Achieving Process Understanding and Control in Film Coating

Coating tablets involves selecting a specific color and gloss that are used to distinguish a drug product and build brand identity for that product. In addition to the issue of color and gloss, the coating must have appropriate surface-slip properties and adequately adhere to the tablet. The coating process and the materials used in the coating system can be optimized using a quality-by-design (QbD) approach to meet these requirements. The authors describe a QbD study that was performed to optimize a coating system.

Critical quality attributes (CQAs) for a tablet coating are color, slip, gloss, and adhesion. A quality-by-design (QbD) study that uses a statistical design and data analysis can identify the critical process parameters (CPPs) needed to obtain the desired CQAs.

CPPs include both the material composition (i.e., formula) and parameters in the coating process. In this study, quality function deployment (QFD) was used to identify the expected effects of material composition on the CQAs. QFD can be used to minimize the number of experiments, which limits experimental error. Based on the QFD shown in Table I, the amount of titanium dioxide (TiO₂) and a red dye were identified as the CQAs for color, and it was determined that a mixture design was not necessary to test the effect of material composition on color. Instead, a three-factor full factorial with a center point for color effect was chosen as a practical design of experiment (DOE).

Table I: Quality function deployment for effects of material on coating properties (Spectrablend II Pink, Sensient Pharmaceutical Coating Systems). (ALL TABLES ARE COURTESY OF THE AUTHORS)

The coating system compositions were studied. The level of red dye (Red 40 Lakes, 12–14% concentrate) and the level of TiO₂were varied to set the design space for the color effect. Medium-chain triglyceride (MCT) levels were adjusted to maintain the mass balance. Table II lists the temperature and spray-rate process variables used to test each of the nine coating-system compositions.

Table II: Experimental coating conditions.

A 1200-g batch of placebo tablets was coated with each of the nine coating systems. After spray coating, 3 and 5% weight-gain samples were collected and rolled in the coating pan for additional 5 min. The percentage of weight gain was based on theoretical coating solution consumption and did not account for coating efficiency. A total of 45 coating experiments were conducted, and 90 samples were collected. Samples were tested, with five replicates, for color, gloss, slip, and adhesion.

Color quality

Colorimetry was used to measure the color shade. The L* value is the white/black color space. The a* value is the red/green space, and the b* value is the blue/yellow space. Target L*, a*, and b* values are listed in Table III. The total color difference (ΔE*) is calculated from the differences between L*, a*, and b* values and the target by taking the square root of the sum of differences squared. Specifications for L*, a*, and b* values were defined based on ΔE* < 2.0.

Table III: Target colorimetry values (Spectrablend II Pink, Sensient Pharmaceutical Coating Systems).

All 450 samples were analyzed by regression to detect the factors that had a significant L* value. The data analysis revealed that all six factors (i.e., level of titanium dioxide, red dye, and MCT in the coating system, and percentage of weight gain, coating exhaust temperature, and spray rate) had significant effect on the L* value. The regression model had a R² (adjusted) of 92.10% and a lack-of-fit p-value of 0.000. The relative effect of the variables is shown in Figure 1. The data analysis revealed that TiO₂ and red dye had the most significant effect; both exhaust temperature and spray rate also had a significant effect on L*.

Figure 1: Pareto chart of the standardized effects of the experimental variables on the L* value, which describes the white/black color space, Alpha = 0.10. (ALL FIGURES ARE COURTESY OF THE AUTHORS)

A similar regression analysis was performed to identify the factors that may have significant effect on the a* value. The data analysis revealed that the same six factors had significant effect on the a* value. The regression model had a R² (adjusted) of 91.57% and a lack-of-fit p-value of 0.000. The relative effect of the variables is shown in Figure 2. The data analysis revealed that the levels of TiO₂ and red dye had the most significant effect; both exhaust temperature and spray rate also had significant effect on the a* value.

Figure 2: Pareto chart of the standardized effects of the experimental variables on the a* value, which describes the red/green color space, Alpha = 0.10.

Lastly, a similar regression analysis was performed to identify the factors that may have a significant effect on the b* value. The data analysis revealed only four factors (i.e., TiO₂ and red dye levels, percentage of weight gain, and coating exhaust temperature) had significant effect on the b* value. The regression model had a R² (adjusted) of 86.6% and the lack-of-fit p-value of 0.000. The relative effect of the variables is shown in Figure 3. The data analysis revealed that TiO₂ and red dye levels had the most significant effect; both exhaust temperature and spray rate also had a significant effect on b* value.

Figure 3: Pareto chart of the standardized effects of the experimental variables on the b* value, which describes the blue/yellow color space, Alpha = 0.10.

An optimization of the factors targeted for the L*, a*, and b* value was calculated using multiregressional analysis. Optimal material and process factors are highlighted in red in Figure 4, which shows that the colorant level had a larger effect on color shade compared with the process effects (e.g., exhaust temperature and spray rate). For the optimized solution at 3% weight gain, the predicted L*, a*, and b* values were 75.86, 32.21, and 9.07, respectively. The predicted ΔE* was 0.0, which is a perfect match. A process specification of the TiO₂ and red dye was also calculated based on the lower and upper percentage limits of both ingredients. The white area of the plot in Figure 4 is the allowable level of the color ingredient that would meet ΔE* < 2.0. A different color specification can be determined by using the same design space.

Figure 4: Contour plot of calorimetry measurements (L*, a*, and b* values) in relation to red dye and titanium dioxide concentrations in the formulation. Variables held constant were weight gain (3%), medium chain triglyceride concentration (7.0%), exhaust temperature (40 °C), and spray rate (10 g/min).

