Effect of pressure drop on the in vitro dispersion of adhesive mixtures of different blend states for inhalation

In this study, the effect of pressure drop ( Δ P) on the in vitro dispersion of a series of carrier-based adhesive mixtures of different fines-to-carrier proportions, corresponding to the four different blend states of the blend state model, i.e. S1 to S3, was investigated. Four binary and one ternary adhesive mixture consisting of lactose carrier and budesonide fines and lactose fines were prepared. The dispersion was assessed using a next generation impactor (NGI) at Δ P of 0.5, 2 and 4 kPa. For the S1 mixture, where the fines were located in surface cavities of the carrier, the fine particle fraction (FPF) increased nearly linearly with Δ P. For S2 and S3 mixtures, with adhesion layers on the enveloped carrier surface, the FPF-Δ P relationships were bended and approached a plateau. Examination of powder captured in the pre-separator of the NGI led to the conclusion that the dispersion of these adhesive mixtures occurred by erosion of the adhesion layer, i


Introduction
A common formulation principle for dry powder inhalers (DPIs) is the carrier-based mixture (de Boer et al., 2017;Elsayed and Shalash, 2018) which is a mixture type that contains agglomerates formed spontaneously during mixing of larger particles (carriers) with micronsized drug particles, typically referred to as fines (Jones et al., 2010;Kaialy, 2016;Thalberg et al., 2004).Such agglomerates consist of an adhesive layer of fines attached to the carrier and upon inhalation, the drug particles are to be detached from the carrier surface and enter the lung, while the carrier particles themselves impact in the throat (Pilcer and Amighi, 2010).The role of the carrier particles is to aid the aerosolization of the fines and to provide flowability to the powder (Thalberg et al., 2004).
Depending on the size and shape of the carrier and the fine particles and the mass ratio between them different physical states of adhesive mixtures are obtained (Rudén et al., 2018;Young et al., 2011).The blend state theory (Rudén et al., 2018) was developed based on the appearance of adhesive mixtures in combination with indications of their macroscopic mechanical properties, i.e. their packing and flow.It represents and attempt to classify adhesive mixtures into classes, denoted blend states, dependent on the spatial distribution of fine particles in the mixture.
A similar approach to describe the evolution of structure of adhesive mixtures with increased drug load has been presented by Young and coworkers (Young et al., 2011).This concept was developed by studying the appearance of adhesive mixtures in relationship to their aerosolization properties.In both cases, it was proposed that the fine drug particles in a carrier-based mixture can be distributed between two subpopulations of fines, i.e. one fraction that is bonded to the carrier and one fraction that is non-bonded, i.e. self-agglomerates of fines.This segregation was proposed to take place at high drug loads and represented in both models the final class, denoted type V by Young and coworkers and type 3 in the work by Rudén and co-workers.At lower drug loads, before self-agglomeration (i.e.segregation) occurred, the fines were predominantly bonded to the carrier and a key aspect of both models is how the structure of the adhesion layer evolve with increased drug load.Young and co-workers proposed four structural classes for the E-mail address: goran.alderborn@farmbio.uu.se (G.Alderborn). 1 Present address.adhesion layer denoted type I to IV.The blend structures denoted type I to III consist all of monolayers of particles distributed over the surface of the carrier and the different types correspond to different carrier-to-fines adhesion force.The high energy binding sites were referred to as active sites.In type IV, a multilayer of particles is formed and this multilayer adhesion stage was associated with the highest FPF.At the final segregated state, a decline in FPF was obtained due to the collapse of the adhesive mixture.
Rudén and co-workers proposed three structural classes before the segregation state, denoted blend state 1, 2a and 2b.In the first blend state, S1, the fine drug particles adhere to surface irregularities or cavities of the carrier particles.In the second state, S2, the fine particles adhere to the enveloped (outer) surface of the carrier and with increased fines concentration, adhesion layers of increasing volume will be formed.A graphical illustration of the different blend states is published elsewhere (Rudén et al., 2019).
An important aspect of the Young model is the distinction between a monolayer and multilayers of particles that are formed in sequence.In the blend state model, the completion of a monolayer of single particles before formation of a more voluminous adhesion layer is not assumed and the expression of blend state 1 is related to the geometry of surface features of the carrier.Dependent on the breadth and depth of surface cavities, fines can pack in multilayers within such surface cavities before substantial amounts of fines are adhered to the enveloped surface.Furthermore, when adsorption of particles occurs at the enveloped surface, both single particles and micro-agglomerates can be categorized as blend state 2a.With increased fines concentration, the thickness of the adhesion layer will increase and the distinction between state 2a and 2b is proposed to be related to the dynamics of the adhesion layer rather than a transition from mono-to multi-layers, i.e. the transition from 2a to 2b is associated with a more morphological heterogeneous and more mobile layer which is more susceptible to rearrangement while subjected to an external force.Rudén and co-workers later (Rudén et al., 2021) also discussed the link between blend state and the dispersion of the fines.
Thalberg and co-workers (Thalberg et al., 2012) took a different approach, focusing on formulations with 2-15% drug load and thus avoiding the very low as well as the very high drug load regimes.A FPF model based on the cohesivity of the different drugs used and a total fines effect (i.e. the sum of drug and fine lactose particles in the formulation) could be fitted to the data.Thus, no active sites nor transition steps were needed to explain the dispersibility behavior within the range studied.An explanation to the total fines effect was however not given in this paper.
The formation of aerosolized fine particles from an adhesive mixture may involve two principally different processes, i.e. the detachment from the carrier and the deagglomeration of the fines (Dickhoff et al., 2005;Grasmeijer et al., 2013).A dispersed particle is typically defined as having an aerodynamic diameter equal to or below a critical diameter, often 5 μm, and the degree of dispersion is hence typically assessed by the fine particle fraction (FPF), i.e. the fraction of the emitted dose satisfying the definition (Hickey and da Rocha, 2019).The term dispersibility of an adhesive blend is the propensity of the adhesive mixture to disperse in air (Thalberg et al., 2012).Since the dispersion of adhesive mixtures during in vitro impactor experiment depends on several factors, including the properties of the blend and operating factors such as air flow velocity and type of inhaler, the in vitro dispersion of a powder blend must in practice be determined under standardized conditions (Hickey and da Rocha, 2019).The use of a defined pressure drop, i.e. the pressure difference created by the vacuum source in reference to the atmospheric pressure, is critical.It is pointed out (Grasmeijer and de Boer, 2014) that it is preferable to investigate the dispersion at a series of flow rates in order to get good insight into the dispersion performance of an adhesive mixture.In addition, an understanding of how different pressure drops affect the aerosolization performance of a deviceformulation combination is obviously clinically important since; firstly, patients produce different pressure drops when inhaling through an inhaler dependent on their lung capacity (Clark et al., 2020) and; secondly, different inhalers have different inherent resistance dependent on the device design.It is for example reported (Buttini et al., 2016) that the impact of flow rate on the delivered drug dose varies between inhalers.Hence, recommendations to demonstrate the importance of flow rate for the delivered drug dose from a DPI can be found in official guidelines for inhalation products, for example (EMA, 2006).
In a previous study (Rudén et al., 2021), the dispersion of three active pharmaceutical ingredients (APIs) in adhesive mixtures of different blend states was studied with a Fast Screening Impactor (FSI).A ScreenHaler device was used and the dispersion was studied at a single, fixed air flow rate.From the results, it was concluded that all three APIs formed the aforementioned blend states and that the dispersiblity could be linked to the different blend states.In this follow up study, dispersion profiles, i.e.FPF and MMAD as a function of pressure drop, were derived for adhesive budesonide mixtures using three different pressure drops: 0.5, 2 and 4 kPa.Four binary adhesive blends were prepared, one for each blend state, and one additional ternary mixture, in which lactose fines and carrier were premixed before admixing of budesonide.The FPF and the MMAD were determined using a Next Generation Impactor (NGI) with a ScreenHaler device connected to a Turbuhaler mouthpiece.The aim of the study was to investigate the effect of pressure drop on the in vitro dispersion of a series of adhesive mixtures of different fines-to-carrier proportions corresponding to the four different blend states of the blend state model, i.e.S1 to S3.The study is intended to provide improved mechanistic understanding of the dispersion process of adhesive mixtures of different blend states.

