Continuous manufacture of polymeric cellular dosage forms

Chemical Engineering Journal 320 (2017) 549–560

Contents lists available at ScienceDirect

Chemical Engineering Journal journal homepage: www.elsevier.com/locate/cej

Continuous manufacture of polymeric cellular dosage forms Aron H. Blaesi ⇑, Nannaji Saka Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

h i g h l i g h t s
A continuous, microfluidic melt process for the manufacture of polymeric cellular dosage forms is presented.
The size and volume fraction of gas-filled cells in the dosage forms are precisely controlled.
Disintegration rate and density of the dosage forms can be tailored by altering the volume fraction of cells.

a r t i c l e

i n f o

Article history: Received 28 September 2016 Received in revised form 17 January 2017 Accepted 9 February 2017 Available online 11 February 2017 Keywords: Continuous pharmaceutical manufacturing Cellular dosage forms Microfluidics Pharmaceutical tablets Cellular solids Dosage form microstructure

a b s t r a c t The most prevalent pharmaceutical dosage forms at present are granular solids in the form of oral tablets and capsules. While effective in releasing drug rapidly upon contact with gastrointestinal fluid, their manufacture, which relies on particulate processing, is fraught with difficulties associated with the unpredictable inter-particle interactions. Such difficulties, however, could be easily overcome by transitioning to a liquid-based process. Therefore, we have recently introduced melt-processed polymeric cellular dosage forms. The drug release behavior of the cellular forms was tailored by altering the microstructure; yet their preparation relied on an inefficient batch method comprising gas dissolution, and nucleation and growth of microscopic gas bubbles in the melt. In this study, therefore, we present a continuous microfluidic melt extrusion and molding process. The cellular dosage forms are produced by injecting gas bubbles directly into the melt stream in a micro- or milli-fluidic channel, followed by molding and solidification of the cellular structure. A model is developed to illustrate the effects of the width, frequency, and pressure of the gas injection pulses, and the flow rate and viscosity of the melt, on the microstructural parameters of the dosage forms produced. Experimental results show that the size and volume fraction of gas-filled cells (or voids) are predictable. They also confirm that the dosage form disintegration rate and density can be tailored by altering the volume fraction of voids. It is thus demonstrated that polymeric cellular dosage forms with predictable drug release properties, and density, can be readily manufactured by a continuous process. Ó 2017 Published by Elsevier B.V.

1. Introduction For decades, the most prevalent dosage forms for delivering drugs, the oral tablets and capsules, have been compacted, granular solids [1]. A granular dosage form typically permits percolation of gastrointestinal fluid to the interior after ingestion, and then the bonds between the granules are severed so that the dosage form disintegrates into its constituent drug and excipient particles. If the drug particles are small, they have a large specific surface area, and thus rapid drug dissolution upon contact with gastrointestinal ⇑ Corresponding author at: Department of Mechanical Engineering, Room 35104, Massachusetts Institute of Technology, Cambridge, MA 02139, USA E-mail address: [email protected] (A.H. Blaesi). http://dx.doi.org/10.1016/j.cej.2017.02.057 1385-8947/Ó 2017 Published by Elsevier B.V.

fluid is promoted. This enables that a large fraction of the ingested drug is absorbed by the blood stream, and available for distribution to the disease-specific target sites in the human body [2–6]. Although the granular dosage forms effect rapid drug release, their manufacture, which relies on particulate processing, is fraught with numerous difficulties [7–12]. For example, mixing drug with the carrier, or excipient, particles is hampered by particle segregation and agglomeration, and dispensing and compacting the particulates is complicated by their uneven flow. Moreover, because the theories elucidating particulate behavior are still incomplete it is difficult to predict and control manufacturing processes [13,14]. As a consequence, both resource-intensive and time-consuming batch processing (e.g., particulate mixing, granulating, drying, milling, screening, tableting, and coating) is required

550

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

Nomenclature Ap b Ca c c0 cg cs D D0 Dcell Deq fopen g H0 j k kb kg kl ks ku Lc Lcell Lm Ln Lsolid lcell dMf/dt N Ncell Patm Pb Pcap Pg Pm,F ð0Þ

Pm;F ð1Þ Pm;F

Pm,H p Qg Qm Qm,0

area of syringe’s piston bond length along a polymeric chain capillary number specific heat interfacial concentration of dissolving excipient specific heat of gas phase specific heat of solid phase diffusivity of excipient in the dissolution medium dosage form diameter diameter of a cell equivalent diameter of a sphere fraction of open cells acceleration due to gravity thickness of dosage form flux of eroding polymer thermal conductivity Boltzmann’s constant thermal conductivity of gas lower-bound for thermal conductivity thermal conductivity of solid material upper-bound for thermal conductivity channel length total length covered by the cells along test line(s) length of melt-filled channel segments length of hypodermic needle total length covered by solid along test line(s) average intercept length of a cell mass flow rate of particulates in the syringe number of bonds along a polymeric chain number of cells intersecting a test line atmospheric pressure bubble pressure at hypodermic needle exit capillary pressure gas pressure in the gas reservoir melt pressure at point F melt pressure at point F (melt-filled channel) melt pressure at point F (channel filled with bubbles and melt) melt pressure at point H pressure volumetric gas flow rate volumetric melt flow rate at channel entrance volumetric melt flow rate in melt-filled channel

Qm,1 Qm,s Rc Re Rec Rn Scell T T0 Tc Tw t t0.8 tdis Vb Vb,F Vb,H Vcell Vg Vm Vb

vm,0 vp

x

c

k kcell ud ue um,c uv

q qd qe qf qg qm qs lf lg lm sg sm X

to produce granular dosage forms [15,16]. Additionally, the lead-times for developing and scaling up new formulations are unduly long, limiting flexibility in product development and timely delivery of dosage forms for clinical trials [17,18]. In contrast to the unpredictable processing of granular matter, mixing, dispensing and molding liquids in laminar flows are highly predictable and repeatable. The streamlines follow deterministic pathways and the flow rates are calculable from simple ‘‘constitutive” models [19–21]. Therefore, we have recently introduced cellular dosage forms prepared from polymeric melts [22–25]. A solid disk was first placed in a container in a high-pressure oven, melted for several minutes, and charged with an inert gas at a temperature of 70–130 °C and a pressure of 5–34 MPa. After saturation of the molten disk by the gas, the pressure was reduced to induce nucleation, growth, and coalescence of gas bubbles. Finally, the temperature of the mixture was lowered to solidify the cellular structure.

volumetric melt flow rate at entrance of channel filled with bubbles and melt (during valve closing) volumetric flow rate of melt-filled channel segment radius of the channel Reynolds number critical Reynolds number radius of the hypodermic needle surface area of a cell temperature initial temperature center temperature of the dosage form wall temperature time time to dissolve 80% of the drug content disintegration time of dosage form bubble volume bubble volume at point F bubble volume at point H cell volume volume of gas-filled segment volume of melt-filled segment terminal rising bubble velocity mean velocity of melt in channel velocity of the syringe’s piston axial coordinate surface tension ‘‘free” distance between bubbles mean free distance between cells volume fraction of drug in solid particle bed volume fraction of excipient in solid particle bed volume fraction of melt in the channel volume fraction of voids density density of solid drug density of solid excipient density of dissolution medium density of gas density of melt average density of non-porous solid material viscosity of dissolution medium viscosity of gas viscosity of melt duration of valve opening duration of valve closing angular velocity of basket

