International Journal of Pharmaceutics 530 (2017) 79–88

Dual centrifugation – A new technique for nanomilling of poorlysoluble drugs and formulation screening by an DoE-approach

Martin Hagedorna,b,*, Ansgar Bögershausenb, Matthias Rischerb, Rolf Schuberta,Ulrich Massinga,c

aAlbert-Ludwigs-Universität Lehrstuhl für Pharmazeutische Technologie und Biopharmazie, Hermann-Herder-Straße 9, D-79104 Freiburg i. Br., Germanyb Losan Pharma GmbH, Otto-Hahn-Straße 13, 79395 Neuenburg am Rhein, GermanycAndreas Hettich GmbH & Co KG, Engesserstr. 4a, D-79108 Freiburg, Germany

A R T I C L E I N F O

Article history:Received 22 May 2017Received in revised form 3 July 2017Accepted 15 July 2017Available online 18 July 2017

Keywords:NanomillingDesign of experimentsDual centrifugationNanosuspensionWet ball millingNanocrystalline suspensionNanoparticles

A B S T R A C T

The development of nanosuspensions of poorly soluble APIs takes a lot of time and high amount of activematerial is needed. In this publication the use of dual centrifugation (DC) for an effective and rapidAPI-nanomilling is described for the first time. DC differs from normal centrifugation by an additionalrotation of the samples during centrifugation, resulting in a very fast and powerful movement of thesamples inside the vials, which � in combination with milling beads � result in effective milling.DC-nanomilling was compared to conventional wet ball milling and results in same or even smallerparticle sizes. Also drug concentrations up to 40% can be processed. The process is fast (typical 90 min)and the temperature can be controlled. DC-nanomilling appears to be very gentle, experiments showedno change of the crystal structure during milling. Since batch sizes are very small (100–1000 mg) andsince 40 sample vials can be processed in parallel, DC is ideal for the screening of suitable polymer/surfactant combinations. Fenofibrate was used to investigate DC-nanomilling for formulation screeningby applying a DoE-approach. The presented data also show that the results of DC-nanomillingexperiments are highly comparable to the results obtained by common agitator mills.

© 2017 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

International Journal of Pharmaceutics

journal homepage: www.elsev ier .com/locate / i jpharm

1. Introduction

The increasing focus of drug screening towards more lipophilictargets has led to a significant higher number of potential newactive pharmaceutical ingredients (APIs), which are characterisedas poorly soluble in accordance to the established BCS system(Amidon et al., 1995). Almost 90% of APIs in current developmentstudies can be classified as poorly soluble (Loftsson and Brewster,2010), which reduces their suitability for oral application in manycases. However, oral delivery of drugs offers several advantagesand is clearly the preferred route of drug application.

Orally administered APIs have to be dissolved in the gastroin-testinal tract to be passively absorbed (Liversidge and Conzentino,1995) or to be delivered by a transporter system. To increase thesolubility of poorly soluble compounds one of the suitable

* Corresponding author at: Losan Pharma GmbH, Otto-Hahn-Straße 13, 79395Neuenburg am Rhein, Germany.

E-mail addresses: martin.hagedorn@losan.de (M. Hagedorn),ansgar.boegershausen@losan.de (A. Bögershausen), matthias.rischer@losan.de(M. Rischer), rolf.schubert@pharmazie.uni-freiburg.de (R. Schubert),ulrich.massing@hettichlab.com (U. Massing).

http://dx.doi.org/10.1016/j.ijpharm.2017.07.0470378-5173/© 2017 Elsevier B.V. All rights reserved.

technologies is to reduce their particle size, thus increasing theeffective particle surface area, which leads to a higher rate ofdissolution and an oversaturation effect (Noyes Whitney-Equation(Buckton and Beezer, 1992)). The resulting increase of bioavail-ability allows a faster dissolution (decrease of t-max), higher AUC(Area under Curve) and therefore lower doses, which helps toreduce adverse effects as well as food effects (Liversidge andConzentino, 1995; Juenemann et al., 2011; Jinno et al., 2006;Junghanns, 2008).

One approach to produce small drug particles is to mill anaqueous suspension containing the poorly soluble drug, as well aspolymer(s) and surfactant(s) for stabilization by avoiding agglom-eration. Since the milling process is usually supported by millingbeads, it is often called wet ball- or pearl milling (Junghanns,2008). The advantages of wet ball milling technique are a waterbased approach (no organic solvents) and low batch to batchvariations (Chin et al., 2014). There are several FDA approvedproducts on the market which have been produced by thistechnique (e.g. Rapamune1 (Sirolimus, Wyeth), TriCor1 (Fenofi-brate, Abbott) and Megace1 ES (Megestrol acetate, Par Pharma-ceuticals))(Kesisoglou et al., 2007).

80 M. Hagedorn et al. / International Journal of Pharmaceutics 530 (2017) 79–88

After wet ball milling, the resulting nanosuspensions should beconverted into a dry state to prevent Ostwald ripening andsubsequent agglomeration or chemical changes by hydrolysis.Therefore, the suspensions have to be transferred into a solid formvia lyophilisation, spray-/freeze drying, layering or granulation(Van Eerdenbrugh et al., 2008) to finally achieve solid dosage formslike tablets, capsules or stick packs in case of granulates or pelletsas interim step.

The development of a suitable suspension of (nano-) particles(nanosuspensions) of a poorly soluble drug stabilized by anoptimal combination and ratio of different co-polymers andsurfactants is very time-consuming. For every drug compound ahigh number of different combinations/ratios of polymers andsurfactants have to be tested by milling and at least the particle sizedistribution (PSD) and the particle stabilities with respect toagglomeration of each combination have to be measured.

