The use of near-infrared spectroscopy (NIRS) has been widely documented within pharmaceutical quality and manufacturing functions as a unit operation, but much less has been written about it when it comprises a central part of a pharmaceutical Quality Management Strategy. With the increased attention from industry and regulators relating to PAT and QbD principles, this article will discuss the benefits of the strategic application of this philosophy when they are integrated with risk mitigation tools, design of experiments and careful selection of equipment and methods for a near-infrared method on continuous processes.
Introduction
Near-infrared Spectroscopy (NIRS) has become established as a key technique within the Process Analytical Tool (PAT) toolbox and there are many examples within the literature of applications for conventional pharmaceutical laboratory and manufacturing functions [1-4]. However, much less has been published about NIRS methods that utilize Quality-by-design (QbD) principles throughout the development of continuous processes to achieve higher quality methods.
The continuous manufacturing of a pharmaceutical preparation requires frequent monitoring of its critical quality attributes (CQAs) and typically this will require the removal and analysis of sub-samples. In conventional manufacturing, a batch is a self-contained unit, from which sub-samples will statistically reflect the whole. In a continuous process, the beginning and end of the batch are not necessarily related. Therefore, testing procedures in a continuous process are required to ensure homogeneity throughout the duration of the manufacturing run. In the stringent financial environment that most quality control laboratories operate within, resourcing this testing is an additional challenge. As Pharma continues to explore continuous manufacturing as an alternative to conventional unit operations, the ability to work within the QbD/PAT framework issued by regulators allows the application of good science and can lead to innovation and cost effectiveness. Consequently, an automated continuous monitoring system is desirable.
NIRS is cost effective, flexible, has a non-destructive nature and ability to collect spectral information quickly. A NIRS method designed with QbD principles has the potential not only to use as an on-line monitoring method, but as a system of control that can provide greater understanding of the product and process and has the ability to reduce variability of the CQAs. This enables process control in real time through process understanding - the true essence of PAT.
Qualitative or Quantitative Design Space
In the QbD/PAT framework the concept of a design space is important. The design space represents the changes to the CQAs response after changes to manufacturing or testing conditions. This may occur when key conditions are varied either deliberately or unintentionally through instrument failure. For instance, in the validation of High Performance Liquid Chromatography (HPLC), changes to the flow rate are critical to retention times and therefore the method is tested with deliberate variations to the parameter to assess the effect. Therefore, a programmed study of the design space can identify critical variables and enable strategies to be put in place that mitigate their risk. Conversely, conditions that are not significant can be identified and varied to produce better quality product and optimize yield. Once the design space has been accepted by the regulatory authorities changes, within the design space can be made without notification [5], enabling rapid deployment. Initially, the selection of a qualitative or quantitative design space must be made. The benefits and validation requirements differ between the methods, as does the dimensionality within the design space.
Quantitation provides information that is comparable to that of the reference method and allows the method to be more easily accommodated within Manufacturing, Quality Assurance and Regulatory functions. For this reason it is advantageous as an alternative test method to have quantitative methods filed with the regulatory authorities. However, it has a restricted design space, resulting in a reduced ability to compensate for process and material changes, which may result in greater prediction errors. Development and validation of a quantitative near-infrared can be expensive because of its requirement to calibrate against a single parameter. This is due to the necessity to have NIRS development samples that extend the parameter of interest beyond the range of normal manufacturing variation, and with the exception of reaction monitoring, are not readily available.
Qualitative methods which trend spectral features or match an unknown spectra against that of a reference spectra can be much quicker to implement and do not require as many range samples to be prepared. For example, a qualitative or Principal Component analysis plot that uses information within the current design space, may meet the same goal to assure quality, without requiring the time and expense that quantitation necessitates. However, as they are not calibrated, their design space can be significant and uncontrolled, leading to difficulties in correlation to the parameter of interest.
The integration of QbD approach to NIRS methods seeks to consolidate the strengths from both approaches by expanding the quantitation design space to incorporate the relationship between the parameter of interest, the material and manufacturing conditions. This will facilitate understanding of the critical process parameters that affect the CQAs and improve robustness.
Existing Knowledge Review
The design of a NIRS method with QbD principles begins with a review of the validation guidelines and specifications. These reflect the CQAs of the product, and thus, are the minimum criteria to meet. Additional CQAs which are product specific may also exist. The sample concentration range requirements will be dictated by either existing or planned specifications or compendia limits e.g., European Pharmacopoeia (EP) assay limits 95.0% - 105.0% of the label amount. Careful planning and flexibility is paramount, as the breadth of data required is usually obtained from a reference method, which may involve negotiations with other departments to prioritize the use of equipment and personnel who would normally have other duties and tasks to perform. A good understanding of the prospective requirements that will allow maximum efficiency and clarity to be obtained from these negotiations is therefore advisable. Every effort to leverage the extent of current knowledge from subject matter experts (SMEs) and published literature, whether located internally within an organization or externally, should be made before practical studies begin.
