Johannes G. Khinast is professor at the Institute of Process and Particle Engineering, Graz University of Technology, Austria

Johannes G. Khinast

The pharmaceutical coating process is a well-established unit operation, but accurately measuring the endpoint of the coating process remains a challenge. Knowing the endpoint is crucial, especially for functional coatings, because quality attributes, such as coating layer thickness, have a direct impact on product performance, specifically the dissolution rate. 

	This article summarizes results of a study that looked at optical coherence tomography (OCT) as a way to monitor pharmaceutical spray coating processes in-line for tablets and pellets. The study measured coating thickness as a function of time for tablets in a drum coater and for pellets in a fluid bed coater. It also looks into the possible influence of dye, which was found to have no impact on measurements. Coating thickness was determined automatically based on OCT images rather than via chemometric calibration models. In-line data and off-line three-dimensional mapping revealed additional facets of tablet and pellet coating quality (i.e., intra- and inter-particle coating variability).

	Results of the study indicate that an industrial-ready OCT system can improve process understanding and assure product quality in pharmaceutical functional coating applications, assisting pharmaceutical scientists with process development, scale-up, transfer, and troubleshooting. In the future, OCT could become a new tool for quality control (QC) release, replacing unreliable and time-consuming testing procedures. One of the technology’s key strengths is that it allows detailed understanding of the source of and presence of defects and can identify poor coating quality, both within a batch and between batches.

	Functional coating is a common way to improve patient compliance, prevent counterfeiting, and enhance bioavailability-and with it, the overall functionality of solid oral dosage forms. In some cases “active coatings” are applied that contain one or more APIs to mitigate interactions between different drugs or to account for different release behaviors in a single dosage form. Enteric coatings are typically applied to tablets in pan coating equipment, while beads, mini-tablets, and pellets are typically coated in fluid-bed coating equipment (e.g., for taste-masking applications in pediatric dosage forms). The coating process must be controlled, based on coating thickness, thickness variability (both between different processing lines and within a single coating line), and coating defects (e.g., coloration or surface blemishes and cracks).

	Problems with coating are common during development and manufacturing, especially when scaling-up a process or transferring it from one lab or manufacturing site to another. Thus, having technology available that can monitor in real time the critical quality attributes of coatings (including thickness, thickness variability, morphology, porosity, defects, and cracks) can reduce overall development time, accelerate process scale-up, and enable greater precision in determining the root-cause analysis of coating weakness.

	A number of techniques are currently used to study oral solid-dosage form coating quality, including optical inspection of cross-sectional cuts and measurement of tablet diameters or weight gain.

	However, these approaches are time consuming, error prone, and can only provide approximate results. In addition, these methods cannot be applied in-line to allow for process monitoring or control. 

	To overcome these shortcomings, in-line process analyzers have been developed to monitor and assess coating quality non-destructively, each with its own strengths and weaknesses. 

		Near-infrared (NIR) (1,2) and Raman spectroscopy (3,4) 

		Terahertz (THz) sensing (5,6), a powerful approach, but one that is not easy to implement in real time

	In contrast, optical coherence tomography (OCT) allows coating quality of translucent functional coatings to be determined rapidly, in real time, without the need for calibration. It also eliminates the need to develop and maintain chemometric models in order to interpret the data. These models are required for both NIR and Raman (1). Compared with other process analytical technology (PAT) approaches, OCT offers much higher scanning speeds (i.e., up to 250,000 measurements per second vs. approximately 30 for THz and less than one for Raman). Furthermore, OCT provides very high axial and lateral resolution (more than one order of magnitude better than all other approaches).

	OCT is an interferometric technique that is used to generate cross-sectional depth-resolved images of coating layer(s). The physical setup and operation of OCT systems have been described in the literature (9–11), with a functional distinction being made between time- and spectral-domain OCT systems. During image acquisition, a light source with high spatial and low temporal coherence is focused on the coating surface. Most light is reflected or diffracted, but a substantial part penetrates the surface and is reflected from interfaces of different materials with distinctive changes in refractive index. The reflected light is detected with a spectrometer (for spectral-domain OCT). By measuring the optical path length between the reflections, it is possible to determine the distance between the interfaces considering the refractive index of the material.

	Several studies have already demonstrated the high performance of OCT systems for measuring the coating thickness of pharmaceutical solid dosage forms (12–14). OCT can also be used to analyze coating thickness variations within single particles (intra-particle variation) and between particles (inter-particle variation) during the coating process (15). Recently, OCT has been commercialized as a monitoring technology for good manufacturing practice (GMP) applications, and an ATEX model is available for industrial use. The research examined in this article evaluated the GMP commercial device.

Throughout all experimental work, a commercial spectral-domain OCT system (OSeeT, Phyllon, Austria) was used for measurements and data recording. The OSeeT system works at a central wavelength of 832 nm with a spectral bandwidth of 75 nm, leading to a theoretical axial resolution of 4 µm. 

