: effect of formulation and pre-lyophilizationconditions (equilibration during loading)

Pilot studies were performed to evaluate a potential correlationbetween fogging and formulation (ingredients and surface ten-sion), as well as vial type. Solutions of interest were freeze-driedin two types of vials that were used ‘‘as is’’: siliconized (S) vials(baked-on silicone) from Vendor 1 and unsiliconized (US) Duranvials from Vendor 2. ‘‘As is’’ refers to untreated vials. Furthermore,freeze drying was performed in unsiliconized Duran vials fromVendor 2 obtained from one commercial production site pre-sub-jected to standard depyrogenation treatment (washing followedby depyrogenation at 320 �C) to get an initial indication if vialtreatment impacts fogging behavior.

All formulations displayed some level of fogging when freeze-dried in unsiliconized vials (Table 2). Occurrence of fogging with pla-cebo 1 (surfactant-free formulation with high surface tension) mayindicate that contact angle and/or adsorption to hydrophilic glasssurfaces play a role in the appearance of the phenomenon withnon-surface active solutions. In unsiliconized vials, fogging also oc-curred in the presence of surfactant, regardless of surfactant typeand level tested (P0.01% w/v). Fogging occurred regardless if thevials were used ‘‘as is’’ or if they were ‘‘treated’’ before-hand (wash-ing followed by depyrogenation). Most protein formulations containa surfactant (e.g., Polysorbate 20 or Polysorbate 80) in order to min-imize or avoid (potential) aggregation or precipitation aftermechanical or freeze/thaw stress [15,16]. Surfactants are known topromote solution wetting on hydrophilic surfaces, lower contact an-gle, and perhaps even adsorb themselves onto hydrophilic glass sur-faces. The extent of fogging changed in formulations with bothprotein and surfactant (BSA, mAb1 DP, mAb2 DP). Compared to theirequivalent placebo formulations with surfactant, a greater percent-age of vials were observed where fogging either reached ‘‘criticalheight’’ inside the vial (potential reject) or reached the shoulder (re-ject). This increase in (potential) rejects may be associated with asynergistic protein–surfactant effect in worsening fogging appear-ance in freeze-dried formulations. Furthermore, fogging of the sameformulation (BSA, mAb1 DP, mAb2 DP) in ‘‘treated’’ unsiliconizedvials appeared worse compared to the same formulation in un-treated vials. As described by Schwarzenbach et al, vials before thewashing step contain certain types and levels of what we would con-sider as contaminants arising from the vial forming process [12]. Thepresence of these contaminants may contribute to reducing the ex-tent of fogging. Preliminary analysis by ToF-SIMS of the unwashedvials used in this study revealed phthalate-like contaminants (re-sults not shown). Finally by freeze drying all solutions in siliconizedvials (baked-on silicone), a significant reduction in or disappearanceof fogging was observed (Table 2). Decreasing surface energy byusing hydrophobic inner surface treatment of vials has been pro-posed as a solution to reduce/eliminate, adsorption, and creeping.This could prevent adsorption of formulation components to theglass walls and hence progression of creeping and resulting foggingafter freeze drying. Thus, the use of containers with a hydrophobicinner surface would minimize or prevent cosmetic defects due tofogging [1]. An additional advantage of a hydrophobic coating insidethe vial is that it improves cake elegance and also (slightly) increasesextractable volume.

In a follow up pilot study to further confirm previous observa-tions with more vials per lyophilization run (53 vials per formula-