A critical area of the existing knowledge review is the applicable guidelines issued by Regulators, as ultimately these will be used to ensure that the method complies with these and thereby ensures the highest standards of quality, safety and efficacy. The task of recommending best practice guidelines which can be successfully integrated within the quality or Chemical Manufacturing and Controls (CMC) section of the marketing license is performed by various working groups. In Europe this function is performed by the Committee for Medicinal Products for Human use (CHMP), the International Conference of Harmonization (ICH), and the EP. In the US, it is commonly the Code of Federal Regulations issued by the Food and Drug Administration (FDA), the United States Pharmacopoeia (USP) or the American Society for Testing and Materials (ASTM).
The development and validation of qualitative and quantitative NIRS methods is described in a CHMP guideline [6] and is heavily influenced by the existing ICH analytical method criteria of accuracy, linearity, reproducibility, repeatability, specificity and robustness [7]. There are also ASTM guidances and a USP chapter which describes the use and application of NIRS and these documents supplement existing chapters on validation [8,9].
In 2004, the FDA issued its Process Analytical Technology (PAT) guidelines as part of its good manufacturing practices for the 21st Century, with the intention of encouraging the exploration of the product design space for new and existing ventures within Pharma [10]. This document sought to encourage the growth of knowledge and understanding of product attributes through the application of new technologies and methods.
More recently, QbD has been supplemented by guidelines such as ICH Q8, Q9 and Q10 which describe how integration of risk analysis,pharmaceutical quality systems and exploratory development within the design space can lead to an increase in understanding [5,11,12]. These guidelines were issued for the specific purpose of encouraging methods of this type and those which can be applied as part ‘of a system of process and controls’ which could provide greater assurance to existing product quality.
Design Deconstruction to Meet Specification
Targeted deconstruction of the validation requirements from validation criteria can be translated into actions within the development process. For example, ICH guidelines for near-infrared, state that the precision of the NIR method is equivalent or superior to that of the reference method. However, as continuous manufacture may involve heterogeneous dynamic solid-state particles of different sizes and shapes, compared with that of a homogenous solute in solution on a HPLC, the difference in precision between these systems is obvious. Deconstructing this requirement demonstrates the benefit of accessing existing data. Quantitative validation data should exist from the reference methods precision and intermediate precision studies and this data can be used to determine a standard error of the laboratory (SEL). The SEL is a measure of the natural variability or noise that the reference method contributes to the measurement of the parameter of interest i.e., common cause variation and is a base-line measurement of how the reference method performs. As the NIR method incorporates this typical error from the determination of the reference results, it is possible for the NIR to be limited in its accuracy and precision by the reference method. However, when combined with information about the repeatability of replicate spectra obtained from the feasibility study, the number of replicate scans required to achieve superior precision to the reference method can be determined. This process also defines other requirements such as scan time and the scale of scrutiny. These, in turn, can define the wavelength range, co-averaging, resolution (if changeable) and pathlength.
NIRS Sample Interface
Introduction of development samples to the NIR presents significant engineering challenges. The ideal solution is to engineer a bypass loop with the NIR interface, where a representative sample of the product flow can be diverted, whilst allowing for isolation from the main flow and introduction of small amounts of development samples. This reduces the risk, time and expense of sample preparation. In this development configuration the pump or other movement system within the loop should seek to match the flow characteristics that will result when the system is in its normal operating mode for routine production. This will minimize the differences in the resulting spectra. A successful strategy of this type excludes the need for calibration transfer and significantly simplifies the calibration process. This is only possible in systems where divergence of the flow will not affect the product characteristics.
Unfortunately, this ideal situation is rarely found in practice. For systems which interface to the manufacturing equipment directly through a site glass, or are within the normal product flow, then the challenge of manufacturing development samples outside the normal product range at full scale can be cost prohibitive. In the event that this process is financially feasible it will need to address the challenges of cleaning the residue from hyper-potent batches. This consideration is especially critical if a normal manufacturing campaign follows the development work as cleaning methods may not be validated to recover materials at the upper concentration range of the development samples.
