Segregation in inhalable powders: Quantification of the effect of vibration on adhesive mixtures

The objective of this investigation was to study the effect of induced vibrations on adhesive mixtures containing budesonide and salbutamol sulphate as active pharmaceutical ingredients (APIs) and InhaLac 70 as carrier. A series of adhesive mixtures with varied API concentration (1 – 4%) was prepared for each API. Half of the adhesive mixture was stressed on a vibrating sieve under conditions resembling hopper flow. Based on scanning electron micrographs, it was concluded that InhaLac 70 contains particles of two distinct shapes, one irregular with groves and valleys and the other more regular with well defined edges. The dispersibility of the control and stressed mixtures was studied using a next generation impactor. The stressed mixtures containing 1 and 1.5% API displayed a significant reduction in fine particle dose (FPD) compared to the control. The reduction in FPD resulted from a loss of API from the adhesive mixture during vibration and as a consequence of restructuring and self agglomeration resulting in reduced dispersibility. However, no significant difference was observed for mixtures with larger weight fractions of API (2 and 4% API) but these have a drawback of reduced fine particle fraction (FPF). It is concluded that vibrations induced on the adhesive mixtures during handling potentially have a significant effect on the dispersibility of the API and the total amount of drug delivered to the lungs.


Introduction
Drug delivery to the lung is commonly used to treat diseases such as asthma and chronic obstructive pulmonary disease [1] and it also has the potential to deliver biologics [2] such as insulin [3] and antimicrobial peptides [4].The lung has numerous physiological advantages such as a large surface area, low enzymatic activity and high permeability, which are prime factors for potentially easy and successful delivery of locally as well as systemically acting drugs [5].Numerous inhaler types have been developed over the years and the Dry Powder Inhalers (DPIs) are steadily gaining impetus [6][7][8].
Most DPIs use a special type of mixture known as an adhesive mixture for the delivery of Active Pharmaceutical Ingredients (APIs) to the lungs [9,10].(So-called carrier-free formulations, often consisting of loose aggregates of particles, are an interesting alternative but will not be considered further in this work.)Adhesive mixtures consist of micronized API (fine particles) of size < 5μm and comparatively larger sized carrier particles (∼ 100μm).This work focuses on adhesive mixtures with negligible amounts of inactive fines, and we will for simplicity therefore refer to the fine API particles simply as 'fines' or 'fine particles'.Owing to their high surface-to-volume ratio, van der Waals' forces have a significant effect on the fine particles, leading to high cohesiveness and agglomeration, resulting in low flowability [11,9].Carrier particles with adhered fine particles improve flowability and dispersibility [11].During inhalation, the particles in the adhesive mixture are exposed to air flow and impacts, aiming to detach API particles from the carriers.Thus, the aerosol performance is strongly dependent on the balance between adhesive and cohesive interparticle forces [12].Due to the composition of the mixtures, micronized drug particles may interact with numerous neighbours, comprising other drug particles as well as coarse, and sometimes also fine, lactose particles.There is thus a possibility that agglomerates with varying size, structure and composition exist in the mixture.This highlights the importance of trying to control parameters which could impact powder structure, aiming to gain the most optimal performance of end product [13].A parametric understanding of adhesive mixtures would provide knowledge to improve formulation performance [14].
Production of commercially qualified adhesive mixtures comes with requirements of stable and robust manufacturability and performance [15].As for all pharmaceutical products, dose uniformity is of utmost importance, and this requires adequate mixing and distribution of the typically very small fraction of API over the (surfaces of the) carriers.
Maintaining homogeneity during handling is thus of subsequent importance.The assurance of production of adhesive mixtures with high content uniformity does, however, come with a number of challenges, related to the nature of the fine particles, such as adhesion to blender walls and containers, sieves and impeller wings [16].The risk of loss of fine particles during production with subsequent risk of failure to achieve intended drug content therefore has to be analysed and taken into consideration [16][17][18].The behaviour of adhesive mixtures is complex but can be rationalized in terms of structural changes of the mixture that take place when the fraction of fine API is increased.For example, Rudén and co-workers have presented a conceptual scheme referred to as a blend state map [19][20][21].According to this scheme, adhesive mixtures can, depending on the localization of the fines on the carrier and the structure of the adhered layer of fines, be classified into up to four types, denoted S1, S2a, S2b and S3.In S1, the fines are predominantly present in cavities.An adhered layer of fines is formed during S2a that increases in thickness and becomes less tightly bound during S2b.Finally, S3 is characterised by the formation of self-agglomerates.Similar ideas have been presented by Young et al. [22] and Hertel et al. [23].
Segregation may potentially occur during packaging and handling of adhesive mixtures, which compromises the aerosolization performance, whereby affecting quantitative and qualitative properties of the mixture [17].The increased interest in creating mixtures of higher drug loads raises an important question about the mechanical stability of adhesive mixtures.The purpose of this work is to study the effect of small vibrations, resembling those in a hopper, on the stability of pre-formulated adhesive mixtures.The adhesive mixtures were formulated by mixing the lactose carrier InhaLac 70 with different mass fractions of API (micronized budesonide or salbutamol sulphate).The performance of the thus formed adhesive mixtures was analysed using a next generation impactor (NGI) and fine particle fraction and fine particle dose were reported.

