Fernando J. Muzzio, PhD, is a professor and the director of the pharmaceutical engineering program at the Department of Chemical and Biochemical Engineering, Rutgers University, 98 Brett Road, Piscataway, NJ 08854-8058, tel. 732.445.3357.

Fernando Muzzio, PhD

Loss-in-weight (LIW) or gravimetric feeding is the most commonly used continuous feeding method for pharmaceutical powders (1–4). With the commercial adoption of continuous manufacturing of solid oral pharmaceutical dosage forms in recent years, there has been increasing focus on LIW feeding for these applications, looking at both equipment design and utilization within the process. Feeding is one of the most important unit operations within a continuous manufacturing process, and while this can be said regardless of the type of process (e.g., wet granulation, roller compaction, direct compression), it is most critical for continuous direct compression processes. In a direct compression process, the material attributes of the formulation ingredients being fed can have a major effect on the downstream process and attributes of the final product. Materials that are challenging to feed may lead to higher variability of the final product, or in extreme cases, the need to pause operations and attend to the feeder. Therefore, it is important that LIW feeder performance is optimized during process development, particularly for poorly flowing materials.

Previous work has demonstrated that the operational range of LIW feeders depends on the feeder size, tooling, and the attributes of the material being fed. Cohesive materials that tend to agglomerate or entrap air within the powder bed will have a narrower range of operation as compared to well-flowing materials. Additionally, irregularity in bed properties can result in unstable flow rates, and erratic mass flow behavior can be transferred down to successive unit operations downstream of the feeders (5–12) and to the final product. Recent work has been performed to correlate LIW feeder performance to material properties (13–16). Li 

 identified a correlation between the feedfactor for Coperion K-Tron KT-20 LIW feeders to conditioned bulk density obtained from the Freeman FT4 Powder Rheometer (13), and Escotet-Espinoza 

 found a similar correlation between conditioned bulk density and the feedfactor for a GEA Compact LIW Feeder (14). Escotet-Espinoza 

 also found that using principal component analysis (PCA), a correlation can be established between bulk material properties (e.g., free-flowing particles versus soft compressed agglomerates) and how the material discharges from the feeder (14). Tahir 

 also found correlation between bulk material properties and feeder performance using partial-least square regression and PCA similarity grouping of materials (15). The recent focus of research has further highlighted the need to understand and optimize LIW feeder performance for the further development of pharmaceutical continuous manufacturing processes of solid dose products (16).

Even though LIW feeder systems may seem simple in principle, the behavior of feed materials can be complex (3). Given their importance in a continuous pharmaceutical solid dose manufacturing process, LIW feeders need to be tested and validated during process development to understand the performance capabilities of a given material–feeder combination. The ability to feed powder consistently and continuously is a critical requirement of a continuous manufacturing process (17,18). This requirement becomes especially relevant when feeding APIs, which cannot always be selected based on their suitability for the process such as an excipient potentially can.

In this study, a previously defined method for characterizing and classifying materials was used to identify a range of materials, from poorly flowing through well flowing, for experimentation (18). Within this classification system, group 1 materials are poorly flowing with a tendency to stick to surfaces, group 2 materials have better granular flow and a lesser tendency to stick to surfaces, and group 3 materials are considered to be well flowing. These materials were then characterized so that their material properties could be used to determine the initial set-up of a LIW feeder. The feeder set-up was then optimized for each material, including the screw, screen, and configuration for refill.

Materials from the three groups were chosen for this investigation. Microcrystalline cellulose (MCC) (Avicel 102, DuPont) was a group 1 material. Another MCC (Avicel 101, DuPont), croscarmellose sodium (AcDiSol, DuPont), and Prosolv 50 (JRS Pharma) represented materials from group 2. Powder grade acetaminophen (Mallinckrodt), magnesium stearate (Mallinckrodt), and Lactose 450m (DFE Pharma) represented materials from group 3. The particle sizes (d10, d50, and d90) for all seven materials were obtained from a laser diffraction particle sizer (LS 13-320, Beckman-Coulter) with a tornado dry powder module. Aerated and tapped density were recorded from a tapped density test performed according to 

 <616> standard procedure (19) using an automated tap density measurement instrument (Autotap, Quantachrome). The Hausner ratio was calculated from the ratio of tapped density to aerated density. A rheometer (FT4, Freeman) was used to characterize the powder packing and flow properties. Compressibility, permeability, flow energy, and shear cell tests were performed according to the Freeman FT4 protocol. The conditioned bulk density (cBD), compressibility percent at 15 kPa normal stress (CP% at 15 kPa), and density at 15 kPa were obtained from the FT4 compressibility test. Permeability at 15 kPa was obtained from the FT4 permeability test. Basic flow energy (BFE) was obtained from the FT4 stability and variable flow rate test. Cohesion, unconfined yield strength (UYS), major principal stress (MPS), angle of internal friction (AIF), and flow function coefficient (FFc) were obtained from the FT4 3-kPa shear cell test. The material characterization values appear in 

The Compact Feeder (GEA) is a twin-screw, LIW feeder with an integrated refill system that is designed specifically for use in pharmaceutical applications. The small footprint design allows for a high number of units to be placed around the inlet of subsequent unit operations without the need for sloped hoppers or multiple feeding stages. The feeder is made up of two main sections: a swappable top section, which can be removed and replaced, consisting of the powder pump and hopper and a bottom section containing the load cell and servo motor used to drive the agitator and screws.

The feeder uses a 2.6-L flat-bottomed hopper with a four-blade agitator used to fill the screws. The agitator can either be flat or fitted with vertical bridge-breaker elements to prevent cohesive materials from bridging in the hopper. Typically, the feeder is expected to deliver mass flow rates accurately in the range of hundreds of grams per hour to tens of kilograms per hour. Delivery for different ranges is achieved by using change parts, such as different servomotor gearboxes (63:1, 235:1, and 455:1) and screw designs (10-mm and 20-mm concave).

The compact feeder uses an integrated refill system used to accurately and repeatably refill the feeder hopper with material. The refill system consists of a ball valve and hopper (either configured with a vacuum system or gravity feed). The ball valve volume can be changed by using inserts with final volumes of 1.6, 1.2, 0.8, and 0.4-L achievable. The refill volume is typically chosen based on the feedfactor profile for a given material. The feedfactor is the maximum amount of powder which can fit within the screw flight per unit time.

 The feeder was mechanically configured in its optimum set-up for each material and then filled to its maximum load with an expected volume of powder of 2 L after refill. The maximum load is a function of the expected refill volume and the bulk density as shown in 

In this work, the feeder was primed by running the screws until the mass flow rate of material leaving the screws was at a steady state. Steady state was confirmed by performing three static feed factors tests at a screw speed close to the screw speed required for the mass flow rate set-point and checking that the measured feed factor was reproducible . The feeder was then refilled back to its maximum load and then started under gravimetric control. After the screws ran empty, the feeder was stopped. The whole procedure was repeated in triplicate. The data were recorded directly from the programmable logic controller using software (TwinCat Scopeview, Beckhoff Automation).