	The base unit can be combined with a one-dimensional (1D) sensor with 14-µm lateral resolution for in-line measurements. This sensor is also part of the additionally available at-line sampling device. A three-dimensional (3D) sensor with 10-µm lateral resolution is available for off-line measurements. Sensors and peripherals can easily be changed due to the use of standard connections such as fiber channel/angled physical contact (FC/APC) interface for the sensors and optical peripherals, and Universal Serial Bus Version 3.0 (USB 3.0) for electrical interfacing. 

	This research used a 1D sensor head (Phyllon) to monitor and validate a pan-coating process, and a custom-made 3D sensor (16) in two-dimensional (2D) operation mode (i.e., with one galvo mirror disabled) to monitor the fluid-bed coating experiments. The 3D sensor was also used to measure pellets and tablets periodically drawn from the process, off-line.

	Sensor exposure time was set to 15 µs for in-line measurements of tablets (1D) and 30 µs for in-line measurement of pellets (2D) and off-line measurements (3D). The idle time (for read-out and digitalization) was set to 1.9 µs for all measurements. The exposure and idle times resulted in acquisition rates of 59.2 kHz (1D) and 31.3 kHz (2D/3D). The acquisition rate corresponds to the number of single-depth scans per second, resulting in a frame rate of 57.8 frames per second (fps) for the 1D device, and 30.6 fps for the 2D and 3D devices.

As a model use case, pan coating was evaluated on tablet cores containing acetylsalicylic acid (ASA) similar to the commercial product Thrombo ASS. Coating was performed in a lab-scale pan coater (ProCepT, Zelzate, Belgium) with a reservoir capacity of 1 L, and a 0.8-mm Schlick spray nozzle was used to apply the coating suspension. The batch consisted of 350 g uncoated tablet cores containing 50 mg acetylsalicylic acid, lactose monohydrate, microcrystalline cellulose, highly dispersed silicone dioxide (SiO

) starch, talc, and triacetin. The biconvex-shaped tablet cores had a diameter of 7.15 mm, a height of 3.7 mm, and a radius of cubature of 7.9 mm.

	A common enteric coating (Eudragit L30 D-55, Evonik Industries AG, Darmstadt, Germany) was used, and blue colorant Liquitint Blue HP (Milliken & Company, Gent, Belgium) was added to the coating solution to provide a visual indicator of applied coating homogeneity. Following the coating manufacturer’s instructions (17), 504.9 g of coating solution were prepared, and the amount of added spray solution was calculated (18) to achieve a targeted coating thickness of approximately 70 µm.

	A total of 150 g of coating suspension was applied to the cores, resulting in an additional 30 g of dry mass on the tablets. Coating parameters and targeted coating thickness were recommended by GL-Pharma, who manufacturers commercial product. The pan speed was set to 40 rpm, and the coating was performed at a spray rate of 3.2 g/min, at an inlet airflow rate of 0.4 m³/min, and a temperature of 50 °C.

	Tablets were monitored (19) through the holes of the perforated drum of the pan, and OCT images were continuously acquired and saved for post-processing throughout the whole coating run. An automated evaluation algorithm (19) was applied on all saved images, including:

		Extraction of air/coating and coating/core interfaces

		Correction of distortions due to tablet speed, oblique orientation, and curvature

	The refractive index was assumed to be 1.48, as validated for the same coating material (20). Due to the insignificant influence of appearance-altering colorants on the refractive index, a negligible impact on measurements was expected.

	Additional samples of 10–15 tablets were drawn every eight minutes and further analyzed using the OCT 3D sensor head configuration. Results of in-line and off-line measurements, as well as weight of applied coating were compared using Matlab software.

This use case aimed to coat relatively large pellets completely and monitor thickness. Pellet coating was performed in an lab-scale air flow technology coater (Romaco Innojet VENTILUS V-2.5, Pharmatechnik GmbH, Karlsruhe, Germany) on extruded calcium stearate pellets, which were composed of a matrix carrier of 75% (w/w) calcium stearate, 20% (w/w) paracetamol (API), and 5% (w/w) glycerol monostearate as plasticizer (21).

	The pellets had a mean particle size of 400–2000 µm. 450 g of these pellets were coated with molten Dynasan 118 glyceryl tristearate (Cremer Oleo, Germany) in three replicated experiments (B01–B03), all of which were run at the same process conditions.

	The molten coating temperature was roughly 90 °C during all experiments, and the refractive index was assumed to be 1.438 (22). Preliminary tests showed that a 100-µm coating thickness was sufficient to generate a smooth and cohesive film. The coating process lasted for 32 minutes until 218 g of coating mass were sprayed onto the pellets to achieve a theoretical coating thickness of 100 µm. Process conditions were set to a spray rate of 7 g/min, 1.5 bar atomizing pressure, 60 m³/h air flow rate and 35 °C inlet air temperature.

	Pellet coating was monitored using the 3D sensor in 2D operation mode (13). At the inspection window, the sensor was protected by a thin sheet of plastic foil during all runs. In-line measurements of pellets were recorded every other minute for a duration of 60 seconds. 

	The research then compared automated sampling data to results from manually drawn samples, which were analyzed off-line and evaluated based on 3D-OCT measurements and particle-size analysis using a particle analyzer (Qicpic, Sympatec GmbH, Germany). Results of in-line, 3D, and particle-size measurement were then compared using Matlab software.