Materials
The powders used in this study were a spray-dried α-lactose monohydrate carrier (Lactopress SD, DFE Pharma, The Netherlands) and two fine powders, i.e. micronized Budesonide (AstraZeneca Gothenburg, Sweden) and micronized α-lactose monohydrate (produced by Astra-Zeneca, Södertälje, Sweden).Some particle and powder properties of the three materials have been reported earlier (Rudén et al., 2021) but for convenience to the reader, relevant particle and powder characteristics of the carrier and the fines are also presented here (Table 1).
For the UPLC analysis (see section 2.4 and 2.5), LC grade methanol and acetonitrile were used (Merck KGaA, Darmstadt, Germany).MilliQ water was produced using a Purelab Flex (ELGA LabWater, United Kingdom) operated at 18.2 Ohm.Fluocinolone acetonide, used as an internal standard, as well as the orthophosphoric acid, sodium dihydrogen phosphate and trifluoroacetic acid (TFA) were purchased from Merck/Sigma Aldrich (Merck KGaA, Darmstadt, Germany).The NGI coating agents (Brij 35 and glycerol) were also acquired from Merck/ Sigma Aldrich.

Preparation of adhesive mixtures
Four binary adhesive mixtures consisting of the carrier and budesonide fines were prepared, one for each type of blend state (S1, S2a, S2b and S3).The binary mixture compositions used have previously been shown to give the blend states aimed at in this study (Rudén et al., 2021).Prior to mixing, a 200 mL glass vessel was filled with carrier and API powders in two steps, i.e. firstly 40-60 g of carrier powder was filled into the vessel and then the required amount of API powder was added on top of the carrier powder, giving approximately 50% of the vessel volume filled with powder.Thereafter, the powders were mixed using a Turbula T2F mixer (Willy A. Bachofen AG, Switzerland) for one hour at 46 rpm.The aim of the mixing process was to achieve a high blend homogeneity of the adhesive mixtures.It is previously reported (Nyström and Malmqvist, 1980) that a blending time of 100 mins was required to provide complete deagglomeration of the fines of an adhesive mixture using the same mixer type as in this study and it was concluded that the deagglomeration of the fines was a slow process.Based on this report, rather than on pre-trials, a mixing time of one hour was used in this study.
In addition, one ternary adhesive mixture consisting of a carrier and two fines, i.e. fines of lactose and budesonide, was prepared using the same mixing conditions as for the binary mixtures.The ternary mixture was prepared by a two-step mixing process.In step one, carrier and lactose fines were mixed with a proportion of lactose fines of 1.83%, a proportion that has been reported (Rudén et al., 2018;Rudén et al., 2019) to be equivalent to the amount required to fill the surface cavities of the carrier for this combination of carrier and lactose fines, i.e. to complete the S1 state.Thereafter, 2.41% budesonide fines was added to the binary mixture and admixed for an additional one hour, i.e. the total the mixing time for the ternary mixture was two hours giving a total percentage of fines of 4.24%.By this procedure, budesonide particles were assumed to be adhered to the enveloped surface of the carrier and to the top layer of lactose fines filling the surface cavities.The blend state of the ternary mixture is henceforth denoted S2a only with reference to the localisation of the budesonide fines.The total percentage of fines in the ternary mixture was slightly lower than the percentage of fines in the S2a binary mixture (4.24% vs 4.83%).
The composition of all mixtures and the corresponding surface coverage ratios, calculated as described earlier (Rudén et al., 2018) are reported in Table 2.The SCR is a theoretical number indicating the degree to which the carrier surface is covered with fines that are adsorbed as single particles and a value of 100% corresponds hence to a carrier surface completely covered with a monolayer of single particles.
All powders and mixtures were pre-conditioned before further use by storage in a climate-controlled room for at least 24 h at a temperature of 21-24 degreesC and a relative humidity (RH) of 30-34%.