However, what makes screening moreover time-consuming isthat a lab-milling procedure has to be applied which is predictiveand comparable to the milling process which is used later in thelarge scale drug manufacturing process. Thus, planetary wet ballmilling is used very often (Malamatari et al., 2015; Laaksonen et al.,2011; Palo et al., 2015; Tuomela et al., 2015). Most planetary ballmills (e.g. Pulverisette 7, Fritsch) are equipped with millingzirconium oxide bowls. Using these devices, it is necessary to millmixtures of water, drug, polymer(s) and surfactant(s) and millingbeads over several hours if API concentrations of 10% or higher areused. Since this device has no cooling system, the process has to beinterrupted by several breaks to allow the suspensions to cooldown to room temperature. Thus, the screening is very time-consuming, only a small number of different formulations can betested in a certain time.

In addition, even small planetary ball mills (if not modified withspecial inserts) need minimum batch sizes of 10 ml, which makes abroad screening of formulations containing expensive APIs almostimpossible � or at least limits the number of formulations whichcan be tested. In literature further milling approaches arementioned, e.g. (Möschwitzer, 2015) in which suspensions arestirred with an simple magnetic stirrer in 20 ml bottles or (Romeroet al., 2015) in which group stacked stirring bars (3 bar) plus millingbeads are used in 2 ml glass vials for up to 120 h. Also screeningprocedures using a 96-well plate were already described (VanEerdenbrugh et al., 2009). Here, 0.5-mm yttrium-stabilizedzirconium milling beads were added and the plate was shakenover 24 h. Although these systems allow the processing of a highnumber of samples, they have disadvantages like a maximum drugload of 5% and/or long milling times. Therefore the resultingparticles sizes are not comparable with those resulted from theagitator mills used in large scale drug production.

Dual centrifugation (DC) is to a certain extent related to the wetball milling. DC differs from normal centrifugation by an additionalrotation of the samples during the centrifugal process. Thus, thedirection of the strong centrifugal forces inside the vials changescontinuously. The resulting powerful movement of the samplesinside the vials finally results in their rapid homogenization ormilling.

DC-based-processes has so far been used for preparation of lipidnanoparticles like liposomes (Massing et al., 2008) by homogeniz-ing aqueous phospholipid-drug-mixtures in small containers usingthe same zirconium oxide beads as used for the production ofdrug-nanosuspension in common ball mills. Beside its extremelyhigh power, the most important advantage of DC is that allhomogenisation, mixing and milling processes can be done inclosed disposable containers of different volumes, very small onesincluded. In addition, up to 40 vials (2 ml) can be processed in onerun and due to the high input of energy the resulting milling time isvery short. Furthermore, the most actual DC-device allows efficient

cooling of the samples during the DC-process which allowcontinuous milling without the need of cooling breaks.

Thus, DC is as a promising tool for a very rapid and broadscreening of suitable polymers and surfactants for nanomilling ofpoorly soluble drug compounds. In this article, the development ofa nanomilling screening protocol using the new DC-apparatusZentriMix 380R (Andreas Hettich, Tuttlingen, Germany) forsimultaneous milling of 40 samples in small disposable containers(2 ml) is described. Furthermore, nanomilling using the new DC-system as well as a conventional ball mill is compared. Forhandling of the high number of data points resulting from the DC-nanomilling approach a design of experiment (DoE) approach wasutilized to classify and quantify all the critical formulationvariables with respect to the gained particle size distributions(PSD). Afterwards, design spaces were obtained in which particlesizes can be predicted.

2. Materials and methods

2.1. Materials

HPMC 3 mPas was purchased from Shin-Etsu Chemical (Tokyo,Japan). PVP 25 K, PVP VA 64, Poloxamer and SDS were purchasedfrom BASF SE (Ludwigshafen, Germany). Tween 80 was orderedfrom Merck (Darmstadt, Germany) and Sodium-Docusate (DOSS)from Cytec (New Jersey, USA). As milling beads (for DC-milling aswell as planetary wet ball milling) Yttrium oxide-stabilizedzirconium oxide beads (0.1–0.2 mm, Sigmund Lindner GmbH,Warmensteinach, Germany) were used. Highly stable 2 ml DC-Twist-Top-vials were purchased from Andreas Hettich GmbH & CoKG, Tuttlingen, Germany.

2.2. Preparation of API-suspensions

The API-suspensions for the milling experiments were preparedas follows: weigh milling beads and API in the DC-Twist-Top-Vialor the planetary ball milling bowls. Polymer and surfactant are pre-dissolved in purified water and added. Afterwards the suspensionswere further diluted with purified water. Sample preparation forthe investigation of the DC-system and the planetary ball milldiffered only in the batch size. All amounts of the formulationcomponents are given as percentage by mass (%w/w).

2.3. Dual centrifugation (DC)

DC was performed using a ZentriMix 380 R (Andreas HettichGmbH und Co KG, Tuttlingen, Germany). Milling parameters aregiven in the corresponding chapters. For every milling trial thecooling device was set to 0 �C (measured in the rotating chamber,which results in sample temperatures of approx. 18 �C after 90 minmilling at 2000 rpm/1000 mg milling beads). Since this is themaximum DC-speed and milling time used in this investigation, itcan be assumed that the process temperature was always below18 �C in all experiments.

2.4. Planetary wet ball milling

Planetary ball milling was done with a Pulverisette 7 (FritschGmbH, Idar-Oberstein, Germany). For each milling trial two 45 mlbowls (Fritsch GmbH, Idar-Oberstein, Germany) each filled with10 g of API-suspension were used. This is the minimum reasonablebatch size for this equipment using the mentioned 45 ml bowls(note: smaller bowls are available as well). Milling conditions were750 rpm over 14 cycles of 30 min interrupted by cooling breaks of5 min (total milling time: 7 h). The defined parameters reflect astandard process often used in industry as well as academia.