 The refill system was configured in its optimum set-up for each material. The system was then filled to the specified maximum fill level. In order to test the reproducibility of the refill system, 10 refills were conducted, and the mass of each refill was recorded using an external balance. When the level fell below the specified minimum fill volume, the system was refilled manually back to the maximum fill volume. The mean and variability for the 10 refills were calculated for each material.

Different material classes with optimized feeders

Prior to use, LIW feeders need to be tested and validated to understand the performance capabilities of a given material–feeder combination. A feeder is first calibrated to obtain an initial feed factor, and then it is run, without interferences, for a given period of time. The feeder data are then analyzed to quantify the accuracy of mass flow rate dispensed by the feeder.

In this experiment, the standard hopper and standard impeller were used for the materials except magnesium stearate and acetaminophen. The baffled hopper was used with the asymmetric impeller for the magnesium stearate to break up the powder flow within the hopper. For the group 1 materials, magnesium stearate, lactose 450m, and powder-grade acetaminophen, a screen was used at the exit of the feeder screws. The screen creates a slight backpressure to improve uniformity of powder filling the screws and reduce any carving effects from the materials leaving the screws. A 2-mm mesh screen was used for magnesium stearate, whereas a 8-mm mesh screen was used for the lactose 450m and the powder-grade acetaminophen. Additionally, the gearbox, which dictates the maximum screw speed from a gear reduction of the motor speed, was varied based on the desired target set point. Higher gear reduction ratio is required for lower set points. Each set point was chosen based on an assumed weight percentage for each of the materials that was likely to occur in a given formulation with a total operational throughput of 50 kg/hr. For instance, the set point for powder-grade acetaminophen was calculated to be 5 kg/hr, assuming it would be 10 w/w% in a representative formulation. Therefore, a 63:1 ratio gearbox number was used for Avicel 101, Avicel 102, power-grade acetaminophen, and Prosolv 50; a 235:1 ratio gearbox number was used for lactose 450m; and a 455:1 ratio gearbox was used for AcDiSol and magnesium stearate. Concave screws with a 20-mm pitch were used for Avicel 101, Avicel 102, AcDiSol, Prosolv 50, Lactose 450, and powder-grade acetaminophen. Since the target throughput for magnesium stearate was very low, 10-mm pitch concave screws were used. The 10-mm pitch screws have a lower volumetric throughput than the 20-mm screws, resulting in better control of throughput at low set points. The conditions used for the feeder performance tests are shown in 

. All experiments were performed in a controlled environment with an average relative humidity of 33% and temperature of 19.7 °C.

For each material, the feedfactor calibration was performed three times, prior to each of the three runs, and the average and relative standard deviation (RSD) were calculated from the nine values. The calibration was performed to obtain the maximum throughput for the specific feeder, screw, and powder. The feeder was then filled and run until it was empty. The GEA control software can also calculate and record the feedfactor in bins, separated by every 10% of hopper fill. All feedfactor data appear in 

. For all materials, the feedfactor decreases as the hopper fill diminishes; the decrease is more pronounced for different types of materials, due to the flowability and the head pressure of the powder sitting above the screws. Prosolv 50 and AcDiSol maintained a consistent feedfactor throughout the run, with a feedfactor above 90% of the initial feedfactor. Avicel 102 and Avicel 101 maintained a feedfactor above 90% of the initial feedfactor, down to 10% fill; however, noticeable change occurred in the final 10% hopper fill. The group 1 materials showed more significant changes in feedfactor than both groups 2 and 3. Lactose 450m and powder-grade acetaminophen maintained 90% of the initial feedfactor until 40% hopper fill. Below 40%, both feedfactors decreased significantly, with Lactose 450m ending with a feedfactor that was 63% of the initial feedfactor, and powder-grade acetaminophen ending with a feedfactor that was 44% of the initial feedfactor. Magnesium stearate showed the most significant decrease in feedfactor, dropping below 90% initial feedfactor at 50% hopper fill and ending with a feedfactor that was 58% of the initial feedfactor. The changes in feedfactor assist the user with determining when a refill should occur, as the refill should be performed prior to significant changes in feedfactor.

From the one-second interval feeder data of the complete dispensing of each of the powders, the average, standard deviation and RSD were then calculated for the mass flow and screw speed. The theoretical weight percent, set point, average mass flow rate, mass flow rate RSD, average screw speed, and the screw speed RSD are presented in 

 show that the feeder accurately dispenses the mass flow on target with high consistency. The feeder mass flow RSD values for all materials were less than 4%, whether the material was from group 1, 2, or 3. The most change observed was for the feeder screw speed when dispensing magnesium stearate. This value shows that there is a change in bulk density of the magnesium stearate because the screws had to change speed more frequently than for the other materials. However, even with the changing in screw speed, the feeder was able to accurately meet the target of 0.1 g/s with a mass flow RSD below 2%.

The refill system was set up for an optimal discharge for each material. Based on past experimental experience with acetaminophen, the straight hopper was used for powder-grade acetaminophen because it results in less bridging. For the remaining six materials, the Volkman hopper type was used. Each hopper type had a vibration unit attached, which was controlled by an air supply with 2 Bar of pressure and run from 3 to 5 seconds after each refill.

To characterize the consistency of powder refill, each material was dispensed 10 times, and the mass recorded for each instance was used to calculate an average mass and RSD (see 

). For the majority of the materials tested, the RSD was below 1%. Magnesium stearate and powder-grade acetaminophen had the highest RSD values of 2.7 and 5.5%, respectively, which was to be expected due to the more cohesive nature of these materials. However, in our experience, these values of mass flow rate RSD are small to moderate and easily handled by typical continuous blenders.

For successful and robust process development of sold-dose products by continuous pharmaceutical manufacturing, LIW feeders need to be tested and validated to understand the performance capabilities of a given material–feeder combination. The authors have demonstrated a robust process for testing individual materials in LIW feeders to obtain the ideal configuration and operational parameters that will provide the minimal RSD and refill perturbation at the exit of the feeder. A range of materials were selected, including challenging compounds that are not well flowing. Under the optimized conditions, nearly all of the materials were fed with acceptable RSD values. 

Because the unit operations downstream of the LIW feeders within the process will influence the RSD of the final product, it remains important to consider the LIW feeder operational RSD as well as the refill perturbation of each ingredient in combination with these downstream unit operations. Continuous blending will typically follow feeding within one or two unit operations and can serve to further mitigate feed-rate variations and disturbances (20,21). This mitigation is particularly important for poorly flowing group 1 materials within this study, such as powder-grade acetaminophen and magnesium stearate.

Overall, the proper set-up and optimization of a given feeder–material combination will ensure that the most robust continuous pharmaceutical sold-dose manufacturing process is developed. The proper time and effort should be put into examining feeder performance as early as possible during process development. While it may be difficult to do this examination for APIs based on available quantity and associated cost, it should be feasible to test excipients early on and include feedability as one of the criteria for their selection for a continuous process.