tion), flexible bulk (FB) solutions of mAb2 was freeze-dried with orwithout surfactant (0.01% w/v Polysorbate 20). Solutions werefreeze-dried in unsiliconized Duran vials from Vendor 2 pre-sub-jected to depyrogenation (washing followed by depyrogenationat 320 �C). Results are summarized in Table 2 and representativepictures of lyophilized product are shown in Fig. 2a–d. Even with-out surfactant, fogging occurred in presence of protein-only solu-tions (Fig. 2a and c). Most proteins are surface active andcharged [17] and are well known to adsorb onto borosilicate glasssurfaces. Borosilicate glass, in aqueous solution, is typically presentin a charged state. Adsorption of proteins to hydrophilic glass sur-faces is driven to a great extent by charge–charge interactions[18,19]. Fogging was difficult to assess in protein only solutionsand was characterized as being random and discontinuous (un-even) fogging around the vial (Fig. 2a). As previously observed,the addition of surfactant seemed to change fogging behavior(Fig. 2b and d). In the presence of surfactant, fogging in freeze-dried mAb2 was non-uniform in terms of height and intensity overcircumference around the vial and close to 20% of vials displayed‘‘feathering’’ behavior on the walls (Fig. 2b). ‘‘Feathering’’ refersto freezing pattern of water to ice where pattern appears as icebranching and appears as feathers to the observer. Additionally,fogging was more opaque and dense than with protein only(Fig. 2a and b). However, there was a notable decrease in percent-age of vials where fogging either reached ‘‘critical height’’ insidethe vial or reached the shoulder.

During the pilot studies to assess the effect of formulationingredients on fogging, we also observed a ‘‘processing’’ effect.mAb2 bulk solution (FB) discussed in the previous paragraph (withand without surfactant) were freeze-dried with a modification.Pre-freezing conditions were modified before start of the freezedrying run where vials were equilibrated on the shelves at 5 �Cfor 10 h before start of freezing (as opposed to only 1 h for all pre-vious runs). A significant change in fogging behavior was observedafter freeze drying by applying this low temperature equilibrationstep prior to freezing, in what can only be described as ‘‘receding’’of fogging. In both absence and presence of surfactant, fogging wasmore uniform around vial circumference and more even/less ran-dom in height (Fig. 2c and d). Additionally, there was a significantdecrease in the percentage of vials where fogging either reached‘‘critical heights’’ inside the vial (rendering the vials as potential re-jects) or reached the shoulder (rendering the vials as definite re-jects), regardless of presence or absence of surfactant. Also less‘‘feathering’’ was observed with the surfactant vials pre-cooledfor 10 h (around 5% of the vials) versus those pre-cooled for only1 h (around 20% of the vials). Fogging had a more opaque/denserappearance, though, after 10 h of the pre-cooling treatment thanwith only 1 h, regardless of presence or absence of surfactant.

In summary, preliminary assessments showed that formulationparameters tested (surface tension and ingredients), sterilizationtreatment, vial inner surface, and exposure of solutions to low tem-perature for sufficient amount of time prior to freeze drying all im-pacted fogging appearance in freeze-dried product. Overall, foggingseemed to improve with siliconized vials and when shelves withvials containing DP were pre-cooled for relatively long periods oftime. Process parameters and vial inner surface were further inves-tigated in the sections to follow.

3.2. Development studies: effect of tunnel depyrogenation conditions

This section describes the outcome of freeze-drying mAb1 DP inunsiliconized Duran vials from Vendor 2 as well as in siliconizedvials (baked-on silicone) all pre-subjected to different tunnel dry-ing conditions for depyrogenation. Batch size for mAb1 DP was50–53 vials per batch.

Table 2Solution surface tension measurements and visual inspection for fogging in freeze-dried placebo formulations with different levels of surfactant and in drug product (DP) formulations of mAb1 and mAb2. Unsiliconized vials that wereused were obtained from Vendor 2.

Solution Surfactanttype andlevel (% w/v)

Surfacetension ofsolutionsa

(mN/m)

Siliconized vials Unsiliconized untreated vials Unsiliconized treated vials

% Vials with fogging % Vialswithfogging

% Vials with foggingaround entirecircumference

% Vials withfogging to/atshoulder

% Vials with sub-shoulder criticalfogging height

% Vialswithfogging

% Vials with foggingaround entirecircumference

% Vials withfogging to/atshoulder

% Vials with sub-shoulder criticalfogging height

Placebo 1 None 72.3 NP 100 40 0 0 100 15 0 0Placebo 2 0.01% PS20 53.7 0 100 0 0 10 100 60 0 0Placebo 3 0.03% PS20 48.6 0 100 50 0 0 100 10 0 0Placebo 4 0.05% PS20 45.2 0 100 20 0 0 100 90 0 0Placebo 5 0.01% PS80 57.5 0 100 10 0 0 100 80 0 0Placebo 6 0.05% PS80 49.9 0 100 60 0 0 100 80 0 0Placebo 7 0.2% P188 56.7 0 90 0 0 0 100 100 0 0Placebo 8 0.5% P188 51.8 0 100 10 0 0 100 0 10 0