It can therefore be advisable to avoid the commercial manufacturing route entirely and manufacture small development batches on the laboratory or pilot scale equipment, where these issues are not as critical. This decision poses challenges as the change of scale may significantly affect the product characteristics being monitored by the NIR; typically variability at the smaller scale will be different to that at the full-scale deployment. In addition, the near-infrared instrumentation may not necessarily be that used for commercial work and so the robustness and reproducibility of the equipment becomes important. This can be solved by purchasing instruments claiming transferability or simply by relocating the original instrument between locations. The decision to use either a single or multiple NIR instruments should be recognized at an early stage to allow standardization of the operating parameters before embarking on spectral collection. Once standardized, the spectra from the multiple instruments should be verified for their comparability.
Feasibility Study
The importance of a well-planned and controlled feasibility study cannot be overstated as it can specify instrumental specific parameters that will ensure that the spectra collected are of the right quality to ensure the success of the overall objective.
The feasibility study should provide the opportunity to examine the method parameters that will be required in the validation. It should also allow for the expected growth in process and system understanding that results during the development phase of the method. Other potential issues that can be highlighted by the feasibility study are the non-linearity of detector response or changes to the formulations physiochemical properties i.e., solubility, particle size, water content or rheology. These may change significantly with increasing concentrations and may result in the theoretical specification not being able to be achieved in practice. Consequently, these limitations and not the regulatory guidelines will define the calibration range of the NIRS method, as opposed to that recommended by the guidelines and will need to be determined in practice and justified. The change in such parameters may also affect the formulations physical interaction with the NIRS equipment, where fouling of the probe and mechanical attrition on the sapphire windows may result.
Specific tools such as the Ishikawa (fish-bone) diagram can help define and record the results from different threads of discussions with the SMEs and also leads to the identification of areas where supplementary methods or controls will be required. Risk mitigation tools such as Failure Mode Effects Analysis (FMEA) can be useful in understanding the areas of risk, before embarking on a gap assessment and remediation. In some cases these variables have existing engineering or other supporting controls that can mitigate their ability to influence the manufacturing process and these should be assessed for their robustness and ability to withstand scrutiny. The result of this process is a fewer number of critical variables which should be included within the experimental design.
The NIRS sample viewing area should measure a pharmaceutical dose. In dynamic systems where the material is moving, this allows collection of multiple spectra which can be averaged to complete a unit dosage. Determination of the scale of scrutiny of the NIR will relate to the speed of the material, scan time and pathlength and may entail performing depth penetration studies in solid materials, or understanding if laminar or turbulent flow is prevalent in the liquid system.
The SEL can be used to calculate the specifics (number, interval and range) of the development samples required for a quantitative approach. The sample range must be broad enough to allow sufficient selectivity of the concentration vectors. An inability to obtain these development samples will result in the sample range being extremely small in relation to the SEL. This will result in a much greater perpendicular spread of data relative to the concentration vector, reducing the correlation coefficient and the signal-to-noise response for the calibration. However, the range should not be extended so far that truncation of the absorbance axis due to saturation results. This can be confirmed in the feasibility study by using the extremes of the development range samples only.
The feasibility study will also indicate which wavelength regions of the formulation’s spectra are within the working range of the detector at the pathlength selected. Preliminary off-line NIRS studies may have been suggestive of the selection of pathlength where changes to either the pathlength or viewing area are easy to apply. This will allow different areas of the spectra to be explored within the working range of the detector. Cross-referencing of wavelength regions with the selective absorption of known functional group will identify regions that should be selected. The pathlength can then be adjusted to allow maximum response from this region.
As NIRS is selective rather than specific, it is appropriate to use other supporting Chemometric methods to achieve overall specificity. This can take the form of identification before quantification e.g.,Mahalanobis distance or wavelength correlation to verify that the spectra conform to the normal formulation. It will also serve to demonstrate that the method is robust and that the method collection parameters are acceptable.
This feasibility study will also define the ability of the instrumentation to function practically within the environment and address implementation issues without affecting routine production, whilst simultaneously beginning documentation of training for the system end users. It also provides an opportunity for a constructive dialogue that may result in more favorable operating characteristics, thus increasing the sense of ownership of the system and equipment.
The spectra should be cross-referenced by time to the analytical laboratory samples, to enable the concentration data and the spectra to be matched. Consequently, it is important that where a significant difference exists between the NIRS interface and the sampling point, that any time lag be recorded. This will allow the correct spectra to be matched. Special care should be taken to avoid time differences between different sources of time e.g., GMT and local time. In applications where averaging of the spectra are required to meet the targeted precision requirements, a spectral window around the matched sample point of spectra is required.