Assessment of unsettled bulk density
The unsettled bulk density was determined as previously described (Rudén et al., 2018) using a steel sample container with a volume of 20.05 mL manufactured by AstraZeneca Gothenburg (Mölndal, Sweden).In short, the measurement involved the use of an inner cylinder placed inside the sample container, which was slightly higher than the container.The inner cylinder was filled with powder and thereafter lifted so that the powder flowed into the sample container.After gently levelling the sample cylinder, the powder was weighed and the bulk density subsequently calculated.All bulk density experiments were performed in the same climate conditions as used during conditioning of the mixtures.

Assessment of mixture homogeneity
To assess the mixture homogeneity, samples were drawn from each mixture and the content of API in the samples was subsequently determined by liquid chromatography.A Waters Acquity UPLC system (Waters Corp, Milford, USA) equipped with a C18 BEH 1.7 µm 2.1x50 mm column and a photo diode array (PDA) detector was used to analyse the API content.The mobile phases used were water and acetonitrile, both with 0.03% TFA.The internal standard used was fluocinolone acetonide in approximate concentrations of 20 mg/L.The UV wavelength was set to 254 ± 10 nm.The limit of quantification (LOQ) was determined to be 5 µg budesonide per 20 mL of internal standard.The amount of API at each stage was determined from a calibration curve in the concentration range 0.5-87 μg/mL and a correlation coefficient R 2 of at least 0.999.
For each of the selected mixtures, 10 samples of 15-20 mg were drawn by scoop sampling from randomly selected places of a bed of the mixture held in the same container as used during powder mixing.Similar positions within the mixture bed from which the samples were drawn were used for all mixtures.The samples were weighed and then transferred into a vial in which the samples were dissolved in an internal standard solution (see section 2.5).The concentration of API of each sample was calculated using calibration curves of the response factors in the concentration range.The concentration of API was then normalized to a constant sample weight of 15.0 mg and the variation in API concentration between the samples, expressed as the relative standard deviation, was used as an indication of blend homogeneity.

Assessment of mixture dispersion
In the dispersion experiments, a Next Generation Impactor (NGI) (Copley Scientific, UK) was used together with a ScreenHaler device coupled with a Turbuhaler mouthpiece (Thalberg et al., 2016).The ScreenHaler is a passive inhaler intended for formulation screening and consists of a cylindrical tube, 60 mm long and 7 mm in inner diameter, which has a 90 degree bend.Each dose is loaded from the top with a spatula and weighed, where after the air flow is connected and the dose is withdrawn.In the current study, a Turbuhaler mouthpiece was added with the aim to increase the deaggregation power of the device.For this set up, a flow rate of 50 L/min corresponds to 2 kPa, which means an intrinsic resistance value (R-value) of 54.
The volumetric rate of airflow during measurement was set to match a pressure drop of 0.5, 2 and 4 kPa and the suction time was adjusted to correspond to a total suction volume of 4 L using a Triggbox model III (AB FIA, Lund, Sweden).In addition, the rise-time through the impactor was assessed using a STIG rise-time measuring device (AB FIA, Lund, Sweden).The rise-time was found to be in the range of 220-240 ms in all experiments.Before each dispersibility experiment, the NGI cups were coated with a solution containing ethanol (51%), Brij 35 (15%) and glycerol (34%) (Hertel et al., 2020) to reduce the risk of particle bounce.The dispersion experiments were performed at ambient conditions, i.e. room temperature and ~ 30-45% RH.
Prior to testing, the inhaler device was manually filled with a dose varying between 15.0 and 17.0 mg for each single dispersion experiment.Several actuations, or dose withdrawals, were used in order to reach quantifiable amounts in the chemical analysis (a target emitted dose of around 3.2 mg was used).The number of required actuations was dependent on the API concentration in the mixture and varied between 3 and 12.The emitted dose (ED) was defined as the sum of masses of API collected at all stages of the impactor (1-MOC), at the throat and at the pre-separator.Samples were collected from each stage by first adding 20 mL of an internal standard solution.Following this, the throat and pre-separator were set to shake using a Sample Preparation Unit (Copley Scientific, UK) for 20 min, while the NGI cups were placed on a shaking table for the same duration.After the sample preparation was completed, 0.5 mL from each stage was transferred into separate LCvials.Before the analysis, 0.8 mL of phosphate buffer solution (pH 3.2) was added to the LC vials to improve the resolution of the peaks in the chromatogram.The same LC procedure was used as for the homogeneity analysis described above.
The amount of API at each stage was determined from the response factors using calibration curves in the concentration range.The mass median aerodynamic diameter (MMAD) was calculated for every dispersion experiment using Microsoft Excel 2016 (Microsoft, Redmond, USA) as instructed by the European Pharmacopoeia v9.0 (Council-of-Europe, 2017).In short, the logarithmic values of the cut-off diameters at the specific flow rate was plotted against the inversed normally distributed values of the cumulative mass fractions (Norm.invfunction in excel).The resulting straight line was used to calculate the MMAD.Using the same calculations, the fine particle dose (FPD, µg), which is the total mass of particles equal to or below a defined cut-off diameter, was determined.The degree of dispersion was assessed by the fine particle fraction (FPF, %) calculated as follows: Three different aerodynamic particle cut-offs of 2, 3 and 5 μm was used in the calculation of FPD and FPF.