Materials
InhaLac 70, an inhalation grade α lactose monohydrate, was procured from Meggle (Wasserburg, Germany) and used as carrier.Micronized budesonide and salbutamol sulphate were obtained from AstraZeneca (Gothenburg, Sweden) and used as API.For brevity, budesonide is henceforth referred to as BUD and salbutamol sulphate is referred to as SBS.
Ethanol (Solveco, Sweden) along with fluocinolone acetonide (Sigma Aldrich, USA) was used as an internal standard for BUD.Milli-Q water along with orciprenaline sulphate (Sigma Aldrich Com St. Louis, USA) was used an internal standard for SBS.Phosphate buffer was prepared using ortho-phosphoric acid (Sigma Aldrich GmbH, Switzerland) and sodium dihydrogen phosphate (Sigma Aldrich Switzerland).Milli-Q water was used in BUD sample preparation.Brij 35 (Sigma Aldrich, Ontario, Canada) was used along with glycerol (Sigma Aldrich, Germany) and ethanol to prepare Brij solution.

Basic characterisation of starting materials
Owing to their highly cohesive nature, micronized BUD and SBS tend to form large self-agglomerates and were therefore sieved prior to characterisation (710 μm aperture size).InhaLac 70 has a good flowability and larger self-agglomerates were not observed.All powders were conditioned in a climate controlled room for at least 24 h prior to any experiments or mixture preparation.The temperature and relative humidity (RH) were maintained between 21-24 • C and 30-34% RH, minimizing the risk of powders being exposed to any changes in surrounding conditions during storage.

Particle size
The particle-size distributions of the powders were measured using a Beckman Coulter (Brea, California, USA) LS230 Small Volume Module laser diffractometer.Lactose was dispersed in isopropanol and BUD was dispersed in a solution 0.1% Tween 80 in Milli-Q water.SBS was difficult to measure owing to its high dissolution rate in water and most organic solvents; therefore information about particle size was obtained from [21], in which the size of SBS from the same batch was measured using dispersion in air rather than liquid.The measurements were repeated in triplicates and the Fraunhofer model [24] was used for analysis.The span of particle-size distribution was calculated using, where D x denotes the xth percentile.The span indicates how far the 10th and 90th percentiles are apart on a scale determined by the median.

Particle shape
Particle shape was assessed using Scanning Electron Microscopy (SEM; Leo/Zeiss 1530 microscope, Jena, Germany).The SEM sample holders were coated with a thin layer of Au/Pd under argon using a sputter coater (Polaron Quorum Technologies Ltd., Newhaven, United Kingdom).An in-lens detector was used with a magnification of 100× for carrier and 20 000× for API.An acceleration voltage of 2.0 kV and a working distance between 1.6 -2.4 mm were used.

Specific surface area
The external specific surface area of the carrier and APIs was determined using air permeability.A transient Blaine permeameter was used for the API [25], whilst a steady-state permeameter was used for the carrier particles [26].The specific surface area was calculated by the Kozeny-Carman equation with (API) or without (carrier) slip flow correction [26].

Particle density
The apparent particle density was measured using Helium pycnometry (AccuPyc 1330 Micrometrics Instruments, Nordcross, USA).The steel cylinder was filled with powder to a fill volume of 50%.Each sample was measured in duplicates, and each trail was analysed five times in succession.

Bulk density
A Freeman FT4 Powder Rheometer (Freeman Technology, Tewkesbury, UK) was used to measure the unsettled bulk density (UBD) of the starting materials.The sample cylinder was filled with powder, thereafter the powder was gently conditioned by a steel blade, followed by removal of excess powder.The term unsettled refers to a measurement with only gravity in effect and no external applied force.

Preparation of adhesive mixtures
A range of adhesive mixtures were prepared with four different weight fractions of BUD and SBS; see Table 1.The adhesive blends were prepared by initially mixing the carrier particles and micronized API manually using a sandwich layer technique in a 250 mL glass vessel with an approximate fill volume of 50%.The final blend was formed using a Turbula T2F mixer (Willy A. Bachofen AG, Switzerland) with a mixing time of 1 h at 46 rpm [19,20].API was added in the same proportion for both BUD and SBS systems as summarised in Table 1.The percentage of fine particles could also be represented as a surface coverage ratio (SCR), which was calculated rather than measured as follows, Here S w,c and S w,f are the weight specific surface area of the carrier and API (fines) respectively, and m c and m f are the corresponding masses of carrier and API.The constant α was set to 1/π, which corresponds a rectangular closed packed structure of spherical particles [27] and is in line with previous work by Rudén et al. [19].