Gloss quality

Gloss is a measure of the proportion of light that has a specular reflection from the surface. A smooth surface has a high gloss, while a rougher surface has less, because the light reflected is diffused. High percentages of solid in the dry blend can make a rougher surface that can affect the gloss. Process conditions, such as coating temperature and spray rate, may also have a significant effect on the film surface rheology, which can affect the gloss. A novo-curve gloss meter was used to measure the film-coating gloss on the tablet coating with 60° light reflection.

Regression analysis showed a weak model with R² (adjusted) of 40.45% even though the lack-of-fit p-value was 0.000. A low R² (adjusted) means that other significant factors were not accounted for or that the process is not in control. A weak model resulted in low predictability. It is known, however, that the gloss can be induced by the batch weight and the pan rolling time after the spraying is completed.

The statistical summary of the gloss data showed a broad range of gloss results. Analysis of variables (ANOVA) was used to deduce the main effect. Based on the ANOVA result, high levels of TiO₂ and red dye had a negative effect on the gloss. High percentage of weight gain had a positive effect on gloss. High temperature and high spray rate had a positive effect on gloss but had no effect above a 45 °C exhaust temperature and a 20 g/min spray rate. High solids percent is expected to cause more roughness, which can affect the light reflection. MCT level did not show a significant effect on gloss. The requirement for color matching constrains the TiO₂ and red dye levels, but a minimum coating exhaust temperature of 45 °C and spray rate of 20 g/min are recommended to achieve the maximum gloss.

Slip quality

Slip (i.e., surface friction) of the film coating was measured using a texture analyzer. Five tablets were measured for each experiment (i.e., a total of 2250 samples) to obtain a statistical justification, and the data were analyzed by regression. The R² (adjusted) of 68.80% was marginally acceptable. Regression analysis indicated that spray rate, exhaust temperature, and level of MCT also have a significant effect on slip. The regression model predicted that a minimum slip of 27.56 g-force could be achieved with the recommended material composition and process conditions. Because the material composition must keep constant to match the target color, the optimal response was adjusted using weight gain and titanium dioxide and red-dye levels.

Adhesion quality

Adhesion quality was determined by measuring the tension force on a tablet. A tablet was placed on a piece of double-sided tape adhered to the top of the texture analyzer's flat platform. Another piece of tape was pressed flat to the bottom of a 25-mm, stainless-steel, cylindrical probe, compressed to 800-g force onto the tablets for 10 s, then pulled apart at 1 mm/s. Regression analysis had a R² (adjusted) of 59.04%, which suggested unaccounted-for factors or high experimental error. The data distribution for the adhesion data was a non-Gaussian distribution, which was fitted as a 3-parameters Weibull distribution. The probability of failure based on the design space was estimated to be low (0.94%), which indicated that adhesion was not sensitive to material factors (i.e., TiO₂, red dye, and MCT levels) or process factors (i.e., exhaust temperature and spray rate).

QbD roadmap

A typical QbD exercise is to outline the process flow diagram and construct the process mapping to identify the key process input variables (KPIVs) and key process output variables (KPOVs). The relative importance of the KPIVs and KPOVs were assessed with a cause-and-effect (C&E) matrix. The KPOVs with the highest scores (e.g., in the top 20th percentile) were considered to be CQAs for which process specifications needed to be defined. The C&E matrix result showed that the blending uniformity before and after the addition of MCT was critical. The blend uniformity before MCT addition determined the blend uniformity after MCT addition. Therefore, there was only one CQA in the blending process. The detection sensitivity needed to determine the blend uniformity was based on visual detection by obtaining a blend drawdown. Visual detection under magnification, which can further ensure uniformity quality, is recommended. From the C&E matrix, the KPOVs with high scores (i.e., raw material charge and sample checking for blend uniformity) were defined as the CPPs. It is important to control the CPPs to achieve desirable quality. A failure mode effect analysis (FMEA) was performed based on the CPPs identified.

Coating QbD

Based on the FMEA assessment, the risk-priority number (RPN) was generated, and the raw-material charge had a high RPN. Corrective action plans were developed to control the raw-material charge procedure and ensure accuracy and execution of the production procedure. Similarly, the drawdown step was also a CPP. The RPN was not as severe compared with the raw material charge step, and a process-control procedure was also proposed. The C&E assessment indicated a completely dispersed Spectrablend II pink in water was a CQA. This CQA could be controlled by visual testing. All KPIVs were CPPs. The coating process had six process variables (i.e., coating temperature, spray rate, pan speed, atomization pressure, AIR pattern, and air flow).

Relative coating effect

The C&E assessment suggested coating efficiency, drying efficiency, and coating quality were the CQAs. Drying efficiency is a visual observation during coating and is dependent on the operator's experience. Coating quality can be quantitatively measured by colorimetry and other rheological tests. The C&E assessment suggested all six KVIPs had a significant effect. A FMEA assessment was performed.

FMEA for coating process

Tablet charge, air flow, coating temperature, spray rate, and pan speed needed to be controlled. Tablet-charge amount should be maximized according to the pan-coater design. Air flow was correlated to coating temperature and should be adjusted to the optimal setting to stabilize the coating temperature. Pan speed should be adjusted to have an ideal flow without excessive tumbling to cause tablet or coating chipping. The coating temperature and spray rate effect were more subjected to the coating material as previously described.

Conclusion

Critical quality attributes, such as color shade, gloss, slip, and adhesion, can be controlled by proper material and process specification. A QbD approach is a practical tool to understand and control quality and the process.

Eric Van Ness* is technical support manager, Beverly Schad is technical sales manager, Thomas Riley is formulations manager, and Brian Cheng is formulations scientist, all at Sensient Pharmaceutical Coating Systems, Saint Louis, MO, tel. 314.286.7135, Eric.VanNess@sensient.com.

*To whom all correspondence should be addressed.