• 沒有找到結果。

62

A batch size can also be defined as a range. For example, a batch size range can be established by 63

defining a minimum and maximum run time.

64

3. SCIENTIFIC APPROACHES 65

3.1. Control Strategy 66

The development of a successful control strategy for CM is enabled by a holistic approach, 67

considering aspects specific to CM (discussed below) and the principles described in ICH Q8–

68

performance and product quality. The condition may vary, depending on the mode of CM and the 72

specific process steps. For example, a state of control can be demonstrated for some CM processes 73

when a set of parameters (e.g., process parameters, quality attributes) are within specified ranges, 74

but the processes are not necessarily in a steady state condition. Elements of the control strategy 75

monitor a state of control and, when necessary, take appropriate actions to maintain control of the 76

process. It is important to have mechanisms in place to evaluate the consistency of operation and 77

to identify situations in which parameters are within the specified range yet outside historical 78

operating ranges, or they are showing drifts or trends. The latter situation may indicate that the 79

process is at risk of operating outside the specified operating range and warrants evaluation and, 80

when necessary, corrective action.

81

3.1.2. Process Dynamics 82

Knowledge of process dynamics is important to maintaining state of control in CM. Specifically, 83

understanding how transient events propagate helps to identify risks to product quality and to 84

3

develop an appropriate control strategy (see Section 3.1.5 for process monitoring and control 85

considerations). Transient events that occur during CM operation may be planned (e.g., process 86

start-up, shutdown and pause) or unplanned (e.g., disturbances).

87 88

Characterisation of the residence time distribution (RTD) can be used to help understand process 89

dynamics. RTD characterises the time available for material transport and transformation, and it 90

is specific to the process, composition/formulation, material properties, equipment design and 91

configuration, etc. Understanding process dynamics (e.g., through the RTD) enables the tracking 92

of material and supports the development of sampling and diversion strategies, where applicable.

93

In addition, such understanding is of importance from a process performance perspective. For 94

example, process dynamics may impact process characteristics, such as selectivity in the 95

manufacture of chemical entity drug substances and viral safety in the manufacture of therapeutic 96

protein drug substances.

97 98

Process dynamics should be characterised over the planned operating ranges and anticipated input 99

material variability using scientifically justified approaches. Appropriate methodologies (e.g., 100

RTD studies, in silico modeling with experimental confirmation) should be used to understand the 101

impact of process dynamics and its variation on material transport and transformation. These 102

methodologies should not interfere with the process dynamics of the system, and the 103

characterisation should be relevant to the commercial process. For example, when conducting 104

RTD studies, the tracer used to replace a constituent of the solid or liquid stream should have 105

highly similar flow properties as those of the constituent replaced. A tracer should also be inert to 106

the other components of the process and should not alter how processed materials interact with 107

equipment surfaces. Step testing by making small changes to the quantitative composition of the 108

process stream (e.g., small increments of a constituent) is another useful technique to determine 109

the RTD and avoid the addition of an external tracer to the process. Other approaches can be used;

110

the approach taken should be justified.

111

3.1.3. Material Characterisation and Control 112

Material attributes can impact various aspects of CM operation and performance, such as material 113

feeding, process dynamics, and output material quality. Understanding the impact of material 114

attributes and their variability on process performance and product quality is important for the 115

development of the control strategy. Input materials may require evaluation and control of 116

attributes beyond those typically considered for a material specification used in batch 117

manufacturing. For example:

118 119

 For a solid dosage form process, particle size, cohesiveness, hygroscopicity, or specific 120

surface area of drug substances and excipients may impact the feeding of powders and 121

material flow through the system.

122 123

 For a chemically synthesised drug substance process, viscosity, concentration, or the 124

 For a therapeutic protein (e.g., monoclonal antibodies) process, the higher variability of 128

feed stocks such as metal salts, vitamins, and other trace components may adversely impact 129

4

cell culture performance. Prolonged run times may require different lots of media, buffers, 130

or other starting materials for the downstream CM process, potentially introducing more 131

variabilities to the process.