Particle imaging
SEM images of particles, which were collected from the pre-separator after a dispersion experiment, were prepared using a Zeiss 1530 (Carl Zeiss GmbH, Oberkochen, Germany) microscope operated at an acceleration voltage of 2.5 kV with images captured at 5-10kx magnification using an InLens detector.Prior to SEM imaging, the particles were coated with gold/platinum.In order to collect particles from the preseparator of the NGI, a SEM stub was placed in the cup of the preseparator at which some dispersed particles could impact and adhere during a dispersion experiment.After the dispersion experiment was completed, the stub with the adhered particles was gently removed from the pre-separator and subjected to SEM imaging.Using this procedure, particles from all mixtures were captured and inspected after dispersion experiments at the highest pressure drop, i.e. 4 kPa.In addition, particles from the S2b and S3 blends were also captured and inspected using a pressure drop of 0.5 kPa during the dispersion experiment.

Results
The bulk densities of the binary mixtures have been published earlier (Rudén et al., 2021) but are reported also here (Table 3) to facilitate comparison with the bulk density of the ternary S2a only mixture.The bulk density of the ternary mixture was similar to or slightly lower than the S1 mixture but slightly higher than the binary mixture of the S2a state.Thus, the S2a only mixture had a somewhat closer packing of the adhesive units compared to the S2a mixture, indicating some difference in the structure of the adhesion layer.
The variation in drug content (DC) between samples drawn for the homogeneity analysis generally had a relative standard deviation (RSD) below 3% (Table 3).The measured drug contents were generally lower than the theoretical doses (TD, the amount of calculated drug required to reach the target SCR), i.e. they were 10-17% lower than the TD with the largest difference obtained for the S3 mixture.The main cause for this loss of drug was likely adhesion to the walls of the mixing vessel.
For all three definitions of fine particle fraction, the relationships between pressure drop (ΔP) and fine particle fraction (FPF) were similar (Fig. 1A) and hereafter results are presented only for a cut-off of 5 μm (Fig. 1B etc.).The FPF -ΔP relationships were dependent on the blend state (Fig. 1B).For the S1 mixture, all three FPF -ΔP profiles were nearly linear and the values of FPF were generally lower than for the other blend states.For S2 and S3 mixtures, the FPF increased markedly from 0.5 to 2 kPa but thereafter the rate of increase became lower and the FPF pressure drop profiles tended to level out between 2 and 4 kPa.In the pressure drop range 0.5 to 2 kPa, the rate of increase in FPF with pressure drop increased with blend state and was most marked for the S3 mixture.Thus, independent of the definition of FPF, the degree of dispersion increased with blend state at the two highest pressure drops used.At the lowest pressure drop, blend state S3 showed a lower FPF than the S2 mixtures.
In Fig. 2, FPF according to the 5 μm level is plotted as a function of the surface coverage ratio (SCR).A change from S1 to S2a increased the FPF for all pressure drops.At the lowest pressure drop, FPF was thereafter more or less constant, while at the higher pressure drops, FPF tended to increase slightly with SCR.A pressure drop of 0.5 kPa gave generally the lowest values of FPF while the FPF profiles obtained at 2 and 4 kPa were similar.
For the mass median aerodynamic diameter (MMAD), a reverse relationship was obtained (Fig. 3).At the lowest pressure drop the highest MMAD values were obtained and when the pressure drop was increased to 2 kPa the median diameter decreased and thereafter levelled out.The S2 mixtures displayed nearly parallel relationships to the S1 and S2a only mixtures but with generally lower MMAD.The S3 mixture showed a different type of relationship with a more marked initial drop in MMAD than the other mixtures before levelling out.
The distribution of drug represented by the fraction of emitted dose (ED) at three different stages of the impactor (throat, pre-separator and NGI stages) is shown in Fig. 4. For each type of mixture and each pressure drop the following stage distribution pattern was obtained: the largest fraction of ED was deposited in the pre-separator, varying between about 0.5 and 0.8, followed by deposition in the throat, about 0.15 to 0.4, followed by the lowest deposition in the 1-MOC stages (denoted the NGI stages), about 0.05 to 0.25.
The effect of pressure drop on the relationship between blend state and stage distribution was generally small.The 2 kPa and 4 kPa pressure 4 TD = Theoretical dose in 15 mg of mixture drops gave similar stage distributions for each blend state.A pressure drop of 0.5 kPa gave a somewhat deviating pattern with lower fractions deposited in the NGI stages and thus higher in the throat and preseparator for all blend states.However, the blend state (the drug load) affected the distribution of ED between the stages.Regarding the deposition in the throat and preseparator, the fraction of ED deposited in the throat increased with increased drug load while for the deposition in the pre-separator the opposite effect was obtained, i.e. a decrease in deposited drug with increased drug load.Thus, a shift from predominant deposition in the pre-separator towards more equal deposition between throat and preseparator with increased drug load (higher blend state) was obtained.This deposition pattern between the throat and pre-separator applied to all three pressure drops.For the fraction of ED deposited at the NGI stages, a similar pattern was obtained as in the throat for the 2 and 4 kPa pressure drops, i.e. the fraction of deposited drug increased with increased drug load.At the lowest pressure drop of 0.5 kPa, a different trend was obtained with a low fraction of deposited drug for the S1 mixture and higher and similar fractions deposited for the S2a and S2b blends and a tendency to a reduced fraction for the S3 blend.
For the ternary mixture, S2a only , the deposition pattern, i.e. the distribution between throat, pre-separator and NGI stages, were similar to the S2a mixture.There seems to be a tendency to a lower deposition at the NGI stages, corresponding to higher deposition at the other stages, for the S2a only mixture compared to the S2a mixture.One may note that the concentration of drug fines (drug load) was higher for the S2a mixture compared to the S2a only mixture.
By the use of a stub placed in the pre-separator, particles impacted in the pre-separator were collected and the appearance of these inspected by SEM (Fig. 5).For all blend states, the captured particles were adhesive units, i.e. carriers with adhered fines, which is consistent with the finding that a large fraction of ED was deposited in the pre-separator (Fig. 4).The number and density of fine drug particles attached to the carrier particles increased with increased SCR.The large amounts of drug attached to the carrier indicate that most of the fine particles did not detach during aerosolization.The effect of pressure drop on the appearance of particles captured from the S2b blend was minimal.
However, for the S3 blend, a larger number of small agglomerates appeared to be present in the adhesion layer of the carrier particles captured at 0.5 kPa than at 4 kPa.In addition, a crushed large selfagglomerate (Fig. 5H) was captured in the pre-separator indicating that these large agglomerates did not disintegrate completely into primary particles during aerosolization.The total dose recovered from the impactor stages was generally lower than the measured drug content of the different mixtures, i.e. the ED was about 55-80% of the drug content (Fig. 6).The emitted fraction tended to increase with an increased pressure drop and was affected by the drug load (blend state), i.e. the emitted fraction vs pressure drop relationships were shifted in parallel with increased blend state.

