Effective and reliable optical system for cleaning validation in pharmaceuticals manufacture

Final Report Summary - OPTI-CLEAN (Effective and reliable optical system for cleaning validation in pharmaceuticals manufacture)

Executive Summary:
The OptiClean technology is being developed to support the process of cleaning verification in the pharmaceutical and biotechnology industries. Cleaning is an integral part of the process for the production of therapeutic drug product. The challenge for industry is the verification of the cleanliness of equipment prior to its use for manufacturing. Currently the process of validation and verification involves the generation of a complex swabbing plan, the verification of the recovery method and the use of expensive and time consuming HPLC methods to verify the residual quantities of active pharmaceutical ingredients and chemical cleaning aids.The primary objective of the project was to develop a technology which would facilitate the determination in real time of the cleanliness of pharmaceutical manufacturing equipment.
The development of the OptiClean pre competitive prototype was achieved through a systematic process of theoretical and technical evaluations, starting with an assessment of the market requirements and specifications for a cleaning verification technology. In addition there was an assessment of the status and availability of potential process analytical technology for the real time verification of the cleanliness status of process and packaging equipment. The process of developing the technology began with a series of industry consultations as well as a questionnaire distributed to the wider pharmaceutical industry. This process served to generate the specifications for the final pre-competitive prototype. The determination of the component specifications for the portable device was achieved through the building of a NIR-CI test rig using a MCT sensor and a tunable Fabry Perot Interferometer coupled with a combination of band pass filters, to cover a wavelength range of 900nm to 2500nm. The laboratory based testing of the system demonstrated the suitability of NIR for the detection residual substances whether as raw materials or materials as part of typical blends. Using the bench top application a lower limit of detection of 1.0mg/cm2 was achieved.
The design of the portable device was progressed using an extended InGas sensor. This was due to the cost and supply lead time for MCT sensors. The extended InGas sensor has a spectral wavelength range of 900nm to 2200nm, which was more than adequate for detection as typical spectral peaks are in the area of 1700nm to 2100nm. The portable prototype was built successfully however during the first phase of testing there were issues with the configuration and calibration of the Fabry Perot. The Fabry Perot was returned to VTT and the system recalibrated and returned to IRIS for installation. During this second stage of testing, there were some issues with the generation of images for effective evaluation of images. The system demonstrated a capability for limits of detection of at least 50mg/cm2, however this does not match the limits of detection achieved during the laboratory testing phase. Following consultations with the sensor provider it was suggested that image process software would solve the problem. Sample images were sent to Photonic Sciences to determine if the image clean up software could improve the quality of the images. Following this evaluation a decision was taken to return the sensor to Photonic Sciences for investigation. Following a lengthy assessment of the sensor it was determined that there was a faulty logic board in the sensor. This fault was difficult to diagnose as normally under these circumstances it would not be possible to generate images. The sensor was returned to Innopharma on 24th Feb 2014 for re-installing in the device for further testing. The issues encountered during the testing of the portable device prohibited the consortium from completing evaluation in an industrial environment. However the SME partners remain committed to the completion of the technology and once the system has been tested successfully, industrial trials will be scheduled.

Project Context and Objectives:
The proposal to develop a NIR based portable cleaning verification technology arose as there is a regulatory and industry need to develop a rapid, on-line technology for cleaning process verification within the pharmaceutical manufacturing sector (industry statement). This innovation will lead to the following improvements:
1. A reduced risk of cross contamination, product recalls and risk to patient health
2. A reduction in time and cost of pharmaceutical product manufacture coupled with a greater use of equipment for multi products, better efficiencies. This could lead to an increased spend in R&D, which is a key industry goal.

