OR WAIT 15 SECS
Adeola O. Grillo, PhD, is an assistant professor in the department of pharmaceutical sciences at the Feik School of Pharmacy, University of the Incarnate Word
Martin Kane is an associate director of process statistics in the department of biostatistics, in biopharmaceutical development at Human Genome Sciences
Melissa Perkins, PhD, is the director of the department of drug product sciences, all in biopharmaceutical development at Human Genome Sciences
NeÃ§ois Penn is a bioprocess associate in the department of drug product sciences in biopharmaceutical development at Human Genome Sciences
Design of experiments is a valuable tool for identifying aspects of a formulation that are critical to product quality.
Design of experiments (DOE) is a valuable tool for identifying aspects of a formulation that are critical to product quality. The formulation design space can be characterized by performing excipient robustness studies that use DOE. This paper presents considerations for performing robustness studies as well as two case studies in which DOE was used to determine the robustness of protein formulations to changes in protein, excipient, and pH levels. The results from the DOE studies identified formulation components that must be tightly controlled and showed that variations had a minimal impact to product in formulation component levels within the formulation design space.
The US Food and Drug Administration's Quality by Design (QbD) initiative encourages pharmaceutical manufacturers to use modern tools that facilitate the implementation of robust manufacturing processes and reliably produce pharmaceuticals of high quality.1 Filing a new drug application (NDA) or biologics license application (BLA) under the QbD initiative may reduce the extent of regulatory oversight and may result in faster review times. Implementing a QbD approach in the development and characterization of manufacturing processes and products provides several advantages. These include a more thorough understanding of the manufacturing process and the product, as well as the potential for increased process and product robustness and process efficiency. The International Conference on Harmonization (ICH) Q82, Q93, and Q104 guidelines describe principles, tools, and examples for implementing QbD. One of these tools is the use of formal experimental designs or design of experiments (DOE) to characterize and establish a functional design space.
Human Genome Sciences
DOE is a tool that can be used during formulation development to screen for stabilizing excipients, determine excipient levels that provide optimal stability with adequate robustness, and characterize the formulation design space. The design space is defined in ICH Q8 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." The formulation design space would thus define limits for the active pharmaceutical ingredient (API), excipients, pH ranges, and other critical characteristics in the formulation that maintain product stability. Because multiple formulation components are screened simultaneously, the combinations that provide optimum stability and interactions among formulation components can be identified. For example, in protein formulations that contain sodium chloride (NaCl), an interaction between pH and NaCl concentration is likely to be observed because both modulate electrostatic interactions. A DOE formulation screening study can thus help to identify optimum levels of pH and NaCl for stability.5 Similarly, in lyophilized or spray-dried protein formulations, an interaction may be observed between protein and cryo/lyoprotectant levels because an optimum ratio of the stabilizer to protein is required for cold and dehydration-induced denaturation.6,7 DOE also can be used to identify effective preservatives for multi-dose protein formulations at optimal concentrations for antimicrobial efficacy and protein stability, to characterize the effect of stress conditions and degradation pathway mechanisms in forced degradation studies, and to characterize the effect of formulation components on drug delivery systems.8–13
The robustness of the formulated product can be characterized further in excipient robustness studies that use DOE.14,15 Excipient robustness studies performed using DOE can:
Excipient robustness studies can, hence, be used to evaluate and characterize the formulation design space (Figure 1).
When designing an excipient robustness study, the following factors should be considered to appropriately select the formulation ranges, analytical methods, and experimental design used in the study:
Stage of development: The extent of robustness studies required is dependent on protein availability and resource commitment. Thorough robustness evaluations typically are performed in late-stage development, as the commercial formulation and configuration are being locked.
Manufacturing experience: Protein, excipient, and pH levels in historical lots can help give guidance for selecting appropriate ranges for the study. This is especially important for biotechnology products because purification processes such as the ultrafiltration–diafiltration (UF–DF) step typically used to formulate the product may result in pH and excipient levels different from the formulation buffer.16,17 The number of lots and manufacturing campaigns used to assess the manufacturing history also should be considered. If there is limited manufacturing experience before starting a robustness study, the ranges of the formulation components used in the study can be set relatively wide to cover potential future ranges. Ideally, the excipient robustness study is performed early enough in product development so that changes can be made to the target level and acceptable formulation ranges if warranted.