Though adequate for preparing experimental dosage forms, such a process is not optimal for continuous manufacture of pharmaceuticals for various reasons. First, the diffusivity of gas in the polymer at the process temperatures and pressures is of the order of 1010–109 m2/s limiting the rates at which the several millimeters thick raw material can be charged with gas. Second, high pressure is required to add the required amount of gas to the material. Third, relatively high temperatures are needed to achieve the optimal rates of nucleation, growth, and coalescence of the gas bubbles. High temperatures degrade many kinds of drugs. The above limitations could be overcome if the gas is delivered directly into the melt stream, eliminating the gas dissolution, bubble nucleation, and bubble growth steps altogether. For example, a gas-releasing blowing agent could be dispersed in the melt, as in the manufacture of foods and polymeric foams [26,27]. But the conversion of a small volume of a solid or a liquid blowing agent into a large volume gas bubble is difficult to control, and typically

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

results in structures with non-uniform, macroscopic voids. Furthermore, residuals that could adversely affect the properties of the drug may remain in the dosage forms. A better concept, therefore, would be to dispense the bubbles directly into the liquid. This method is employed in the manufacture of cellular metals and other foamed materials from low-viscosity liquids [28–30]. The bubbles are injected in a liquid-filled container, rise to the top, form a foam and solidify. This particular technique, however, has poor control over the resulting microstructure, and, in any case, is not applicable to such high-viscosity fluids as polymeric melts in which bubble rise is too slow. The manufacture of polymeric cellular dosage forms, therefore, requires a moving liquid stream into which bubbles are injected at fixed time intervals and at discrete locations. Such systems could achieve fully predictable microstructures, for both the flow rate of the polymeric liquid and the formation and motion of the gas bubbles are controllable. Indeed, recently several microfluidic devices have been proposed and discussed as tools for producing foams of aqueous or other low viscosity liquids, but the concepts have not been developed yet to produce polymeric cellular solids [31–34]. In this work, accordingly, a novel micro- or milli-fluidic process is presented for the continuous manufacture of polymeric cellular dosage forms with predictable microstructure and desirable drug-delivery characteristics.

2. Theory 2.1. Process overview In the manufacturing process presented here the input is a particulate mixture of a solid drug and a solid thermoplastic polymeric excipient at a determined drug-to-excipient mass (or volume) ratio. The powder is filled in a syringe of uniform barrel diameter at point A (Fig. 1). By controlled displacement of the piston of the syringe pump, the powder mixture is fed through a hopper into an extruder at B. The extruder integrates multiple unit steps into a continuous process. First the solid granules are conveyed forward from the inlet by the rotating screw. Additionally, the granules are mixed further as they are conveyed along the screw, and if the barrel temperature is kept above the melting temperature of the excipient, the excipient melts and fluidizes the mixture as it

551

is transported forward. The mixture is fluidic at point C, and in section CD the pressure of the melt increases towards the end of the screw at D. The high pressure generated by the extruder then pushes the melt first through a converging nozzle from D to E, and subsequently through a capillary channel into which a hypodermic needle is inserted at F. The hypodermic needle is connected to a gas delivery system at G, comprising an electrical shut-off valve and a high-pressure gas tank. By controlling the gas pressure at the outlet of the tank and the opening and closing times of the valve (i.e., the pulse width and repetition rate), a train of bubbles of well-controlled size and spacing is introduced into the fluid stream at F. The two-phase fluid stream is subsequently dispensed in a mold at H. The cellular material is then shaped to form the final tablet, solidified, and ejected from the mold at I.

2.2. Powder and melt flows in the screw Because the inputs are granular materials, powder handling is unavoidable at the early stages of the process. At present, there are no rigorous mathematical models that adequately describe the flow behavior of granular matter. Thus the models presented here are heuristic. The mass flow rate of solid particles fed by the syringe pump is:

dMf ¼ ðqd ud þ qe ue ÞAp v p dt

ð1Þ

where qd and qe, respectively, are the densities of the solid drug and excipient particles, ud and ue their volume fractions in the particle bed, Ap is the cross-sectional area of the syringe barrel, and vp the translational velocity of the piston. It has been reported that from the point where the barrel temperature exceeds the melting temperature of the solid, at about 2– 4 times the pitch of the screw forward, a small melt pool forms [21]. (For further details of material flow in the extruder, Reference 21 may be consulted.) The flow is constricted at the end of the extruder screw, and thus the melt pressure increases as the material is transported forward along the screw. The pressure is greatest at point D, and it then decreases as the melt flows from D to E to F. At point F, the melt pressure depends on the melt flow rate and the flow resistance of the channel.

Fig. 1. Schematic of the proposed process for continuously manufacturing cellular dosage forms. A syringe pump delivers the drug and excipient granules to an extruder where the material is fluidized by melting and the pressure increased. The high pressure at the extruder outlet drives the melt through a micro-or milli-fluidic channel where gas bubbles are injected. The multi-phase mass (solid drug particles, liquid excipient, and gas bubbles) is subsequently dispensed in a mold where it is cooled and solidified to the final dosage form.

552

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

2.3. Melt flow in the channel There are two cases to be considered. In the first case, the gas pulse valve is off, i.e., no gas is injected, and the channel is filled with a Newtonian viscous, incompressible drug-laden polymeric melt with no gas bubbles, as shown in Fig. 2a. In laminar flow, the pressure drops linearly along the channel, and the pressure difference between points F and H can be expressed by the HagenPoiseouille equation: ð0Þ

Pm;F  Pm;H ¼

8lm Lc Q m;0

ð2aÞ

pR4c

ð0Þ

where Pm;F is the melt pressure at F, Pm,H the pressure at H, lm the viscosity of the melt, Lc the length of the channel from F to H, Rc the channel radius, and Qm,0 the volumetric melt flow rate. The melt pressure at point H is atmospheric, thus Pm,H = Patm. The melt flow rate in the channel with no gas bubbles, Qm,0, can be obtained from the mass flow rate, dMf/dt, of solid particles fed to the extruder and the density of the melt, qm, as:

Q m;0 ¼

1 dMf

ð2bÞ

qm dt

For the parameters and conditions of this work (listed in Table A1 of ð0Þ

the Appendix), Qm,0 = 1.5 mm3/s, and, by Eq. (2a), Pm;F = 0.199 MPa. It may also be noted that the average velocity in the channel, vm,0 = Qm,0/pR2c = 1.9 mm/s and the Reynolds number Re = 2qmvm,0Rc/lm = 3.8  105. This is much smaller than the critical Reynolds number, Rec 2300, and thus the flow is laminar throughout the channel. Furthermore, for a surface tension of the melt, c = 0.045 N/m, the capillary number Ca = lmvm,0/c = 4.9. At this value of the capillary number, the capillary forces are negligibly small compared with the viscous forces, which supports the validity of Eq. (2a) under the assumptions given. In the second case, the gas pulse valve is again off, but gas bubbles of appreciable volume fraction are present in the channel as shown in Fig. 2b. Because the viscosities of the drug-laden melt and the gas are vastly different, the pressure gradient, dp/dx, is different in the melt- and gas-filled segments of the channel. Hence the difference of pressures at points F and H cannot be determined by Eq. (2a), and must be found by integration of dp/dx along the melt- and gas-filled segments of the channel. In a melt-filled segment, the pressure gradient is approximately:


8lm Q m;s dp

¼ dx
m pR4c

where Qm,s is the melt flow rate in the melt-filled channel segment. In the gas-filled segments, the pressure gradient inside the bubbles is small due to the low viscosity of the gas. Furthermore, the shear stresses at the bubble-melt interfaces may be assumed essentially equal to zero. Thus the pressure gradient in the gasfilled segments, dp=dxjg  dp=dxjm , and the flow resistance offered by the bubbles to overall flow can be neglected. Assuming that the melt flow rate in a melt-filled channel segment, Qm,s, is the same as the melt flow rate into the two-phase channel at point F, Qm,1, the pressure drop in the channel, by integration of Eq. (3), is: ð1Þ

Pm;F  P m;H ¼

8lm Lm Q m;1

ð4Þ

pR4c

ð1Þ

where Pm;F is the melt pressure at point F of the two-phase channel and Lm the length of the melt-filled fraction of the channel. Lm may be written in terms of the volume fraction of melt in the channel, um,c, and the total channel length, Lc, as

Lm ¼ um;c Lc

ð5Þ

Under the conditions of this work, um,c is between about 0.31 and 0.83, and Qm,1 is about 1.03–1.15 times greater than Qm,0 (as shown ð1Þ

in Section 2.5 later). Thus P m;F is in the range 0.139–0.191 MPa when the gas valve is closed but gas bubbles are present in the channel (as listed in Table 1 of section 4). 2.4. Gas bubble injection and growth To inject gas bubbles into the melt stream, the bubble pressure, Pb, at the exit of the hypodermic needle should be greater than the sum of the melt pressure at F, Pm,F, and the capillary pressure, Pcap. Thus

Pb > Pm;F þ Pcap ¼ P m;F þ

2c Rn

ð6Þ

where c is the tension of the gas-melt interface and Rn the inner radius of the hypodermic needle. For Rn = 25 mm and c = 0.045 N/ m, Pcap = 3.6 kPa, which is much smaller than the values of Pm,F calculated above. Thus gas can be injected into the melt at about Pm,F. ð0Þ

(Note that Pm,F = P m;F for a melt-filled channel without gas bubbles ð1Þ P m;F

Fig. 2. Schematics of melt-driven flows in the capillary channel: (a) melt flow without any gas bubbles, (b) melt-driven flow of drug-laden melt and gas bubbles when the gas pulse valve is off.

ð3Þ

for a channel with gas bubbles). and Pm,F = If the gas pressure is only slightly greater than the melt pressure, very small gas bubbles will be released slowly from the hypodermic needle into the melt stream. Melt flow in the channel will be essentially unaffected by the injection of gas in this case. The object of this study, however, is to create dosage forms with considerable volume fraction of cells or voids. This can be accomplished only if the gas pressure in the bubble during gas injection is considerably greater than the melt pressure, and the bubble radius assumes a value of the order of the channel radius. In that case, melt flow at the channel entrance will be virtually blocked as gas is injected. Thus the driving force for fluid flow in the channel alternates between the melt pressure at the channel entrance when the valve is closed, and the gas pressure in the bubble when the valve is open. Fig. 3 is a schematic of gas flow through a hypodermic needle into a channel in which fluid flow is driven by the gas pressure. Neglecting the flow resistance of the gas-filled regions in the channel, and assuming that the pressure gradient is the same in each of the melt-filled segments, the bubble expansion rate, dVb/dt, may be expressed as:

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

553

where sg and sm, respectively, are the times when the valve is open and closed in a pulsing cycle. For the pulsing parameters listed in Table 1, Qm,1 = 1.03–1.15  Qm,0. The volume of a melt-filled segment, Vm, is determined by Qm,1 and the time sm during which melt flows into the channel in a pulsing cycle (i.e., the time when the valve is closed):

V m ¼ Q m;1 sm

ð11Þ

The melt may be assumed incompressible, and thus Vm is constant as the melt-filled segment flows forward along the channel. Accordingly, the length of a melt-filled segment, or the mean ‘‘free” distance between the bubbles in the channel, k, is:

k¼

Fig. 3. Schematic of gas flow through a hypodermic needle connected to a pulse valve and a gas reservoir at pressure Pg. The schematic is drawn at the end of the gas injection pulse.

dV b pR4c ðPb  Patm Þ ¼ 8lm Lm dt

ð7Þ

The bubble pressure, Pb, can be calculated from the gas pressure drop in the hypodermic needle. Assuming that the gas is incompressible in the needle, by Hagen-Poiseouille’s equation the pressure drop is:

Pg  Pb ¼

8lg Ln Q g

p

R4n

ð8Þ

where Pg is the gas pressure in the reservoir, lg its viscosity, Ln and Rn, respectively, are the length and radius of the passage in the hypodermic needle, and Qg is the gas flow rate which must be equal to dVb/dt. By combining Eqs. (7) and (8), and inserting the parameter values representative of this work (Table A1), it is found that Pg  Pb  Pb  Patm. Thus Pb  Pg. Accordingly, the gas flow rate, Qg, or bubble expansion rate, dVb/ dt, when the valve is open, is:

Qg ¼

dV b pR4c ðPg  P atm Þ ¼ 8lm Lm dt

Q m;1

pR2c

sm

ð12Þ

For the present parameters, k = 879–1184 lm (Table 1). The gas flow rate into the channel at point F, Qg, by Eq. (9) is constant when the valve is open, provided that um,c is again assumed time-invariant. When the valve is closed, the gas flow rate is zero for a needle volume so small that the amount of compressed residual gas in it is negligible. Thus also the profile of the gas flow rate versus time into the channel roughly follows a square wave as shown in Fig. 4b. Unlike the melt-filled segments, however, the gas-filled segments cannot be assumed incompressible. The volume of a gas-filled segment increases as it flows forward along the channel and the pressure is decreased. A rough estimate of the average volume of a gas-filled segment, Vg, in the channel is the average of the segment volumes at points F and H, respectively, and may be written as:

Vg ffi

  1 Pg Pg þ Patm Q g sg þ Q g sg ¼ Q g sg 2 Patm 2Patm

ð13Þ

where sg is the time during which the gas flows into the channel in a pulsing cycle (i.e., the time when the valve is open).