Effect of surface coverage ratio on fine particle fraction
In a previous study (Rudén et al., 2021), the effect of surface coverage ratio (SCR) on the dispersibility of the same type of binary    adhesive mixtures of budesonide as used here was investigated.The SCR concept is an indication of the theoretical distribution of fines at the carrier surface.In this study, the blend state was determined after emptying of the mixing vessel while the SCR calculation was based on added rather than residual amounts of drug after mixing.The assessment of blend state was hence not affected by any wall loss of drug during mixing (see above) while wall loss may have slightly affected the calculated SCR values, i.e. the SCR allocated to a blend state transition may be shifted along the blend state axis depending on the drug loss.Since firstly, the SCR concept is based on an ideal distribution of discrete fine particles on the carrier surface and secondly, accurate values of the effective surface areas of carrier and fines to be used in the calculation are typically not known, the SCR values should be considered as approximate values.Hence, any discrepancy between calculated and physically accurate SCR values is difficult to define.However, the SCR concept may anyway provide improved understanding of mixture properties in the comparison of adhesive mixtures of differently sized carriers.
The type of bended FPF-SCR relationship obtained in this study (Fig. 2) was similar to the one reported earlier, especially for the profile obtained at a pressure drop of 0.5 kPa.In S1, the fine particles are located in surface cavities of the carrier and thus shielded from the airflow and hard to detach, which resulted in a very low FPF.With increased fines concentration, the FPF increased markedly, i.e. the detachment of fines became easier due to a gradual filling of the surface cavities.At a SCR of about 1, an abrupt transition in the FPF-SCR relationship occurred, i.e. at the point at which an adhesion layer started to be formed on the enveloped carrier surface.Thereafter, the FPF continued to increase at a considerably lower rate or became more or less constant.
With an increased pressure drop, the absolute level of FPF was affected by the SCR.Thus, the flow rate sensitivity of these adhesive mixtures during aerosolization was dependent on the SCR (Grasmeijer et al., 2013;Rudén et al., 2021;Young et al., 2011).It should be noted that in this study, a different impactor set-up is used compared to the earlier study, i.e. a NGI instead of a FSI and a ScreenHaler device with rather than without a Turbuhaler® mouthpiece.The Turbuhaler® mouthpiece provides a stronger capability of the inhaler to disperse the adhesive mixture.Nevertheless, the character of the FPF-SCR relationship was similar between the two studies indicating that the blend state theory may represent a rational approach to study the manufacturing and aerosolization performances of adhesive blends.