, the measured coating thickness of tablets steadily increases with process time. Due to the 4.5-µm axial resolution limit of the 1D sensor (23), reliable thickness measurements are available only after seven minutes. For each analyzed tablet, at least 50 measurements per tablet were performed. Tablets with fewer values were automatically rejected to guarantee accuracy. In total, 1385 tablets were analyzed during 48 minutes of process time and the coating thickness approached a final value of 69.1 ± 4.9 µm, which is a good result compared to the target 70 µm coating thickness.
	

The collected data also permit the time-dependent analysis of coating thickness variations on single tablets (i.e., the intra-tablet as well as inter-tablet coating variability). This analysis was done based on the relative standard deviations (RSD) of the coating thickness as illustrated in 

	The high RSD at the beginning of the process was due to the inhomogeneous distribution of coating mass on the total number of tablets and on single tablets.  As can be seen, the RSD decreased with process time, and thus, the coating uniformity was enhanced while approaching the process end. A final weight gain of 150 g and an RSD of mean coating thickness (inter-tablet coating variability) of 7.2% could be achieved.

	Manually drawn samples at different stages of the process enabled analysis of polymer film integrity and homogeneity on the tablet using off-line evaluation via the 3D sensor. Results agreed with in-line-measurements, which are shown as red circles in 

. The RSD of the coating thickness of the 3D-mapped tablets is of the same magnitude as the RSD of the measured in-line data. This shows that deviations of in-line data originate mainly from differences of the coating, whereas other effects, such as random movement of tablets or different parts of the tablets facing the sensor, are of minor impact.

The quality of real-time OCT images during fluid-bed coating was not as good as it was for the off-line images, due to rapid movement of the pellets and signal attenuation originating from scanning through the protection foil. The results were also affected by a systematic error (13), due to the curved surface of the pellets, which needs to be corrected for. To minimize errors, the algorithm used for automated evaluation of pellet coating thickness must consider these effects. To correct for these effects, a spherical shape was assumed for all pellets, and an ellipsoid fit of the captured OCT images was transformed to an equivalent circle radius. After this correction, the air-coating and coating-core interfaces could be evaluated in the same way as the tablet measurements. Due to the fact that pellets have a more curved surface than typical tablets, the algorithm evaluated only the upper part of the pellets, which are facing the sensor head within an angle of ±30° around the highest point in the image.

	Results of the automated data evaluation and reference measurements are shown in 

. While OCT data were acquired for the entire process time, only a subset of data are shown for the sake of clarity. A total number of 15,685 pellets were automatically evaluated with an additional number of 1395 pellets analyzed by particle size analysis (shown as green filled squares in the figure) to provide an independent reference. In-line OCT measurements were only accepted if at least 30 measurements per pellet could be performed.
	

), results for all coating runs agreed with results from the manual evaluation and showed even better correlation with the results from particle size analysis. As illustrated in 

, the plots of inter-particle coating variability showed good correlation for all runs. All results of endpoint coating thickness were close to the target coating thickness of 100 µm and correlated well among the applied methods. As expected, the RSD of the coating thickness between particles decreases during the process, following a function of √(1/t) as described by Turton (24). In general, the variability between runs is more pronounced in the beginning, due to higher uncertainty of measurements originating from the resolution limits of the sensor and spectrometer setup. 

	The plots of intra-particle coating variability in 

also show a decrease in process time for all automatically evaluated runs, in contrast to the results from manual evaluation, which did not follow this trend. This behavior can be explained by the fact that the manual evaluation does not need to compensate for the experimental setup and the unknown velocity of the pellets. Additionally, fewer samples could be evaluated manually compared to the automatic evaluations. Therefore, these values should be considered with a high error probability compared to the automated evaluated results. Results of the coating experiments for the different evaluation and reference methods are summarized in 

Results of the automatic evaluation of in-line OCT data during tablet coating showed a linear increase of coating thickness with process time, which is in excellent agreement with the reference method (sprayed coating solution) and the off-line 3D mappings. Inter- and intra-tablet coating variabilities were found to decrease with process time, as expected. A minimum of 50 thickness measurements for every tablet guarantees a reliable basis for all automated evaluations and additional statistical results. Note that the non-zero coating thickness during the first 10 minutes originates from a high uncertainty of the measurements as a consequence of the resolution limit of the system (<10 µm). Additionally, only a few tablets were coated in the first few minutes, leading to a much higher spread of thickness readings.

Colorant had no impact on measurements. 

The addition of 1‰ of a liquid colorant to the spray suspension showed no negative effect on the quality of the readings, nor on the results of in-line OCT measurements, although the visual appearance of the tablets changed significantly with the color changing to dark blue. Note that the blue colorant used was not pharmaceutical grade, and therefore, not suitable for ingestion, but was used to allow visual inspection, and to demonstrate that OCT can also be used to analyze colored coatings.