Inducing stress on adhesive mixtures
A method to induce stress on adhesive mixtures was designed and used to study the effect of stress on the stability of adhesive units and to measure the segregation of the mixtures during handling.The stress on the adhesive mixtures was induced using a Retsch 200AS (Verder Scientific GmbH & Co. KG) vibratory sieve shaker with conditions similar to those in a hopper during packing [28][29][30].The adhesive mixture was poured onto a sieve with an aperture size of 65 μm and vibration with an amplitude of 1 mm and a frequency of 50 Hz was induced for 120 s.A schematic representation of the setup is provided in Fig. 1.It was observed that longer vibration time (150 s or more) did not result in a significant difference compared to 120 s.The sample was transferred back to the original container and conditioned in a climate controlled room for 24 h.

Mixture Morphology
The mixture morphology was assessed using SEM.The procedure was the same as for particles (Section 2.2.2) except that a different magnification was used (between 500× and 1000 × ).

Quantification of API: UPLC analysis
An ACQUITY UPLC system (Waters Corp, Milford USA) equipped with a C18 BEH (Ethylene Bridged Hybrid) 2.1 mm × 50 mm column (1.7 μm particle size) and a photo diode array (PDA) detector was used for quantification of the API.Internal standard for BUD was prepared by adding 40 mg of fluocinolone acetonide in 1 L of 95% ethanol and for SBS by adding 25 mg of orciprenaline sulphate in 1 L of Milli-Q water.The calibration curve based on the reference solution with 4 data points had an R 2 of 0.995 and 0.998 for BUD and SBS, respectively.The Ultra Violet (UV) wavelength of the detector was set to 254 nm.The following UPLC methods were used to analyse the concentration of API: Budesonide.Mobile Phase A -0.03% TFA in Milli-Q water; Mobile Phase B -0.03% TFA in acetonitrile; Ratio -65% A and 35% B; Seal Wash -LC grade isopropanol; Needle wash -50:50 methanol and Milli-Q water; Flow rate -0.6 mL/min; Retention time -1 min.

Mixture Homogeneity and Carrier Residue
Homogeneity analysis was performed by taking 10 samples from each powder blend (both control and stressed) with effective weight ranging from 13.5 to 17.7 mg except for the S 1% BUD sample where somewhat larger samples (28.5-32.5 mg) were taken to ensure adequate quantification.The samples were collected from the surface, edges, centre and bulk of the powder mixtures in an attempt to capture maximum variability.The samples were scooped onto a glass petri-dish, and dissolved in 20 mL internal standard.The scooping technique used was similar to the one used by Rudén et al. [21] in their study of the blend state map.The collected samples were set to shake on a shaking table for 20 min at 60 rpm to dissolve the sample.0.5 mL of each sample was added to LC vials and 0.8 mL phosphate buffer (pH 3.2) was added to samples containing BUD to increase the peak resolution.The concentration of API in each sample was quantified using UPLC (Section 2.6).The calculated API mass was normalised to a nominal sample weight of 15 mg.The blend homogeneity was evaluated based on the relative standard deviation across 10 sample points.
Segregation induced by vibrations of adhesive mixtures could lead to loss of API from the powder blend.To account for the loss, the average masses of API in the control and stressed sample were compared.The carrier residue (CR), representing the percentage of API remaining on the carriers, was defined as Mass of API in the stressed sample Mass of API in the control sample × 100, where the same nominal weight of the stressed and control samples was used.

Impactor experiments
Dispersion experiments were performed using an NGI (Copley Scientific, UK) together with a Screenhaler device [31] coupled to a Turbuhaler mouthpiece.The volumetric flow rate was set to 60 ± 0.3 L/min, with a suction time of 4 s, corresponding to a total suction volume of 4 L. The actuation and flow rate were controlled using a Trigger Box model III (AB FIA, Lund, Sweden).To avoid particle bouncing, the NGI stages were coated with a solution of ethanol, Brij 35 and glycerol; 51:15:34 wt%.The pre-separator was filled with 20 mL of internal standard (Section 2.6).A total of 6 actuations were performed with 15 mg of powder per actuation.The experiments were performed in triplicates.
After the actuations were performed, 20 mL internal standard was added to each NGI stage.The NGI plate was set on a shaking table for about 20 min at 60 rpm to ensure that all API dissolved.The throat along with 20 mL internal standard and pre-separator were set in motion using a Sample Preparation Unit (Copley Scientific, UK) for 20 min.Samples for UPLC analysis were prepared by adding 0.5 mL solution from each stage, throat and pre-separator to LC vials.In order to improve the resolution of peaks in the chromatogram, 0.8 mL phosphate buffer solution (pH 3.2) was added to the solution containing BUD. b Mass added to obtain a total weight of 50.0g.