132

3.1.4. Equipment Design and System Integration 133

The design of equipment and their integration to form a CM system impacts process dynamics, 134

material transport and transformation, output material quality, etc. When developing a CM process 135

and its control strategy, it is important to consider the characteristics of individual equipment as 136

well as those of the integrated system that can affect process performance. These include the 137

system’s ability to maintain a continuous flow of input and output materials, manage potential 138

disruption to CM operations (e.g., filter changes), and complete the intended transformation of the 139

material stream within the respective planned operational ranges of the equipment. Examples of 140

design considerations are given below:

141 142

 Design and configuration of equipment (e.g., compatibility and integrity of equipment 143

components for the maximum run time or cycles; geometry of constituent parts to promote 144

the desired transformation; spatial arrangement of equipment to facilitate material flow and 145

avoid build-up or fouling) 146

147

 Connections between equipment (e.g., use of a surge tank between two unit operations to 148

mitigate differences in mass flow rates) 149

150

 Locations of material diversion and sampling points (e.g., selection of locations for a 151

diverter valve and sampling probe without interrupting material flow and transformation) 152

153

Furthermore, appropriate design or selection of equipment for a CM process may enable process 154

simplification, facilitate process monitoring and material diversion, and improve process 155

capability and performance. For example, in a drug substance process, reactor design can 156

effectively reduce formation and build-up of impurities, resulting in fewer purification steps.

157

Similarly, for therapeutic protein drug substance manufacturing, system design can enable process 158

intensification and reduce cycle times.

159

3.1.5. Process Monitoring and Control 160

Process monitoring and control support the maintenance of a state of control during production 161

and allow real-time evaluation of system performance. Common approaches to process monitoring 162

and control—including establishment of target setpoints and control limits, design space, and 163

specifications for attributes being measured—are applicable to CM.

164 165

Process analytical technology (PAT) (ICH Q8) is well-suited for CM. Example applications 166

include in-line UV flow cells to monitor therapeutic protein concentration information, in-line 167

near-infrared spectroscopy to assess blend uniformity, and in-line particle size analysis to monitor 168

the output of a crystalliser. The use of PAT enables disturbances to be detected in real time.

169

Therefore, CM is readily amenable to automated process control strategies based on, for example, 170

active control such as feedforward or feedback control. Principles of control strategy as described 171

in ICH Q8 and ICH Q11 can be applied to CM processes.

172 173

5

An appropriate sampling strategy is an important aspect of process monitoring and control. The 174

variables monitored, monitoring method and frequency, amount of material sampled (either 175

physical sampling or data sampling using in-line measurement), sampling location, statistical 176

method, and acceptance criteria depend on the intended use of the data (e.g., detection of rapid 177

changes such as disturbances, assessment of quality of a batch when real-time release testing 178

(RTRT) (ICH Q8) is used, analysis of process trends or drifts) and process dynamics. Another 179

important consideration is the avoidance of measurement interference with the process.

180

Assessment of risks associated with data gaps (e.g., PAT recalibration, refill of a feeding system, 181

failure of system components) should inform whether contingency methods are required.

182

3.1.6. Material Traceability and Diversion 183

CM processes may include periods when non-conforming materials are produced, for example, 184

during system start-up and shutdown and when disturbances are not appropriately managed and 185

mitigated. The ability to divert potential non-conforming material from the product stream during 186

production is an important characteristic of CM and should be considered in developing the control 187

strategy.

188 189

Understanding the process dynamics of individual unit operations and integrated systems over 190

planned operating conditions enables tracking of the distribution of materials over time. This 191

allows input materials to be traced throughout production. Material traceability, understanding 192

how upstream disturbances affect downstream material quality, and the use of appropriate 193

measurements (e.g., PAT) allow for real-time determination of when to start and stop material 194

collection or diversion. The amount of material diverted can be influenced by several factors, such 195

as process dynamics, control strategy, severity (e.g., magnitude, duration, frequency) of the 196

disturbances, and location of the sampling and diversion points. Additionally, it is important that 197

the diversion strategy accounts for the impact on material flow and process dynamics when 198

material is diverted. Criteria should be established to trigger the start and end of the diversion 199

period and restart of product collection.

200

3.1.7. Process Models 201

Process models can be used for development of a CM process or as part of a control strategy for 202

commercial production, including the diversion strategy. Process models may also be used to 203

predict quality attributes in real time, enabling timely process adjustments to maintain a state of 204

control. During development, process models can support the establishment of a design space by 205

explaining how inputs (e.g., process parameters, material attributes) and outputs (e.g., product 206

quality attributes) are related. Through use of in silico experimentation, process models also 207

enhance process understanding and can reduce the number of experimental studies.