Development and Design of Experiments
A design-of-experiments is a series of structured experiments that will define a design space. The size and type of any experimental design is dependent on time, cost and the value of the information required. The critical process parameters that have been provided by the SMEs after categorization and ranking from the FMEA, will be used as the control variables to develop the design. The CQAs will be the response variables which measure the extent of the change.
A design-of-experiments can be used to build directly the NIRS calibration model, therefore reducing the cost of the overall study. However, some care should be taken in this approach to ensure that a correlation not based on the parameter of interest is established. In those cases where this occurs, it can be preferable to supplement the experimental design with samples of increasing parameter concentration and spectra from the batches of full-scale material to aid in verifying the selectivity.
An additional use of the experimental design is in calibration transfer, where the spectra from development samples that are made in a laboratory or pilot scale can be used on the full-scale manufacturing process. In this case, the experimental design can help to discover process conditions that produce material that is qualitatively and quantitatively identical. The discovery of a set of conditions which produces material identical to that of a full-scale unit removes a large source of variation from the transfer and significantly eases calibration transfer. In those circumstances where development samples are performed on the pilot scale then once the scaled down conditions have been determined the materials within the design can be placed through the NIR system.
A full-factorial experiment provides the most useful information, but may be costly and time consuming to perform. Reduced factorial experiments that retain some of the statistical power may be sufficient. Once the design and levels of the variables are established the design can be implemented. The development samples should be of nominal concentration and should be introduced into the NIR sample interface as part of a randomized block, allowing sufficient time for the process to reach steady state conditions. This will eliminate start-stop characteristics, which may not reflect normal operation and will obscure the statistical analysis. As conditions in the design are varied the response variables alter, the resulting statistical breakdown will indicate those factors that are significant to the process. Laboratory analysis with the primary method should be made throughout the processing run to see the within run processing variability and then compared between runs. A statistical sampling plan to ensure that representative sampling occurs throughout the study conditions should also be designed to exclude sampling bias. The resulting laboratory analysis of samples at each condition can confirm the conditions which produce material that is identical to full-scale manufacturing. In addition to searching for instrument-scalable conditions, this will also determine a method design space based on engineering variables and determine the robustness of the calibration when engineering parameters are altered.
Once the design-of-experiments has been analyzed it will demonstrate those parameters which are critical to the NIRS design space. This can be assessed by comparing the difference between the actual and predicted results (residuals). These must be less than SEL x 2; equivalent to a 95% confidence in the center of the range that the prediction is unaffected by a processing condition. If the residuals are significantly greater than this, the NIRS method is affected by this processing condition and as such the choice is either to include these spectra within the calibration or to adjust the calibrations based on the critical process parameters.
For the latter case, this will require a second level of modeling which combines the NIR model with the output of a critical process parameter(s) to adjust the design space in real-time for those changes that may occur. The benefits to robustness of the resultant calibration model are clear. Secondary benefits are much more subtle; as a source of variability to the calibration is controlled the calibration’s predictive ability often improves providing more accurate predictions. However, this should be reviewed in terms of the existing controls using risk analysis principles. If the existing controls already maintain the critical process parameter tightly within the design space because of other product issues such as yield, then the benefit to the NIRS is purely in obtaining more precise predictions which may not justify this additional expense.
Once successfully installed the NIRS can provide, in some cases, the ability to control and monitor degradation. It should be clearly stated that NIRS is not a suitable tool for measuring very low-level impurities because of its lack of differentiation and sensitivity, but if correctly applied it can be advantageous to have an indirect method of measurement, by measuring the parent molecule as a surrogate. In certain circumstances where the underlying mechanism and kinetics of degradation are understood, it may be possible to determine a mechanistic model based on first principles and supplemented by the NIRS as a means of control. It may also demonstrate a correlation between the degradation products to engineering parameters. Thus, by controlling the critical process parameters, monitoring the surrogate and applying multivariate statistical process controls to monitor process trajectories, a system which continually controls degradation quality in real-time is possible. Having established the scalable process conditions during development the full range of development samples can now be run and tested and their spectra collected. The calibration range should be symmetrical and centered about the nominal concentration of 100% of the label claim; as some multivariate techniques seek to minimize the variance around the center of the range e.g Partial Least Squares Regression. Existing guidelines recommend a 2:1 ratio of calibration to independent validation batches and it is also sensible to ensure that duplicates at the extremes of the range are included within the validation set. The spectra can be matched against the resulting reference analysis results and NIRS calibration can begin. A number of different parameters should be explored to determine the most effective calibration, before proceeding to the validation. This may require a number of iterative attempts with various pre-treatment and transformation properties before assessment against the validated criteria.