M. Hagedorn et al. / International Journal of Pharmaceutics 530 (2017) 79–88 81

However some groups work also with different milling speed,milling bowls and milling times, but the optimization theplanetary wet ball milling process was not the topic of this work.

2.5. Design of experiment (DoE)

DoE approaches were performed with a full factorial designcontaining all possible combinations between the factors and theirlevels. In addition a centre point was investigated (n = 3). AllDoE-plots and calculations were done by using MODDE 11software (MKS Data Analytics Solutions, Malmö, Sweden).

2.6. Laser diffraction (LD)

Particle sizes were measured by using LD (Mastersizer2000/Hydro 2000S; Malvern Instruments GmbH, Worcestershire,UK) the measurements were performed using purified water asdiluent at room temperature. Three runs for each sample weremeasured with a volume based approach using Mie-Theorycalculation. The refractive index of the API and of dispersingmedium was fixed at 1.6 and 1.33, respectively. The average of thethree runs is reported.

2.7. Cryo-Transmission electron microscope (cryo-TEM)

Leo 912 V-mega (Leo Elektronenmikroskopie GmbH, Oberko-chen, Germany) cryo-TEM was used to investigate particle shapesin the suspensions. Pictures were obtained by a Proscan camera(HSC 2 Oxford Instruments, Abingdon, USA). The suspensions arebrought onto a grid and the small liquid film is rapidly frozen to90 K by fluid ethanol, so that ice crystals do not form and the frozenwater film remains transparent in an amorphous state. Thetemperature during the measurements was <�170 �C.

2.8. X-Ray-Powder-Diffraction (XRPD)

Reflex pattern are obtained with an Stoe Stadi P XRPD (STOE &Cie. GmbH, Darmstadt, Germany) equipped with an Ge-(111)-monochromator using a copper-K-alpha1-radiation. For samplepreparation the suspensions were desiccated in a drying cabinet,grinded and applied on a tape (Scotch Brand, St. Paul, USA).Transmission was measured. Every sample is measured n = 8 andreflexes are summarized. For better visualization a manualbackground correction was performed. The software WinXPow(STOE & Cie. GmbH, Darmstadt, Germany) was used.

Table 1Particle sizes (D90-values) of a Fenofibrate suspension (10% Fenofibrate, 1% HPMC, and

Mixing speed 700 mg milling beads 1000 m

Milling time D(90) Milling

1000 rpm 30 min 695 nm 30 min60 min 374 nm 60 min90 min 294 nm 90 min120 min 277 nm 120 mi

1500 rpm 30 min 356 nm 30 min60 min 290 nm 60 min90 min 271 nm 90 min120 min 260 nm 120 mi

2000 rpm 30 min 265 nm 30 min60 min 247 nm 60 min90 min 242 nm 90 min120 min 254 nm 120 mi

3. Results

3.1. Nano milling of APIs using dual centrifugation

To investigate if nanomilling by dual centrifugation is possible,a known formulation of Fenofibrate (10% Fenofibrate, 1% HPMC,0.075% SDS), already used in other milling approaches (Azad et al.,2015) was processed by DC. The influence of milling time,DC-speed and the amount of milling beads per vial wasinvestigated. The formulation, vial-type (disposable 2 ml contain-ers), batch size (1000 mg, containing 100 mg Fenofibrate) and typeof milling beads (0.2 mm milling beads) were kept unchanged. Theexperiments were evaluated with respect to the particles size. InTable 1 the D90-values are summarized.

In all cases DC-milling leads to visually homogeneous nano-suspensions. The D90-values of all samples were well below 1 mm,in the range of 237–695 nm. In addition, some expectedcorrelations could be drawn: increasing runtimes, a higher amountof milling beads per vial and a faster milling speed results insmaller particles (lower D90-values). The results show thatDC-nanomilling is possible in reasonable runtimes and can becarried out with small amounts of API and excipients. Based onthese results the following parameters were selected as suitablefor further experiments: milling speed: 1500 rpm, amount ofmilling beads per vial: 1000 mg; and a runtime of 90 min. Theseparameters were judged to be sufficient to obtain small particlesbut are still mild enough to allow milling of potentially sensitivecompounds (see also Table 1).

In literature the preparation of a Fenofibrate nanosuspensioncontaining the identical ingredients by using an agitator wet ballmill process has been described by Azad (Azad et al., 2015) using aMicroCer-mill (Netzsch Premium Technology LLC, Exton PA USA;milling chamber: 80 ml, agitator speed: 3200 rpm, 50 ml Yttrium-stabilized zirconium oxide beads (0.4 mm)). To show if DC-nanomilling can reproduce the results using the agitator mill, theparticle sizes of Fenofibrate suspensions produced by both meanswere compared (DC-nanomilling: milling beads: 1000 mg; run-time: 90 min; milling speed: 1500 rpm, s. Table 1) is compared tothe outcome of Azad-study (Table 2). The D90-values obtained byboth milling procedures were almost similar. D10- and D50-valueswere somewhat lower for the DC-milling procedure.

To prove if DC-milling is also feasible for other compounds andformulations, already known nanosuspensions of different APIswere produced by DC-milling: a Fenofibrate suspension (stabilizedwith HPMC/SDS), Naproxen (stabilized with HPMC/DOSS),

0.075% SDS) milled by DC under different conditions.

g milling beads 1300 mg milling beads

time D(90) Milling time D(90)

514 nm 30 min 383 nm 323 nm 60 min 282 nm 285 nm 90 min 279 nmn 273 nm 120 min 259 nm

337 nm 30 min 337 nm 278 nm 60 min 329 nm

258 nm 90 min 313 nmn 304 nm 120 min 303 nm

248 nm 30 min 248 nm 250 nm 60 min 245 nm 237 nm 90 min 250 nmn 254 nm 120 min 254 nm

Table 2Particle sizes of Fenofibrate (10%) suspension produced by agitator mill (obtainedfrom Azad et al., 2015) and DC-milling measured by laser diffraction.