J.G. Osorio et al., “Continuous Powder Mixing,” in 

, P.J.Cullen, R.J. Romãnach, N. Abatzoglou, C.D. Rielly, Eds. (John Wiley & Sons, Ltd. Chichester, UK, 2015) pp 59-67.

D.H. Wilson and K.W. Bullivant, Loss-in-weight gravimetric feeder. Google Patents: 1986.

W. Engisch, M. Ierapetritou, and F. Muzzio, “Hopper Refill of Loss-in-weight Feeding Equipment,” 

Proceedings of the 2010 AIChE Annual Meeting

A.U.Vanarase, J.G. Osorio, and F.J. Muzzio, 

T. Li et al., Identifying a Loss-in-Weight Feeder Design Space Based on Performance and Material Properties. 

F. Tahir et al., Development of feed factor prediction models for loss-in-weight powder feeders. 

 General Chapter <616> Bulk Density and Tapped Density of Powders (2015).

S. Oka et al., “Design of an Integrated Continuous Manufacturing System” in 

Continuous Manufacturing of Pharmaceuticals, Advances in Pharmaceutical Technology

James Scicolone is research professor, and Douglas B. Hausner is associate director at C-SOPS, both at Rutgers, the State University of New Jersey; John Palmer is process modeling specialist at GEA; Andrew Birkmire is process development manager at GEA; James Holman is Technology Management director, Pharma Solids, at GEA. Fernando J. Muzzio* is director at C-SOPS and a Distinguished Professor at the Department of Chemical and Biochemical Engineering, Rutgers, The State University of New Jersey, Piscataway, NJ 08854, USA, Tel.: +1-848-445-3357; fax: +1-848-445-6758, muzzio@soe.rutgers.edu.

*To whom all correspondence should be addressed.

 Supplement: APIs, Excipients, and Manufacturing
September 2020
Pages: 24–28

When referring to this article, please cite it as J. Scicolone et al., “Optimizing Loss-in-Weight Feeding of Poorly Flowing Materials,” 

Articles

Prior to use in a continuous manufacturing system for oral solid dosage forms, loss-in-weight feeders need to be tested and validated to understand the performance capabilities of a given material–feeder combination. In this article, the proper strategy for set-up and optimization of a loss-in-weight feeder is demonstrated for a range of materials. The optimized set-up of the feeders was demonstrated to provide suitable performance for even the most challenging, poorly flowing materials.

Optimizing Loss-in-Weight Feeding of Poorly Flowing Materials

Granular mixing can be generally classified either as macro-mixing or micro-mixing, based on the scale of mixing. Macro-mixing is defined as the macro-scale transport of particles or particle assemblies from one spatial coordinate to the other. Macro-mixing is responsible for most of the reduction in variance in unmixed powder streams. Micro-mixing is defined as mixing at the scale of a single particle. Micro-mixing is always associated with a change in the properties of individual particles (1).

	The role of micro-mixing becomes particularly important when it involves shear-sensitive ingredients, such as magnesium stearate or other lubricants, which are added to the formulation to reduce sticking in the tablet press. Magnesium stearate also improves the packing efficiency of coated materials (i.e., increased bulk density) and improves flow properties (2).

	Lubricants, however, can also cause problems in the formulation. Magnesium stearate, generally present as thin nanoparticles, is known to coat larger, shear insensitive excipients and active ingredients (3). This coating reduces the bonding ability of particles, and consequently causes a loss in the tabletability of the material (2). Magnesium stearate is also hydrophobic in nature, so coating active ingredient particles with magnesium stearate increases their hydrophobicity, resulting in a lower dissolution rate for the active ingredient (4, 5). 

	The coating extent of the shear-sensitive material and, consequently, its associated effects, are directly proportional to the total amount of strain experienced by the blend. The greater the strain, the more pronounced the extent of coating and, consequently, the more pronounced the associated effects (2–5).

	Close attention is thus paid to the extent of lubricant mixing in traditional batch manufacturing to ensure that the lubricant is not subjected to excessive strain levels during the blending operation. Typically, shear-sensitive ingredients are added in a second stage, after the mixing of other ingredients, to minimize the strain they experience. The bulk ingredients are blended for a few hundred revolutions in the blender before the lubricant is added to the system, and the lubricant is generally mixed for a few tens of revolutions.

	The design and operation of a continuous paddle blender differ significantly from that of a batch tumble blender. Moreover, the material is in the continuous blender for a much shorter time. In a continuous blending process, the material spends typically between a few seconds to a few minutes in the blender and experiences anywhere between 50 to a few hundred impeller revolutions (6).

	The number of impeller passes or revolutions is defined as follows:

	Number of impeller passes = τ (min) * impeller speed (rpm)

	where τ is the mean residence time of the material in the blender.

	How do strain levels differ in the continuous blender compared to the batch blender? The answer to this question will determine whether or not shear-sensitive ingredients must be added at a second stage.

	Comparison of strain energy in batch and continuous processes

	Razavi et al. (7) performed a study to quantify the amount of strain that the material experiences in a full-scale continuous process, and compared that to the strain resulting from a laboratory-scale batch blending operation. The researchers quantified the difference in the speed of sound in compressed tablets due to density differences between materials subjected to the variable degrees of strain. The authors found that the total amount of strain experienced by the material in a full-scale continuous process was less than that in a laboratory-scale batch operation.

	In this experiment, the continuous process was operated at a total mass flow-rate of 20 kg/hr. The continuous blender was run at 200 rpm, while the feed-frame was operated at 25 rpm. The material thus experienced strain in both the continuous blender and the feed-frame (7). The same material was blended in a laboratory-scale V-blender, with magnesium stearate, for 50 revolutions and compressed in a tablet press emulator. The tablet press emulator does not use a feed-frame to feed powder in compression dies. The powder processed by the batch operation is thus only subjected to strain during the batch blending operation. Despite the presence of the additional shear provided by the feed-frame in the continuous process, it was found that, overall, the material experienced less strain in the continuous process than it did in the batch blending operation (7).

	These results suggest that the amount of strain experienced by the material while it traverses a full-scale continuous blender is negligible compared to that experienced by the material in a laboratory-scale batch device. There are two reasons for this effect:

	Material in the continuous process spends a shorter amount of time in the equipment than it does in the batch process, and thus the total time of exposure to the shearing environment is much lower.

	The total mass of material in the continuous process, at any time, is significantly lower than it is in the batch process. A lower net mass results in lower consolidation and frictional stresses, leading to a lower overall degree of strain.

	Thus, the typical strain level in industrial-scale horizontal convective blenders is much lower than it is in benchtop batch devices, and most certainly lower than it is in full-scale batch blenders.

	Results also suggest that there needn’t be as much concern about potential over-lubrication in continuous blending operations, because they are gentler, so two-staged ingredient addition may not be necessary.

	It is also interesting to note that, in a continuous blender, the ratio of free volume in the system (i.e., volume not swept by the rotating blades) to the total blender volume is much smaller than it is in a batch blender (where the entire volume of the blender is considered free).