Solution Surfactanttype andlevel

Surfacetension ofsolutionsa

(mN/m)

Fiolax vials fromVendor 1 (baked-onsiliconized vials)

Duran vials from Vendor 2 (untreated) Duran vials from Vendor 2 (treated)

% Vials with fogging % Vialswithfogging

% Vials with foggingaround entirecircumference

% Vials withfogging to/atshoulder

% Vials with sub-shoulder criticalfogging height

% Vialswithfogging

% Vials with foggingaround entirecircumference

% Vials withfogging to/atshoulder

% Vials with sub-shoulder criticalfogging height

BSA Placebo 9 0.01% PS20 58.4 80 100 80 10 0 100 90 10 10mAb1 DP 0.01% PS20 55.8 0 100 90 0 0 100 100 0 10mAb2 DP 0.01% PS20 52.2 40 100 100 0 0 100 100 10 0mAb2 FB None NPb NP NP NP NP NP 100 83 32 37mAb2 FB 0.01% PS20 NP NP NP NP NP NP 100 78 11 13mAb2 FBc None NP NP NP NP NP NP 100 84 3 5mAb2 FBc 0.01% PS20 NP NP NP NP NP NP 100 82 2 3

a n = 4–6 measurements per sample. Standard deviation of surface tension measurements did not exceed 2.4%.b NP = not performed.c Loading conditions were modified for this batch before start of the freeze drying run (vials kept on shelves at 5 �C for 10 h before start of freezing).

A.M

.Abdul-Fattah

etal./European

Journalof

Pharmaceutics

andBiopharm

aceutics85

(2013)314–

326319

Fig. 2. Images of vials of mAb2 reformulated flexible bulk freeze-dried in Duranvials from Vendor 2 (washed and depyrogenated): (a) without surfactant (1 h at5 �C pre-freezing), (b) with surfactant (1 h at 5 �C pre-freezing), (c) withoutsurfactant (10 h at 5 �C pre-freezing), and (d) with surfactant (10 h at 5 �C pre-freezing).

320 A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326

Vials depyrogenated under different conditions were character-ized by test inks to determine the respective surface energy and byToF-SIMS measurements. Surface energy values are summarized inTable 3. Surface energy measured for untreated unsiliconized vialsand all vials washed and depyrogenated at 280 �C was <72 mN/m.Vials washed and depyrogenated above that temperature hadmixed values. Several samples had surface energies higher or lowerthan 72 mN/m, indicating inhomogeneities between the surfaceconditions of the samples. Note that vials surfaces are known to

be very heterogeneous in structure and composition, thus resultingin much variability. In contrast, all siliconized (baked-on silicone)vials consistently exhibited low surface energy values (�22 mN/m), regardless of the depyrogenation treatment. All unsiliconizedvials showed relatively clean and comparable surfaces since ToF-SIMS spectra were dominated by substrate peaks (see example inFig. 3a). Similarly, all unsiliconized vials showed comparable sur-faces by ToF-SIMS. ToF-SIMS spectra for siliconized vials werefound to be dominated by silicone peaks (see example in Fig. 3b).The absence of substrate peaks indicates a closed silicone layer.

Fig. 4 compares the progress of three representative lyophiliza-tion cycles performed in development with different pre-freezingequilibration times at 5 �C (1 h, 5 h, and 22 h). Average producttemperature and chamber pressure for all three runs were compa-rable. Furthermore, average product temperature was well belowshelf temperature (and product collapse temperature) during pri-mary drying. Thermocouple data show that primary drying in allcases ended well before progress to secondary drying. Cakes pro-duced from freeze drying were elegant with excellent retentionof cake structure. Cake moisture levels in different vial treatmentsand types tested were well below 2% w/w (see Table 3).