D10 [nm] D50 [nm] D90 m[nm]

Agitator mill (Azad et al., 2015) 96 150 233DC-milling 69 127 237

82 M. Hagedorn et al. / International Journal of Pharmaceutics 530 (2017) 79–88

Ibuprofen (stabilized with HPMC/SDS) and LOS1 (stabilized withPoloxamer/TPGS). DC-milling was performed by using theparameters defined before. All API-excipient-mixtures were alsomilled with a not modified planetary wet ball mill (Pulverisette 7),which is often used to generate API-nanosuspensions in lab-scale.Particles sizes of all suspensions are shown in Fig. 2.

In all cases DC-nanomilling resulted in small particles with anarrow particle size distribution. For Naproxen (A) and LOS1 (B)the results are almost similar for DC- and planetary wet ballmilling. Ibuprofen (D) and especially Fenofibrate (C) suspensionsshow smaller particle sizes and narrower particle size distributionswhen using DC-milling compared to the planetary wet ball milling.

In general, during the early phase of development of newnanosuspensions, most milling trials are performed with less than10% API in the suspensions. For different down-processingtechnologies e.g. lyophilisation, coating or granulation, a higherdrug load can be beneficial. Since processing of suspension with ahigher API-load is known to be difficult using planetary wet ballmills, those formulation trials have to be performed with morepowerful, large scale mills, which needs a rather large amount ofAPI. Therefore one important question was if formulationscreening by DC-milling is also possible with higher drug loads(>10%). To test this, suspensions containing different amounts ofNaproxen (5–40%) were produced by DC-milling and planetary wetball milling. Only the amount of Naproxen was adjusted while theamounts of polymer/surfactant as well as the milling conditionswere kept unaltered. In Fig. 3 the comparison of the particle sizedistributions of the Naproxen particles getting from the twodifferent milling approaches are presented.

DC-milling experiments with all chosen drug loads result inidentical small particle sizes and the same particle size distribu-tions. Using planetary wet ball milling, small Naproxen particlescould only be obtained by using lower concentrations of Naproxen(5 and 10%). Higher concentrations resulted in larger particles andat 40% Naproxen-load, virtually no particles smaller than 1 mmcould be detected. However, an API-load of 40% is not very usualsince the high viscosity of the suspensions lead to difficulties in themilling process also at larger scales. This experiment wasperformed to examine the limits of the new milling technique.However, the used process conditions for planetary wet ballmilling were standard parameters. It was not further investigated ifan optimization (e.g. increase of milling time) would providesmaller particles.

Fig. 1. ZentriMix 380R (left picture), sample adapter unit and corr

To further reduce the amount of API necessary for formulationscreening, down-scaling of the batch size was investigated. To doso, the standard DC-milling batch size of 1000 mg (containing100 mg API) was reduced to 500 mg, 200 mg and 100 mg,corresponding to (50, 20 and 10 mg API). For this set of experi-ments the API LOS1 was used as model drug compound.

To ensure comparable milling conditions for all batch sizes theratio of batch size to amount of milling beads were kept constantwith respect to the ratio used in the standard batch size (1000 mgmilling beads for batch size of 1000 mg). The formulation itself,size of vials (2 ml), type and size of milling beads (0.2 mm) as wellas the milling conditions (90 min; 1500 rpm) were kept constant.As shown in Fig. 4, also smaller batch sizes resulted in practicallythe same particle sizes and distributions with respect to D10, D50and D90-values.

Due to the high energy intake it is important to make sure thatno change of the crystal structure of the API takes place during DC-milling. Thus, the Fenofibrate suspension (10%) already shown inFig. 2 and the Naproxen suspension (30%) of Fig. 3 were milled byDC and changes of the crystal structures were investigated byXRPD. Figs. 5 and 6 show the reflexes of the dried suspensionbefore (black) and after DC-milling (red). In both cases no changesof the crystal structure could be observed.

That there is virtually no change of the crystalline structure isalso illustrated by cryo-TEM pictures of Fenofibrate- (upper figure)and Naproxen formulation (figure below) after DC-milling (Fig. 7).

3.2. Formulation screening by using a DoE-approach

Statistical planning of experiments is widely used to gain abetter process understanding and identify interactions betweenprocess parameters within a limited number of well-definedexperiments. Moreover, the ICH guidance for industry Q8 (R2)stated that a quality by design approach and especially the use of adesign space is highly recommended during the pharmaceuticaldevelopment. To fulfil these requirements a design of experiment(DoE) approach was applied to the DC-based formulationdevelopment described beforehand. As model API, the drugsubstance Fenofibrate was used. The aim was to produce ananosuspension providing reduced particle sizes in the nanometrerange (<1 mm) by using DC-milling combined with a DoE-approach.

DC-based formulation screening starts with a critical evaluationof the characteristics of the API and excipients (polymer andsurfactant). For the API, log p, pka and the solubility with respect topH-value have to be considered. For the excipients, criticalparameters are the charge, sterically characteristics or viscosity.Not only experience, but also an intense literature screening and acomparison with already produced nanosuspensions are funda-mental. But also processes triggered during the milling procedure

esponding sample vial (2 ml DC-Twist-Top-Vial/right picture).

Fig. 2. Comparison of DC- and planetary wet ball milling. (red: DC-milling/green planetary wet ball milling). All suspensions produced by DC were manufactured in batchsizes of 1000 mg containing 100 mg of the corresponding API and planetary wet ball mill suspension with a batch size of 10 g containing 1 g API.

Fig. 3. Comparison of particle size distributions of Naproxen-suspensions with different drug loads produced by DC-nanomilling and planetary wet ball milling.