	Analyzing the recorded 3D maps in more detail revealed that the coating layer partially compensated for small surface irregularities of the tablet cores, while larger cavities in the tablet surface persisted throughout the whole coating process as shown in 

Thus, OCT can be used to develop coating formulations designed to smooth tablet cores, if needed. Defects in the coating and existence of poorly coated regions (e.g., at the edges) can be readily detected, enabling closer engineering and control of the coating process (e.g., by increasing the coater temperature).

All three runs showed good correlation with each other and the reference method (dynamic imaging analysis). They are in an even better agreement with results from the manual evaluation, where only three points per pellet were measured and evaluated. This emphasizes the fact that the presented OCT system is capable of extracting the correct interfaces of coated pellets and interprets the data in a reproducible way.

	The improved algorithm, fitting an ellipsoid, worked well for the tested experiments, demonstrated by consistent measurements compared with the reference methods. As listed in 

, the comparison of results from automatic and manual evaluation reveals the advantage of the high number of measured values, with at least 30 thickness measurements per pellet for automated evaluation, because the standard deviation (SD) values are only half of the manual evaluation. This highlights the advantages offered by OCT in terms of precision of automatic thickness measurements due to a higher significance compared to results based on only three manual measurements per pellet. All runs met the targeted coating thickness of approximately 100 µm, which represents a coating amount of 32.6% (weight), which is typical for taste masking of multi-particulates (25) for similar coating systems.

	In contrast to the manually evaluated data, OCT data enabled statistical information to be developed that showed intra- and inter-pellet coating variability and the variability decay with process time. In this study, all three experiments showed the same behavior in general, with a decrease of RSD for the inter-pellet coating variability over time. The intra-pellet coating variability of the manual evaluation only deviated from the automated evaluated data in the beginning of the run due to the small number of samples used for the manual evaluation.

	This study demonstrated that different coating processes could be easily analyzed by means of an industrial OCT system in real-time, both for tablets and pellets, in a coating thickness ranging from 10 µm up to 100 µm. OCT could be applied for inline measurements ranging from lab- to production-scale coaters. It can accurately detect the coating end-point based on coating thickness, independent of spray efficiency, process parameters, and scale. For long-term inline measurements, air purging of the optical window of the 1D sensor worked well, and all in-line equipment is available in hygienic design for use in GMP applications.

	In this work, the authors focused on investigating the growth of the coating layer, demonstrating that OCT is a suitable PAT tool. No dissolution data were generated. However, by comparing tablets used in these experiments with results for commercially available Thrombo ASS, research showed that the coating thicknesses matched perfectly, with differences of less than 1 µm in coating thickness between the commercial product and our showcase tablets.

	Work is now underway to correlate OCT data and dissolution behavior, and will be the focus of future articles. However, preliminary data correlating OCT measured coating thickness and dissolution behavior were presented at the International Forum on Process Analytical Chemistry (IFPAC) meeting in Washington, DC in 2018 (26).

	Despite OCT’s strengths, the tested OSeeT system showed some limitations.  For example, coatings that contain a high concentration of pigments that are frequently used in cosmetic coatings (e.g., titanium dioxide as a whitener or iron oxide as a colorant) cannot yet be measured. Moreover, thin layers (<10 µm) cannot be accurately resolved. Development efforts are now underway to overcome these limitations.

	In addition to results discussed in this article, the 3D-OCT system has been used successfully to test other commercial products, as shown in 

All examples could be measured and evaluated automatically. Coating thicknesses for the given examples were in the range of 10 to 250 µm.
	

	The presented results highlight the ability of OSeeT, as the first commercially available OCT system for the pharmaceutical industry, to generate significant statistical data for the mean coating thickness, as well as inter- and intra-tablet variability, as a function of coating time. In addition, inner structure (e.g., porosity, morphology) and defects can be studied, as shown in 

	All of these properties, within the current quality assurance frameworks, can be (and are) analyzed daily. The number of dosage forms tested is low, however, and testing is time-consuming and costly.

	In contrast to the current picture, OCT allows a detailed understanding of the critical quality attributes of coatings in real-time, including the source of variability and the presence of defects or poor coating quality, both intra and inter batch. Thus, OCT technology can support pharmaceutical scientist in rationally developing coating formulations and the associated processes, allowing straightforward process transfer, scale-up, and troubleshooting.

Figure 3: Reconstruction of a complete 3D optical coherence tomography (OCT) scan of one of the tablets under test. The coating partly compensates for irregularities in the core-coating interface.

	This work has been funded by the Austrian Competence Centers for Excellent Technologies (COMET) program, under the auspices of the Austrian Federal Ministry of Transport, Innovation and Technology (BMVIT), the Austrian Federal Ministry of Economy, Family and Youth (BMWFI) and by the State of Styria (Styrian Funding Agency SFG). COMET is managed by the Austrian Research Promotion Agency FFG.