	The authors hypothesize that, in continuous blenders, a small amount of material is more uniformly subjected to the convective action of the blades, resulting in a more uniform nature of strain energy compared to what is found in batch systems.

	To quantify the effect of strain rate on blend properties, experiments were performed in which 300 g of material were subjected to 50, 100, 500, and 1000 revolutions with 1% magnesium stearate (w/w), in a modified continuous blender, albeit at different strain rates. The lubricity of the blend was quantified by measuring its contact angle with water, as described by Denesuk et al. (8).

	The blend consisted of 65% anhydrous Compap L acetaminophen and 34% Avicel PH-101, by weight. It was added to a Gericke GCM-250 continuous blender whose blades were adjusted such that they do not promote any axial motion of the material. The exit of the blender was sealed so that no material could leave the blender tube. Strain rates of 50, 150, and 250 rpm were investigated.

 shows the effect of the strain rate addition on the lubricity or hydrophobicity of the final blend. Blend lubricity is quantified as the cosine of the blend contact angle with water; a smaller cosine value indicates greater hydrophobicity. It can be observed that the blend hydrophobicity is not dependent on the rate at which strain is applied to the material, but only dependent on the total strain it experiences.

	The total amount of strain experienced by the in-process material in a batch tumbling blender is greater than that in a continuous paddle blender, under typical operating conditions. The higher strain experienced in batch processing necessitates the addition of shear-sensitive materials, such as magnesium stearate, at a later stage in the blending operation, resulting in two stages of ingredient addition. Two-stage addition may not be necessary in gentler continuous operations. The material experiences a more uniform degree of shear in a continuous blender compared to a batch blender.

	Lastly, the rate of strain addition (i.e., rotation speed) does not have an impact on the lubricity of the final blend. Lubricity depends only on the total amount of strain the blend experiences in either batch or continuous processing.

	Sarang Oka, Sara Moghtadernejad, Zhanjie Liu, Douglas Hausner and Fernando Muzzio* are all from the Department of Chemical & Biochemical Engineering at Rutgers, The State University of New Jersey, fjmuzzio@yahoo.com.

	*To whom all correspondence should be addressed.

	When referring to this article, please cite it as S. Oka, et al., "Lubrication in Continuous Tubular Powder Blenders," 

Shear-sensitive ingredients in a tableting formulation experience different conditions in a batch tumble blender than they would in a continuous paddle blender.

Lubrication in Continuous Tubular Powder Blenders

	The pharmaceutical industry has recognized the value of implementing a systematic approach to drug product development where quality is built into the product and process. The FDA initiative on quality by design (QbD) promotes the design of the product and manufacturing process using principles of chemistry, engineering, material science, and quality assurance to ensure acceptable and reproducible product quality and performance throughout a product’s lifecycle. Product quality is achieved through design of robust processes that are controlled and optimized using product and process knowledge (1, 2). In the QbD paradigm, mathematical models can potentially be used at every stage of drug product development and manufacturing (3). Modeling can help establish a predictive framework using experimental data and scientific principles to create mathematical representations of the system. Predictive models aid process design by evaluating the impact that operations, equipment, and inputs have on product attributes 

. Predictive models also provide a framework for risk assessment, process control, and optimization, where accurate predictions of the system are required (4).
	


	In this article, the authors focus on the use of flowsheet modeling, a process system engineering tool for the design, development, and integration of pharmaceutical processes. More specifically, they discuss application of flowsheet models for process risk assessment and design of control strategies.
	
	

Process development paradigm and novel methods

 Major components of the QbD approach to development include assessment of process risk and establishment of a design space. Risk is defined as “the combination of the probability of occurrence of harm and the severity of that harm” (5). Risk assessment is a science-based process used in quality risk management to identify and rank parameters (e.g., process, equipment, input materials) with potential to have an impact on product quality. Once the significant parameters are identified, they can be further studied to enhance process understanding, which could lead to the establishment of a design space (5). Design space is defined as “the multidimensional combination and interaction of input variables (e.g., material attributes) and process parameters that have been demonstrated to provide assurance of quality” (6). In general, a good understanding of potential risks when defining a design space can potentially reduce process uncertainty and increase process sustainability. Such knowledge can be used to establish a quantitative framework to measure how process failures impact product quality and determine a risk mitigation approach to reduce process-derived patient hazards.
	
	

Pharmaceutical process systems engineering. 

As the pharmaceutical industry modernizes its manufacturing practices and increasingly incorporates more efficient processing approaches such as continuous manufacturing, it is important to assess the process design elements that affect product quality for these emerging pharmaceutical manufacturing approaches. In designing continuous flow systems, while the analysis and optimization of individual process equipment remain important, the primary objective is to identify and evaluate design elements that pose a potential risk to product quality for the fully integrated system, leading to effective risk management. It is also important to consider the multivariate nature of such systems in process design (7). Within this context, process systems engineering (PSE) tools have been implemented with the goal of facilitating effective and efficient process design. PSE is the application of computer-aided systematic science and engineering approaches to the modeling, design, analysis, control, optimization, and operation of process systems.  
	
	PSE tools can provide insight to pharmaceutical development as a means of evaluating processes 

 (i.e., using a computer). Mathematical models embedded in the PSE tools can potentially supplement expensive and time-consuming 

 experimentation throughout process development (8). Furthermore, predictive mathematical models, once validated, can be used to study the process dynamics in detail, to help achieve high process efficiency, and to attain the desired product quality. These models can facilitate the design of processes where consistent product quality is achieved at every step of manufacturing within the framework of QbD and process analytical technology (PAT) (9). 

 summarizes the PSE tools (10) and their potential utility in pharmaceutical process development.


	In an integrated process, individual pieces of equipment (i.e., unit operations) are connected in series. In such a process, a train of multiple units, one after the next, is connected with piping to perform powder-to-tablet manufacturing sequentially, without isolation of intermediates. The output of a preceding unit becomes the input of a subsequent one, with material continuously flowing between them. Mathematically, process integration follows the same logic. Individual equipment models are combined by taking the results from a preceding model and using it as the inputs of a subsequent one. The integrated process models are called flowsheet models, as the flow of information between the unit models resembles the flow of material(s) between unit operations. 

 shows an example of a flowsheet model for a continuous direct compression system developed at the Engineering Research Center for Structured Organic Particulate Systems. Several flowsheet modeling software packagies (e.g., ASPEN Plus, ChemCAD, gPROMS) have been effectively demonstrated for predictive modeling and design of fluid-based processes and are already widely used across the chemical and petrochemical industries (11). Flowsheet models have been recently developed for continuous pharmaceutical manufacturing schemes and have been shown to effectively capture integrated process dynamics (3, 12).

Application of flowsheet models for process development


	Flowsheet models, such as the one shown in 

, can be used as a tool for process design, optimization (13, 14), risk assessment (15), control strategy analysis (16, 17), and monitoring of a continuous pharmaceutical process.