In those instances where the developmental reference analysis is supplemented by full-scale routine analysis an understanding of the inter-laboratory error between the two locations should be obtained and may require some analysis of the development material in parallel. This is especially important when the calibration is expected to be transferred to different sites. Where in-line quantification is required then calibration transfer will be required to compensate for spectral and laboratory differences. In those instances samples from the development study should be sent to the appropriate local QC laboratory for testing. This will indicate the presence of any systematic or gross biases between the source and receiving laboratories that may require correction.
Validation and Implementation
Once the calibration has been developed it should be validated. As each validation criterion has been systematically deconstructed, the validation reflects the QbD approach applied during its design and development and as such should be met with ease. In those events where deviations in development from the guidelines do occur then an appropriate scientific rationalization has already been assessed, documented and can be stated to Regulators.
The use of independent batches is highly recommended and should be tested as validation set. They also provide greater opportunity to capture differences in manufacturing variability. Once validation is completed there will be a period of commercial parallel monitoring which will allow familiarity to be obtained in the technique and provide further evidence of the predictive ability and robustness to be made.
Conclusion
As Pharmaceutical companies seek to explore the benefits of continuous manufacturing using QbD and design space exploration principles, the requirement to have fast, agile, flexible and robust control systems such as NIRS becomes increasingly important. Application of these principles to development of NIRS ensures that provision within the method is made from the earliest design stages, resulting in a more robust measurement system which furthers the ultimate aim of feedback control. This allows changes within the design space to be made and accurately measured without significant regulatory oversight. As ultimately the measure of success of any test method may be judged by its application and acceptance by the regulatory authorities, the benefit to Regulators, Pharma and the patient is obvious.
References
- Quantitation of the active compound and major excipients in a pharmaceutical formulation by near infrared diffuse reflectance spectroscopy with fibre optical probe, Blanco,M.; Coello,J.; Iturriaga,H.; Maspoch,S.; De la Pezuela,C., 1996,Analytical Chimica Acta, Vol. 333,No. 147 p. 156.
- Application of near Infrared reflectance to the analytical control of pharmaceuticals assay of ranitidine chlorohydrate and water content in tablets, Corti,P; Dreassi,E.; Corbini,G; Lonardi,S; Viviani,R.; Mosconi,L.; Bernuzzi,M, 1990, Pharmaceutica Acta Helvetiae, Vol 65, pp 28 -32.
- Near infrared transmittance analysis for the assay of solid pharmaceutical dosage forms, Corti,P; Ceramelli,G.; Dreassi,E.; Mattii,S., The Analyst, Vol. 124, pp 755 - 758.
- A Process Analytical Technology Approach to Near-Infra Red Process Control of Pharmaceutical Powder Blending Part 1: D- Optimal Design for Characterisation of Powder Mixing and Preliminary Spectral Data Evaluation, El-Hargasy,A.; D’Amico,F.; Drennen,J.K. ,2006, Journal of Pharmaceutical Sciences, Vol. 95, No. 2, 392 - 406.
- Pharmaceutical Development Q8, R2, International Conference on Harmonisation, August 2009.
- Committee for Medicinal Products for Human use – Note for Guidance on the use of Near-Infrared Spectroscopy by the Pharmaceutical Industry and the Data Requirements for New Submissions and Variations..
- Validation of Analytical Procedures: Text and Methodology Q2, R1. October 1994.
- Standard Practices for Infrared Multivariate Quantitative Analysis, E1655-05, American Society for Testing and Materials, 2005.
- United States Pharmacopoeia National Formulary, Near Infra-red Spectrophotometry, USP 29 edn. pp 2720-2722.
- PAT – A framework for Innovative Pharmaceutical Development, Manufacturing and Quality Assurance – Guidance for industry, Center for Drug Evaluation and Research, 2004.
- Quality Risk Management Q9, International Conference on Harmonisation, November 2005.
- Pharmaceutical Quality System, Q10, International Conference on Harmonisation, June 2008.
Author Biography
Mark Morton has 22 years of technical experience including quality control, PAT and Regulatory affairs. He has presented and published in the areas of NIRS and PAT. He gained his Doctorate from the London School of Pharmacy and is a 6-sigma Black belt. He currently divides his time between consultancy and contract work. Mark can be reached at [email protected]
This article was printed in the September/October 2011 issue of American Pharmaceutical Review - Volume 14, Issue 6. Copyright rests with the publisher. For more information about American Pharmaceutical Review and to read similar articles, visit www.americanpharmaceuticalreview.com and subscribe for free.