	1. C. Möltgen, T. Herdling and G. Reich, 

., 85, 1056–63 (2013).
	2. P. Wahl, G Fruhmann, et al., 

., 87, 271–8 (2014).
	3. S. Romero-Torres, J Pérez-Ramos et al., 

, 41, 811–9 (2006).
	4. J. Müller, D. Brock D, et al., 

,80, 690–7 (2012).
	5. R. May, M. Evans et al., 

, 100, 1535–44 (2011).
	6. H. Lin, Y. Dong, et al., 

., 106, 1075–84 (2017).
	7. D. Wiegel, G. Eckardt, et al., 

, 301, 261–7 (2016).
	8. V. Naidu, R. Deshpande, et al., 

., 1–6 (2017).
	9. J de Boer, R. Leitgeb et al., 

., 20, 251–76 (2010).
	12. D. Markl, G. Hannesschläger, et al., 

, 26, 1–12 (2015).
	13. D Markl, M. Zettl, et al., 

., 125, 200–8 (2015).
	14. Y. Dong, H Lin, V. Abolghasemi, et al, 

, 106, 546–53 (2017).
	15. D. Markl, G. Hannesschläger, et al., 

, 55, 58–67 (2014).
	16. D. Markl, P. Wahl, et al., Int.

, 536, 459–66 (2018).
	17. Evonik Industries AG, “

  www.otomed.co.kr/english/img/evonik_im/evonik-quickstart-eudragit-l-30-d-55-enteric-coating-with-gms-as-anti-tacking-agent-q7.pdf, cited 25 January 2018.
	18. M. Zettl, Thesis, Technical University Graz, 2017.
	19. D. Markl, G. Hannesschläger, et al., 

, 104, 2531–40 (2015).
	20. D. Koller, G. Hannesschläger et al., 

, 44, 142–8 (2011).
	21. E. Roblegg, E Jäger et al., 

., 79, 635–45 (2011).
	22. M.O’Niel, M. Smith, 

, 13th ed., The Royal Society of Chemistry, (2001).
	23. D. Markl, G Hannesschläger et al., 

, 181, 186–94 (2008).
	25. K. Becker, E. Saurugger et al., Int. 

, 497, 136–49 (2016).
	26. E. Gartshein, J. Timmermans “Optical Coherence Tomography (OCT) to Support the Coating end Point Determination,” a presentation given at the International Forum for Process Analytical Chemistry (IFPAC), January 2018.


	Vol. 43, No. 3
	March 2019
	Pages: 36, 38, 40, 42–44, 46–47

	When referring to this article, please cite it as M. Wolfgang, et. al

 “Real-Time Measurement of Coating Film Thickness," 

are senior scientists at the Research Center for Pharmaceutical Engineering GmbH, Graz, Austria; 

 is senior manager at Pfizer Global Supply, Peapack, NJ, USA; 

* is professor at the Institute of Process and Particle Engineering, Graz University of Technology, Austria, 

Articles

Optical coherence tomography can improve quality control and development of coated dosage forms by allowing film thickness to be measured in real time.

Real-Time Measurement of Coating Film Thickness

The gluing pills technology has recently been developed for the manufacturing of multilayer tablets by compressing single layers separately and joining them together using approved pharmaceutical excipients as a gluing agent. 

	This novel technology was developed in response to challenges involved in conventional manufacturing of multilayer tablets, including in-line control of the tablet weight, the tendency to delamination, direct contact between the two tablet layers, and cross contamination. The technology can easily be implemented on miniaturized equipment with reduced footprint for real-time production of individualized fixed-dose combinations and applied to both translational pharmaceutics and personalized medicine.

	The aim of the gluing pills technology (GPT) is to offer solutions to the increasing demand for flexible fixed-dose combinations (FDC) for improving patient adherence to therapy. A high level of patient adherence is a major success factor in the development of effective and cost-saving therapeutic strategies, especially for complex disease states that require polypharmacy such as cardiovascular diseases, pain therapy in cancer patients, metabolic diseases (e.g., diabetes mellitus type 2), HIV/AIDS, and multidrug-resistant infections.

	The most widely used FDC dosage forms are multilayer tablets. The development of multilayer tablets also suits life-cycle management (LCM) of pharmaceutical products. In addition to the conventional purpose of multilayer tablets, which is to have different APIs in different layers (e.g., to avoid stability problems), this dosage form can also be designed with the same API in the layers but with each layer having a different API release profile. 

	Currently, the established manufacturing technology for multilayer tablets poses substantial challenges. Multilayer tablets are typically produced via rotary tablet presses. Raw materials supplied as powders or granules are filled into a die separately and compacted either simultaneously or consecutively. Both methods have drawbacks. First, weight control via tablet thickness or compression force measurement is rather complex. Moreover, the adhesion of layers is often complicated by an elastic-modulus mismatch between the layers. These effects can lead to improper layer weights or layer ratios, as well as delamination. Relatively low production efficiencies and high rejection rates are unfavorable consequences. Additionally, the cross contamination of incompatible APIs and the missing of a clear boundary layer can pose a problem (1).

	A novel technology for manufacturing multilayer tablets, GPT, has been developed in response to these challenges. Each layer is compacted separately, and tablets are attached to each other in an automated process with a binding agent, which at the same time acts as a barrier against cross-contamination and interactions between the tablet layers. GPT offers more flexibility with regard to API combinations, reduced waste, and higher process efficiency compared to conventional multilayer tablet manufacturing. Moreover, different layers of different thickness can rapidly be combined in one dosage form without redevelopment of the process. GPT is, therefore, ideally suited to fulfill the demands of individualized medicine.