 From a design standpoint, flowsheet models can potentially provide great value as a tool to evaluate equipment configurations and manufacturing schemes 

 at a much lower cost than the equivalent experimental investigation. Using flowsheet models, the major routes for drug product manufacturing can be evaluated, and challenges with process scale-up can be anticipated and resolved (9, 12). Flowsheet models for continuous direct compression, dry granulation, and wet granulation, shown in 

, have already been developed and demonstrated for their potential use in process design and optimization (15). Such multiplicity of models leads to a flexible flowsheet modeling platform that can streamline the design process compared to the relatively iterative and expensive, experimental and empirical-based process design approach.  

	Using flowsheet models, process engineers can study the system 

 and obtain information about process conditions that would lie outside the range of acceptable outcomes, and, thus, narrow the scope of subsequent experimental investigations. This information would naturally lead to more focused development efforts, which could eventually lead  to a higher level  of process understanding and establishment of the design space. The reduction in experimentation due to 

 evaluation reduces materials usage (e.g., API, excipients), waste, development time, cost, and personnel exposure, while it can potentially improve product quality. During process optimization, flowsheet modeling can potentially be used to determine optimal values for high-risk process parameters.

 Understanding the impact upstream operations have on the process is one of the critical aspects for assessing risk to product quality.  Flowsheet models can aid risk assessment through the use of sensitivity analysis. This tool elucidates the impact different process variables and parameters have on the overall system performance and product quality. Case studies considering disturbances and their impact on the process performance using sensitivity analysis have been previously discussed (12, 15, 18). A simple example of risk assessment relates to whether a continuous blender can sufficiently reduce the disturbances introduced by the feeders. In the example illustrated in 

, flowsheet models are used to study the scenario addressing the question: “Will the blender keep the concentration within specified limits if one of the feeders overfeeds material for a short time?” The scenario was simulated by modeling the output blender concentration after the feed of API surges, leading to a change in concentration of the material entering the comil (i.e., the conical mill between the feeder and blender) for 10 seconds. For comparison, two impeller configurations (“all blades forward” and “one-third blades back”) for the blender were simulated using pre-existing blender models to demonstrate the application of flowsheet models in the risk assessment.

	The simulation illustrates the impact of the feeder and blender on the API concentration in the blend as a function of time (

). Before the comil, the output of the feeder surged,  causing a 10% increase in the concentration of API, which was well above the (arbitrarily selected) permissible 2% upper specification represented with the green dotted line. After the blender, only one of the blade configurations (i.e., one-third configuration) was capable of mitigating the perturbation sufficiently to bring the output API concentration back within the specified limit. The operational space where the blender, given a blade configuration, would be able to dampen similar process disturbances is shown in 

. This type of analysis allows process engineers to understand the risk associated with each process parameter (e.g., blade configuration). Edges of failure, process sensitivity, and flexibility can also be studied using a similar approach (15).

. In the pharmaceutical industry, it is imperative to assure consistent manufacture of the desired product quality. To achieve this goal, material properties and process parameters need to be maintained within predetermined ranges. Deviations from the established ranges increase the risk of producing poor quality products. A process control system can be implemented to automatically adjust the process in response to disturbances to ensure that the quality attributes consistently conform to the established ranges. The design and implementation of an efficient control system is an interactive procedure that involves identification of critical controlled variables; coupling of the controlled variables with suitable actuators (manipulated variables); selection of monitoring tools; selection of a process-control approach followed by controller tuning; model-based, closed-loop performance assessment; and finally, implementation at the manufacturing plant through the available sensing and control platform integrated with control interfaces (19, 20). Integrated flowsheet models can facilitate the design, implementation, and tuning of process-control systems (21). 

 identification and evaluation of process variability sources (e.g., blending) aid the selection and location of appropriate monitoring and control methods.

	Using the same scenario modeled in the previous section, a case study was conducted to identify a suitable  process control approach for mitigating a deviation in API feed rate. Using the flowsheet models, the process parameters for which adjustment can help mitigate the process deviation (i.e., potential risk) can be identified. The analysis conducted in this case study indicates that reducing the flow rate and, counter-intuitively, the impeller rotational speed can increase the mixing ability of the system. The increased mixing is able to mitigate the simulated disturbance, bringing  the product back within specification as shown in 

. This analysis therefore identifies blade speed and flow rate as potential control variables.

	Flowsheet models can also provide knowledge regarding the process dynamics through estimation of the residence time distribution to determine the adequate in-process measurement frequency for process monitoring and control application. Furthermore, controller tuning and testing can be evaluated using flowsheet simulations prior to being implemented in the manufacturing plant.

Pharmaceutical process development using flowsheet models.

 Given the potential benefits of flowsheet models, a methodology for their use in design, optimization, control, and future process assessment is proposed in 

	The process begins with the characterization of materials, to assess whether the mathematical models for individual equipment are capable of predicting the behavior of the ingredients and intermediates (e.g., blends) in the process. If the material properties are not within a unit’s studied range, an experimental evaluation should then be performed to characterize the powder behavior in the unit, and the resulting information should be incorporated into the model.
	
	Once individual models have been tested, 

 design spaces for these units can be created. Combination of the design spaces of individual unit operations using flowsheet models can then be used to propose a set of manufacturing process conditions that best suits the system. Subsequently, the process design space can be created and optimized. Target operating conditions and process control approaches can then be formulated. Once the process conditions and process control approaches are selected, it is recommended to experimentally verify model predictions. The experimental data collected can be used to further tune and improve model predictions as appropriate. After confirmation of the model’s accuracy, risk assessment and process sensitivity are performed to ensure process robustness. The flowsheet models created during process design and development can then be applied to evaluate strategies for process monitoring and control.

Potential regulatory application of flowsheet models


	Risk-based regulatory approaches increase the efficiency and effectiveness of review and inspection activities by directing resources to focus more on the assessment of high-risk areas for products and processes. Quality risk assessments require product and process knowledge to evaluate potential sources of harm (e.g., failure modes of a process and sources of variability) and probability of detection of problems (5).

	As discussed, continuous processing, compared with batch processing, offers a greater opportunity to develop and better use process models to gain process knowledge, because governing equations can generally be derived based on physical and chemical principles. Integrated process models can support a quantitative initial risk assessment through sensitivity analysis by examining the relative magnitude of the impact of variation in process parameters and/or material attributes on quality attributes.

	As an example, in a published study, the most significant sources of variability for a particular continuous tablet-manufacturing process were found to be the mean particle size and bulk density of the raw materials (12). This type of analysis can then be used to focus the regulatory assessment on whether a proposed control strategy is appropriate for mitigating the identified high-risk areas (e.g., variability in raw material particle size and density). Sensitivity analysis can also be used to guide the evaluation of advanced process control approaches employed by identifying control and manipulated variables that should be incorporated into the control strategy (22). The process control strategy can be further assessed through the use of case studies, such as examining the processes’ ability to mitigate the impact of disturbances (e.g., feeder refills). These case studies can aid regulatory review and inspection activities by identifying types of disturbances that may have a significant impact on product quality.