Analysis of impactor data
The Mass Median Aerodynamic Diameter (MMAD) was calculated using instructions from the European Pharmacopeia [32].The fine particle fraction (FPF), representing the percentage of fine particles below 5 μm diameter, was calculated from the following equation: where FPD is the fine particle dose and ED is the emitted dose.ED is defined as the total dose emitted which includes throat, pre-separator and the sum of 7 stages of NGI and Micro-Orifice Collector (MOC).
The FPD is the mass of API below 5 μm aerodynamic diameter.

Stage grouping
The stages of an NGI could be grouped based on the aerodynamic diameter of the particles.In order to increase the data resolution the following grouping based on Roberts et al. [33] with a flow velocity of 60 mL/min was considered: Non-sized: Particles from the throat and induction port.Extra-fine: Particles with an aerodynamic diameter < 1 μm, collected from Stages 5, 6, 7 and MOC.

Flowability and packing
The Freeman FT4 Powder Rheometer was used to assess flowability and packing of the blends.The unsettled bulk density was determined as described in Section 2.2.5.The compressed bulk density (CBD) was measured by compressing the powder at normal stress of 30 kPa.The Hausner ratio was subsequently calculated as the ratio between the compressed and unsettled bulk density, i.e. as the ratio between CBD and UBD.The measurements were repeated in triplicates.

Statistical analysis
To better understand and represent the data, multiple t-tests were performed with a p value of 0.05 between the control and stressed samples.

Basic particle and powder characteristics
Basic characteristics of the particles and powders are summarized in Table 2.Although BUD is slightly larger than SBS, their median particle size are both about 2 μm.However, the span of particle size distribution, calculated using Eq. ( 1), was almost twice as large for SBS than for BUD.
The size of the carrier InhaLac 70 was about 250 μm.Consistent with these results, the weight-specific surface area was somewhat larger for SBS than for BUD, and these values were more than two orders of magnitude larger than that obtained for InhaLac 70.Moreover, the bulk density is as expected considerably smaller for the micronized APIs than for the relatively large lactose carriers.This finding is mostly caused by differences in particle size, but it can also be noted that the APIs have somewhat lower apparent density than the carriers.
Table 3 summarizes the SCRs of the adhesive mixtures, as obtained from Eq. (2) using the weight fractions in Table 1 and weight-specific surface areas in Table 2.The weight fractions were chosen so as to obtain SCRs of 0.5, 0.75, 1 and 2 for adhesive mixtures containing BUD.Since the same weight fractions were used for SBS, the SCRs differ.Specifically, the SCRs are for each mixture type (S 1% etcetera) about 35% larger for SBS than for BUD.
Scanning electron micrographs of the initial materials are shown in the Fig. 2. It could be observed from Fig. 2a that InhaLac 70 exhibits different particle shapes, ranging from flatter particles with well-defined geometrical edges to more irregular particles with cavities and deeper groves.Fig. 2b and c show the SEM images for BUD and SBS, respectively.The SBS particles present an elongated shape whereas the BUD particles tend to be more spherical.e Unsettled bulk density.