ð9Þ

In the present work, the gas pressure is constant at Pg = 0.35 MPa. Thus Qg is between 4.27 and 11.35 mm3/s for Rc, lm, and Lc as shown in Table A1, and Lm/Lc = um,c = 0.31–0.83 as derived below. 2.5. Microstructure of the fluid in the channel The volume fraction of melt in the channel, um,c, is equal to the ratio of melt volume to total volume in the channel. It can be derived from the volumes of the melt- and gas filled segments which are in turn determined by the melt- and gas flow rates. If melt flow into the channel is blocked during gas injection, and ð1Þ

P m;F (the melt pressure at F when the valve is closed) and um,c are assumed constant for a given set of conditions, the profile of the melt flow rate versus time at point F follows a square wave as shown in Fig. 4a. The melt flow rate into the channel when the valve is closed, Qm,1, may thus be derived from the timeaverage volumetric melt flow rate, Qm,0, as:

Q m;1 ffi Q m;0

sg þ sm sm

ð10Þ

Fig. 4. Melt- and gas-driven flow rates at the channel entrance: (a) melt is delivered when the valve is ‘‘off” (closed), and (b) gas is delivered when the valve is ‘‘on” (open).

554

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

Now the volume fraction of melt in the channel can be derived. At steady-state, um,c is the same as the melt volume fraction in an average pair of melt- and gas-filled segments. Thus

Q m;1 sm Vm ¼ V m þ V g Q m;1 sm þ Pg þPatm Q g sg 2P atm  1 Pg þ P atm Q g sg ¼ 1þ 2Patm Q m;1 sm

um;c ¼

ð14Þ

This result and Eqs (4), (5), and (9) show that the microstructure of the fluid in the channel and the gas- and melt flow rates are interdependent. Nonetheless, Eqs. (5), (9), (10), and (14) can be combined to give the following expression for the bubble volume produced per pulse at point F, Vb,F, for a needle volume that is negligible small:

V b;F ¼

dV b sg dt

pR4c ðPg  Patm Þsg Pg þ Patm pR4c ðPg  Patm Þ sg ¼ 1 8lm Lc Q m;0 sg þ sm 8lm Lc 2Patm

!1

ð15Þ As shown in Table 1 of section 4, Vb,F = 0.233 mm3 for the parameters of this work with Pg = 0.35 MPa, sg = 40 ms and sm = 500 ms. We may note that the dimensionless term on the right hand side of Eq. (15) is equal to 1/um,c, which is between 1/0.31 and 1/0.83 for the parameters of this work.

At the exit of the channel the melt pressure is equal to atmospheric pressure. Accordingly, the bubble volume at H, Vb,H, from Boyle’s law, is:

Pg V b;F P atm

ð16Þ

Assuming that the bubble or cell shape in the liquid foam is spherical, the cell size, Dcell, is:

Dcell ¼

 1 6V b;H 3

ð17Þ

p

Here Dcell = 833–1664 lm (as listed in Table 1). Finally, the volume fraction of gas (or voids), uv, in the liquid foam at atmospheric pressure is:

uv ¼

 P g Q g sg Patm Q m;1 ¼ 1þ P atm Q m;1 sm þ Pg Q g sg Pg Q g

sm sg

The liquid foam is subsequently filled into a mold cavity with the shape of the dosage form (Fig. 5). A low viscosity melt would be desirable to spread and distribute the material evenly in the mold. However, such a system also exhibits fast bubble dynamics. For example, the Hadamard-Rybczynski equation for the terminal rising bubble velocity, vb, of a gas bubble with density, qg  qm, and viscosity, mg  mm, is [35,36]:

vb ¼

1 D2cell g qm 3 4lm

ð20Þ

By Eq. (20), in a low-viscosity melt with qm = 1150 kg/m3 and mm = 2 Pas (e.g., PEG 8k at 70 °C), a bubble of diameter Dcell = 800 mm assumes a terminal rising velocity of 0.3 mm/s. Thus if bubble rise must be less than about 300 lm so as to preserve the microstructure of the liquid foam (i.e., to avoid such problems as bubble assembly at the top of the dosage form, and bubble coalescence and bursting [37–40]), the melt must solidify within 1 second after discharge from the capillary channel. This is much faster than the typical cooling time of the liquid foam, as we shall see in the next section, and hence the low viscosity melts should be avoided. The bubble rise and bursting and coalescence time can be delayed by increasing the viscosity of the melt sufficiently. Higher viscosity fluids (and fluids with large bubble content) may, however, grow vertically in the mold and not spread evenly. But this problem can be overcome by positioning the mold close to the channel exit, and by moving it in the horizontal plane. 2.8. Cooling and solidification of the liquid foam

2.6. Microstructure of the liquid foam near the channel exit

V b;H ¼

2.7. Mold filling

Nonetheless, to preserve the microstructure of the liquid foam, and to achieve high process rates, cooling and solidification of the liquid foam should be fast. Assuming that the liquid foam is a circular disk (or a right circular cylinder) with temperatureindependent, isotropic thermal properties, and no heat of fusion, the heat conduction equation for the axisymmetric problem may be written as [41]:

qc

     @T k @ @T @ @T ¼ r þ r @t r @r @r @z @z

ð21Þ

where q is the density of the molded liquid foam, c the specific heat, and k the thermal conductivity. Initially, the liquid foam is at the temperature T0. The boundary condition to Eq. (21) is that the mold wall temperature, Tw, is kept constant. As illustrated in Fig. 6, the liquid foam or dosage form considered has a diameter greater than the thickness, i.e., D0 > H0. In this

1 ð18Þ

By combining this equation with Eqs. (5), (9), (10), and (14), an intricate term is obtained for uv, which is found to be between 0.24 and 0.77 for the parameters of this work (Table 1). Thus the microstructure of the liquid foam is determined by wellcontrollable parameters in the system presented. Furthermore, it ð1Þ

may be noted that if P m;F were measured and controlled, a simple term for uv could be derived from Eqs. (4), (9), and (18) as:

uv ¼

ð1Þ

Patm Pm;F  Patm 1þ Pg Pg  P atm

sm sg

!1 ð19Þ

This underscores the benefit of monitoring and controlling the melt pressure at F for achieving a liquid foam with readily predictable microstructure.

Fig. 5. Schematics of foam formation in a mold. The melt stream with gas bubbles is dispensed in the mold where a liquid foam builds up: (a) melt with a small volume fraction of gas bubbles and (b) foam with a large volume fraction of gas.

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

555

limited here to the calculation of an upper bound, ku, and a lower bound, kl, of the effective thermal conductivity. The upper bound is:

ku ¼ ð1  uv Þks þ uv kg

ð27Þ

and the lower bound may be written as:

1 ð1  uv Þ uv ¼ þ kl ks kg

Fig. 6. Cooling and solidification of the liquid foam in the mold. The initial temperature of the foam is T0 and the temperature of the mold walls are kept at a temperature Tw.