100% of vials were visually assessed for fogging. Results aresummarized in Table 3. For interpretation of the labeling systemin this and in the following paragraphs (for example, US25320),we remind the reader to refer to Table 1. As is consistent with sur-face energy and ToF-SIMS results, no vials with fogging were ob-served in any siliconized (S) vials, regardless of whether theywere untreated or treated (washing followed by depyrogenation).However, fogging was observed in all unsiliconized (US) vials –at varying degrees – regardless of whether they were untreatedor treated. Overall, best results in terms of improvement in foggingappearance when using US vials was observed with untreated vialsand with US60320 vials. Contamination could be one factor to ex-plain less fogging with untreated vials, but we currently have noexplanation for improvement in fogging appearance withUS60320 vials. Except for mAb1 DP in untreated vials andUS60320 vials, there were different levels of vial rejects due to fog-ging reaching shoulder height. % rejects were highest with vials la-beled US60300. Overall, there were poor correlations betweensurface energy measurements and % vial rejects due to foggingreaching shoulder height. In the remaining vials that were not re-jected due to fogging at shoulder level, % vials with sub-shouldercritical fogging height were higher with vials labeled US60280,US60300, and US25320. A second notable difference betweenmAb1 DP lyophilized in US60320 vials versus the rest of treat-ments is that fogging appeared receded (or thinned out) in 89%of DP freeze-dried in US60320 vials. However, one side effectwas that small patches and spots still remained and appeared atshoulder level in up to 48% of the vials. Some receding was also ob-served in mAb1 DP freeze-dried in US60280 and US60300 vials, butto a much lower extent.

The same freeze-drying runs were also performed on solutionsof placebo 2, placebo 9 (BSA DP), and mAb2 DP in different vialtypes. Results are summarized in Table 4. Consistent with mAb1DP results, no vials with fogging were observed in any siliconizedvials, regardless of depyrogenation treatment. With placebo 2,most vial rejects in unsiliconized vials due to shoulder level fog-ging was observed with freeze-dried product in US60300 vials.Similarly, fogging was also significant when placebo 9 and mAb2DP were freeze-dried in US60280 and US60300 vials. Appearancewith regard to fogging significantly improved in placebo 9 whenfreeze drying was performed in US60320 vials.

Diminished fogging in US60320 vials compared to other treat-ments could however not be explained efficiently by either ToF-SIMS or surface energy measurements. Differences between heattreatments are known to cause differences in glass thermal history,

Table 3Impact of depyrogenation treatment of vial and vial type on characteristics of freeze-dried mAb1 DP.

Inner vialsurface

Vialtreatmenta

Surfaceenergy(mN/m)

mAb1 DPmoisturecontent (%w/w)

Equilibrationtime at 5 �C priorto freezing stage(h)

Visual inspection for freeze-dried mAb1 DP

% Vialswithfogging

% Vialswithfogging to/at shoulder

% Vials withsub-shouldercritical foggingheight

% Vials withfogging aroundentirecircumference

% Vialswithrecededfogging

% Vials withpatches orspots atshoulder

Unsiliconized(US)(Duranvials)

Untreated <72 NPb 1 100 0 0 90 0 0US25280 <72 0.6 ± 0.1 1 100 18 44 83 0 0US60280 <72 1.8 ± 0.0 1 100 25 64 92 13 0US60300 Mixed 1.8 ± 0.0 1 100 52 84 72 8 0US25320 >72 0.7 ± 0.0 1 100 25 63 80 0 0US60320 672 1.8 ± 0.0 1 96 0 4 41 89 48US25320 >72 0.5 ± 0.1 5 100 8 51 100 94 4US25280 <72 NP 22 100 8 17 98 79 4US25320 >72 0.5 ± 0.0 22 100 2 12 88 90 51US60300 Mixed 0.5 ± 0.1 22 100 0 22 84 92 53

Siliconized (S)vials(baked-onsilicone)

Untreated �22 NP 1 0 No further assessment necessary due to absence of foggingS25280 �22 0.6 ± 0.0 1 0 No further assessment necessary due to absence of foggingS60280 �22 NP No freeze drying runs performedS60300 �22 NP No freeze drying runs performedS25320 �22 NP 1 0 No further assessment necessary due to absence of foggingS60320 �22 1.4 ± 0.4 1 0 No further assessment necessary due to absence of fogging

a Refer to Table 2 for a description of depyrogenation treatment for each vial labels.b NP = not performed.

