Find a Sales office 
Have us call you
Email the sales teamMultivariate tools for defining the Design Space
Fundamental to the PAT and QbD initiatives is a statistical toolkit for better understanding the hidden information contained in complex process data outputs. So important to the design of better processes, the tools of Design of Experiments (DoE) are the underlying principle of the recent Design for Six Sigma (DFSS) initiative. This was the first real recognition that the fundamental tools of six sigma were not capable of fully understanding complex processes.
DoE enforces a logical and systematic approach to designing and performing experiments in order to maximize the information obtained. DoE can be used as a first investigational step into isolating what the CQA’s of a system are and the corresponding CPP’s that influence them.
Use Design of Experiments (DoE) to
- Screen out the critical parameters influencing quality from a large number potential parameters using,
- Understand the important interactions in a smaller number of screened out variables using,
- Full Factorial or
- High resolution Fractional Factorial Designs
- Optimize the process using
- Develop new formulations and optimize their manufacturing processes using
- Mixture Designs
- Process/Mixture Designs
The Unscrambler® X contains the necessary tools for the above steps.
The Unscrambler Optimizer can also be used to target and refine drug formulations to meet desired criteria.
Once the CQA’s and CPP’s have been defined and understood using DoE, the next phase of a QbD project usually involves matching a process analytical technology capable of monitoring the CQA’s. This may be as simple as numerous temperature or pH probes located strategically in the process right up to complex spectroscopic or imaging systems. This is where the tools of MVA become extremely important.
Use Multivariate Data Analysis (MVA) to
- Define process signatures in time evolving processes using Principal Component analysis (PCA) combined with process and spectroscopic outputs in applications such as Fluid Bed Drier monitoring or fermentation reaction monitoring.
- Predict quantitative CQA values of blending, mixing or coating operations using regression methods such as,
- Multiple Linear Regression (MLR)
- Principal Component Regression (PCR)
- Partial Least Squares Regression (PLSR)
- Isolate pure protein fractions during bioprocessing operations using Multivariate Curve Resolution and process chromatography.
- Classify batch operations as good or bad using,
- PCA Classification (SIMCA)
- Linear Discriminant Analysis (LDA) or
- Support Vector Machines (SVM) Classification.
- Holistically monitor a number of unit operations using Multiblock modelling and the new Unscrambler Online (soon to be released).
The new Unscrambler Online takes the principles of MVA to the next level, helping pharmaceutical and biopharmaceutical manufacturers worldwide achieve the Pharmaceutical Quality System (PQS) defined in ICH Q10, Pharmaceutical Quality System . For more information, see the section on MVA for Process Monitoring and Control.
 |
 |
 |
 |
 |
 |