Morphology of adhesive mixtures
The physical structure of carrier-API complexes for control and stressed (Section 2.4) samples was studied qualitatively using SEM imaging.Images of each sample (with a magnification between 500× and 1000 × ) are shown in Figs. 3 and 4 for BUD and SBS, respectively.For each weight fraction, represented in Fig. 3a to d and Fig. 4a to d, control samples (i and ii) are compared to stressed samples (iii and iv).An attempt has been made to present particles with an irregular surface (i and iii) and a mostly flat surface (ii and iv).The morphological differences between the powder blends are next discussed.To ease interpretation, magnified views of Fig. 3a-d -S8.S 1% Fig. 3a and 4a show the SEM images of adhesive units formed by mixing carrier particles and with 1.009% API by weight.No significant visual difference between the control (i) and the stressed (iii) samples is observed for irregular shaped adhesive units with valleys or cavities, as the fine particles are captured in the grooves of the cavities and are well protected during vibrations.Fig. 3a (ii) and (iv) show carrier particles with flat surfaces.It could be observed from visual inspection that the number density of fine particles on the flat surface of the stressed particle (iv) is less than that of the control particle (ii).
S 1.5% The adhesive blend was formed by adding 1.513% of API to carrier particles by weight and could be observed in Fig. 3b for BUD and Fig. 4b for SBS.It is seen that the coverage of API is marginally higher compared to that of S 1% .On further analysing the sub-images (i) and (iii) it could be observed that for controlled and stressed samples, the API particles are well protected in the cavities for irregular shaped carriers.In contrast, for carriers with flat surfaces, the stressed sample (iv) appears to have a lower fine particle density compared to the control (ii).S 2% Figs.3c and 4c show SEM images of adhesive units with an API concentration of 2.018% for BUD and SBS, respectively.With increasing weight fraction of API, a larger coverage of the carrier surfaces is observed, with the formation of secondary and tertiary layers of fines.Adhesive units with irregular shaped carrier particles with groves and cavities (i and iii) show a similar morphology for both control and stressed samples.For adhesive units with carrier particle with flat surfaces, the control sample (ii) has a higher fine particle density compared to the stressed sample (iv).The secondary and tertiary layers tend to be lost due to vibration whereas the primary layers are adhered stronger to the carrier surface.
S 4% Fig. 3d and 4d show an adhesive unit from a blend containing 4.036% of API for BUD and SBS, respectively.Similar to Figs. 3c and 4c, secondary and tertiary layers of fine particles could be observed.On increasing the coverage to twice the surface area of the carrier, loose API particles start to appear in the background as also noted by Rudén et al. [19].The irregular shaped carriers (i and iii) have well-protected API particles compared to the carriers with flatter surfaces (ii and iv).However, the losses from the surface of the carrier particles is not significant as the amount of loose API is high and thus there is a constant exchange of API instead.The morphological changes are not very clear owing to a higher density of fine particles in the background.

Mixture homogeneity and carrier residue
The homogeneity or the variation of distribution of API in the sample, expressed in terms of the relative standard deviation (RSD) calcu-  lated from UPLC results, is presented in Table 4.A RSD < 5% is commonly regarded as acceptable (e.g.[34]), although a stricter requirement (< 3%) has also been used [35].The RSD is for most samples below 5%, i.e. acceptable.Some notable exceptions are stressed S 1% and S 4% samples for BUD and control and stressed S 4% samples for SBS but overall the RSD is < 10%.A higher RSD is expected for samples with higher API concentration.A higher weight fraction of API tends to increase the segregation tendency of the mixture.
The overall effect of vibration on the adhesive mixtures could be inferred from the carrier residue, calculated by Eq. ( 3) and summarised in Table 5.The CR ranges from 90 to 99% for BUD and from 94 to 98% for SBS.For both APIs, the largest losses (lowest CRs) were obtained for the intermediate amounts of fines (S 1.5% and S 2% ).More detailed information is obtained from the aerosolization experiments, whose results are reported next.

Aerosolization performance
Aerosolization studies were performed for stressed and control samples for different weight fractions of API using an NGI.The results are reported in terms of FPF and FPD/ED in Figs.5-8 for S 1% , S 1.5% , S 2% and S 4% , respectively.The FPD and ED correspond to a nominal sample weight of 15 mg.S 1% .Fig. 5 summarizes the results obtained for an API weight fraction of 1.009%.For BUD, the ED was significantly different (p < 0.05) between the control and stressed samples (Fig. 5b).For SBS, the mass was in all stages significantly lower for the stressed compared to the control sample (Fig. 5d).The effect of stress on the FPF for BUD and SBS was not significant (Fig. 5a and c).
S 1.5% .Fig. 6 represents the results obtained for an API weight fraction of 1.513%.For BUD, the < 5 μm stage and ED are significantly different between the control and stressed samples (Fig. 6b).For SBS, there is a significant difference observed between the control and stressed samples for all stages (Fig. 6d).Hence, the effect of vibration is more pronounced for SBS compared to BUD.It could be observed that FPF for the stressed SBS sample is shifted towards lower values relative those of the control (Fig. 6c).A similar trend is seen for BUD (Fig. 6b), but in this case there is no significant difference between the control and the stressed sample.S 2% .Fig. 7 represents the adhesive mixtures with an API weight fraction of 2.018%.The ED is significantly different between the control and stressed samples for both BUD (Fig. 7b) and SBS (Fig. 7d), which otherwise have a similar aerosolization performance (p > 0.05).Although a slight shift of the FPF could be seen, similar to the one observed for S 1.5% , no significant difference was obtained (Fig. 7a and c).S 4% .Finally, results for adhesive mixtures with an API weight fraction of 4.036% are provided in Fig. 8.The profiles for both FPF and FPD are similar for the stressed and control samples for both BUD and SBS.The high weight fraction of API results in loose API in the mixtures and there are no significant differences observed except for the ED of BUD (Fig. 8b).