M. Hagedorn et al. / International Journal of Pharmaceutics 530 (2017) 79–88 83

have to be reflected e.g. the change of pH, change of viscosity due tothe increased surface area of insoluble particles, a possibletemperature stress and interaction of components with the millingbeads. The selection of appropriate excipients is one importantaspect during the formulation development and creates the basisfor the DoE-screening.

In the following case study, the polymers and surfactants usedfor the DoE-guided formulation development were commonexcipients and do not reflect a broad literature screening. In thefirst DoE-stage two qualitative factors � the type of polymer and

the type of surfactant � and two quantitative factors � the amountof polymer and the amount of surfactant � were investigated.Based on the conventions of DoE-calculations, the qualitative andquantitative factors were grouped in different levels (Table 3).According to the factors and levels presented in Table 3, in total 36samples were prepared, DC-milled. PSD (D90-values) wereanalysed by laser diffraction. To also test the reproducibility ofDC-nanomilling, a centre point (randomly defined with respect tothe qualitative factors: 1% SDS; 2% HPMC) (n = 3) was investigatedas well. The screening was performed keeping the following

Fig. 4. DC-Milling of a LOS1-formulation (10% API; 2% Poloxamer; 1% TPGS; water ad.) using different batch sizes (n = 3).

Fig. 5. XRPD measurement of a dried Naproxen suspensions (stabilized by HPMC/DOSS) before (black/lower line) and after (red/upper line) DC-nanomilling.

84 M. Hagedorn et al. / International Journal of Pharmaceutics 530 (2017) 79–88

parameters constant: (i) API concentration 10%, (ii) batch size1000 mg (100 mg API) (iii) amount of milling beads 1000 mg and(iv) the milling parameters runtime (90 min) and speed(1.500 rpm). All screening steps: sample preparation, DC-millingand the subsequent determination of particle size distributionscould easily be performed within one day.

In Fig. 8 the DoE-generated contour plots of the screeningapproach are shown. Every column of plots represents onepolymer and every row one surfactant (quantitative parameters).Thus, each of the nine boxes depicts the result the variation of acertain polymer/surfactant combination. The axes of each boxpresent the qualitative factors (y-axis: amount of polymer andx-axis: amount of surfactant). The different colours of the contour

plots give information about the D90-values of the particles (red:larger particle sizes, green smaller particle size).

Reproducibility was confirmed by investigate the particles sizesof the centre points which results in an RSD of �1.6% with regard tothe D90-values. All polymer/surfactant mixtures show a similartrend towards smaller particle sizes with decreasing amounts ofsurfactant and increasing amounts of polymer. Most formulationsshow D90-values above 1500 nm. The system PVP-K25/Tween 80(central box) shows no D90-value lower than 2500 nm. Thepolymer-surfactant mixture HPMC/SDS (left bottom corner)results in the smallest particle sizes. With a high amount ofHPMC (3%) and a low amount of SDS (0.5%), D90-values smallerthan 750 nm were observed.

In the next stage of the DoE-guided formulation screening themost promising excipient-combination (HPMC/SDS) was investi-gated in more detail. Five different levels of the SDS-concentrationwere examined and three levels of HPMC-concentration and thelevels were defined with respect to the findings of the firstDoE-experiment. Since it was found that decreasing SDS-levelsresulted in smaller particles, in the second DoE-experiment theSDS-concentration was investigated down to 0.1%. Furthermore,since it was found that increasing HPMC-concentrations resultedin smaller particles, the highest HPMC level for the secondDoE-experiment was increased to 4% HPMC. In addition to theseadjustments, also the effect of increasing concentrations of themodel API Fenofibrate (10%, 20% and 30%) was investigated. TheDoE-factors and corresponding levels are summarized in Table 4.Again, sample preparation, DC-milling and the subsequentdetermination of particle size distributions were performed withinone day.

The results (D90-values) of the second DoE are shown in Fig. 9.The axes reflect again the amount of the surfactant (y-axis: amountof SDS) and polymer (x-axis: amount of HPMC). Every box

Fig. 6. XRPD measurement of a dried Fenofibrate suspensions (stabilized by HPMC/SDS) before (black/lower line) and after (red/upper line) DC-nanomilling.

Fig. 7. cryo-TEM pictures of (above) Fenofibrate- (stabilized by HPMC/SDS) and(below) Naproxen-suspensions (stabilized by HPMC/DOSS) after DC-nanomilling.

Table 3DoE-Factors investigated for the screening of Fenofibrate suspensions (first DoE).

Qualitative factors Level 1 Level 2 Level 3

Type of polymer HPMC 3 mPas PVP 25K PVP VA 64Type of surfactant SDS Tween 80 Sodium-Docusate

Quantitative factors Level 1 Level 2

Amount of polymer 1% 3%Amount of surfactant 0.5% 1.5%

M. Hagedorn et al. / International Journal of Pharmaceutics 530 (2017) 79–88 85

represents one Fenofibrate concentration (left 10%; middle 20%;right 30%). For each Fenofibrate concentration, the previouslyfound trend to smaller particles by increasing amounts of polymerand decreasing amounts of surfactant could be confirmed. For largeregions within the second DoE-approach (green coloured boxes forall formulations) and up to 30% Fenofibrate-loading the milling ofnanosuspensions by using DC is possible. In all runs homogeneous

suspensions with D90-values smaller than 1000 nm are obtained.Furthermore higher amounts of HPMC and lower amounts of SDSshow again the smallest particle sizes and obviously lead to robustformulation systems.