A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326 321

as well as surface restructuring (e.g., bonding reactions takingplace on glass surface silanol groups). It is considerable that afterremoval of contamination from glass vials by the rigorous washingprocesses followed by depyrogenation, and as a result of differentsurface restructuring as a function of thermal history, the newly

Fig. 3. Positive ToF-SIMS spectra for the following vials: (a) unsiliconized, untreated, andfrom Vendor 1. (For interpretation of the references to color in this figure legend, the re

formed inner glass surface interacts differently with the same DPsolution, and hence the extent of creeping after filling in the glassvials and subsequent extent of fogging after freeze drying. At themore aggressive depyrogenation treatments, different factorscould be relevant for the way inner glass surfaces of vials react

washed Duran vial from Vendor 2 and (b) siliconized untreated (baked-on silicone)ader is referred to the web version of this article.)

Fig. 4. Progress of freeze drying from three representative runs with different equilibration times at 5 �C pre-freezing: purple 1 h, red 5 h, orange 22 h. Dotted lines representpressure data from capacitance manometer (CM), dashed lines represent set point temperatures (Shelf SetPt), and solid lines represent average product temperature (Tp) forall three runs.

322 A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326

with aqueous solution of drug product. First, structural relaxation(i.e., annealing) of the glass surface is expected to happen duringheat treatments to different levels, as a function of the heat treat-ment. Second, differences in rates of condensation of surface sila-nol groups are expected and hence differences in kinetics ofdehydroxylation at glass surfaces as a function of heat treatment.As a result, differences in rehydration kinetics after heat treatment,as a function of thermal history, are also expected. The condensa-tion of silanol groups to form three membered silicate rings isknown to occur at temperatures �200–300 �C, whereas the forma-tion of edge-shared two-membered tetrahedral rings is known tooccur at temperatures as low as 300–400 �C [20,21]. In a study

Table 4Impact of depyrogenation treatment of vial and vial type on characteristics of freeze-dried5 �C prior to freezing stage.

Freeze-driedproduct

Inner vial surface Vialtreatmenta

Visual inspection for freeze

% Vials withfogging

% Vials witto/at should

Placebo 2 Unsiliconized (US)(Duran vials)

US25280 100 27US60280 100 23US60300 100 68US25320 100 8

Siliconized (S) vials(baked-on silicone)

S25280 0 No furtherS25320 0

Placebo 9(BSA DP)

Unsiliconized (US)(Duran vials)

US60280 100 50US60300 100 42US60320 100 4

Siliconized (S) vials(baked-on silicone)

S25280 0 No furtherS60320 0

mAb2 DP Unsiliconized (US)(Duran vials)

US60280 100 50US60300 100 65

Siliconized (S) vials(baked-on silicone)

S25280 0 No furtherS25320 0

a Refer to Table 2 for a description of depyrogenation treatment for each vial labels.

by D’Souza and Pantano [20], heat treatment of the same glass atdifferent temperatures was found to result in similar initial silanolconcentrations. However, the samples heat treated for longer timeswere found to show slower rehydration kinetics which indicated agreater degree of structural relaxation having taken place on thesurface. Rehydration kinetics was described as having an initialphase with rapid increase in silanol concentration followed byslower rehydration kinetics for longer times. The authors attrib-uted this observation to different rehydration kinetics of thestrained silanol surface groups. The two reaction rates observedon rehydration are believed to correspond to the rehydration ofthe two types of strained siloxane bond defects described previ-

placebo 2, placebo 9 (BSA DP), and mAb2 DP. All samples were equilibrated for 1 h at

-dried DP

h fogginger

% Vials with sub-shouldercritical fogging height

% Vials with fogging aroundentire circumference

100 3290 4075 3879 46

assessment necessary due to absence of fogging

38 6940 530 19

assessment necessary due to absence of fogging

38 9244 78

assessment necessary due to absence of fogging

Fig. 5. Dehydroxylation of silanol groups on glass surfaces to produce strained solixane bonds belonging to (a) three-membered silicate rings and (b) edge-shared two-membered tetrahedral rings (adapted from D’Souza et al and from Michlaske et al). (For interpretation of the references to color in this figure legend, the reader is referred tothe web version of this article.)