This research consortium will develop a mobile chemical imaging technology that will rapidly quantify chemical residues from pharmaceutical manufacturing equipment surfaces. This is possible through the combination of the latest innovations in spectrograph components (patented by VTT, PCT/FI2010/050258) coupled with the recently filed patent on “A method of validating a cleaning process” I. Jones, Irish Patent Applications filed 11/18/2010 by the coordinator INNOPHARMA .
The development of this technology will generate revenue in excess of €21 M within 5 years of market entry, combined with 21 jobs within the participating SMEs alone – providing a compelling commercialisation justification for this research activity.
In pharmaceutical manufacture, plants or individual pieces of equipment may be dedicated to individual products or indeed used in multi-product manufacture. Cleaning should be carried out as soon as practical after the end of processing and should leave the plant in a suitable condition for next use.
Cleaning can generally be defined as the removal of unwanted contaminants. Unwanted contaminants may be considered as active ingredients, detergents and microbes. To this end, the objective of a cleaning process is primarily to ensure that the safety, efficacy and quality of the product subsequently manufactured on the same equipment. The impact of inadequate cleaning procedures is that any of a number of contaminants may be present in the next batch manufactured on the equipment, such as: precursors to the Active Pharmaceutical Ingredient (API); by-products and/or degradation products of the API; the previous product; solvents and other materials employed during the manufacturing process; microorganisms; cleaning agents and lubricants.
Cleaning validation in the pharmaceutical industry is the verification that the cleaning procedure used to remove contaminants from equipment surfaces is successfully reproducible. Cleaning validation is executed on all product contact equipment. It is the responsibility of the manufacturer to demonstrate that the level of cleaning and validation performed is adequate. Cleaning validation poses a real challenge to pharmaceutical manufacturers, whereby in addition to being required to comply with regulatory standards, the safety of pharmaceuticals, feasibility and manufacturing efficiencies are key aspects.
The cleaning procedure is currently validated by swabbing the product-contact surfaces following the clean. These swabs are subsequently analysed, predominantly using High-performance liquid chromatography (HPLC). TOC and PH methods of analysis may also be used. This swabbing and analytical activity is time consuming. For example, each HPLC analysis takes about 10 minutes and multiplied by the number of samples run to validate a cleaning method (typically 100-200), capital equipment may be quarantined for as much as two days while cleaning validation is performed . However, in some cases it may even take up to three days to generate approved results. Usually the equipment cannot be used until approved results are available, posing enormous economic burdens on manufacturers. In addition to plant downtime, additional limitations with current techniques include the need for trained personnel to develop cleaning validation protocols and reports, and to conduct the swabbing and the subsequent laboratory analysis, the regulatory burden of conducting and maintaining cleaning validation activities on all product contact equipment on site, as well as the risk of cross contamination as a result of not conducting continuous verification of the cleaning process following each clean, which is manifested in the number of product recalls originating from cross contamination of products. Such incidences involve a serious threat to consumer health and safety, in addition to proving very costly to manufacturers as well as severely undermining consumer confidence.