An important aspect of a formulation development is theinvestigation of the stability of the resulting nanosuspensions withrespect to particle size distribution. Very often, agglomeration and/or Ostwald ripening takes place during storage. Hence, theformulations with the highest and lowest amounts of surfactantand polymer (corresponding to the corners of the boxes in Fig. 9)and the middle concentration of polymer and surfactant for everyFenofibrate-concentration (10, 20, 30%) (corresponding to thecentre of every box in Fig. 9) were investigated for its storagestability

Due to the minimal amount of sample material necessary forthe stability tests, no additional sample preparation was necessary.The samples were stored at room temperature (20 � 5 �C) and day-light in the PP-vials already used for the milling process. In total 15samples were stored and after 1, 3 and 6 weeks the particle sizedistributions were measured again by laser diffraction.

After one week a few changes of the particle sizes could be seen.The most important parameter in terms of stability appears to bethe amount of SDS � while suspensions carrying only low amountsof SDS appear stable, suspensions with higher amounts starting toagglomerate. However, the suspensions with high amounts SDSbut also high amounts of HPMC appear stable. The drug load doesnot influence the particle size. Suspensions containing 10, 20 or30% of Fenofibrate but having an identical HPMC/SDS-load showidentical particle sizes over time.

After three weeks of storage all formulations containing 10%Fenofibrate show an increase of the particle sizes. In contrast, twosuspensions containing 20 and 30% Fenofibrate, respectively, andstabilized with high amount of HPMC (4%) and low amount of SDS(0.1%) are still stable (D90 < 240 nm). These two suspensions werestored for additional 3 weeks, and no increase of the particle sizeswas found (D90 < 240 nm). The PSD profiles of both suspensions(20% Fenofibrate left; 30% Fenofibrate right) at the beginning of thestability study (red curve) and after 6 weeks of storage (greencurve) are presented in Fig. 10.

4. Discussion

4.1. DC-nanomilling

DC-milling in small and disposable vials is a new and innovativetechnique which can successfully be used to prepare stablenanosuspensions of poorly soluble APIs with small particle sizesand narrow particle size distributions. The comparison with acommonly used agitator mill as well as a wet ball mill showed thatDC-milling resulted in nanosuspensions with similar or evensmaller particles sizes. The superior effectiveness of DC-nano-milling can be explained by the characteristic movements of themilling beads within the sample vials. The beads are allowed toaccumulate very high kinetic energy while accelerating along thetube shaped milling vials (Fig. 1) because of the horizontalorientation of the vials in the dual centrifuge (Fig. 11).

At a first glance, it was surprising that DC-nanomilling cansuccessfully be performed in disposable plastic (PP) vials, which �in contrast to zirconium oxide milling bowls of planetary ball mills� have flexible inner walls. Nevertheless, other groups’ grinded APIsuspensions in glass vials using magnetic stirrers equipped withplastic stirrer bars. But in contrast to the very fast DC-millingprocess, the milling times of these processes were up to 120 h(Möschwitzer, 2015). Thus, energy intake is much higher duringDC-milling and it can further be concluded that DC-millingpredominantly takes place between the (numerous) beads when

Fig. 8. DoE generated contour plot of PSD (D90-values) from screening investigations of suspension containing 10% Fenofibrate milled under constant milling conditions toinvestigate the mixture producing the minimal particle size.

Table 4Factors of the detailed investigation of the system HPMC/SDS (second DoE).

Factor level 1 level 2 level 3 level 4 level 5

Amount of SDS 0.1% 0.25% 0.5% 0.75% 1.0%Amount of HPMC 2% 3% 4% -/- -/-Amount of Fenofibrate 10% 20% 30% -/- -/-

86 M. Hagedorn et al. / International Journal of Pharmaceutics 530 (2017) 79–88

clashing against the top or bottom of the vials. This assumption isfurther supported by the finding that an increase of the amounts ofmilling beads � and thus the increase of the (grinding) surface area� makes the milling process faster and more efficient. However, a

Fig. 9. Contour plot of the detailed investigation of the system HP

wearing of PP from the vessel should be taken into account but wasnot investigated so far.

Basically, the high impact of the milling beads duringDC-milling depends on the strong centrifugal acceleration, andthus on the high centrifugal speed of the dual centrifuge. However,at least as important seems to be the horizontal orientation of thetube-sized milling vials placed in the rotating disks of the dualrotor. This unique orientation of the vials results in a recurringprocess involving (i) transport of the milling beads against thecentrifugal acceleration, (ii) the following acceleration over thewhole length of the vial � and thus the chance to accumulate highkinetic energy, and (iii) the resulting impact at the top or bottom ofthe vials (Fig. 11). At a rather low speed of the dual centrifuge of

MC/SDS containing different amounts of the API Fenofibrate.

Fig. 10. Storage stability of the most stable Fenofibrate-suspensions. Storage conditions were room temperature (20 �C) for 6 weeks.

Fig. 11. Schematic process of milling process within dual centrifugation process.

M. Hagedorn et al. / International Journal of Pharmaceutics 530 (2017) 79–88 87

1500 rpm this high-energy impact takes place about 1000 timesper minute or 17 times per second. Despite the high energy intakeit was shown that no alteration within the crystal structureoccurred.

In addition to the aspects discussed above, an at least evenimportant aspect might be that the bead-bead-interactions areextremely frequent, since virtually all beads accelerate as “one”cloud and also clashes at once to the bottom or top of the vials,including the sample material (cloud milling). This means thatnearly every bead is involved in the milling process, which is incontrast to a typical planetary ball mill, where the beads aredistributed around the walls of the milling device with only a smallportion of beads actively involved in the milling process. Thatnearly all beads are involved in the milling process does not onlyincrease the effectiveness of this milling procedure, but also helpsto prevent high sample temperatures, possibly by preventing bead-wall interactions as known from ball mill approaches. This “cloudmilling” phenomena might also contribute to the finding thatDC-milling is effective even at lower centrifugal forces (e.g.1000 rpm).