A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326 323

ously (see Fig. 5a and b). Since the degree of strain of Si–O–Si bondsin edged-shared tetrahedra exceeds that of the three-memberedrings (14 vs. 4 kcal/mole), the rehydration rate for two memberedrings is expected to be greater than for three membered rings[20,22]. The same study by D’Souza et al reports that at moreaggressive heat treatments, dehydration becomes progressivelyirreversible and silanol concentration obtained after rehydroxyla-tion is lower [20].

In summary, no changes were observed with siliconized glassvials regardless of the depyrogenation conditions evaluated. Fur-thermore, no vials with fogging were observed when using hydro-phobic (siliconized) vials, regardless of depyrogenation treatment.With using unsilionized vials, depyrogenation had a variable im-pact on glass inner surface. In terms of fogging, the pre-warmingtemperature alone did not seem to have a significant effect onimprovement of appearance. However, with higher depyrogen-ation temperatures, we observed an improvement in appearancewith regard to fogging (US60320). Though the root cause is notfully clear, this may be due to reduction on any contaminants withimpact on fogging.

3.3. Development studies: effect of shelf cooling prior to freeze drying

Significant improvement in visual appearance with regard tofogging was observed when vials containing DP were pre-cooledprior to freezing for sufficient amounts of time (Fig. 6a and b, Ta-ble 4). Equilibrating the vials at low temperature (5 �C) on thelyophilizer shelves before freezing caused a reduction in fogging,similar to observations with small scale runs reported earlier in pi-lot studies with mAb2 DP. Regardless of the depyrogenation treat-ment, % rejects due to fogging on shoulder significantly decreasedafter 5 h (US25320 vials) and even more after 22 h (US25320,US25280, and US60300 vials) of vial equilibration at 5 �C comparedto the same vials with only 1 h of equilibration time at 5 �C prior tofreezing. Moreover, receding of fogging was observed at varyinglevels manifested as appearance of a gap between top border ofcake and beginning of fogging. Additionally, fogging was thinneddown to spots and patches in most vials, especially after 22 h ofpre-cooling at 5 �C. However, the pre-freezing treatment did notprevent appearance of spots and or patches at the vial shoulder,which rendered the vials as rejects (up to �50% of the vials).

No further studies were performed to look at the relationshipbetween creeping of the solution upwards at the glass wall andlow temperature. However, one can speculate that receding of fog-

ging as a function of low temperature exposure of the vials con-taining the DP solution for sufficient periods of time is likely aresult of reduced solution creeping at low temperature.Temperature dependence of surface tension, contact angle,adsorption, and Marangoni effect in solution are all reported[2,5,6,8,11,14,23]. By reducing temperature, one would expect adecrease in all aforementioned parameters, but contact angle, lead-ing to reduced creeping behavior in solution. Some reports also de-scribe this as a ‘‘thermophoresis’’ effect [14].

3.4. Commercial scale runs: effect of depyrogenation conditions andshelf cooling prior to freeze drying

This section describes the outcome of visual inspection, with re-gard to fogging, of mAb1 DP freeze-dried in production in unsilic-onized Duran vials from Vendor 2 all pre-subjected to differentdepyrogenation treatments similar to the ones in development.The lot number of vials used for this study was different than theone used in development lyophilization studies. Cakes producedfrom freeze drying were all elegant with excellent retention of cakestructure. Cake moisture levels in different vial treatments andtypes tested were well below 2% w/w, similar to results fromdevelopment. 100% of mAb1 freeze-dried DP vials were visually as-sessed using the same visual inspection machinery used for prod-uct freeze-dried in development, however by a different inspector.The new inspector assigned a score (fogging factor) for each vial inan attempt to make a semi-quantitative comparison betweenproduct freeze-dried in different vials based on the following crite-ria: fogging (2 points for yes, 0 for no), density/opaqueness of fog-ging (1 point for weak, 2 for medium, and 3 for high), approximatearea covered by fogging on vial (1 point for 25%, 2 for 50%, 3 for75%, and 4 for 100%), and height of fogging in vial (1 point fornon-critical height, 2 for mid-critical height, and 3 for criticalheights including fogging at shoulder). So for example, if a vial isassigned 2 for fogging, 2 for fogging density, 2 for the vial area cov-ered by fogging and 2 for fogging height, then the fogging score is2 � 2 � 2 � 2 = 16.