208 209

For general considerations regarding models (including implications of model impact to validation 210

requirements), refer to Points to Consider: ICH-Endorsed Guide for ICH Q8/Q9/Q10 211

Implementation. For CM applications, additional considerations are discussed below.

212

 Model development requires an understanding of the underlying model assumptions (e.g., 217

plug flow versus mixed flow systems) and when these assumptions remain valid. Risk 218

6

assessments, sound scientific rationales, and relevant data are needed to select model inputs 219

and model-governing equations. It is important to determine the relevant inputs that affect 220

the model performance, based on appropriate approaches such as sensitivity analysis.

221 222

 Model performance depends on factors such as mathematical constructs and the quality of 223

model inputs (e.g., noise, variability of data). When setting acceptance criteria for model 224

performance, the model’s intended use and the statistical approaches that account for 225

uncertainty in the experimental measurement and model prediction should be considered.

226 227

 Model validation assesses the fitness of the model for its intended use based on 228

predetermined acceptance criteria. Model validation activities are primarily concerned with 229

demonstrating the appropriateness of the underlying model assumptions and the degree to 230

which sensitivity and uncertainty of the model and the reference methods are understood.

231 232

 Monitoring of model performance should occur on a routine ongoing basis and when a 233

process change (e.g., input material, process parameter change) is implemented. A risk-234

based approach to assess the impact of a model change (e.g., optimisation of model 235

performance, change of the model’s intended use, change of underlying model 236

assumptions), scope of model development, and model validation criteria enables effective 237

and efficient lifecycle management of models. Depending on the extent of a change and its 238

impact on model performance, a model may need to be redeveloped and validated.

239

3.2. Changes in Production Output 240

Several considerations associated with some common approaches to production changes are 241

changes may require process modification and process validation.

245 246

 Change in run time with no change to mass flow rates and equipment: Issues not 247

observed over shorter run times may become visible as run time increases. Additional risks 248

and constraints should be considered and may include, for example, process drift, increased 249

heat, material build-up, exceeding the performance limit of components (e.g., validated in 250

vitro cell age, resin cycle number, measurement system calibration status), material 251

degradation, membrane or sensor fouling, and microbial contamination. Decreasing 252

production output (below the longest run time previously validated) should not imply 253

additional risks, given the same equipment, process and control strategy are used.

254 255

 Increase mass flow rates with no change to overall run time and equipment: The risks 256

associated with this approach may impact output material quality and are related to changes 257

in process dynamics and system capability to handle increased mass flow rates. Therefore, 258

this approach may require re-evaluation and modification of the control strategy, including 259

process parameters and controls, material traceability, RTD, sampling, and diversion 260

strategies.

261 262

7

 Increase output through duplication of equipment (i.e., scale-out): Considerations for 263

two commonly used scale-out approaches are provided below.

264 265

o Replication of production lines (like-for-like): Replicating the integrated CM 266

production line (i.e., same equipment and setup as the original CM system) can be 267

used to increase production output. The replicate production lines follow the same 268

control across parallel unit operations. Aspects to consider are maintenance of 273

uniform flow distribution among the parallel operations, re-integration of parallel 274

flow streams, changes to process dynamics, and material traceability.

275 276

 Scale up by increasing equipment size/capacity: Depending on the process and 277

equipment design, increasing production by increasing equipment size may be possible.

278

General principles of equipment scale-up as in the case of batch manufacturing apply. As 279

elements such as RTD, process dynamics, and system integration may change, various 280

aspects of the control strategy may be impacted. The applicability of the original control 281

strategy should be assessed at each scale and modified where needed.

282

3.3. Continuous Process Verification 283

In CM, frequent process monitoring and control can be achieved through use of PAT tools, such 284

as in-line/online/at-line monitoring and control, soft sensors and models. These tools allow real-285

time data collection for parameters relevant to process dynamics and material quality, and hence 286

ensure the state of control for every batch. Additionally, since CM can facilitate changes to 287

production output without increasing equipment size, there is an opportunity to generate 288

development knowledge at the same scale intended for commercial manufacturing. These tools, 289

together with the system design and the control strategy, facilitate early execution of process 290

validation activities and the adoption of continuous process verification (ICH Q8) as an alternative 291

to traditional process validation.

292

4. REGULATORY CONSIDERATIONS