ð28Þ

For the material properties listed in in the Appendix, and a liquid foam with H0 = 6 mm and uv = 0.5, the time sc to reduce the temperature at the center from T0 = 80 °C to Tc = 70 °C with Tw = 25 °C is between 15 and 23 s. This result can now be combined with the model for the terminal rising bubble velocity introduced in the previous section. By Eq. (20), a bubble of diameter Dcell = 800 mm rises by less than 300 lm in 23 s if the melt viscosity is greater than about 46 Pas. Thus to preserve the microstructure reasonably well after the foam is dispensed in the mold, the viscosity of the melt must be greater than about 46 Pas. In the present work, therefore, PEG 35k is used as the excipient. The shear viscosity of PEG 35k is 115 Pas at 80 °C. 3. Materials and methods

case, the spatial derivatives of the temperature gradients are greater axially than radially. Thus Eq. (21) may be reduced to the one-dimensional heat conduction equation with the same initial and boundary conditions:

qc

@T @2T ¼k 2 @t @z

ð22Þ

The solution to this problem is [41]: 1 TðzÞ  T w 4 X ð1Þn ¼ T0  Tw p n¼0 2n þ 1

! k ð2n þ 1Þ2 p2 t ð2n þ 1Þpz  exp  cos 2 2H0 qc 4H0

ð23Þ

It can be readily shown that the first term in the series with n = 0 is the dominant term. Thus the cooling time sc (i.e., the time required for the center of the dosage form to reach a temperature Tc) may be approximated as:

sc ¼



qc H20 4 T0  Tw ln k p2 p Tc  Tw

 ð24Þ

The molded foam, however, is not a homogeneous material because it comprises molten excipient, solid drug particles, and gas bubbles. Thus to calculate sc by Eq. (24), the effective thermal properties (i.e., q, c, and k) of the composite medium must be derived. The excipient and drug particles may be assumed to have the same thermal properties, but their combined density, specific heat, and thermal conductivity are vastly different from those of the gas bubbles. The molded foam is therefore considered a twophase material consisting of the solidifying drug-excipient structure and the gas-filled cells. The density and heat capacity are scalar properties, thus their effective values are equal to the average values

q ¼ ð1  uv Þqs þ uv qg

ð25Þ

and

c ¼ ð1  uv Þcs þ uv cg

ð26Þ

The thermal conductivity, however, is affected by the microstructural details of the molded foam and a calculation of the exact effective value is highly involved. Thus the analysis is

3.1. Apparatus for preparing solid and cellular dosage forms The dosage forms were prepared by a self-made desktop screw extruder with the following components: a granule-feeding unit, an extrusion screw and barrel, a gas delivery unit, and a molding unit. The granule-feeding unit consists of a syringe with uniform barrel diameter and a syringe pump to deliver the granules into a hopper. The hopper then directs the granules into the extrusion screw and barrel. The extrusion screw is 244 mm long, has an outer diameter of 10 mm, a helix angle of 17.65°, a channel height of 1 mm, and a channel width of 8.5 mm. The barrel is surrounded by a resistance heater coil to set the temperature of the barrel. The material exiting the extruder is delivered into a capillary channel with 1 mm diameter. The channel is made of Plexiglas (to facilitate imaging) and heated by an infrared lamp. A 25 mm long hypodermic needle with an inner diameter of 50 mm is inserted into the channel. The needle is connected to a pulse valve (on/off valve) in line with a gas regulator and cylinder. The material exiting the channel flows downward into a cylindrical mold. Fig. 7 is a photograph of the apparatus. 3.2. Preparation of solid and cellular dosage forms The dosage forms were prepared by first mixing 5 wt% of solid Acetaminophen particles with 95 wt% granules of PEG 35,000 (PEG 35k). The particles were mixed and loaded in the granulefeeding unit (a syringe in a syringe pump) which was set do deliver 1.73 mg/s. The rotation rate of the screw was about 3 rpm and the barrel and nozzle temperatures were set to 80 °C. The temperature of the ambient air and the mold were 25 °C. For preparing the non-porous solid dosage forms, the pulse valve was closed. The mold was filled with the effluent stream until a height of 6 mm was reached in the mold. The material was then left in the mold for about two minutes to solidify. Subsequently, the final dosage form was ejected from the mold. For preparing the cellular dosage forms, the pulse valve was operated at various combinations of sg and sm. The range of sg was 20–60 ms and that of sm 400–600 ms. Again, the mold was filled with the effluent stream until a height of 6 mm was reached. The material was then left in the mold for about 2 min to solidify and afterwards, the cellular dosage form was removed from the mold. All the dosage

556

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

Fig. 7. Photograph of the apparatus for continuous manufacture of the dosage forms.

forms prepared in this work were cylindrical disks with the same dimensions: 9 mm in diameter and about 6 mm in thickness.

4. Results and discussion 4.1. Melt flow and gas delivery

3.3. Imaging of the fluid streams The process steps that were imaged include gas injection in the channel, transport of gas bubbles to the channel exit, discharge of the fluid stream from the channel exit to the atmosphere, and deposition and build-up of the foam on a horizontal surface. Images were taken by a Nikon DX camera. 3.4. Volume fraction of voids of the dosage forms The volume fraction of voids in the solidified dosage forms, uv, was determined from their density, q, as:

uv ¼ 1 

q qs

ð29Þ

where qs is the density of the non-porous solid. The densities of the dosage forms were determined by measuring their weight and volume. The volume fraction of voids of the dosage forms was measured in triplicate. 3.5. Scanning electron microscopy For obtaining a scanning electron micrograph of the dosage form microstructure, a cross sectional surface of the dosage form was first prepared by cutting the dosage form with a thin microtome blade (MX35 Ultra, Thermo Scientific, Waltham, MA). The cross sectional surface was then imaged by a Zeiss Merlin High Resolution SEM with a GEMINI column. Imaging was done with an in-lens secondary electron detector. An accelerating voltage of 5 kV and a probe current of 95 pA were applied to operate the microscope. 3.6. Dissolution testing Drug release of the dosage forms was tested by a USP dissolution apparatus 1 (from Sotax AG) filled with 900 ml of 0.05 M phosphate buffer solution (using sodium phosphate monobasic and sodium phosphate dibasic) at a pH of 5.8 and at 37 °C. The basket was rotated at 50 rpm. The concentration of dissolved drug was measured versus time by UV absorption at 244 nm using a fiber optic probe (Pion, Inc.). Dissolution experiments were done in triplicate.

Images of the melt stream after gas injection are presented in Figs. 8a–d for different pulsing parameters of the gas flow. Fig. 8a shows the melt stream in the channel without any injected gas. The stream is bubble-free. A representative image of the melt stream with gas pulses of sg = 20 ms and sm = 600 ms (condition B) is shown in Fig. 8b. The average bubble volume, Vb = 0.15 mm3, and the distance between the bubbles k = 1.397 mm (Table 1). Figs. 8c and d demonstrate that Vb increases if sg is increased, and k decreases if sm is decreased. For example, if sg = 60 ms and sm = 400 ms, Vb = 0.766 mm3 and k = 676 mm (condition D, Fig. 8d and Table 1). These results are compared with the values calculated by the theory presented in section 2. The bubble volume right at the exit of the needle is calculated by Eq. (15) (Table 1). The calculated and measured values agree if a 10–40% volumetric expansion of the bubble (due to a drop in pressure along the channel) is assumed between the needle exit and the location of the bubble in the channel shown in Figs. 8b–d. The ‘‘free” distance between the bubbles, k, is calculated by Eq. (12). As shown in Table 1, the calculated values deviate by less than 30 percent from the experimental values.