	The level of detail required for describing a model in a regulatory submission depends on the impact its implementation has in assuring the final product quality. Integrated process models used to support process development and initial risk assessments by industry may be considered “low-impact models” because they are not used to assure the final product quality. Documentation for low-impact models should include a discussion of how the models were used to make decisions during process development. Integrated process models used for operating space determination or process control design may be classified as “medium-impact models.” The International Conference on Harmonization Quality Implementation Working Group Points to Consider gives recommendations on documenting higher impact models (23).    

	The use of models by FDA to support review is not new (e.g., pharmacokinetic and drug adsorption modeling to support regulatory decisions with regards to bioequivalence and quantitative structure activity relationships models to risk-assess the potential toxicity of impurities). It is recognized that, although there have been significant advancements in the modeling and simulation of continuous pharmaceutical manufacturing processes, the technology is not yet sufficiently mature to aid regulatory assessment. To address this gap, FDA has sponsored two grants for the development of process simulation and model tools for the continuous manufacturing of solid oral dosage forms to facilitate the risk assessment of manufacturing process and control strategies. The goal is for these projects to lead to a collaborative platform for process simulation that builds on the process modeling knowledge developed in academia, industry, and regulatory bodies. The use of common risk assessment approaches and tools can facilitate the communication of risk mitigation approaches between industry and regulatory bodies.

design through flowsheet models has had major impact in other industries, and it is reasonable to expect that it will also be transformative for pharmaceutical manufacturing. It could have major impact on the design process through the implementation of better, less wasteful, and smarter process design. It could also facilitate process optimization and process control, while minimizing development time. From a regulatory perspective, use of predictive models can enable a quantitative risk assessment, facilitating the quality assessment of manufacturing processes. It can also support the evaluation of control strategies by demonstrating system capabilities to handle multiple sources of variability, either individually or in combination.

	While the technology is relatively new to pharmaceutical manufacturing, its potential is evident. The authors would welcome an active dialogue on how to accelerate the development of such capabilities across the spectrum of relevant processes.


	This work is supported by the National Science Foundation Engineering Research Center on Structured Organic Particulate Systems, through Grant NSF-ECC 0540855, and FDA.
	
	All figures are courtesy of the authors. 

 25(4) 781-791 (2008).
	3. T. Kourti et al., 

 34(6) 1-21 (2014).
	4. A. Rogers, A. Hashemi, and M. Ierapetritou, 

Guidance for Industry: Q9 Quality Risk Management

Guidance for Industry: Q8(R2) Pharmaceutical Development

 (Rockville, MD, 2009).
	7. A. Rogers and M. Ierapetritou, “Challenges and Opportunities in Pharmaceutical Manufacturing Modeling and Optimization,” in 

, J.D.S. Mario, R. Eden, and P.T. Gavin, Eds. (Elsevier, 2014), pp. 144-149.
	8. K.V. Gernaey and R. Gani, 

, 65(21) 5757-5769 (2010).
	9. K.V. Gernaey, A.E. Cervera-Padrell, and J.M. Woodley, 

 42(0)15-29 (2012).
	10. F. Boukouvala, “Integrated Simulation and Optimization of Continuous Pharmaceutical Manufacturing,” in 

 (Rutgers, The State University of New Jersey, New Brunswick, NJ, 2013).
	11. M.R. Othman et al., Ind.& Eng.Chem. Res. 49(17) 7870-7881 (2010).
	12. F. Boukouvala et al.,  

  42(0) 30-47 (2012).
	13. F. Boukouvala and M. Ierapetritou, 

 8(2) 131-145 (2013).
	14. M. Sen, et al.,  

 102, 56-66 (2013).
	15. A.J. Rogers, C. Inamdar, and M.G. Ierapetritou, 

 53 (13) 5128-5147 (2013).
	16. R. Singh et al., 

 66(0) 186-200  (2014).
	17. R. Singh et al., 

 473(1-2) 38-54 (2014).
	18. M. Sen et al., 

 80(0) 349-360 (2012).
	19. R. Singh, M. Ierapetritou, and R. Ramachandran, 

Eur. J. Pharmaceutics and Biopharmaceutics

 85(3, Part B) 1164-1182 (2013).
	20. R. Singh, M. Ierapetritou, and R. Ramachandran, “Hybrid advanced control of flexible multipurpose continuous tablet manufacturing process via direct compaction,” in 

, K. Andrzej and T. Ilkka, Eds. (Elsevier, 2013), pp. 757-762.
	21. R. Singh et al., 

 9(1) 16-37 (2014).
	22. Lakerveld, R., et al., 

 59(10) 3671-3685 (2013).
	23. ICH Quality Implementation Working Group, 


	M. Sebastian Escotet-Espinoza is graduate research assistant, Ravendra Singh is assistant research professor, and Maitraye Sen is graduate research assistant, all of the Chemical and Biochemical Engineering Department at Rutgers, The State University of New Jersey, Piscataway, NJ 08854. Thomas O’Connor is chemical engineer, science staff, Sau Lee is acting associate director for science, and Sharmista Chatterjee is acting branch I chief/division 1 of process assessment/Office of Process and Facilities, all from the Office of Pharmaceutical Quality, Center for Drug Evaluation and Research, FDA, 10903 New Hampshire Ave., Silver Spring, MD 20993. Rohit Ramachandran is assistant professor, Marianthi Ierapetritou is professor and chair, and Fernando J. Muzzio* is distinguished professor, all of the Rutgers Chemical and Biochemical Engineering Department. Dr. Muzzio is also director of the Engineering Research Center for Structured Organic Particulate Systems (ERC-SOPS).

	*To whom all correspondence should be addressed.  

 When referring to this article, please cite it as M.S. Escotet-Espinoza, “Flowsheet Models Modernize Pharmaceutical Manufacturing Design and Risk Assessment,” 

In-silico design facilitates process optimization and evaluation of process control strategies.

Flowsheet Models Modernize Pharmaceutical Manufacturing Design and Risk Assessment

Currently, there is a high level of interest in the pharmaceutical industry in continuous-manufacturing strategies, integrated with online-monitoring tools, that are designed, optimized, and controlled using advanced, model-predictive systems. These strategies can accelerate the full implementation of the quality-by-design (QbD) paradigm for the next generation of pharmaceutical products. In addition to its flexibility and time- and cost-saving features, continuous manufacturing is intrinsically steady and therefore easily amenable to model predictive design, optimization, and control methods. These methods have proven to be effective approaches to improve operational efficiency and have been widely used in various process industries. Excitingly, in the pharmaceutical industry, the application of the model-predictive design, optimization, and control is virgin territory, wide open to researchers and technology providers. 