The assigned ‘‘fogging factors’’ for DP vials from different treat-ments are compared in Fig. 7. Consistent with results from devel-opment scale, depyrogenation treatment at a higher temperature(320 �C) had lower fogging scores, i.e., it had a positive influencein terms of improvement in fogging appearance. Furthermore, vialsdepyrogenated at 320 �C that were first introduced into the lyoph-ilizer (i.e., equilibrated at longer times at a low shelf temperature

Fig. 6. Impact of low temperature equilibration time of mAb1 DP solution in unsiliconized vials (Vendor 2) prior to freezing step on fogging behavior: (a) DP in US25320 vialsand (b) DP in US60300 vials.

324 A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326

of 5 �C on the lyophilizer shelf before freezing) consistentlyshowed a lower ‘‘fogging factor’’ compared to the same batch ofvials loaded last (i.e., 1 h before the start of freezing).

In summary, the outcome from the development studies thatwe performed using unsiliconized vials subjected to different ster-ilization/heat tunnel treatments were reproducible at a commer-cial setting. The slight modification of the lyophilization cyclethat was necessary to be used in production was not impactingthese results.

4. Conclusions

Fogging has always occurred – though to varying degree andheight – with lyophilized drug products studied to date, whenusing glass with hydrophilic inner surface. Our studies have alsorevealed a potential impact of processing conditions on fogging.Processing conditions, including depyrogenation, and low temper-ature equilibration of DP vials for extended periods of time prior tofreezing step in freeze drying, and/or annealing process for the vial,

Fig. 7. A comparison of ‘‘fogging factor’’ of mAb1 freeze-dried in production in vials subjected to different depyrogenation treatment and pre-freezing equilibration times at5 �C (unsiliconized vials from Vendor 2). Solid black bars: Vials washed at different temperatures, depyrogenated at 320 �C, drug product holding time in vial at 5 �C for 1 hbefore freezing step, Vertical striped bars: Vials washed at different temperatures, depyrogenated at 320 �C, drug product holding time in vial at 5 �C for 22 h before freezingstep, Mesh bar: Vials washed at 60 �C, depyrogenated at 300 �C, drug product holding time in vial at 5 �C for 1 h before freezing step, Horizontal striped bars: Vials washed atdifferent temperatures, depyrogenated at 280 �C, drug product holding time in vial at 5 �C for 1 h before freezing step.

A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326 325

etc. may change the appearance of fogging, though, are unlikely tofully avoid the phenomenon. Furthermore, pre-cooling at low tem-peratures (2–10 �C) for extended periods of time prior to freezingmay not be a commercially viable solution as it would increasethe overall freeze-drying cycle time. Taking into account the re-sults from the studies, it seems not being possible to fully excludethe fogging effect in primary packaging with an inner hydrophilicsurface.

Vials with a hydrophobic inner surface (such as provided by sil-iconization) are considered a robust means to minimize or elimi-nate fogging phenomena in lyophilized drug products.

Acknowledgments

We would like to acknowledge Dr. Uwe Rothhaar and Mr. Vol-ker Scheumann at Schott, Mainz, Germany, for their support onperforming and interpreting contact angle, surface energy, and so-lid state characterization measurements on glass vial inner surfaceas well as fruitful discussions on root causes of fogging. We wouldlike to acknowledge Dr. Wigand Weirich (Roche, Basel) and Schott,Müllheim, for their support on preparing and providing siliconizedvials (baked-on silicone). We would also like to thank Stefan Kunzeand Domenico Currenti from Pharmaceutical Bulk Operations Par-enterals, Kaiseraugst, Switzerland, for helping organize and set uplyophilization runs on commercial scale. Finally, we would like toacknowledge Dr. Wigand Weirich, Dr. Phillip Lam (Genentech,San Francisco, CA), and Dr. Michael Pikal (University of Connecti-cut, Storrs, CT) for discussions about creeping and fogging.

References

[1] I. Bauer-Dauphin, H.C. Mahler, A method for avoiding glass fogging ofpharmaceutical containers during freeze drying, in: PCT Int. Appl. 2010, F.Hoffmann-La Roche AG, Switz., pp. 26.

[2] R. Tadmor, Marangoni flow revisited, Journal of Colloid and Interface Science332 (2) (2009) 451–454.