	The entire GPT is carried out in an automated process consisting of several steps, as schematically illustrated in 

. The main steps are feeding of the first tablet layer; application of gluing agent; feeding of the second tablet layer; hardening the gluing agent between the two tablet layers, which results in the production of bilayer tablets; and unloading of bilayer tablets. More tablet layers can be applied as well. 

 shows bilayer tablets produced using a prototype of this automated process.

	The system was developed based on quality-by-design principles, enabling the application of process analytical technology (PAT) tools for the in-line monitoring of the various manufacturing steps. This system includes the assurance of thickness and continuity of the binding layer. The correct amount of gluing agent and the adjustment of this layer on the tablet are monitored within the process. Therefore, the specified thickness and continuity of the binding layer is guaranteed.

	The nozzle system for glue application is a high-speed micro-dispensing system based on piezo technology. It has a heating unit for melting the gluing medium in the nozzle. Both aqueous and organic solutions of binding agents can be applied. Molten polymers can be used as a gluing agent by using the nozzle, combined with the heating unit.

	Despite a longer evaporation time, aqueous solutions are preferred in the pharmaceutical industry because they are environmentally friendly and fully comply with “green chemistry” measures (2, 3). The case study described in the following illustrates the use of GPT with aqueous binding solutions to produce bilayer tablets. 
	The automated GPT is available on two scales, ML7 and ML240, suitable for both industrial production and specific small-scale applications (see 

). The technical properties of ML7 and ML240 equipment are listed in 

 The bonding and the mechanical properties (e.g., Young’s modulus, Poisson ratio) of the API and selected excipients play a crucial role in the deformation layers of a conventional bilayer tablet. Because elastic mismatch of the layers can cause radial stresses and delamination of tablet layers (1), careful adjustment of the compression force is essential for conventional manufacturing of multilayer tablets. It has been reported that applying high compression force to the first layer can result in tablets with low porosity and a bonding strength of the layer that is higher than the interfacial strength of two layers, which may also cause delamination (4). 

	Using GPT, the tablets are compressed separately, and controlling the compression force is easier than in conventional multilayer manufacturing. Tablets composing different formulations can be produced with predefined strength and surface porosity. By gluing the layers in an extra step, the interfacial strength of two layers is only a function of the selected binding agent and is no longer the function of bonding properties of the formulation and the applied compression force. At the same time, the gluing agent acts as a separation layer to prevent direct contact between the tablet layers and cross contamination.

 The tack, in term of adhesive strength, is dependent on the viscoelastic properties of the gluing agent, the evaporation time of the solvent, and the thickness of the applied gluing layer. These parameters should be screened, and a suitable range must be selected carefully. Moreover, the glass transition temperature of the gluing agent is an additional factor that determines the interfacial strength of the tablet layers after production and during storage. Surface properties, such as surface and interfacial tensions and surface roughness of tablet layers, affect the tack properties of the gluing agent. The primary type of interfacial bonding mechanism between the tablet layer and gluing agents with polar functional groups is the hydrogen bond formation. Dipole-dipole and dipole-induced dipole interactions are also possible (5).

 Ibuprofen free acid and caffeine monohydrate were used as model APIs for manufacturing bilayer tablets. Blends of each API (30% w/w) with Tablettose 70 (69.5% w/w) and magnesium stearate (0.5% w/w) were prepared in a blender (Stephan machinery GmbH). The blending time was 10 minutes, at a blending speed of 100 rpm. Blends were directly compressed with a Stylcam 200R with a compaction force of 10 kN for each mixture to generate thin layers of the respective API-excipient mixtures. Bilayer tablets composed of a layer of ibuprofen-free acid and a layer of caffeine monohydrate were produced via GPT using 5 µL of a 40% w/w aqueous solution of fish gelatin with a viscosity of 2.4 Pa-s. The friability and delamination tendency of these tablets were examined. Moreover, bilayer tablets were produced by gluing ibuprofen-free acid tablets and placebo tablets with fish gelatin as a gluing agent. They were used to investigate the API penetration through the gluing layer to confirm if the gluing layer can also act as a separation layer.

 show the hardness and friability of individual layers, and the delamination tendency of bilayer tablets, respectively. The hardness and friability tests were carried out according to the 

) 8.0, 2.9.7., and 2.9.8., respectively, both for single layers and bilayer tablets using 10 bilayer tablets (6). The delamination tendency was determined by considering the number of delaminated bilayer tablets after completing the friability test (Pharmatest) for bilayer tablets. As can be seen from 

, the friability of bilayer tablets was less than 1 % w/w and in full agreement with the 

 requirements. No delamination was observed after performing the friability test.