Moreover, swab sampling is an indirect method and does not cover the entire equipment surface area, rather swab sites are selected based on worst case locations on the equipment and the results are then extrapolated to account for the total product contact surface area. Such techniques tend to be supplemented with a visual inspection, which relies on operator expertise and training, and are thus limited by human subjectivity.
The industry is in great need of rapid on-line detection techniques that would be sensitive enough to detect low concentrations of API and detergent residues on the commonly used surfaces in the pharmaceutical industry (stainless steel, glass) in real-time, thereby saving pharmaceutical manufactures time and money, and ensuring high levels of safety and reduced incidences of cross-contamination.
The proposed project will build on the very promising results of preliminary research that has been carried out using Near Infra Red Chemical Imaging (NIR-CI) technology, which provides both spectral and spatial information from an object, to detect on surfaces low concentrations of APIs and detergents used in pharmaceutical manufacture. The results have shown the feasibility of such technology to be used to provide accurate information in real-time, thereby opening up the possibility for the development of a custom-made version of the technology that could be trialled and validated for its use in the pharmaceutical industry as a cleaning validation tool. To this end, over the course of the OPTI-CLEAN project, a portable imaging device will be designed and built and tested on commonly used APIs and detergents in real pharmaceutical environments in order to validate its effectiveness and reliability. To date NIR-CI has lent itself to industrial applications in the areas of in-line sorting of paper, real time quality control during andalusite production, on-line classification of synthetic polymers, medical devises. However its use in the pharmaceutical sector is still emerging, and in particular its application for cleaning validation is completely novel.
The impact of the uptake of the proposed NIR-CI OPTI-CLEAN system would represent a major breakthrough and holds massive benefits for rapid turn-around times, increased through-put and profitability in European pharmaceutical plants, as well as increased safety standards, which are paramount to safeguarding the health and safety of drug product consumers.
The overall objective of this project will centre on the development of a Near Infra Red Chemical Imaging (NIR-CI) technology to rapidly quantify the contaminant levels remaining on product contact equipment following the execution of cleaning activities. It is envisaged that the proposed OPTI-CLEAN device will be mobile and will be capable of rapidly identifying and quantifying active and detergent residue levels on common surfaces in use in the bulk and finishing pharmaceutical, biopharmaceutical and medical device industries. Typical residue levels would be between 1 and 500µg/25cm2 area.
Chemical Imaging (CI) is an emerging platform technology that integrates conventional imaging and spectroscopy to attain both spatial and spectral information from an object. The rapid, non-destructive and non-invasive features of CI mark its potential suitability as a process analytical tool for the pharmaceutical industry, for both process monitoring and quality control in the many stages of drug production. Vibrational spectroscopic methods, such as Near Infrared (NIR), combined with imaging are particularly useful for analysis of biological/pharmaceutical forms .
Multiple instrumental techniques are available today for recording hyperspectral data of solid samples for chemical imaging applications. Stationary samples are frequently measured using staring or tunable filter imagers (e.g. LCTF), which record complete 2D images sequentially for each wavelength. Push-broom imagers are useful for measuring moving sample material, thanks to capturing instantaneous line images with full spectral information linked to every pixel. Today push-broom imagers provide an alternative high speed approach for studying laboratory samples too, by using sample movement stages synchronized to imager operation. Furthermore, VTT Technical Research Centre of Finland is currently developing high speed hyperspectral techniques using Fabry-Perot interferometer, aiming for high speed applications without the need for sample movement.
The OPTI-CLEAN project will build on the results of feasibility work that was commissioned by Innopharma Labs and conducted through DIT and VTT on excipient, active and detergents used in pharmaceutical manufacture and which proved the concept of the use of a NIR Chemical Imaging technology for the detection of such materials at low concentrations¬2. Moreover, the non-destructive, rugged, and flexible nature of Chemical Imaging (CI) makes it highly attractive for such a purpose. To this end, a pre-competitive system will be designed and built over the course of the 24 month duration of the work programme, and its performance and reliability will be validated in industrial manufacturing sites, with a view to the results going to market with 12 months of completion of this project.
The operation principles of the device will centre on an ergonomic and robust analytical method, whereby the imaging camera will be placed over the surface area to be analysed and an image will be captured (automatically activating illumination source). An appropriate algorithm will then be used to interpret the image, whereby the algorithm takes into consideration the background surface (stainless steel, glass, Perspex, etc.) and excludes this from the calculation. The amount of target residue present is then calculated based on spectral data per pixel. The cleaning activity will finally be approved depending on whether residue concentration meets pre-determined specifications.

In order to realise this, the scientific and technological objectives, and corresponding Performance Indicators, that will be fulfilled during the OPTI-CLEAN project are provided below. Every effort has been made to ensure that these objectives are S.M.A.R.T. (Specific, Measurable, Achievable, Realistic and Timely) and clearly linked to the WPs and linked to deliverables and milestones:
1. To implement a ‘bottom-up’ approach whereby the needs and specifications of companies from the pharmaceutical industries, biopharmaceutical industry and medical device industries will be consulted and the findings used to define the specifications for the OPTI-CLEAN system. (WP1). Performance Indicator: By M3 a public report on the findings of the bottom-up research will be delivered (Deliverable 1.1) and by M4 a confidential report outlining the industrial specifications for the OPTI-CLEAN system (Deliverable 1.2) will be available for project use in order to guide the evolution of the R&D work to ensure that the OPTI-CLEAN system meets with market needs. At this stage Milestone 2 will have been achieved.