That the impact of the beads � and thus the milling power � ishigher than in a usual planetary ball mill is supported by theobservation that the particle sizes obtained from DC-milling arenot strongly affected by an increasing viscosity of the millingmedia. This is in contrast to the findings of Nakach (Nakach et al.,2014) who proposed that at low stabilizer concentrations theparticles size is limited by the stabilizer amount corresponding tofull coverage of the API particles and that at high stabilizerconcentration the resulting particle sizes are limited by themechanical energy of milling system (high viscosity). While theeffect of the low stabilizer concentration can be confirmed in thepresented data (Fig. 8, 1–3% HPMC), a particle size increase due to ahigh concentration of the stabilizer is not visible (Fig. 9, up to 4%HPMC), demonstrating a very high energy input during DC-milling.

In contrast to the usual wet ball milling used for theinvestigation of nanomilling conditions, which needs a millingtime of 7 h and 1000 mg API per milling trial, DC-nanomillingallows the same process with much lower batch sizes. While in thisstudy the standard API amount was 100 mg per vial, it could beshown that DC-milling is also possible with 10 mg withouteffecting the resulting particle sizes, meaning that with less than500 mg API a clear image of an optimal formulation can be drawn.Furthermore, the DC-instrument used in this study allowsprocessing of 40 vials at once in a highly reproducible mannerand a very short processing time and is therefore ideal for rapidformulation screening.

4.2. DC-DoE-approach

Since DC-milling allows a lot of parallel experiment in a shorttime, not only an effective formulation development is possible,but also the implementation of the quality by design approach (ICHguidance for industry Q8 (R2)).

The presented first DoE-approach proves that one DC-run with40 samples is sufficient to select a suitable polymer/surfactant-system, interactions of polymers and surfactants can be seen at aglance. With a second one-day screening it is possible toinvestigate the best polymer/surfactant-system in more detailand optimize the formulation. Since the second DoE-run focusseson only one polymer/surfactant-system, a higher number of levelsof each parameter compared to the first screening can beinvestigated. The second DoE-run proves also the reproducibility,since all particle sizes and trends of the first DoE were confirmed.There is also the possibility to investigate higher API loads (10, 20and 30%), which gives additional information about theAPI/polymer/surfactant-system. For the chosen API/polymer/surfactant-system, it could be shown that a higher API load resultsin smaller particle sizes, which can explained by additional

88 M. Hagedorn et al. / International Journal of Pharmaceutics 530 (2017) 79–88

collisions of drug particles with themselves during milling (Kumaret al., 2015).

With these two consecutive DoE-approaches it was shown thata formulation development including the manufacture of sufficientmaterial for an ongoing stability evaluation is possible within2 days. Even a more complex formulation approach seem to be easyto perform (e.g. by testing a higher number of polymer/surfactant-combinations) and can be finalized within a few days. Thedescribed screening procedure was already used for the formula-tion development of over 20 different APIs. Nevertheless the initialselection of the excipients is one of the most important factors for asuccessful DC-DoE-approach. Here, a data base containingsuccessfully produced nanosuspensions graded with respect tothe APIs and those objectives can be a helpful tool and limits thenumber of potential excipients and following the number ofmilling trials.

The presented DC-DoE approach can be used as a tool forindustry to perform fast formulation develop with a broad processunderstanding and furthermore fulfil the requirements of theauthorities with respect to QbD (quality by design). The academiccommunity can use the DC-DoE approach to further investigateinteractions of different excipients and APIs.

5. Conclusion

This is the first report that shows that DC-nanomilling of poorlysoluble drug compounds is easy, provides superior milling result, isfast, and temperature of the samples can be controlled as well.DC-nanomilling can be performed with only minimal amounts ofAPI and with 40 samples in parallel, which allows efficientformulation screening.

The superior milling power seems to be the result of a uniquecombination of very special features of DC-milling. Milling ispossible in plastic vials and milling preferentially takes placebetween the beads. This increase the effectiveness of the processand the reduction of bead-wall-interactions helps to avoid highsample temperatures. Finally, the use of cylindrically shaped vialsand their horizontally orientation in the dual rotor especially helpsthe beads to accumulate high kinetic energy while moving alongthe inner cylinder of the vials.

The combination of a DoE-approach with the new DC-formulation screening leads to a very rapid and deep understand-ing of formulation aspects. This can in addition be achieved with aresulting design space which can help to fulfil the current ICHguideline Q8 (R2). For the first time the examination of complexformulation systems with a reasonable workload is possible due tothe high number of available data points.

The presented data show that the results of DC-millingexperiments with known APIs are highly comparable to theresults described in literature obtained by common agitator mills.Hence DC-milling can be assumed as predictive for milling in alarger scale, which has to be investigated in further studies.

Acknowledgements

The authors would like to thank Dr. Thilo Ludwig from thedepartment of Inorganic Chemistry of the Albert-Ludwig-Univer-sity Freiburg for his assistance in XRPD analysis and Mrs. SabineBarnert from the Department of Pharmaceutical Technology andBiopharmacy of the Albert-Ludwig-University Freiburg for provid-ing the cryo-TEM pictures.

References

Amidon, G.I., Lennernäs, H., Shah, V.P., 1995. A theoretical basis for abiopharmaceutic drug classification: the correlation of in vitro drug productdissolution and in vivo bioavailability. Pharm. Res. 12, 413–420. doi:http://dx.doi.org/10.1023/A:1016212804288.

Azad, M., Afolabi, A., Bhakay, A., Leonardi, J., Davé, R., Bilgili, E., 2015. Enhancedphysical stabilization of fenofibrate nanosuspensions via wet co-milling with asuperdisintegrant and an adsorbing polymer. Eur. J. Pharm. Biopharm. 94,372–385.

Buckton, G., Beezer, A.E., 1992. The relationship between particle size and solubility.Int. J. Pharm. 82, 7–10. doi:http://dx.doi.org/10.1016/0378-5173(92)90184-4.