Target level and acceptable range: Development, characterization, and historical manufacturing data typically are used to set target levels and normal ranges in manufacturing batch records and a product specification before performing a robustness study. The formulation ranges selected for the study must go beyond the acceptance criteria defined in batch records and product specifications. Selecting wider ranges helps to confirm that variations within current normal operating ranges do not statistically or practically affect product stability. By evaluating wider than normal operating ranges, one also can define the acceptable range that does not effect the product (design space). The ranges selected for the study are limited by levels that may result in significant degradation (see below) or affect patient safety. Previously established levels of generally regarded as safe (GRAS) excipients can be found in the FDA inactive ingredients database and used as a guide. Results from the robustness study can then be used to justify the batch record ranges and product specification.
Results from development and characterization studies: Development experience can be used to ensure that ranges selected are not so wide that high levels of degradation are observed making interpretation of the data difficult. Data from previous studies that provide information on product impact from variations in formulation ranges (i.e., results from formulation screening studies) can be assessed and used to select appropriate ranges for the excipient robustness study.
Material availability to perform the study: Because parameters are evaluated simultaneously in a DOE study, the material requirements can become very large. The study design must be evaluated against the material availability and modified appropriately (if possible). Options to reduce material needs include the use of scale-down models of the bulk drug substance (BDS) and final drug product (FDP) container-closure configuration, reducing the testing frequency on long-term storage, and using accelerated stability data to establish robustness at earlier time-points. Screening designs also can be used (see below).
Experimental design: Examples of designs include screening, full factorial, fractional factorial, and response-surface designs. Screening designs are used to screen the effect of factors on different responses and are amenable to experiments with a large number of factors. A full factorial design allows the identification of the formulation components that have a significant effect on stability and all the interactions among the formulation components. However, depending on the number of formulation components, a fractional factorial design that also allows for the identification of significant factors, but not all the interactions, may be more appropriate. To identify curvature (for example, an optimal range for stability), a response surface design is required. Examples of response surface designs include central composite design (CCD) and Box-Behnken design. Central composite designs can be used when evaluating at least two factors, whereas Box-Behnken designs can be used only for three or more factors.
Central composite designs contain points in a factorial design (cube points) with additional points on the axes that allow curvature to be estimated (Figure 1). Box-Behken designs, on the other hand, contain points on the center of the edges of the cube and can allow for fewer design runs while still being able to analyze curvature in the responses. The main benefit of a Box-Behken design is that none of the factorial design points get run. This means combinations of extremes, for instance the high-high-high or low-low-low combinations in a 3-factor design, do not get run. Table 1 gives examples of the number of runs required for full-factorial and fractional factorial screening, central composite design, and Box-Behnken designs as a function of the number of factors being evaluated.
Table 1. Number of runs1 as a function of the number of factors (formulation components) evaluated for each type of experimental design
It is typical to use a screening design when there are a large number of factors that must be reduced to a manageable few. Sometimes follow-up screening designs are needed to separate confounded (confused) interaction terms from one another. After the important few factors are determined, a response surface design is used to get a complete understanding of the response and how it is correlated to the factors. Because of the use of DOEs, where the changes in the responses are directly correlated with the changes in the factors, the analysis of response surface designs provides a mathematical model that can be used in manufacturing as a control scheme.
In any design, it is important to run some replicate points—points that are run multiple times—to be able to asses the random error in the response models. It is typical, but not necessary, to make these replicates the center-point condition and to run somewhere between 3 and 6 center-point runs. The other benefit of using centerpoints for the error analysis is that these runs can be used to determine process stability through the use of a statistical process control (SPC) analysis.
Analytical capability: Analytical methods used must be sensitive and appropriate for detecting changes in the stability of the product within the experimental design space. In addition to the standard methods used to release the product, characterization methods provide a more thorough evaluation of the formulation design space. Similarly, replacing qualitative methods (such as visual inspection) with quantitative (such as turbidimetry and the USP tri-stimulus method for color analysis) also help to provide a more thorough understanding of the design space. Analytical method attributes such as repeatability, intermediate precision, the limit of detection (LOD), and limit of quantitation (LOQ) determined during method qualifications must be considered when interpreting results obtained from statistical analysis of the experimental design (see below).
Statistical software capability: Results obtained from statistical analysis of the experimental data are based on the input parameters. Accurate interpretation of the data may require consultations with statistical experts. In addition, the formulator's knowledge of degradation pathways and analytical method capabilities is needed to properly interpret the data. For example, methods with high data precision may result in identifying many statistically significant factors while the practical significance of the results observed requires the formulator's knowledge of quality attributes and what constitutes an acceptable change.