	Cross-contamination between two layers. To confirm that the glue layer between the individual tablet layers can act as a separation layer, bilayer tablets were produced as follows: a layer of ibuprofen free acid was glued to a placebo tablet containing Tablettose 70 and 0.5% w/w magnesium stearate without API. Fish gelatin was used as the gluing agent. The cross-section of bilayer tablets was analyzed via Raman microscopy. 

 shows such a cross-section where the gluing layer (blue color) also acts as a separating layer. It was observed that the API (green color) does not penetrate the gluing layer, demonstrating that the placebo layer is not contaminated.
	 

	GPT is a straightforward method of producing personalized medicine. It offers an efficient solution for flexible and individual combinations of APIs and doses and can be used in a host of novel applications. A major advantage of GPT is that it requires only a relatively small footprint for the real-time production of multilayer tablets and can easily be integrated in a continuous production line. This feature enables a rapid development and production of new combinations of APIs or individualized API FDCs. Examples are the development of new therapeutic strategies for treatment of HIV/AIDS and multidrug-resistant infections such as tuberculosis. The individualized FDC can be produced real-time on miniaturized equipment by non-governmental organizations in their missions. Real-time production on miniaturized equipment can be also accomplished in the field of translational pharmaceutics, immediately prior to clinical testing, creating the opportunity to modify dose and formulation compositions in response to emerging clinical data. 

 461 (1-2) 549-58 (2014). 
	2. D.J.C. Constable et al., 

 9 (5) 411-420 (2007).
	3. J.B. Manley, P.T. Anastas, and B.W. Ceu Jr., 

 16 (6) 743-750 (2008).
	4. I. Akseli et al., 

 236 30-36 (2013)
	5. L.A. Felton, J.W. McGinity, 

olid Dosage Drug Development and Manufacturing 2017

	When referring to this article, please cite it as S. Salar-Behzadi et al., "Gluing Pills Technology for the Production of Multi-Layer Tablets," 

olid Dosage Drug Development and Manufacturing

This novel technology was developed in response to challenges involved in conventional manufacturing of multilayer tablets, including in-line control of the tablet weight, the tendency to delamination, direct contact between the two tablet layers, and cross contamination.

Gluing Pills Technology for the Production of Multi-Layer Tablets

In the complex and multi-dimensional environment of modern manufacturing operations, it is a challenge to identify the best manufacturing approach for a specific product, taking into account financial, product, and process related issues, including globalization of manufacturing, regulatory requirements, company-specific strategies and preferences, and a complex process development environment. The authors describe a method based on a techno-economic optimization procedure that supports process development by providing a systematic approach to investigating and comparatively evaluating various manufacturing approaches. The methodology can be applied to both traditional and innovative manufacturing technologies and to both existing and new products.

	Process development in the pharmaceutical manufacturing industry requires the identification and implementation of the best possible manufacturing approach for a specific product. Even more challenging, this task is typically performed in a complex and multidimensional environment with often contradictory technical and economic challenges. Additionally, in pharmaceutical manufacturing, a stringent regulatory framework regarding product quality and patient safety must be considered. Accordingly, new pharmaceutical products are commonly developed and manufactured using standardized recipes and an existing inventory of traditional unit operations. Innovative manufacturing technologies emerging in the process equipment market are often ignored or not considered for a lack of role models, equipment providers, or experience and fear of violating regulatory requirements. Although not all areas of pharmaceutical manufacturing are equally affected, it is certainly the case in the secondary manufacturing of pharmaceutical solid dosage forms. The methodology presented in this article, therefore, was initially developed for this area.

	Interest in innovative manufacturing approaches has grown for a number of reasons. First, cost pressure has been increased by generic-drug companies, the government health system, and the consumers of pharmaceutical treatments. Lean manufacturing concepts have been developed, and supply-chain management has begun to implement concepts, such as just-in-time manufacturing, deployed in other industries. Even more, innovative and continuous manufacturing equipment is gradually penetrating the process equipment market. In process development, concepts and technologies associated with process analytical technology and quality by design are becoming standard. Advanced conceptual approaches, such as real-time release, are slowly becoming a part of operations and appear to deliver their full potential when combined with advanced manufacturing approaches, rather than with traditional batch-based processes.

	In summary, pharmaceutical manufacturing is currently in a situation in which traditional approaches are increasingly questioned and new concepts and technologies are gradually finding their way into manufacturing operations. As such, a holistic evaluation of future manufacturing technologies becomes an increasingly important step in process development.

	The methodology presented here simultaneously considers economic and technological aspects and provides a flexible and holistic platform for the evaluation of pharmaceutical manufacturing processes. The goals were to provide:

		A financial justification of a specific manufacturing approach

		An evaluation of the economic and technological feasibility  of a specific manufacturing approach

		A basis for comparing manufacturing approaches and identifying the best one.

	The manufacturing process evaluation methodology can be applied in three areas, as depicted in 

: Identifying the best of the existing manufacturing approaches. Pharmaceutical products are often produced at different company sites or by different manufacturers. Not all of them may be using the same unit operations or manufacturing approach. For example, a product is manufactured via fluid-bed granulation at one site and high shear or dry granulation at another one. The methodology presented in this paper allows systematic investigation of the different manufacturing approaches to determine the best one among them. This is particularly useful for reducing the number of sites and for centralizing manufacturing.