Chin, W.W.L., Parmentier, J., Widzinski, M., Tan, E.H., Gokhale, R., 2014. A briefliterature and patent review of nanosuspensions to a final drug product. J.Pharm. Sci. 103, 2980–2999. doi:http://dx.doi.org/10.1002/jps.24098.

Jinno, J., Kamada, N., Miyake, M., Yamada, K., Mukai, T., 2006. Effect of particle sizereduction on dissolution and oral absorption of a poorly water-soluble drug,cilostazol, in beagle dogs. J. Control. Release 111, 56–64. doi:http://dx.doi.org/10.1016/j.jconrel.2005.11.013.

Juenemann, D., Jantratid, E., Wagner, C., Reppas, C., Vertzoni, M., Dressman, J.B.,2011. Biorelevant in vitro dissolution testing of products containing micronizedor nanosized fenofibrate with a view to predicting plasma profiles. Eur. J. Pharm.Biopharm. 77, 257–264. doi:http://dx.doi.org/10.1016/j.ejpb.2010.10.012.

Junghanns, J.A.H., 2008. Nanocrystal technology, drug delivery and clinicalapplications. Int. J. Nanomed. 3, 295–309.

Kesisoglou, F., Panmai, S., Wu, Y., 2007. Nanosizing — Oral formulation developmentand biopharmaceutical evaluation. Adv. Drug Deliv. Rev. 59, 631–644. doi:http://dx.doi.org/10.1016/j.addr.2007.05.003.

Kumar, S., Shen, J., Zolnik, B., Sadrieh, N., Burgess, D.J., 2015. Optimization anddissolution performance of spray-dried naproxen. Int. J. Pharm. 486, 159–166.doi:http://dx.doi.org/10.1016/j.ijpharm.2015.03.047.

Laaksonen, T., Liu, P., Rahikkala, A., Peltonen, L., Kauppinen, E.I., Hirvonen, J.,Järvinen, K., Raula, J., 2011. Intact nanoparticulate indomethacin in fast-dissolving carrier particles by combined wet milling and aerosol flow reactormethods. Pharm. Res. 28, 2403–2411. doi:http://dx.doi.org/10.1007/s11095-011-0456-z.

Liversidge, G.G., Conzentino, P., 1995. Drug particle size reduction for decreasinggastric irritancy and enhancing absorption of naproxen in rats. Int. J. Pharm.125,309–313. doi:http://dx.doi.org/10.1016/0378-5173(95)00148-C.

Loftsson, T., Brewster, M.E., 2010. Pharmaceutical applications of cyclodextrins:basic science and product development. J. Pharm. Pharmacol. 62, 1607–1621.doi:http://dx.doi.org/10.1111/j.2042-7158.2010.01030.x.

Möschwitzer, J.P., 2015. Systematic screening of different surface modifiers for theproduction of physically stable nanosuspensions. J. Pharm. Sci. 1128–1140. doi:http://dx.doi.org/10.1002/jps.24266.

Malamatari, M., Somavarapu, S., Bloxham, M., Buckton, G., 2015. Nanoparticleagglomerates of indomethacin: the role of poloxamers and matrix former ontheir dissolution and aerosolisation efficiency. Int. J. Pharm. 495, 516–526. doi:http://dx.doi.org/10.1016/j.ijpharm.2015.09.013.

Massing, U., Cicko, S., Ziroli, V., 2008. Dual asymmetric centrifugation (DAC)-A newtechnique for liposome preparation. J. Control. Release 125, 16–24. doi:http://dx.doi.org/10.1016/j.jconrel.2007.09.010.

Nakach, M., Authelin, J.R., Tadros, T., Galet, L., Chamayou, A., 2014. Engineering ofnano-crystalline drug suspension: employing a physico-chemistry basedstabilizer selection methodology or approach. Int. J. Pharm. 476, 277–288. doi:http://dx.doi.org/10.1016/j.ijpharm.2014.09.048.

Palo, M., Kolakovic, R., Laaksonen, T., Määttänen, A., Genina, N., Salonen, J., Peltonen,J., Sandler, N., 2015. Fabrication of drug-loaded edible carrier substrates fromnanosuspensions by flexographic printing. Int. J. Pharm. 494, 603–610. doi:http://dx.doi.org/10.1016/j.ijpharm.2015.01.027.

Romero, G.B., Keck, C.M., Müller, R.H., 2015. Simple low-cost miniaturizationapproach for pharmaceutical nanocrystals production. Int. J. Pharm. doi:http://dx.doi.org/10.1016/j.ijpharm.2015.11.047 Elsevier B.V..

Tuomela, A., Laaksonen, T., Laru, J., Antikainen, O., Kiesvaara, J., Ilkka, J., Oksala, O.,Rönkkö, S., Järvinen, K., Hirvonen, J., Peltonen, L., 2015. Solid formulations by ananocrystal approach: critical process parameters regarding scale-ability ofnanocrystals for tableting applications. Int. J. Pharm. 485, 77–86. doi:http://dx.doi.org/10.1016/j.ijpharm.2015.03.009.

Van Eerdenbrugh, B., Froyen, L., Van Humbeeck, J., Martens, J.A., Augustijns, P., VanDen Mooter, G., 2008. Drying of crystalline drug nanosuspensions — Theimportance of surface hydrophobicity on dissolution behavior uponredispersion. Eur. J. Pharm. Sci. 35, 127–135. doi:http://dx.doi.org/10.1016/j.ejps.2008.06.009.

Van Eerdenbrugh, B., Stuyven, B., Froyen, L., Van Humbeeck, J., Martens, J.A.,Augustijns, P., Van Den Mooter, G., 2009. Downscaling drug nanosuspensionproduction: processing aspects and physicochemical characterization. AAPSPharmSciTech 10, 44–53. doi:http://dx.doi.org/10.1208/s12249-008-9170-5.