3. Global initiatives References 1. ICH, Q8(R1) Pharmaceutical Development (Geneva, Switzerland, Nov. 10, 2005; Rev. 2008). 2. ICH, Q9 Quality Risk Management (Geneva, Switzerland, Nov. 9. 2005). 3. J. Agalloco et al., "FDA's Guidance for Industry: Process Validation: General Principles and Practices," presented at PDA, Jan. 14, 2009. 4. FDA, Draft Guidance for Industry—Process Validation: General Principles and Practices (Rockville, MD, Nov. 2008). 5. W. Charlton, T. Ingallinera, and D. Shive, "Validation of Clinical Manufacturing," and Validation Chapter, in Validation of Pharmaceutical Process, J. Agalloco and F. Carleton, eds. (Informa Healthcare, New York, 3rd ed., 2008), pp. 542–544.
47. Quality by design (QbD) Process flow: Screening Characterization range Identification of significant parameters Acceptable range Finding parameter ranges Operating range Optimization Finding interactions of parameters Defining models Set point Validation Process design space Identification of CPP Identification of noise factors Process/ product Development: Robust Cost effective Feasible Defining control strategies Production Continuous monitoring and development
51. Screening Parameter selection Physical Chemical Raw material Component/Equipment Process (time, type) Environmental Facility Categorical Continuous
57. Screening Fractional Factorial 23-1 C=AB C is confounding with AB B=AC B is confounding with AC A=BC A is confounding with BC
58. Screening Fractional Factorial PlackettBurman 2 level fractional factorial designs Resolution III design Efficient estimations Interactions between factors ignored Used in Matrix form Multiple of 4 not power of 2 Saturated orthogonal array
60. Screening Lack of fit Before deciding whether to build a response surface model, it is important to assess the adequacy of a linear model The error term ε in the model is comprised of two parts: modeling error, (lack of fit, LOF) experimental error, (pure error, PE), which can be calculated from replicate points The lack of fit test helps us determine if the modeling error is significant different than the pure error
61. Screening Lack of fit Before deciding whether to build a response surface model, it is important to assess the adequacy of a linear model The error term ε in the model is comprised of two parts: modeling error, (lack of fit, LOF) experimental error, (pure error, PE), which can be calculated from replicate points The lack of fit test helps us determine if the modeling error is significant different than the pure error
62. DOE and Experiments RS Model 1 x2 x1 Response surface methodology Input Response Black Boxed System Original System
63. Response surface methodology RSM characteristics Models are simple polynomials Include terms for interaction and curvature Coefficients are usually established by regression analysis with a computer program Insignificant terms are discarded Model equation for 2 factors Y = β0constant + β1X1 + β2X2 main effects + β3X12 + β4X22 curvature + β5X1X2 interaction + ε error Model equation for 3 factors Y = β0constant + β1X1 + β2X2 + β3X3 main effects + β11X12 + β22X22 + β33X32 curvature + β12X1X2 + β13X1X3 + β23X2X3 interactions + ε error Higher order interaction terms are not included
64. Response surface methodology Central composite design (CCD) eg. 2 factor Central composite circumscribed (CCC) 5 Levels α (star point) are beyond levels Central composite face centered (CCF) 3 Levels α (star point) are within levels (center) Central composite inscribed (CCI) 5 Levels α (star point) are within levels Scale down of CCC
65. Response surface methodology Central composite design (CCD) Central composite circumscribed (CCC) 3 factors Total exp: 20 Full factorial 8 Axial points 6 Center points 6 +++ -+- --- +-- -++ ++- --+ +-+
66. Response surface methodology Central composite design (CCD) Central composite circumscribed (CCC) Randomization: To avoid effect of uncontrollable nuisance variables +++ -+- --- +-- -++ ++- --+ +-+
67. Response surface methodology Central composite design (CCD) Central composite circumscribed (CCC) Blocking: To avoid effect of controllable nuisance variables -++ +++ +-+ --+ ++- -+- +-- ---
71. It is possible to estimate main effects and second order terms
72. Box-Behnken experiments are particularly useful if some boundary areas of the design region are infeasible, such as the extremes of the experiment regioneg. 3 factor 12 experiments
73. Response surface methodology Comparison of RSM experiments * One third replicate is used for a 3k factorial design and one-half replicate is used for a 2k factorial design with the CCD for 5, 6 and 7 factors.
77. Robust process development Objective of robust process Smaller-the-Better S/N Ratio = – 10 Log10 ( 1/n Yi2 ) e.g. defects, impurity, process time, cost Larger-the-Better S/N Ratio = – 10 Log10 ( 1/n 1/Yi2 ) e.g. titre, yield, resolution, profit Nominal-the-BestS/N Ratio = – 10 Log10[1/n(YIDEAL- Yi ) 2 ] e.g. target Signal-to-Noise S/N Ratio =10log[μ2/σ2] e.g. trade-off
78. Robust process development Identification of Signal and noise eg: Fermentation Signal: What can be controlled in plant and laboratory Noise: What can not be controlled in plant but in laboratory
79. Robust process development Developing robust process To find a signal settings in presence of noise that minimize response variation while adjusting of keeping the process on target Taguchi approach Signal: Inner array Noise: Outer array