2. To carry out laboratory trials Laboratory Trials with NIR-CI technology for detecting active ingredients and detergents on typical manufacturing surfaces (WP2). Performance Indicator: By M6, a hermetically sealed Fabry-Perot Interferometer will be built and integrated into the test-rig NIR-CI (Deliverable 2.1). Laboratory trials to evaluate the performance and stability of the pre-prototype system (Deliverable 2.2) and trials with NIR-CI for detection of detergents and active ingredients on typical manufacturing surfaces (Deliverable 2.3) by M9. Finally, by M12, the scale-up parameters for developing a precompetitive OPTI-CLEAN prototype for industrial-scale cleaning validation purposes will have been defined (Deliverable 2.4). Milestones 3 and 6 will have been achieved.
• VTT has developed rugged and miniature piezo-actuated Fabry-Perot tunable filter technology platform including a large area tunable filter prototype for NIR imaging applications. Due to limitations in the achieved air gap range, the current Fabry-Perot module can be used in the 2nd and 3rd orders only, limiting the wavelength resolution to 20-25 nm FWHM. The current Fabry-Perot modules have been reported to have a 0.5 nm drift as was seen after executing 558 000 scans in 14 hours. The wavelength drift is expected to be due to temperature and moisture variation in the ambient atmosphere. These environmental sensitivities can be minimised by packaging the Fabry-Perot modules in hermetically sealed window packages. Therefore a hermetically sealed integrated Fabry-Perot Interferometer module will be optimized for improved performance for its use in the NIR-CI test rig and portable prototype developed in WP2 and WP4.
• Setting up of NIR-CI test rig: VTT will design and build the laboratory test rig, of the NIR-CI system. This test rig will be used for laboratory trials in cleaning validation tests and the performance evaluation will enable the performance of the posterior mobile system to be estimated, and suitable specifications for the mobile NIR-CI to be drawn up. The rig will be a fully developed test rig of the OPTI-CLEAN system and will therefore contain same of the main parts as the final system. Also the design phase will involve a series of subtasks: software design, design of light sources, control electronics, optics and mechanics and selection of camera. To maximise the usability of the system in the trials, the wavelength scale and spatial resolution will be made as flexible as possible. To this end significant effort will be required particularly for the optical design of the CI camera optics.

3. To validate the effectiveness of NIR-CI for cleaning validation of active-excipient blends against reference methods and to define the limits of detection of the approach (WP3). Performance Indicators: By M3, a protocol for the controlled contamination of surfaces with various excipients, actives and their blends will be defined (Deliverable 3.1) and by M11 a report on the tests using a bench-top chemical imaging system with typical pharmaceutical contact surfaces (stainless steel, glass and Perspex) will be delivered (Deliverable 3.2) and the limits of detection for the approach will be defined via correlations of NIR-CI against reference swab methods & micro imaging platforms (Deliverable 3.3). Milestones 1 and 4 will have been reached.
• Trials with NIR-CI for detection of active ingredients on typical manufacturing surfaces: The performance evaluation of NIR-CI will be tested with cleaning validation samples. The sensitivity, SNR, selectivity and other key parameters will be tested.
• Similar trials of NIR-CI for detection of detergents on typical manufacturing surfaces is carried out in the performance evaluation testing.
• Laboratory calibration model is created based on test trials. The performance of NIR-CI test rig is analysed and the analysis of results and parameters definition for industrial scale up are performed.

4. To draw up the designs of the precompetitive portable cleaning validation OPTI-CLEAN NIR-CI system in keeping with the industry specifications defined in WP1, as well as the laboratory parameters defined in WP2 and WP3 and to assemble the system hardware (WP4). Performance Indicator: By M14, the designs of the OPTI-CLEAN system will be complete (Deliverable 4.1). Milestone 7 will be complete.
During this design stage, focus will centre on ensuring that the OPTI-CLEAN system is capable of offering the following features:
• Sensitivity- there is a significant challenge in obtaining the resolution required to image minor compounds. The resolution of the various camera systems available and the signal to noise ratio of these sensors will be examined. High resolution cameras, a small field of view and magnification where required will be employed to provide the required spatial resolution. The signal to noise ratio will be addressed by investigating both the NIR to Mid IR range with InGas and MCT cameras.
• Method robustness and Reliability- The system will be based upon knowledge of bench top systems developed by each of the research groups involved which should provide a strong starting point for the envisaged prototype.
• Adheres to industry standards, and meets with regulations and approvals- The system is based upon imaging and therefore will be nonintrusive in nature.
• Portability- The future envisaged end point of this technology would be a handheld system, however for the system developed from this project, a portable system with a mobile imaging head will be build, feeding information back to the central processing unit. Such an approach will facilitate the required flexibility to analyse various manufacturing surfaces including vessels, blenders and granulators.
• Versatility- the proposed system is based upon IR spectroscopy, a universal tool for chemical identification. The system will be trained against a library of IR spectra for excipients and actives alike. Consequently the system will be versatile for identification of all pharmaceutical ingredients.
• Rapid- detection time depends on spatial and spectral resolution required, escalating with increasing resolution. Improvements in recent years in precision and speed in Chemical Imaging have arisen as a result of increased computer processing speeds, improved cameras, faster hardware, more accurate and efficient algorithms. An innovative and patented Fabry–Pérot interferometer will form the basis of the system, facilitating the required process times of 1-2 seconds. KUAVA will employ various data processing strategies to facilitate real time processing.
• Easy to use and interpret- The envisaged system will be calibrated for the selected excipients and actives employed during manufacture. It will map the analysed area in terms of identified ingredients and provide a quantification of each ingredient. The approach will not require the current level of training required by validation personnel.
• Cost effective- It is envisaged that the market price of a unit will be in the region of €60,000, which will make the technology highly competitive in the marketplace (for example the ion mobility spectrometry IONSCAN-LS was selling at $86,000 in 2004 ). Advances in the production of low cost array detectors have meant faster data collection and cheaper Chemical Imaging systems2. Moreover, predictions that the market for NIR-CI instrumentation will increase in the coming years , especially in the pharmaceutical industry due to the Food and Drug Administration’s Process Analytical Technology initiative are expected to drive down the price of imaging spectrometers, opening their way into routine pharmaceutical analysis. Moreover, the development of an average cost imager by combining micro electro-mechanical systems (electrically programmable diffraction gratings) with spectrometers would obviate the need for costly array detectors2.