Effect of pressure drop on fine particle fraction
As a means to further illustrate the interplay between FPF, blend state and pressure drop, the FPF was plotted as a function of the pressure drop for all mixtures (Fig. 1B).In S1, FPF increased linearly with pressure drop while for the different S2 mixtures, non-linear relationships were obtained which approached plateaus in the profiles.Thus, at pressure drops within a range of 0.5 kPa to 2 kPa, the blend state and the pressure drop in combination controlled the FPF.However, above a critical pressure drop (in this study somewhere between 0.5 kPa and 2 kPa) the dispersion of fines was predominantly dependent on blend state and only limitedly affected by the variation in the pressure drop.For the S3 mixture, a similar effect of pressure drop on the FPF was obtained as for S2, however with a higher rate of increase in the lower range of pressure drop.Blend state 3 represents a more complex situation since free self-agglomerates of fines exist together with the adhesive units and thus two sub-populations of fines exist in the blend.In this state, the dispersion of fines may occur in parallel from both sub-populations and the effect of the pressure drop may differ between them.For the S3 blend, the strong effect of pressure drop on the FPF from 0.5 kPa to 2 kPa indicates that the disintegration and dispersion of self-agglomerates increased strongly with increasing pressure drop.
The ternary blend (S2a only ) showed an absolute level of FPF, which was higher than for the S1 blend but lower than for the S2a blend.The increase in FPF for S2a only compared to S1 shows that the detachment of particles attached to the enveloped surfaces of the carrier (outer adhesion sites) was easier than for particles localized in shielded positions in surface cavities (inner adhesion sites) (Grasmeijer et al., 2014;Hertel et al., 2018).Thus, the propensity of fines to detach from the carrier depends on their adhesion sites.The shape of the FPF-pressure drop profile for S2a only was similar to the S2a profile but the S2a only mixture had generally lower FPF than the S2a mixture.The S2a only composition was mixed for twice as long as the S2a mixture and it is previously shown that increased mixing time affects and may reduce the dispersion propensity of adhesive mixtures, e.g.due to increased presson forces during mixing (Dickhoff et al., 2005;Grasmeijer et al., 2014;Hertel et al., 2017;Jones et al., 2010).However, since the lactose fines were pre-mixed with the carrier before addition of budesonide, the mixing time for budesonide was the same for the S2a and the S2a only mixtures.The reason for the reduced FPF for the S2a only mixture may be that budesonide particles are pressed onto the layer of lactose fines located in the surface cavities and form small co-agglomerates of budesonide and lactose fines with strong fine-to-fine adhesion forces.These co-agglomerates may be hard to break and hence do not disintegrate completely during the dispersion of the mixture.
The pressure drop affected also the MMAD (Fig. 3).For all blend states, the highest MMAD was obtained at 0.5 kPa, decreased when the pressure drop was increased from 0.5 kPa to 2 kPa and became thereafter independent of pressure drop.Moreover, the absolute level of MMAD was dependent on the blend state.This evolution in MMAD with pressure drop is consistent with the discussion above that at the lower range of pressure drop, MMAD was controlled by both blend state and pressure drop but at upper range of pressure drop, MMAD was dependent mainly on the blend state.The change in MMAD with pressure drop was especially pronounced for the S3 blend.For all mixtures except S3, the MMAD-pressure drops profiles gave two groups shifted in parallel to  There may be two explanations for the effect of blend state on the MMAD.Firstly, it is earlier shown that the fraction of drug detached from the carrier may be higher than the measured FPF (de Boer et al., 2003) and the authors proposed that the drug was detached from the carrier partly as agglomerates that were unable to reach the lower stages of the impactor and partly as fine particles.It is also possible that the particles that are small enough to reach the lower impactor stages and hence are classified as fine particles can be single particles or microagglomerates consisting of very few drug particles (Dickhoff et al., 2005).A higher MMAD for fines detached from S1 and S2a only blends than from the S2a and S2b blends can be explained by the hypothesis that the proportion of micro-agglomerates in the fine fraction was higher for the former two blend states than for the latter two.A possible reason for this difference in proportion of micro-agglomerates is that particles will be packed denser in the surface cavities and form stronger bonds to each other and thus detach as micro-agglomerates.For S2a only , as discussed above, these micro-agglomerates may consist of both drug and lactose fines, i.e. micro-co-agglomerates (Jones et al., 2008;Kinnunen et al., 2015;Kinnunen et al., 2014).Secondly, it has been stated (de Boer et al., 2003) that adhesive forces are proportional to the particle surface area, i.e. the square of the particle diameter, while dispersive forces due to inertia are proportional to the particle volume, i.e. the cube of the particle diameter.This means that larger particles will be more easily detached from the carrier by an inertial mechanism than finer particles.Since the FPF in the S1 state is less than 3% at 0.5 kPa and less than 10% at 4 kPa, it is possible that only the very largest budesonide particles of the distribution were detached, which explains the high MMAD values in this state.For the S2 state, where the detachment of fines may occur from the outer part of the adhesion layer, other detachment mechanisms than inertia may be involved and the dispersed particles will not preferentially be the largest particles of the distribution.Dominating forces causing detachment may for the S2 mixtures be drag forces caused by airflow past the adhesive units and shearing forces caused by sliding between carriers or between carriers and external obstacles.

Effect of pressure drop on impactor stage distribution
The total emitted fraction of drug content (Fig. 6) tended to increase with the pressure drop, i.e. increased pressure drop facilitated emission of powder from the inhaler for all blend states.However, an increased inhaler retention was obtained with increasing blend state.
The distribution of the emitted dose (ED) at the different stages of the impactor is shown in Fig. 4. The drug deposition occurred mainly in the pre-separator followed by the throat and finally the NGI stages.However, the relative distribution between the stages was dependent on the blend state, i.e. the fraction of deposited drug in the throat and the NGI stages increased with increasing drug load while the opposite trend applied to the deposition on the pre-separator.The consequence is that an increased fines concentration, i.e. higher blend state, gave a more uniform deposition pattern.High deposition of drug in the throat at high drug loads is also reported earlier by Yeung et al (2019).The highest degree of throat deposition was obtained for the S3 blend and can at least partly be explained by the presence of drug self-agglomerates in these mixtures.If the self-agglomerates do not disintegrate while emitted from the inhaler, it is plausible that they have a high tendency to impact in the throat.
The pressure drop had generally a limited effect on the fractions of ED deposited in the throat and the pre-separator.However, for the S2b and S3 mixtures an increased pressure drop from 0.5 to 2 kPa gave a slight reduction in the fraction of drug deposited in the throat and a simultaneous increase in the fraction of drug deposited in the NGI stages.This is consistent with the interpretation that an increased pressure drop increases the detachment of fines and the disintegration of selfagglomerates.