Two case studies are presented in which a DOE approach was used to determine the robustness of protein formulations to changes in protein, excipient, and pH levels. In case study 1, the results indicated that the formulation was robust to wide variations in excipient, protein, and pH levels; much wider than could occur during manufacturing of the product. In case study 2, the results indicated that the formulation was robust to wide variations in excipient and proteins levels, but not pH. Thus, in case study 1, the results were used to support the conclusion that variations in excipient levels had a minimal impact to product whereas in case study 2, results indicated that the pH of the formulation needed to be tightly controlled.
Case Study 1
In case study 1, the robustness of a high protein concentration lyophilized product to protein, excipient, and pH ranges was evaluated. The excipients consisted of a buffering agent, lyo/cryoprotectant, and surfactant. During manufacturing, excipient levels are weighed within 1.5% of target, but some vary wider in the product because of the retention of excipients during the UF–DF process. Excipient levels evaluated in the study were well beyond excipient concentrations expected so that a broad design space could be evaluated. Ranges were ±25% of target concentration for the cryo/lyo-protectant, ±50% of target concentration for the surfactant, and –50% to +100% of target concentration for the buffering agent. The buffering agent was evaluated up to +100% to establish design space at higher buffer concentrations. To accommodate the nonlinear nature of the buffering range, the factor levels were log-transformed, arising in a buffering range of ±0.301 log units of target. It is important to keep the design dimensions equidistant around the center point of interest because this allows for an analysis in which each of the design points gives equal weight to the responses.
Table 2. Screening design for case study 1
Although a full-factorial, response-surface design would have provided information on curvature in the response and all possible interactions, this would have required the preparation of 42 different formulations excluding the center points. To reduce the number of formulations and still be able to determine formulation components and some 2-factor interactions that significantly affected stability, a fractional factorial screening study was used (Table 2). Samples were monitored at intended and accelerated storage conditions by size-exclusion and ion-exchange chromatography, as well as SDS-PAGE, and tested for moisture and potency. Degradation observed was limited to aggregation. Statistical analysis of the stability data using JMP (Cary, NC) showed that protein and cryo/lyo-protectant concentrations were the only formulation components that significantly affected stability. The prediction profile for the impact on stability as a function of formulation components is shown in Figure 2A. Increasing protein concentrations and decreasing cryo/lyoprotectant concentrations negatively affected stability. In addition, a protein:cryo/lyo-protectant interaction was observed such that the impact of the cryo/lyo-protectant on stability was more evident at the higher protein level. Similarly, the impact of protein concentration on stability was more evident at lower cryo/lyo-protectant concentrations (Figure 2B). The robustness of the formulation was demonstrated by the limited degradation observed at 5 °C, the intended storage temperature for the product, even in the least stable formulations containing the highest protein:cryo/lyoprotectant ratio.
Case Study 2
In case study 2, the robustness of a high concentration liquid product to protein, excipient, and pH concentrations was evaluated. The excipients consisted of a buffering agent, tonicifying agent, cryoprotectant, and surfactant. Excipient ranges evaluated in the study were within 20% of cryoprotectant and tonicifying agent target concentrations and 50% of buffering agent and surfactant target concentrations. Because of the number of formulation components, a fractional–factorial screening design was also used in this study. Samples were monitored at intended and accelerated storage conditions by size-exclusion, ion exchange, and reverse-phase chromatography, as well as by potency assay. Degradation was observed by size-exclusion and ion exchange chromatography. The only factor that had a significant effect on stability was pH, resulting in increased aggregation and deamidation (data not shown). The results from the excipient robustness study prompted well-defined pH acceptance criteria in buffer preparation records and on the product specification.
The two case studies illustrate two outcomes that can be obtained from an excipient robustness study:
Irrespective of the outcomes, controls for protein, excipient, and pH levels must be established in the batch records or product specifications. Controls in the batch record may include acceptance criteria for buffer and excipient level weigh-outs for the formulation buffer, the pH of the formulation buffer, as well as pH and protein concentrations of the formulated BDS and FDP. Additionally, controls in the BDS and FDP specifications may include acceptance criteria for pH, protein concentration, and osmolality. The acceptance criteria are selected such that the quality of the drug product is not affected beyond an acceptable level. If the robustness study indicates that the formulation is not robust to wide variations in one or more of the formulation components, controls are put in place to ensure that acceptable levels are achieved consistently.
Quality by Design tools such as design of experiments (DOE) help to provide a more thorough understanding of a product's design space. Excipient robustness studies using DOE are, thus, used to evaluate and characterize the formulation design space. The formulation design space is identified by the ranges established during the excipient robustness study to provide an assurance of quality in the product. Controls included in manufacturing batch records and product specifications ensure that the product is maintained within the formulation design space so that the quality of the product and safety of the patient are ensured.