Flowability and packing
The mechanical properties of the control and stressed adhesive mixtures, reported as the Hausner ratio and unsettled bulk density, are compared in Fig. 9 for both BUD and SBS.It was difficult to formulate and experiment on the S 4% SBS mixtures as clumps were formed; hence only 3 data points are reported.The presented data suggest that vibration of the mixtures does not have any significant effect on the powder mechanics.

Discussion
This paper reports experimental results pertaining to the effect of stress (vibration) on adhesive mixtures.An understanding of the susceptibility of adhesive mixtures to stress is important so that measures can be taken to avoid potential loss of drug being delivered to the lungs.The effects observed and possible causes and interpretations are elaborated upon in the following section.

Homogeneity
As seen in Table 4, the RSD is for stressed samples comparable or higher than for control samples.For high drug loads of SBS (S 4% ), it could be observed that the RSD is between 5-10% in both control and stressed samples.Since the size and density of the API particles are similar (cf.Table 2), these findings most likely have their roots in the type and magnitude of the interparticle forces coupled with differences in particle shape (elongated for SBS and spherical for BUD; cf.Fig. 2).Consistent with our results, difficulties during preparation and handling of SBS mixtures were reported by Rudén et al. [21], who attributed this result to triboelectric charging.
However, although triboelectric charging of SBS mixtures has been addressed by a number of studies (e.g.[36][37][38][39]), it is difficult to find irrefutable support for the claim that triboelectric charging occurs to a larger extent for SBS than for BUD.Telko and Hickey determined the charge on SBS and BUD particles after aerosolization of adhesive mixtures with lactose as carrier and on the contrary found the mean charge on each particle to be almost an order of magnitude higher for BUD than for SBS [40].Earlier work by Byron et al. points in the same direction, but suggests a smaller difference between the two particle types [41].Hussain et al. reported a (positive) charge-to-mass ratio for BUD that was about four times larger in magnitude than the (negative) charge-tomass ratio for SBS.However, the charging tendencies of binary BUDlactose and SBS-lactose mixtures were found to be similar, highlighting the complexity of triboelectric charging [42].
The relative magnitude of cohesive and adhesive forces (the so called cohesive-adhesive balance) could shed further light on the observed differences between BUD and SBS.Begat et al. used SBS, BUD and lactose particles as colloidal probes on flat surfaces prepared from the same materials [12].Interestingly, their results suggest a remarkably weak cohesive interaction between SBS particles, something that is expected to translate into a fragile layer of adhered fines.On the contrary, BUD exhibits a relatively high cohesivity, and is therefore expected to produce tougher adhered layers.The difference in particle shape will likely contribute, since the more elongated SBS particles pack less densely than the more spherical BUD particles [21].The difference in cohesivity between the two APIs is therefore suggested to be the primary reason for the observed results.Nevertheless, this view does not necessarily mean that triboelectrification is unimportant [42].As a result of the low cohesivity between SBS particles, relatively weak electrostatic forces would be required to disrupt bonds between them.Moreover, triboelectric charge would accumulate unevenly for elongated particles, and the local charge density may therefore be higher for elongated than for spherical particles for a given net charge ( [43] and references therein).In addition, smaller particles tend to have higher net charge to surface ratios, so that the effect of triboelectric charging thus is inversely     related to particle size.The smaller size of SBS than BUD could thus increase its susceptibility to triboelectric charging.For high drug loads of BUD (S 4% ), the API detaches from the adhesive units during stressing, forming smaller self-agglomerates leading to the high RSD value reported in Table 4.