Mechanism of dispersion
It has previously been shown that the type of inhaler device used can greatly affect the dispersion, indicating that the dispersion of fine particles mainly occurs in the device (Chew and Chan, 1999;Hertel et al., 2020).Since the fraction of ED impacted in the throat and the preseparator was high (Fig. 4) it seems that the fines after emission from the inhaler predominantly were located in the adhesive units or existed as free large agglomerates.In order to better understand the dispersion process occurring during emission from the inhaler, images of particles captured in the pre-separator were prepared (Fig. 5).The images indicate that substantial amounts of fines were attached to the carriers independent of the blend state.The S1, S2 and S2a only mixtures (Fig. 5A-C) had particles located mainly in carrier cavities and surface irregularities and a limited number of fine particles were adhered to the enveloped carrier surfaces.For S2b and S3 mixtures (Fig. 5D-E), the carriers were more or less covered with an adhesion layer.Thus, the blend state clearly affected the post-impaction appearance of the particles captured in the pre-separator.Moreover, the general appearance of the captured adhesive units were similar to the appearance of the adhesive units directly after mixing (Rudén et al., 2021).For the S2b and S3 blends, particles were captured at two pressure drops, i.e. 0.5 kPa and 4 (Fig. 5D-E and F-G).The images for the S3 mixtures indicate that the structure of the adhesion layer was similar for both pressure drops but the layer thickness was reduced with increased pressure drop.Also for the S2b mixture, the structure of the adhesion layer was the same for both pressure drops but a difference in layer thickness is difficult to observe.Nevertheless, it seems that the adhesion layer was relatively uniformly eroded rather than locally peeled-off from the carrier during emission from the inhaler.In Fig. 5H, the appearance of a selfagglomerate of drug sampled from the cup of the pre-separator from the S3 mixture is presented.The appearance indicates that the selfagglomerate preserved coherency and did not disintegrate at the used pressure drop of 0.5 kPa.
In summary, the particles captured in the pre-separator were to a large extent adhesive units with large numbers of fines still adhering to the carrier.The appearance of these and the structure of their adhesion layer corresponded well to their appearance and structure before the impactor experiment and was hence dependent on the blend state.The physical structure of the adhesive units seems thus to be preserved during emission from the inhaler and it is proposed that the mechanism of dispersion of fines seems to be a removal of single or small clusters of fines (micro-agglomerates) from the surface of the adhesive units.Thus, the detachment mechanism can be described as erosion of the adhesive units rather than peeling-off of relatively large fine particle agglomerates.The single fine particles or the micro-agglomerates of fines hence liberated have the potential to reach the lower stages of the impactor.In S1 and S2a, the fine particles detach from a state of adhesion to the carrier surface while in S2b and S3, the fine particles detach from a state of cohesion at the periphery of the adhesion layer.
There are different potential explanations for the types of evolution of the FPF-pressure drop profiles obtained in this study (Fig. 1B), involving the detachment mechanism or the kinetics of the detachment process.Regarding the detachment mechanism it is earlier proposed (Rudén et al., 2021) that the detachment of fines when situated in S1 occurs predominately as single fine particles while in S2 and S3, the detachment occurs as both single fines and as relatively large agglomerates that will not reach the lower stages of the impactor.In S1, the gradual filling of surface cavities with increased drug load will make the fine particles gradually less shielded which will increase their detachment propensity, i.e. the increase in FPF with pressure drop indicates that the number of fine particles dispersed in air is a direct function of air pressure drop.In S2, agglomerates are located on the carrier surface that may be detached but resist further breakage and this agglomerate detachment may hence occur in parallel to the detachment of single fine particles and micro-agglomerates.The proportion of such large agglomerates may increase with pressure drop in such a way that the amount of fines will be nearly independent of pressure drop.This potential detachment mechanism is consistent with an earlier suggestion (de Boer et al., 2003) that if a fraction of drug is detached from the carrier as large agglomerates, the air-flow rate can affect the discrepancy between the total fraction of drug released from the carrier and the fraction of released fines.Regarding the kinetics of the detachment process, both the rate of detachment of fines from the adhesive mixture (r d ) and the time period during which the detachment and dispersion of fine particles occurs (t d ) can potentially depend on the pressure drop.An increased pressure drop will increase the maximum air-flow velocity during the course of dispersion which can facilitate the breakage of interparticulate bonds within the adhesion layer and liberate single fine particles and micro-agglomerates.At the same time, an increased air flow velocity may increase the rate of transport of the mixture during the course of dispersion and consequently decrease the time of the dispersion process, possibly due to a reduced residence time in the inhaler device.The total degree of dispersion, d d , may consequently be the neteffect of these two processes, i.e. d d = f(r d * t d ).The gradual increase in FPF with pressure drop for the S1 mixture indicates that for particles located in surface cavities, the rate of detachment is the dominating factor for the degree of dispersion, i.e. the magnitude of the forces causing detachment is critical for the achieved FPF.For the higher blend states (S2a and above), the initial increase in FPF with pressure drop indicates that also here the detachment propensity will increase with pressure drop and the rate of detachment dominates the degree of dispersion in a similar way as for S1.The bended character of the S2 profiles after a critical pressure drop suggests that even if the rate of detachment increases further due to stronger detachment forces, it is nearly balanced by a decreased time of the dispersion process.Consequently, the FPF-pressure drop relationship will level out and eventually be almost independent of the pressure drop.
It is proposed that for the combination of adhesive mixtures and inhaler device used in this study, surface erosion is the dominating dispersion mechanism.However, the dispersion process may hypothetically involve other mechanisms dependent on several factors, including the structure of the adhesion layer, the presence of self-agglomerates in the mixture and the type of forces that the powder is subjected to in the dry powder inhaler during emission.It is shown that the specific combination of carrier and fines affects the structure of the adhesion layer (Rudén et al., 2021) which subsequently may affect the dispersion process.The dispersion process from a blend state 3 mixture represents a more complex dispersion situation than from the lower blend states since self-agglomerates of fines co-exist with adhesive units.The dispersion of fines may thus occur in parallel from both these subpopulations of fines with different mechanisms.For an inhaler where the adhesive units are submitted to strong collisions against inhaler walls, e.g. a cyclone-type inhaler (Coates et al., 2007;Hertel et al., 2020), the fines may detach from the carrier by another mechanism than erosion.The relevance of the surface erosion mechanism, proposed to be the dominating dispersion mechanism for the adhesive mixtures used in the study, for the dispersion process for other adhesive mixtures and inhalers remains to be investigated.