At the time of the article's writing, Adeola O. Grillo, PhD, was a senior scientist in the department of drug product sciences, biopharmaceutical development, Human Genome Sciences, Rockville, MD. She is currently an assistant professor in the department of pharmaceutical sciences at the Feik School of Pharmacy, University of the Incarnate Word, San Antonio, TX, 210.883.1099, email@example.comMartin Kane is an associate director of process statistics in the department of biostatistics, Neçois Penn is a bioprocess associate in the department of drug product sciences, and Melissa Perkins, PhD, is the director of the department of drug product sciences, all in biopharmaceutical development at Human Genome Sciences, Rockville, MD, firstname.lastname@example.org.
1. US Food and Drug Administration. Guidance for industry. Pharmaceutical cGMPs for the 21st century—A risk-based approach final report. Rockville, MD; 2004 Sept.
2. International Conference on Harmonization (ICH). Q8(R1), Pharmaceutical Development. Geneva, Switzerland; 2008 Nov.
3. ICH. Q9, Quality Risk Management. Geneva, Switzerland; 2005 Nov.
4. ICH. Q10, Pharmaceutical Quality System. Geneva, Switzerland; 2008 June.
5. Bedu-Addo F, Moreadith R, Advant SJ. Preformulation development of recombinant pegylated staphylokinase SY161 using statistical design. AAPS PharmSci. 2002;4(4):E19.
6. Cleland JL, Lam X, Kendrick B, Yang J, Yang TH, Overcashier D, et al. A specific molar ratio of stabilizer to protein is required for storage stability of a lyophilized monoclonal antibody. J Pharm Sci. 2001;90(3):310–21.
7. Andya JD, Maa YF, Costantino HR, Nguyen PA, Dasovich N, Sweeney TD, et al. The effect of formulation excipients on protein stability and aerosol performance of spray-dried powders of a recombinant humanized anti-IgE monoclonal antibody. Pharm Res. 1999;16(3):350–8.
8. Gupta S, Kaisheva E. Development of a multidose formulation for a humanized monoclonal antibody using experimental design techniques. AAPS PharmSci. 2003;5(2):E8.
9. Fransson J, Hagman A. Oxidation of human insulin-like growth factor I in formulation studies, II. Effects of oxygen, visible light, and phosphate on methionine oxidation in aqueous solution and evaluation of possible mechanisms. Pharm Res. 1996;13(10):1476–81.
10. Maltesen MJ, Bjerregaard S, Hovgaard L, Havelund S, van de Weert M. Quality by design—spray drying of insulin intended for inhalation. Eur J Pharm Biopharm. 2008;70(3):828–38.
11. Guo C, Stine KJ, Kauffman JF, Doub WH. Assessment of the influence factors on in vitro testing of nasal sprays using Box-Behnken experimental design. Eur J Pharm Sci. 2008;35(5):417–26.
12. Dayal P, Pillay V, Babu RJ, Singh M. Box-Behnken experimental design in the development of a nasal drug delivery system of model drug hydroxyurea: characterization of viscosity, in vitro drug release, droplet size, and dynamic surface tension. AAPS PharmSciTech. 2005;6(4):E573–85.
13. Zidan AS, Sammour OA, Hammad MA, Megrab NA, Habib MJ, Khan MA. Quality by design: understanding the formulation variables of a cyclosporine. A self-nanoemulsified drug delivery systems by Box-Behnken design and desirability function. Int J Pharm. 2007;332(1-2):55–63.
14. Campisi B, Chicco D, Vojnovic D, Phan-Tan-Luu R. Experimental design for a pharmaceutical formulation: optimisation and robustness. J Pharm Biomed Anal. 1998;18(1–2):57–65.
15. Gabrielsson J, Sjöström M, Lindberg NO, Pihl AC, Lundstedt T. Robustness testing of a tablet formulation using multivariate design. Drug Dev Ind Pharm. 2006;32(3):297–307.
16. Stoner MR, Fischer N, Nixon L, Buckel S, Benke M, Austin F, Randolph TW, Kendrick BS. Protein-solute interactions affect the outcome of ultrafiltration/diafiltration operations. J Pharm Sci. 2004;93(9):2332–42.
17. Harinarayan C, Skidmore K, Kao Y, Zydney AL, van Reis R. Small molecule clearance in ultrafiltration/diafiltration in relation to protein interactions: Study of citrate binding to a Fab. Biotechnol Bioeng. 2009;102(6):1718–22.