4.2. Deposition of the melt stream on a flat surface The fluid stream exiting the channel was allowed to accumulate on a horizontal surface (with the mold removed) as shown in Figs. 8e– h. Without any added gas, there were no bubbles in the fluid thread. A clear liquid was deposited on the bottom surface as shown in Fig. 8f. When gas was added upstream, the bubbles were visible and assumed the shape of a disk, or cylinder, in the fluid thread (Figs. 8f–h). The foam deposited on the flat surface first grew vertically and then radially before it solidified.

4.3. Microstructures of the dosage forms Scanning electron microscopy images of the solidified material structure are presented in Figs. 8i–m. The cast solid dosage forms (condition A) are essentially non-porous as shown in Fig. 8i. The drug and excipient cannot be distinguished in the image.

557

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

Fig. 8. Fluid structure in the channel, foam formation, and microstructure of the ‘‘manufactured” dosage forms. Volume fraction of gas or voids in the dosage forms: From left to right, 0.01, 0.26, 0.51, and 0.72 (a)–(d) images of the fluid channel after gas injection, (e)–(h) images of the effluent after channel exit (with the mold removed) and foam formation, and (i)–(m) scanning electron micrographs of the resulting microstructures.

Table 1 Summary of process parameters, and calculated and measured process and property values. Process parameters

A B C D

Pg (MPa)

sg

sm

(ms)

– 0.35 0.35 0.35

– 20 40 60

Calculated process and microstructural parameters

Measured microstructural parameters

k (mm)

Vb,F* (mm3)

Deq* (mm)

Dcell (mm)

uv

k (mm)

Vb * (mm3)

Deq* (mm)

Dcell (mm)

uv

kcell (mm)

fopen

(ms)

Pm,F (MPa)

– 600 500 400

0.199 0.191 0.170 0.139

– 1184 1031 879

– 0.085 0.233 0.681

– 551 769 1100

– 833 1163 1664

0.00 0.24 0.50 0.77

– 1397 829 676

– 0.150 0.266 0.766

– 659 798 1135

– 824 1002 1438

0.01 0.26 0.51 0.72

– 1500 748 271

– <0.2 0.55 0.55

The calculated values are derived as follows: Pm,F is calculated by Eqs. (2a) and (4) where dMf/dt = 1.73 mg/s, qm = 1150 kg/m3, mm = 115 Pas, Rc = 0.5 mm, and Lc = 15 mm. k is calculated by Eq. (12). Vb,F is calculated by Eq. (15), Deq is the equivalent diameter of a bubble and is calculated by Deq = (6Vb,F/p)0.33. Dcell is calculated by Eq. (17) and uv is calculated by Eq. (18). All the measured parameters are derived from the images presented in Fig. 8, except uv which is derived as described in section 3.3. It may be noted that uv could also be calculated from the SEM images as: uv = Lcell/(Lcell + Lsolid) where Lsolid is the total length covered by the solid along the test line(s). The results so obtained agree with the tabulated results. * The bubble volume is calculated at the location of gas injection, but the measured bubble volume is an average of the bubble sizes downstream in the channel.

For an analysis of the microstructural details of the cellular dosage forms, a number of lines were drawn randomly on each SEM image of the microstructure (Figs. 8i–m). The mean linear intercept, lcell, which is the ratio of the total length covered by the cells, Lcell, and the number of cells that intersect with the lines, Ncell, was then determined for each microstructure [42]:

lcell ¼

Lcell Ncell

ð30Þ

Assuming that the cells are spherical and of uniform size, according to Tomkeieff [42,43]:

lcell ¼ 4

V cell 2 ¼ Dcell Scell 3

ð31Þ

where Vcell is the volume of a single cell, Scell its surface area, and Dcell the cell diameter. The results of the cell diameter obtained from Figs. 8i–m and Eqs. (30) and (31) are listed in Table 1. Dcell = 824 mm for the dosage

form prepared under condition B, and Dcell = 1438 mm for condition D. The cell size calculated by combining Eqs. (15)–(17) is 833 mm for condition B and 1664 mm for condition D (Table 1). Thus the calculated values agree with the experimental results fairly well. The measured volume fraction of voids in a dosage form was determined by comparing the measured density of the dosage form with that of the solid material as described in section 3.3. The results of the measured uv values are tabulated in Table 1. uv = 0.26 for the dosage form B. It increases to 0.72 for the dosage form D. A plot of the measured versus calculated volume fraction of voids is shown in Fig. 9 (uv is calculated by Eq. (18) and Eqs. (5), (9), (10), and (14) using the parameter values given in Tables 1 and A1). The data can be fitted to a curve y = 0.96⁄x with an R2 value of 0.99. This confirms the agreement between calculations and experimental values. The mean free distance between the cells is [42]:

kcell ¼

1  uv ðLcell þ Lsolid Þð1  uv Þ ¼ Ncell Ncell =ðLcell þ Lsolid Þ

ð32Þ

558

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

Fig. 9. Measured versus calculated volume fractions of voids in the dosage forms.

The values obtained for kcell range from 271 mm for the dosage form D to 1.5 mm for the dosage form B (Table 1). Finally, the fraction of cells that are open, fopen, is determined by counting the number of cells with an open passage to another cell on the SEM image, and dividing this number with the total number of cells. The value for fopen so obtained is approximate, because the SEM images do not show the three-dimensional view of the cells. The estimated fopen is between less than 0.2 for the dosage form B and 0.55 for the dosage form D. Thus the fraction of open cells in the present work is smaller than that reported previously for polymeric cellular dosage forms at similar volume fractions of voids [22–25]. 4.4. Drug release characteristics of the dosage forms Representative results of the fraction of drug dissolved versus time after immersion of the dosage forms in the dissolution medium are shown in Fig. 10a. For all the dosage forms, the fraction of drug dissolved increases steadily with time until the entire amount of drug is dissolved, and the curves plateau out to the final value of 1. The time required to dissolve 80% of the drug content, t0.8, is extracted from these curves and plotted in Fig. 10b. t0.8 is about 60 min for the non-porous solid dosage forms. It decreases with increasing volume fraction of voids and is about 10 min for the dosage forms with uv = 0.72 (condition D). The cellular dosage forms were floating at the top of the basket during dosage form disintegration while the non-porous solid dosage forms settled at its bottom. Dosage form disintegration occurred predominantly by gradual surface erosion of the excipient, similar to the closed-cell structures described in Ref. [22], with occasional removal of visible fragments. Thus, assuming that the streamlines enter the surface dimples due to the cells, the flux of the eroding polymeric excipient can be approximated by a convective mass transfer model that takes the increased surface area due to the cell dimples into account [44,22]: 2 3

j ¼ 0:62  D c0

qf lf

!16

X2 ð1 þ uv Þ 1

ð33Þ

where D is the diffusivity of the eroding polymer in the dissolution medium, c0 the interfacial concentration, qf and mf, respectively, are the density and viscosity of the dissolution medium, and X the angular velocity of the basket. If erosion occurs from both faces of