Recently, the pharmaceutical industry as well as FDA have recognized the need for modernizing pharmaceutical manufacturing and have launched an initiative for enhancing process understanding through QbD and process analytical technology (PAT) tools (1-4). Major goals of these efforts include development of the scientific mechanistic understanding of a wide range of processes; harmonization of processes and equipment; development of technologies to perform online measurements of critical material properties during processing; performance of real-time control and optimization; minimization of the need for empirical experimentation and, finally, exploration of process flexibility or design space (5). In many cases, these goals can be achieved effectively and efficiently by the joint application of designed experiments and modeling tools such as discrete element modeling (DEM), computational fluid dynamics (CFD), statistical models, and population balance models (PBM). DEM and CFD are mechanistic in nature and can effectively capture the motion of particles within equipment or their interaction with a stream of fluid. The DEM approach has been implemented in various pharmaceutically relevant unit operations, such as blending, granulation, and coating (6). The application of CFD has been observed in unit operations, such as mixing, granulation, and crystallization (7). Statistical models, such as response-surface methodologies, have been largely used to determine design space and to a lesser extent for optimization (8). PBMs (hyperbolic partial differential equations representing mesoscopic framework) have also been widely implemented on particle-based unit operations, such as granulation, crystallization, and mixing (9). 

While application of model-predictive design in pharmaceutical applications is only in its infancy, several successes can be reported. Examples of applications at Rutgers University include stirred tank agitators used as bioreactors (10), roller bottle reactors used for mammalian cell culture of viruses (11), and understanding wall losses in the Anderson Cascade Impactor (12). Many other groups have since used CFD tools for many design applications, including chemical reactors of many different scales, crystallizers, liquid-liquid and liquid-gas reactors, and cleanroom design.  

Similarly, use of DEM applications go back well over a decade. Some of the earliest examples of the use of DEM for design purposes in pharmaceutical applications focused on the design of powder blenders, including the V-blender, the double cone, and the bin blender. Since then, many other examples have followed, focusing on hoppers, tablet coating, and tableting, for example. 

An additional modeling tool is the use of response-surface models, which are typically developed using data from designed experiments and subsequently used to select process optima and to design control algorithms. While the data-driven models used in this application are largely empirical and do not require a mechanistic understanding of the processes to which they are applied, they are invaluable as tools to aid understanding of the relative importance of process and formulation variables and thus help narrow down the scope of work required to advance process understanding. 

These tools are increasingly becoming common components of the pharmaceutical process-design toolbox. They have spread from academia to the largest pharmaceutical companies, many of which have formed process-modeling groups in which researchers are using these tools to design and optimize processes 

 prior to expensive equipment acquisition or to reduce the complexity of designed experiments. 

To date, efforts have been piecemeal and typically have focused on individual process components. The emergence of continuous manufacturing as a central focus of attention, however, is now motivating the need to develop modeling frameworks capable of simulating all process components simultaneously, using a variety of tools suited for each specific process. The flowsheet framework meets this need. 

The continuous manufacturing of drugs can be achieved using various routes: direct compaction (DC), roller compaction (dry granulation or DG), or wet granulation (WG), depending on the starting and desired end properties of the formulation. DC is the simplest of the processes mentioned above while DG and WG improve flowability characteristics to prevent ingredient segregation and to increase density.  DC (14), DG (15), and WG (16) routes have been explored using model-predictive flowsheet methodologies. Application of advanced modeling techniques for optimization and control (17, 18) on the overall flowsheet instead of the individual unit operations would enable efficient operation of the continuous process. 

The challenges associated with developing robust and reliable flowsheet models for solids' processes include:  					

Development of models that describe their constituent mechanisms 

Performance of experimental studies for the data acquisition of multi-dimensional key particle properties 

Identification of all the possible manipulated and controlled variables and their interactions (17, 18) 

Integration of process design and control to identify globally valid operating conditions. 

Extensive research is ongoing to identify and develop predictive models for all the unit operations involved in the continuous tablet-manufacturing process. For integrating the various unit operations into a flowsheet, it is crucial to correctly identify the critical connecting properties that communicate across units (17). Simulating the overall flowsheet, the variations in the key properties can also be tracked during the transient states involving process start-up, perturbation propagation, dynamic response to change in settings due to control actions, and process shutdown. Furthermore, through the implementation of various operating scenarios, the flowsheet model can be used for the assessment of different process alternatives (so far achieved by expensive laboratory tests), which are then scaled up to the desired plant size. The developed and validated flowsheet-simulation system can also be used for operator training, since any sequences in operating schedules can be performed virtually and analyzed through a computer screen. Using information obtained from the flowsheet models for plant implementation is the next challenge. Incorporating control systems in the actual plant is a crucial task needed for efficient operation and minimal variation from the setpoint values. 

Of particular interest from a regulatory perspective is the use of integrated flowsheet models to enable identification of the propagation of noise or upsets in a particular unit operation through the entire continuous line (16, 18). This issue is directly relevant to the assessment of robustness and reliability of the continuous manufacturing system. Process optimization can be achieved by implementing optimization algorithms on the overall integrated model. 

 shows a flowsheet model (simulated in gPROMS, Process Systems Enterprise) of a flexible, continuous, tablet-manufacturing process together with the implemented control system. 

Figure 1: Flexible continuous tablet manufacturing process with (1) direct compaction, (2) roller compaction, and (3) wet granulation. (ALL FIGURES COURTESY OF AUTHORS)			

Various control systems can be implemented on the flowsheet model in the form of simple PID loops (15) or with advanced model-predictive control (MPC) (17). Control loops can be implemented by identifying the control-loop pairings and assessing the need for MPC in each control loop (as opposed to using just PID loops). With this information, the PID controllers are designed and implemented to obtain a predictive model of the plant, thereby suggesting the design of the MPC controller. The designed MPC is then incorporated into a general model for model-based performance evaluation. 

As an example, consider integration of control hardware and software in the continuous feeder and blender system, as shown in 

. A PAT system is used to read the near infrared (NIR) spectral data at the blender outlet and communicate it to the multivariable analysis (MVA) model performing principle-component analysis and partial least-squares (PCA/PLS) to provide the API concentration and relative standard deviation (RSD) value. These critical quality attributes (CQAs) are used as inputs to the MPC in the process-control system. The MPC uses the two CQA inputs to drive the feed ratio and the blender speed. The MPC output (feed ratio) gives the feeders' flowrate setpoints, which are then tracked by slave PID controllers. The implemented control scheme utilises a PAT data-management system (synTQ, Optimal) to integrate a digital automation system (DeltaV, Emerson Process Management) with an NIR analyser and MVA model. 

Figure 2: Implementation of a model-predictive controller via a process analytical technology (PAT) datamanagement system (1. SynTQ, Optimal), (2. MATLAB OPC, MathWorks), or (3. SiPAT, PCS7, Siemens). 			

Integrating multiple system parts presented several challenges. A framework, however, is now in place that allows implementation of control architectures for a wide variety of continuous powder processes. 

This work is supported by the National Science Foundation Engineering Research Center on Structured Organic Particulate Systems, through Grant NSF-ECC 0540855. 