: Identifying the best manufacturing approach for a new product. Manufacturing technologies are typically chosen at an early stage in product development. Often, the decision is based on the type of equipment used in the early development or the inventory of process equipment available in operations, rather than on a systematic search for the best manufacturing approach. By applying the proposed methodology, a rational selection of future manufacturing technologies can be made.

: Evaluating the feasibility of changing from a traditional to an innovative manufacturing approach. For pharmaceutical products that are already on the market, a periodic assessment of manufacturing approaches may help to identify cost-cutting potential or to increase competitiveness vis-à-vis generic-drug manufacturers after the product’s patent protection expires. In addition, a simple assessment of an existing manufacturing process can eliminate the existing shortcomings or limitations in terms of yield, product quality, or cost of goods by switching to a different manufacturing process. The methodology presented in this paper can help to identify the differences between established and innovative manufacturing technologies.

	Tools traditionally applied for financial analysis can generally be applied for techno-economic manufacturing process evaluation as well. The most relevant ones are briefly discussed as follows

. A cost-benefit analysis is a common technique to assist decision-making that aims at expressing benefits and costs of a project or investment in monetary terms, preferably as the net present value. The most desirable solution within a given set of options can thus be identified. A cost-benefit analysis, however, is ineffective when the benefits cannot be directly expressed in monetary terms, which is often the case in process engineering applications with complex and multidimensional product- and process-related evaluation criteria. Therefore, in process development, a cost-benefit analysis is primarily applied during the final decision-making steps and not in the early or conceptual phase

. A cost-effectiveness analysis relates the cost of a given option to a specific objective or optimization criterion. For example, if an objective were to reduce the overall manufacturing time in days, a cost-effectiveness analysis could be the first step in comparing the various options, followed by a cost-benefit analysis to produce monetary values for the decision-making process.

. A cost-utility analysis relates costs to a multidimensional measure of effectiveness, referred to as “utility” in economic analysis, as a representation of preferences. The utility value is a value assigned to a project or investment based on the anticipated performance. A cost-utility analysis can be used to measure the financial and technological dimensions of various options and generate a combined economic and technical efficiency.

	Because, from the start, a manufacturing process evaluation contains a number of economic and technical performance or compliance parameters, a cost-utility analysis is the tool of choice for an early or conceptual process evaluation. As process development progresses and risks and technical criteria become specified in more detail, a cost-benefit analysis can be applied if necessary.

	The evaluation methodology for a manufacturing process presented here is derived from classical cost-utility analysis and consists of the six consecutive steps described in

: Definition of ranking criteria and risk-profile elements. The first step is to define which ranking criteria to include in the process evaluation. This selection should be made by a team of technical and financial experts and will comprise cost-related and product- as well as process-related ranking criteria.

		Cost-related ranking criteria can include the following:

		Depreciation (e.g., of building, infrastructure, or process and analytical equipment)

		Direct material costs for primary manufacturing (e.g., raw materials) and secondary manufacturing (e.g., API and excipients, primary and secondary packing materials)

		Utility costs (e.g., energy, water, and HVAC)

		Costs of cleaning, preventive maintenance, and repairs

		Production area and factory overhead costs.

		Product-related ranking criteria can include the following:

		Compatibility with the existing dosage form and composition (regulatory product profile of the existing product, if available)

		Dosage form size, shape, and presentation

		Process-related ranking criteria can include the following:

		Commercial availability of process equipment

	If necessary, additional categories can be included in the evaluation procedure. Manufacturing in different countries, including the availability of trained personnel, wage levels, and energy prices, may be considered, for example.

	Risks are economic, technical, or other (e.g., regulatory) aspects not known at the time of the evaluation but potentially affecting the performance of the selected manufacturing approach. Because risks cannot be quantified, and, thus, cannot be directly incorporated into the process evaluation, each process evaluation includes an associated risk profile to indicate potentially critical unknowns that must be further investigated if the respective process option is ranked favorably.

		Risk profile elements may include the following:

		Changes in the solid state of the API vs. the desired state

		Residuals from chemical manufacturing in the product

		Problems with stability that reduce shelf life

		Technical concerns of the evaluation team.

: Definition of quantitative and qualitative ranking criteria. Each ranking criterion can be included in a quantitative or qualitative ranking procedure. If the former is applied, a direct numerical value is assigned as a measure of performance or compliance. Typical quantitative criteria would be available in monetary terms or in technical units (e.g., a footprint of the unit operation in square meters or the maximum output in doses per hour).

	If a quantitative figure cannot be obtained at the time of the evaluation, a qualitative ranking procedure can be used. Instead of actual numerical values, scores are applied to quantify the estimated magnitude of costs or the degree of technical compliance.

	Costs can be ranked qualitatively, for instance, by assigning the following five score levels:

		A score of 2 if a significant (>30%) cost reduction is expected

		A score of 1 if a minor cost (>10 and ≤30%) reduction is expected

A techno-economic profiling method supports pharmaceutical process development by helping identify the best manufacturing approach.