Fig. 10. Drug release properties of the dosage forms: (a) fraction of drug dissolved versus time of representative dosage forms and (b) measured value of t0.8 versus volume fraction of voids. It may be noted that in this work, the disintegration rate of the dosage form structure is rate-determining. Thus the fraction of drug dissolved per unit time is about equal to the disintegration rate of the dosage form. The full set of dissolution profiles of the dosage forms tested in this work is shown in Fig. S1 of the supplementary information.

the dosage form, the erosion or disintegration time of the dosage form is:

tdis ¼

1 qs ð1  uv ÞH0 2 j

ð34Þ

where qs(1-uv) is the density and H0 the thickness of the dosage form. For the dissolution medium, stirring rate, excipient, and dosage form geometry of this work (Table A1), it is estimated that 0.8  tdis = 52 min for the essentially non-porous dosage form, and 0.8  tdis = 8.4 min if uv = 0.72. The disintegration times of the dosage form structures of this work are much greater than the dissolution times of small Acetaminophen particles that have been released from the dosage forms. Thus the disintegration rates of the dosage form structures are rate-determining in the dissolution of drug by the dosage forms presented here. The above model is supported by the measured data of t0.8 shown in Fig. 10b, which are fairly close to the calculated values of 0.8  tdis.

559

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

5. Summary In the continuous process for manufacturing cellular dosage forms introduced in this work, solid drug and excipient particles are first dosed into an extruder, wherein the material is fluidized by melting, and pressurized to flow through a micro- or milli-fluidic channel. The channel is equipped with a gas injection hypodermic needle connected to an on/off valve and a gas reservoir for injecting gas bubbles of the desired size and at the required frequency into the melt stream. The two-phase melt stream is then dispensed in a mold and solidified to the final dosage form. The flow rate of the melt into the channel is determined by the extrusion and feeding parameters. The volume of the injected bubble, or the cell volume in the liquid foam, is determined by the gas flow rate and the duration of the injection pulse. The gas flow rate is in turn controlled by the injection pressure, while the pulse duration is set by the duration of valve opening. Similarly, from the flow rates of melt and gas, and the valve’s pulsing parameters, the ‘‘free” distance between cells and the volume fraction of voids in the liquid foam can be determined. The cell volume, inter-cell spacing, and volume fraction of voids of the liquid foam are preserved after the foam is dispensed in a mold if the viscosity of the melt is high enough to avoid bubble motion before solidification. For the experimental conditions and parameters of this work, a mixture of molten PEG 35k excipient and drug satisfied this requirement. The cell size of the dosage forms prepared herein was in the range 824–1438 lm, and the volume fraction of voids, uv = 0.01–0.72. The cells were predominantly closed if uv was 0.26 or less, but partially open at higher volume fractions of voids. Dosage form disintegration was primarily by surface erosion and is well described by previously developed models. The time to dissolve 80% of the drug content, t0.8, could be varied in the range 10–60 min by varying the volume fraction of voids. It is expected that faster dosage form disintegration, and drug release, could be achieved by using a faster eroding excipient or by increasing the connectivity of the void space. By contrast, slower dosage form disintegration could be achieved with an excipient that erodes slowly, such as a less soluble or a larger molecular weight polymer. Moreover, because the densities of most polymeric excipients and drugs are similar to that of water, the polymeric cellular dosage forms float over the dissolution medium during disintegration if the volume fraction of voids is sufficiently large and the cells are closed. Thus upon ingestion in the fed state, such low-density cellular forms are expected to float over the gastric contents of the stomach if the patient is in the upright position (as during daily activities). This prevents such dosage forms from being subjected to the vigorous peristaltic movements of the caudad stomach, and uncontrollable gastric emptying, thus providing a longer and more predictable gastric residence time [45–49]. In conclusion, polymeric cellular dosage forms can be efficiently manufactured by a continuous process. Such dosage forms offer predictable drug release rate and density for tailoring the drug release profile in the human body.

Acknowledgements We would like to thank Dr. Geoffrey Grove for providing the dissolution bath from Sotax AG. Thanks are due to Mr. David Kwajewski for making the UV fiber optic probes from Pion, Inc available.

Appendix A Table A1 Values of parameters and properties. Property

Value

b cg cs c0 D D0 H0 kg ks Lc Ln dMf/dt N Patm Pg Qm,0 Rc Rn T

0.154 nm 1005 J/kg K 1000 J/kg K 163.4 kg/m3 1.09  1010 m2/s* 9 mm 6 mm 0.026 W/m K 0.1 W/m K 15 mm 25 mm 1.73 mg/s 2385 101 kPa 350 kPa 1.5 mm3/s 500 lm 25 lm 310 K 0.045 N/m 0.001 Pas 1.98  105 Pas 115 Pas 1000 kg/m3 1.8  105 kg/m3 1150 kg/m3 1150 kg/m3 5.24 rad/s

c lf lg lm qf qg qm qs X

The values of the pulsing parameters sg and sm are given in Table 1. * D is calculated by Zimm’s model as D = 0.192  kbT/ N0.5blf [50].

Appendix B. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.cej.2017.02.057. References [1] A.R. Gennaro (Ed.), Remington’s Pharmaceutical Sciences XVIII, Mack Publishing, Easton, PA, 1990. [2] J.B. Dressman, G.L. Amidon, C. Reppas, V.P. Shah, Dissolution testing as a prognostic tool for oral drug absorption: immediate release dosage forms, Pharm. Res. 15 (1998) 11–22. [3] A.W. Hixson, J.H. Crowell, Dependence of reaction velocity upon surface and agitation: III-experimental procedure in study of agitation, Ind. Eng. Chem. 23 (1931) 1160–1168. [4] J. Siepmann, F. Siepmann, Mathematical modeling of drug dissolution, Int. J. Pharm. 453 (2013) 12–24. [5] G. Levy, B.A. Hayes, Physicochemical basis of the buffered acetylsalicylic acid controversy, New Eng. J. Med. 262 (1960) 1053–1058. [6] M. Rowland, T.N. Tozer, Clinical Pharmacokinetics, second ed., Lea & Febiger, Philadelphia, PA, 1986. [7] H.M. Jaeger, S.R. Nagel, R.P. Behringer, Granular solids, liquids, and gases, Rev. Mod. Phys. 68 (1996) 1259–1273. [8] P.G. De Gennes, Granular matter: a tentative view, Rev. Mod. Phys. 71 (1999) 374–382. [9] T. Shinbrot, F. Muzzio, Noise to order, Nature 410 (2001) 251–258. [10] I.S. Aranson, L.S. Tsimring, Patterns and collective behavior in granular media: theoretical concepts, Rev. Mod. Phys. 78 (2006) 641–692. [11] G. Metcalfe, T. Shinbrot, J.J. McCarthy, J.M. Ottino, Avalanche mixing of granular solids, Nature 374 (1995) 39–41. [12] C.Y. Wu, B.C. Hancock, A. Mills, A.C. Bentham, S.M. Best, J.A. Elliott, Numerical and experimental investigation of capping mechanisms during pharmaceutical tablet compation, Powder Technol. 181 (2008) 121–129. [13] R.P. Cogdill, Case for process analytical technology: Regulatory and industrial perspectives, in: S. Cox Gad (Ed.), Pharmaceutical Manufacturing Handbook: Regulations and Quality, John Wiley & Sons Inc, Hoboken, NJ, 2008, pp. 313– 352.