The authors are from the Engineering Research Center for Structured Organic Particulate Systems (ERC-SOPS), Department of Chemical and Biochemical Engineering, Rutgers, The State University of New Jersey, Piscataway, NJ 08854, USA. 

 6.  F. Boukouvala, Y. Gao, F. Muzzio, and M.G. Ierapetritou, 

 7.  T. Hőrmann, D. Suzzi, S. Adam, and J.G. Khinast, 

 8. F. Boukouvala, F. Muzzio, and M.G. Ierapetritou, 

 9. A. Chaudhury, A. Niziolek, and R. Ramachandran, 

10. E. S. Szalai, P. Arratia, K. Johnson, and F. J. Muzzio, 

11.  D.R. Unger, F.J. Muzzio, J.G. Aunins, and R. Singhvi, R., 

12. P.D. Swanson, F.J. Muzzio, A. Annapragada, and A. Adjei, 

 13.  M. Moakher, T. Shinbrot, and F.J. Muzzio, 

17. R. Singh, M. G. Ierapetritou, and R. Ramachandran, Eur. 

18. R. Singh, M. G. Ierapetritou, and R. Ramachandran, 

Applying model-predictive methods and a continuous process-control framework to a continuous tablet-manufacturing process.

Model-Predictive Design, Control, and Optimization

Pharmaceutical manufacturing in the United States is in rapid decline. To devise a strategy for reversing it, we need to understand the current landscape and develop a coherent plan for re-energizing technological innovation, while making domestic manufacturing economically viable.  

The industry is undergoing wrenching changes. Higher regulatory standards for efficacy and safety, among other reasons, have led to a significant level of difficulty in replacing "blockbusters" developed in the 1980s and 1990s that are now coming off patent. The rapid growth in biological understanding remains a source for current and future drugs, but has also raised the bar on selectivity and toxicology expectations. In the past 10 years, new molecular entities (NMEs) entering the market number less than half of those in the preceding decade. This decrease has caused significant reductions in revenue for the industry, forcing reductions in workforce, a focus on productivity and cost-containment, and a search for new business models.  

Industry is also grappling with issues of quality and compliance in manufacturing. Some key contributing factors are the regulated environment, which has pushed to lock inefficient and poorly understood manufacturing processes too early, the lack of modern process controls, and, ironically, the historic high gross margins of the industry, which until recently often made manufacturing cost a minor component undeserving of serious attention. 

This article focuses on two new key business models being adopted by the industry. First, the emerging distributed model for innovation, where large companies rely for much of the science and execution of research on a network of companies of diverse sizes, sometimes supported by academic laboratories, each one excelling in a given niche (e.g., discovery, clinical testing, product development, regulatory approval). This transition away from the traditional vertically integrated development model has brought a new set of challenges with it, including a reluctance of both large pharmaceutical companies and venture capital firms to fund new drugs and drug-delivery systems that have not yet achieved proof-of-concept (the so called "valley of death"), and the transactional cost associated with operating in this "pharmaceutical ecosystem."  

A second strategy is the increased outsourcing of primary and secondary manufacturing, packaging, and distribution. CMOs have become a major part of this pharmaceutical ecosystem. The trend to outsource has brought a diverse group of CMOs into the picture, ranging from large, well-capitalized companies with a solid understanding of manufacturing, quality, and compliance, to small, inexperienced companies in emerging markets who struggle to understand, much less provide, high quality and compliance standards.  

Furthermore, pharmaceutical sales in emerging markets are expected to grow from 10–40% in the next 20 years, providing most of the worldwide growth, thus driving pharmaceutical companies to invest and create a presence in those markets. Thus, shifting manufacturing to emerging markets is an effective way to achieve this goal, but the net result for the US economy is a massive loss of jobs.  

There are also fundamental reasons why pharmaceutical manufacturing has lagged behind other high-tech industries. Every product is a different molecule, with different characteristics, method of manufacture, analytical methodology, stability characteristics, and so forth. Most pharmaceutical products involve powder processing, which is—at best—a partially understood field. As a result, for every product, the entire product-development process is developed from scratch.  

In contrast, other large-scale manufacturing industries such as petrochemicals, automobile, and microelectronics, deal with a much smaller number of materials and use processes that are better understood. The products and manufacturing methods evolve and improve with time because one can build on what was done before.  

Industry therefore faces a pressing conundrum: how to manage the quality and compliance of pharmaceutical manufacturing, which has lagged behind other industries even in the best of times, during a period of cost-cutting and fragmentation, with a myriad of manufacturers of various degrees of sophistication in a variety of countries and cultures. At the same time, our country seeks to increase manufacturing as a percentage of the US economy as a means of remaining competitive as a nation. We propose here that these two imperatives are related and can be managed in an integrated fashion. 

It is useful to separate pharmaceutical manufacturing, both primary and secondary, into two groups: those operations that have become conventional and essentially commoditized, which are now being executed around the globe, and which are unlikely to return to the US; and those operations which relate to novel, "smart" products and processes that are knowledge-intensive, have IP protections, and are less sensitive to labor costs. 

In the case of the first group, commoditized manufacturing, quality and regulatory compliance remain crucial. Indeed, quality compliance has become more challenging in the outsourced-manufacturing environment because it now must be managed across many more intercompany barriers. Fortunately, a solution has been demonstrated in the chemical industry. The answer is to make manufacturing technology mature, transparent, and portable. The operations involved must be categorized and described in detail, making it possible to harmonize methods, approaches, operations, control algorithms, and quality tests, such that the operations can be understood fundamentally, modeled mathematically, and then executed under predictive-model controls as is done by other high-tech industries. This approach addresses quality concerns while also raising the technological and scientific expectation for producers, making it more likely that manufacturing can once again be performed in the US. 

The later group of "smart" products and processes involve targeted drugs and delivery systems, personalized medicine, medicines prepared locally on demand, complex hybrid biologics, sensors, and new diagnostics. These technologies are enormously promising, but they require long-term fundamental research to create new products, new manufacturing methods, and new technologies to achieve goals like rapid, high-precision microdosing, self-assembled targeted delivery nanoparticles, and so forth. The players in this area are a diverse set of companies and universities from around the world that typically work in a disconnected fashion, perhaps entering into complex, bilateral (and sometimes asymmetrical) agreements.  

In such an environment, it is hard to assess which approaches and products have the most merit, and how they could function together in different and synergistic combinations. The best options may take a long time to come to market or may not be created at all. These are the products that we must ensure get developed and manufactured in the US.  

Given this environment, re-energizing pharmaceutical manufacturing in the US is likely to take place only if a coherent plan is implemented. In the authors' opinion, such a plan must address four crucial elements: strong long-term technology development, a nurturing ecosystem where collaborations among innovation players are facilitated while transactional costs between them are minimized, science-based compliance with high quality standards, and a concerted effort to create and maintain a highly skilled labor pool capable of supplying the needs of industry, government, and academia. Thus, we propose a strategy involving the following components:  

The authors propose a plan to keep the US pharma industry afloat and in the lead.