[3] K. Mayya, M. Sastry, A new technique for the spontaneous growth of colloidalnanoparticle superlattices, Langmuir 15 (6) (1999) 1902–1904.

[4] F. Reincke et al., Spontaneous assembly of a monolayer of charged goldnanocrystals at the water/oil interface, Angewandte Chemie, InternationalEdition 43 (4) (2004) 458–462.

[5] D. Johnson et al., Marangoni flow of Ag nanoparticles from the fluid–fluidinterface, Journal of Physical Chemistry A 112 (39) (2008) 9318–9323.

[6] M. Hennenberg et al., Mass-transfer, Marangoni effect, and instability ofinterfacial longitudinal-waves. 1. Diffusional exchanges, Journal of Colloid andInterface Science 69 (1) (1979) 128–137.

[7] V.J. Levich, Physicochemical Hydrodynamics. Prentice-Hall International Seriesin the Physical and Chemical Engineering Sciences, Prentice-Hall, EnglewoodCliffs, NJ, 1962.

[8] I. Egry, Surface tension measurements in microgravity and their relevance toMarangoni convection, ChemPhysChem 4 (4) (2003) 329–333.

[9] P. Kralchevsky, K. Danov, N. Denkov, Chemical physics of colloid systems andinterfaces, in: K.S. Birdi (Ed.), Handbook of Surface and Colloid Chemistry, CRCPress/Taylor & Francis, Boca Raton, 2009, pp. 197–377.

[10] M. Kawaji et al., Investigation of Marangoni and natural convection duringprotein crystal growth, Journal of Crystal Growth 258 (3–4) (2003) 420–430.

[11] T. Kuroda, The Marangoni effect and its artistic application, Forma 15 (2000)203–204.

[12] M.S. Schwarzenbach et al., Topological structure and chemical composition ofinner surfaces of borosilicate vials, PDA Journal of Pharmaceutical Science andTechnology/PDA 58 (3) (2004) 169–175.

[13] R. Doremus, Glass Science, second ed., John Wiley & Sons, Inc., New York, 1994.[14] S. Iacopini, R. Piazza, Thermophoresis in protein solutions, Europhysics Letters

63 (2) (2003) 247–253.[15] S. Kiese et al., Shaken, not stirred: mechanical stress testing of an IgG1

antibody, Journal of Pharmaceutical Sciences 97 (10) (2008) 4347–4366.[16] P. Kolhe, E. Amend, S.K. Singh, Impact of freezing on pH of buffered solutions

and consequences for monoclonal antibody aggregation, BiotechnologyProgress 26 (3) (2010) 727–733.

[17] H.C. Mahler et al., Surface activity of a monoclonal antibody, Journal ofPharmaceutical Sciences 98 (12) (2009) 4525–4533.

[18] M. Duncan, J. Lee, M. Warchol, Influence of surfactants upon protein/peptideadsorption to glass and polypropylene, International Journal of Pharmaceutics120 (2) (1995) 179–188.

[19] C. Burke et al., The adsorption of proteins to pharmaceutical containersurfaces, International Journal of Pharmaceutics 86 (1) (1992) 89–93.

[20] A.S. D’Souza, C.G. Pantano, Hydroxylation and dehydroxylation behavior ofsilica glass fracture surfaces, Journal of the American Ceramic Society 85 (6)(2002) 1499–1504.

326 A.M. Abdul-Fattah et al. / European Journal of Pharmaceutics and Biopharmaceutics 85 (2013) 314–326

[21] B.A. Morrow, I.A. Cody, L.S.M. Lee, Infrared studies of reactions on oxidesurfaces. 7. Mechanism of the adsorption of water and ammonia ondehydroxylated silica, Journal of Physical Chemistry A 80 (25) (1976) 2761–2768.

[22] T.A. Michalske, B.C. Bunker, Slow fracture model based on strained silicatestructures, Journal of Applied Physics 56 (10) (1984) 2686–2693.

[23] M. Hennenberg et al., Mass-transfer, Marangoni effect, and instability ofinterfacial longitudinal-waves. 2. Diffusional exchanges and adsorption–desorption processes, Journal of Colloid and Interface Science 74 (2) (1980)495–508.