Conclusion
In this paper, the effect of pressure drop on the in vitro dispersion performance of some adhesive mixtures has been studied and the question whether there is a relation to the blend states of the adhesive mixtures is addressed.
For the blend state S1, where the fines are located in shielded positions in cavities (open pores) of the carrier (positions also known as active sites), the fine particle fraction (FPF) increased monotonously in a nearly linear way with the pressure drop (ΔP).For the blend state S2, another type of FPF-pressure drop relationship was obtained with a bended profile which approached a plateau within the range of pressure drop used.Both sub-classes of S2 (S2a and S2b), at which adhesion layers on the enveloped carrier surface exist but of different structure and responsiveness, had a similar dispersion performance but with a slightly higher degree of dispersion for S2b.
The limited increase in FPF with ΔP at the higher pressure drops indicates that a critical pressure drop exists, for the current mixtures about 2 kPa, above which the degree of dispersion is only marginally affected by the pressure drop, as illustrated in Fig. 7.Such an effect is explained to be caused by an effect of pressure drop on the mechanism or kinetics of the detachment process.
Also for the blend state S3, during which self-agglomerates of fines co-exist with adhesive units in the mixture, a bended FPF-ΔP relationship was obtained but with a stronger effect of pressure drop giving a higher FPF at the highest pressure drop used.It is proposed that the dispersed fine particles from the S3 mixture were generated from both sub-populations of fines, i.e. self-agglomerates and adhesive units, and that the dispersibility of the former was lower.Thus, a high pressure drop was required for the self-agglomerates to contribute to the FPF, i.e. the two sub-populations of fines had a somewhat different dependency of pressure drop.
It was observed that the physical structure of the adhesive units was preserved during the impaction experiment but with a reduced volume of the adhesion layer.The dominant dispersion mechanism is thus proposed to be the detachment of single or of small clusters (micro-agglomerates) of fine particles from the periphery of the adhesive units, i.e. a surface erosion process with gradual liberation of fine particles from the adhesive units.
It is demonstrated that the use of a series of pressure drops in in vitro dispersion experiments gives a more nuanced description of the dispersion performance of adhesive mixtures than the use of a single pressure drop.Such experiments can provide an improved understanding of the dispersion mechanism of an adhesive blend as well as information about the sensitivity to variations in pressure drops that can guide the formulation of such blends.The pressure drop sensitivity is also clinically important since it may provide information on the robustness of a formulation towards variation in breathing technique and lung capacity of the patient.It is concluded that the types of FPF-ΔP relationships obtained for different fines concentration in the adhesive mixtures are linked to their blend state and can be explained by the blend state concept.
The sequence of stages that encompass the blend state model is universal but the extension of each stage as well as the transition points between stages will differ between adhesive mixtures dependent on the combination of API and carrier.This shall not be considered a shortcoming of the blend state model but is rather a reflection of system differences.We are currently establishing a decision tree for the determination of blend states that relies on information of the type presented in this and foregoing papers, i.e. a combination of structural information (e.g.imaging) and powder mechanics (e.g.bulk density) information.The blend state model constitutes an attempt to rationalize observed behaviors of inhalation powders in terms of structural differences between them.As demonstrated in the present work, the blend state impacts the performance of inhalation powders in terms of dispersion of fine drug particles and the pressure-sensitivity of this process.
Both Young and co-workers (Young et al., 2011) and Rudén and coworkers (Rudén et al., 2018) used a conceptually similar approach to describe the evolution of structure or state of adhesive mixtures with increased drug load but with different number of stages.The blend state model can be described as simpler, consisting only of three stages, and as more general and physically realistic since it does not assume the formation of a complete mono-particulate layer of fines before multiparticulate structures of fines can be formed.

Fig. 1B .
Fig. 1B.Fine particle fraction (FPF), using a cut-off of 5 μm, of the mixtures at the different pressure drops.Average values with standard deviation (n = 3).Dotted lines are extrapolations to the origin of the axes.

Fig. 3 .
Fig. 3. Mass median aerodynamic diameter (MMAD) of the mixtures at the different pressure drops.Average values with standard deviation (n = 3).

Fig. 4 .
Fig. 4. Fraction of emitted dose (ED) at different stages of the impactor for all mixtures at the different pressure drops.Average values with standard deviations (n = 3).

Fig. 5 .
Fig. 5. SEM images of particles captured from the pre-separator post actuation of one dose from each mixture.For S2b and S3, two pressure drops were used.

Fig. 6 .
Fig. 6.The emitted fraction of the drug content (DC) at the different pressure drops.

Fig. 7 .
Fig. 7. Illustration of the dispersibility of each blend state.Above the S1 state, the dispersibility was largely independent of flow rate once a critical pressure drop has been reached.

Table 1
Particle and powder characteristics of the carrier and the two fines.Average values with standard deviations (n = 3) where applicable.Composition adhesive mixtures.

Table 3
Bulk density, finely dispersed and emitted doses at 4 kPa and drug content.Average values with standard deviation (n = 3).The homogeneity drug content (DC) is an average of 10 measurements with standard deviation and relative standard deviation (RSD).