Aerosolization performance
Section 3.4 details the effect of stress on the aerosolization performance of adhesive mixtures with varying weight fractions of API.It was observed that there is very small or no significant difference between the   FPF of control and stressed samples.On analysing the mass of particles deposited in different stages of the NGI, it could be observed that both the ED and the FPD show a similar reduction and that the FPF consequently remains the same (Fig. 10b and d).Considering the dependence of the FPF on FPD and ED, this result raises the question about the validity of FPF as an efficiency parameter in the current study.The following discussion therefore focuses on FPD.
Overall, the changes in FPD indicate that the susceptibility to stress (vibration) decreases with increasing mass fraction of fines.Significantly different FPDs between the stressed and control samples were obtained for S 1% and S 1.5% , whereas no significant differences were observed for S 2% and S 4% .Moreover, SBS exhibited a larger susceptibility to stress (vibration) than BUD, especially for small weight fractions of fines (< 2%; Fig. 10a and c).Weak cohesive forces between the SBS particles resulting in fragile adhered layers are suggested as the primary explanation for this finding (see Section 4.1 above).One could conclude that adhesive mixtures with increased mass fractions of fines are more tolerant to stress, but with a trade-off due to reduction in flowability and reduction in the percentage of API delivered to the lungs as inferred from the FPF.
Stage Grouping.Different stages in the NGI could be grouped in order to better elucidate potential differences between stressed and control samples.Based on a classification into 'non-sizing', 'coarse', 'fine' and 'extra fine' described in Section 2.8.3, the histograms displayed in Fig. 11a for BUD and Fig. 11b for SBS could be constructed.The 'fine' (or inhalable) particle size, defined as the range 1 -5 μm, is most relevant for pulmonary drug delivery and is therefore analysed further.In the magnified view, the amount of API deposited in the 'fine' stage group for the stressed and control samples are compared.The deposited mass is indicated on the left y axis and the relative decrease on the right y axis.
For BUD, it is seen in Fig. that stressing reduced the deposed mass by about 20% for S 1% and S 1.5% .The relative decrease is smaller for larger weight fractions of BUD (S 2% and S 4% ).In particular, S 2% with an SCR of 1 appears to be optimal, with a loss of no more than 5%.A similar trend is observed for SBS as seen in Fig. 11b.In particular, the largest relative decrease was observed for S 1.5% (nearly 30%) and the smallest for S 2% (about 10%).On the other hand, the overall loss of API from the system, expressed in terms of the carrier residue (Table 5), was largest the intermediate weight fractions (about 5-10% for S 1.5% and S 2% ) and smaller for S 1% and S 4% (1-2% for BUD and < 5% for SBS).It can thus be concluded that the changes in the overall drug content need not necessarily reflect the changes in the FPD.This result indicates that the stress induced via vibrations affects the adhesive mixture in a different manner than the stress caused by aerosolization.This effect seems to be especially pronounced for S 2% with an SCR of 1.0 for BUD and 1.35 for SBS, which for both carrier types exhibited a large overall loss of API but the smallest loss in the inhalable size range.Although the exact cause of this behaviour is unknown, the fact that it occurs at the same weight fraction of API and approximately the same SCR points towards a structural explanation.The blend state map proposed by Rudén and coworkers may therefore be helpful [19][20][21].We remind the reader that the fines are predominantly present in cavities in S1, that an adhered layer of fines is formed during S2a that increases in thickness and becomes less tightly bound during S2b and that self-agglomerates are formed during S3.Moreover, the mode of deaggregation has been claimed to change from predominantly single particles for S1 and S2a to a combination of single particles and self-agglomerates of fines for S2b and S3 [21].The observed behaviour could be rationalized by the blend state model if a transition between two stages would occur when the amount of API is increased from 1.5 to 2% (either between S1 and S2a or between S2a and S2b).According to this scenario, vibrations cause a similar and relatively high overall loss of API for both stages, but a larger fraction of API is aerosolized after the stage transition.Moreover, the relative decrease in the inhalable particle range is generally larger for SBS than for BUD.Again, this finding is attributed to the remarkably weak cohesivity of SBS relative to BUD [12], resulting in fragile adhered layers for SBS (see Section 4.1).

Mechanisms for particle loss
The loss of fine particles/API from different stages/groups could be explained by the following mechanisms.
Loss from the system.During the process of vibration, the fine particles might be lost from the adhesive mixtures due to collisions with other units or collisions with the container.If not transferred to other adhesive units or forming larger agglomerates, the loose API particles might be lost from the system through the sieve.The loss could be quantified based on the Carrier Residue (CR), Eq. (3) and Table 5.It could be observed that the losses are in the range of 1% to 10%.
Restructuring of fine particles resulting in trapping in the groves It could be observed from SEM images that adhesive units formed from irregular carrier particles, characterised by cavities and valleys, tend to be more stable since the fine particles are captured and well protected within these cavities.For carrier particles with flat surfaces, the tendency of fine particles to be dislodged from the carrier surface increases.The quantification of loss of the fine particles from the system and its effect on aerosolization performance is further described in the Section 3.4.The API particles on the surface of the carrier could be captured in the valleys and grooves, making it difficult for them to be released from the adhesive unit during inhalation.This interpretation is consistent with the results from Fig. 5-8.
Agglomerates of API The API particles could detach from the adhesive units and form larger agglomerates, classified as coarse or non-sized particles.These particles contribute to the emitted dose but do not reach the later stages of NGI and therefore result in a reduced FPF.