560

A.H. Blaesi, N. Saka / Chemical Engineering Journal 320 (2017) 549–560

[14] T.A. Bell, Challenges in the scale-up of particulate processes – an industrial perspective, Powder Technol. 150 (2005) 60–71. [15] S.M. Iveson, J.D. Litster, K. Hapgood, B.J. Ennis, Nucleation, growth and breakage phenomena in agitated wet granulation processes: a review, Powder Technol. 117 (2001) 3–39. [16] F.J. Muzzio, T. Shinbrot, B.J. Glasser, Powder technology in the pharmaceutical industry: the need to catch up fast, Powder Technol. 124 (2002) 1–7. [17] H.J. Federsel, Logistics of process R&D: transforming laboratory methods to manufacturing scale, Nat. Rev. Drug Discovery 2 (2003) 654–664. [18] S.M. Paul, D.S. Mytelka, C.T. Dunwiddie, C.C. Persinger, B.H. Munos, S.R. Lindborg, A.L. Schacht, How to improve R&D productivity: the pharmaceutical industry’s grand challenge, Nat. Rev. Drug Discovery 9 (2010) 203–214. [19] T.M. Squires, S.R. Quake, Microfluidics: fluid physics at the nanoliter scale, Rev. Mod. Phys. 77 (2005) 977–1026. [20] M. Prakash, N. Gershenfeld, Microfluidic bubble logic, Science 315 (2007) 832– 835. [21] Z. Tadmor, C.G. Gogos, Principles of Polymer Processing, second ed., Wiley, Hoboken, NJ, 2001. [22] A.H. Blaesi, N. Saka, Melt-processed polymeric cellular dosage forms for immediate drug release, J. Control. Release 220 (2015) 397–405. [23] A.H. Blaesi, N. Saka, On the exfoliating polymeric cellular dosage forms for immediate drug release, Eur. J. Pharm. Biopharm. 103 (2016) 210–218. [24] A.H. Blaesi, N. Saka, Determination of the mechanical properties of solid and cellular polymeric dosage forms, Int. J. Pharm. 509 (2016) 444–453. [25] A.H. Blaesi, The design and manufacture of immediate-release optimal solid dosage forms (PhD thesis), Department of Mechanical Engineering, Massachusetts Institute of Technology, 2014. [26] R. Mezzenga, P. Schurtenberger, A. Burbidge, M. Michel, Understanding foods as soft materials, Nat. Mater. 4 (2005) 729–740. [27] D. Klempner, K.C. Frisch, Handbook of Polymeric Foams and Foam Technology, Hanser Publishers, Munich, Germany, 1993. [28] J. Banhart, Manufacture, characterization and application of cellular metals and metal foams, Prog. Mater. Sci. 46 (2001) 559–632. [29] D. Weaire, S. Hutzler, The Physics of Foams, Oxford University Press, Oxford, UK, 1999. [30] A.E. Simone, L.J. Gibson, Aluminum foams produced by liquid-state processes, Acta Mater. 46 (1998) 3109–3123. [31] M. Hashimoto, P. Garstecki, G.M. Whitesides, Synthesis of composite emulsions and complex foams with the use of microfluidic flow-focusing devices, Small 3 (2007) 1792–1802. [32] J.P. Raven, P. Marmottant, Microfluidics with foams, Soft Matter 5 (2009) 3385–3388.

[33] P. Garstecki, I. Gitlin, W. DiLuzio, G.M. Whitesides, E. Kurnacheva, H.A. Stone, Formation of monodisperse bubbles in a microfluidic flow-focusing device, Appl. Phys. Lett. 85 (2004) 2649–2651. [34] A. Huerre, V. Miralles, M.C. Jullien, Bubbles and foams in microfluidics, Soft Matter 10 (2014) 6888–6902. [35] G.G. Stokes, On the effect of the internal friction of fluids on the motion of pendulums, Trans. Cambridge Philos. Soc. 9 (1856) 8–106. [36] J. Hadamard, Mouvement permanent lent d’une sphere liquid et visqueuse dans un liquide visqueux, Comptes Rendus 152 (1911) 1735–1738. [37] I. Cantat, S. Cohen-Addad, F. Elias, F. Graner, R. Hohler, O. Pitois, F. Rouyer, A. Saint-Jalmes, Foams: Structure and Dynamics, Oxford University Press, Oxford, UK, 2013. [38] G. Debregeas, P.G. De Gennes, F. Brochard-Wyart, The life and death of ‘‘bare” viscous bubbles, Science 279 (1998) 1704–1707. [39] J.D. Paulsen, R. Carmignani, A. Kannan, J.C. Burton, S.R. Nagel, Coalescence of bubbles and drops in an outer fluid, Nat. Commun. (2014). [40] P.G. De Gennes, Introductory Lecture. Mechanics of soft interfaces, Faraday Discuss. 104 (1996) 1–8. [41] H.S. Carslaw, J.C. Jaeger, Conduction of Heat in Solids, second ed., Oxford University Press, Oxford, UK, 1986. [42] E.E. Underwood, Quantitative Stereology, Addison-Wesley Series in Metallurgy and Materials, 1970. [43] S.I. Tomkeieff, Linear intercepts, areas and volumes, Nature 155 (1945) 24. [44] V.G. Levich, Physicochemical Hydrodynamics, Prentice-Hall, Englewood Cliffs, NJ, 1962. [45] L.R. Johnson (Ed.), Gastrointestinal Physiology, eighth ed., Mosby, Maryland Heights, MI, 2013. [46] S.A. Muller-Lissner, A.L. Blum, The effect of specific gravity and eating on gastric emptying of slow-release capsules, New Eng. J. Med. 304 (1981) 1365. [47] L. Whitehead, J.T. Fell, J.H. Collett, H.L. Sharma, A.-M. Smith, Floating dosage forms: an in vivo study demonstrating prolonged gastric retention, J. Control. Release 55 (1998) 3–12. [48] B.N. Singh, K.H. Kim, Floating drug delivery systems: an approach to oral controlled drug delivery via gastric retention, J. Control. Release 63 (2000) 235–259. [49] Y. Sato, Y. Kawashima, H. Takeuchi, H. Yamamoto, In vitro and in vivo evaluation of riboflavin-containing microballoons for a floating controlled drug delivery system in healthy humans, Int. J. Pharm. 275 (2004) 97–107. [50] B.H. Zimm, Dynamics of polymer molecules in dilute solution: viscoelasticity, flow, birefringence, and dielectric loss, J. Chem. Phys. 24 (1956) 269–278.