Conclusion
The effect of vibration on the stability of adhesive units has been systematically studied for adhesive mixtures of two model APIs (budesonide and salbutamol sulphate) and InhaLac 70 as carrier.Four different mass fractions of API were considered (S 1% , S 1.5% , S 2% and S 4% ).From the presented results, the following conclusions can be drawn: • Vibration may result in significant overall losses (∼ 10%) of API from adhesive mixtures.Loosely bound particles are more susceptible to vibration, and as a result, the reduction in the fine particle dose may be even higher (∼ 20-30%).• The overall losses are largest for adhesive mixtures with intermediate amounts of API (1.5-2% by mass, corresponding to surface coverage ratios of 0.75-1.35).Nevertheless, the smallest relative change in the fine-particle dose was observed for mixtures with 2% API by mass.
These results indicate that the mode of particle loss during vibration is different from that of aerosolization and points towards a high sensitivity of the aerosolization performance on the structure of the adhesive unit.• During vibration, adhered particles were preferentially lost from flat surfaces, whereas those present in valleys or groves did not separate.Irregularities in the form of valleys or groves on the carriers thus decrease the susceptibility of adhesive mixtures to segregation during vibration.• The magnitude of the reduction in fine particle mass exhibited a dependence on the type of API, and the largest reduction was observed for salbutamol sulphate.This finding is suggested to be related to weak cohesive forces between the salbutamol sulphate particles, resulting in a fragile adhered layer.• Stressing in the form of vibrations reduces both the fine particle dose and the emitted dose.As a result, the fine particle fraction is largely unaffected.The fine particle fraction is thus not an adequate performance measure in this context and the fine particle dose should be used instead.
The understanding of the effect of vibration on the adhesive unit during handling throws light on the sensitivity of adhesive mixtures and their performance during aerosolization.This study raises further questions about the parametric variations and its effect on the aerosolization performance and further numerical and experimental studies are needed for generalization.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Coarse: Particles with an aerodynamics diameter > 5 μm, collected from pre-separator and Stage 1 of the NGI.Fine: Particle in the inhalable range of 1-5 μm aerodynamic diameter, collected from Stages 2, 3 and 4.

Fig. 1 .
Fig. 1.Schematic representation of the setup used to induce vibrational stress on adhesive mixtures to study their segregation propensity.
are provided as Figs.S1-S4 in the

Fig. 4 .
Fig. 4. Scanning electron micrographs of adhesive mixture formed from salbutamol sulphate and InhaLac 70 with a weight fraction of API represented by (a) S 1% (b) S 1.5% (c) S 2% and (d) S 4% .In each case, i and ii represent control samples and iii and iv stressed samples.

Fig. 5 .
Fig. 5. Aerosolization performance of adhesive mixtures containing 1.009% by weight (S 1% ) budesonide (BUD) or salbutamol sulphate (SBS).The fine particle fraction (FPF) is shown in (a) and (c) and the deposited API mass (corresponding to a nominal sample weight of 15mg) in (b) and (d).

Fig. 6 .
Fig. 6.Aerosolization performance of adhesive mixtures containing 1.513% by weight (S 1.5% ) budesonide (BUD) or salbutamol sulphate (SBS).The fine particle fraction (FPF) is shown in (a) and (c) and the deposited API mass (corresponding to a nominal sample weight of 15mg) in (b) and (d).

Fig. 7 .
Fig. 7. Aerosolization performance of adhesive mixtures containing 2.018% by weight (S 2% ) budesonide (BUD) or salbutamol sulphate (SBS).The fine particle fraction (FPF) is shown in (a) and (c) and the deposited API mass (corresponding to a nominal sample weight of 15mg) in (b) and (d).

Fig. 8 .
Fig. 8. Aerosolization performance of adhesive mixtures containing 4.0363% by weight (S 4% ) budesonide (BUD) or salbutamol sulphate (SBS).The fine particle fraction (FPF) is shown in (a) and (c) and the deposited API mass (corresponding to a nominal sample weight of 15mg) in b) and d).

Fig. 9 .
Fig. 9. Hausner ratio (a and b) and unsettled bulk density (c and d) vs. weight fraction API (fines) for budesonide (a and c) and salbutamol sulphate (b and d).The error bars represent the standard deviation of three repeated measurements.

Fig. 10 .
Fig. 10.Particle Dose (FPD; a and c) and Fine Particle Fraction (FPF; b and d) for control and stressed samples of adhesive blends containing budesonide (BUD; a and b) and salbutamol sulphate (SBS; c and d).

S
. Sarangi et al.

Table 1
Designation and composition of adhesive mixtures containing budesonide (BUD) and salbutamol sulphate (SBS).
a Weight fraction in final mixture.

Table 2
Basic particle and powder characteristics.Mean values from 3 repeated experiments are reported with standard deviations in parenthesis.
a Median particle size.b Span of particle-size distribution.c Weight-specific surface area.d Apparent density.

Table 5
Drug content and carrier residue for adhesive mixtures containing budesonide (BUD) or salbutamol sulphate (SBS) for nominal (15μg) mixture weight.