5. To develop the general software for the operation of the OPTI-CLEAN system, including the synchronized control of the different subsystems, the user application and the analytical software that will process the data obtained by the system. This analytical software will use the algorithms and spectral databases developed during WP3, in order to determine the presence of contaminants and readily display the results via an ergonomic User Interface (WP5). Performance Indicator: By M15 OPTI-CLEAN Control Software (Deliverable 5.1) will be ready and the OPTI-CLEAN User Interface and database (Deliverable 5.2) will be delivered. Milestone 8 will be complete.

6. To integrate the system hardware, software and User Interface in order to provide a pre-competitive prototype that can be validated in industry and to carry out pre-validation tests with the system to ensure its proper functioning before shipping to industry test-sites. (WP5). Performance Indicator: By M16 the prototype will be fully integrated (Deliverable 5.3) and Milestone 9 will have been achieved.

7. To test and validate the OPTI-CLEAN system in commercial pharmaceutical manufacturing sites and in order to assess the efficacy of the developed system to deal with various soils and surface characteristics within a process facility. Typical manufacturing equipment (blenders, granulators, etc.) which deal with both dry and wet materials which be tested before and after validated cleaning steps. Correlations against the standard swab method will be made. Design issues arising from the industrial tests will be fed back into WP4 (WP6). Performance Indicator: Between M18-M24, the prototype will be tested and validated at MERRION in Ireland and SERVIPAST in Spain. By M17, a comprehensive installation and user manual will be edited (Deliverable 4.1) and by M24 a report on the industrial trials will be delivered (Deliverable 4.2). This will contribute to the fulfilment of Milestones 10.

8. To carry out optimisation work on the prototype based on feedback from the field trails, and to carefully outline scaling-up rules and development work for full production (WP5). Performance Indicator: By M24, a report on the optimisation work carried out on the basis of the results of the industrial trials will be delivered; including recommendations for future scale-up (Deliverable 5.4) will be delivered to the participating SMEs. Milestone 12 of the project will have been fully accomplished.

9. To facilitate the uptake of the OPTI-CLEAN results by the participating SMEs as well as a wider audience by carrying out a comprehensive series of knowledge transfer and training activities to on the one hand show the validity for the system for cleaning validation in commercial pharmaceutical, biopharmaceutical and medical device production facilities, and on the other hand to capacitate end-users about its usability and to outline its benefits for facilitating the cleaning validation process, reducing plant down-time, reducing costs and trained operator intervention, as well as increasing the reliability of the cleaning validation process, etc. (WP7). Performance Indicator: Between M18-M24 it is envisaged that at least 3 knowledge transfer and training sessions will be organised at the facilities where the prototypes are being installed and validated in WP6. A report on these knowledge transfer and training activities and materials used, including evaluations and conclusions will be drawn up (Deliverable 7.2) and Milestone 11 will have been met.

Project Results:
The key scientific and technical results and the foreground development are summarised in the following section.
WP1 served to determine the technical specifications for the system as well as determining the industry need and evaluate potential competitive technologies and patents currently focus on residual material identification and quantification.
This was achieved through a series of industry visits and discussions and well as a questionnaire and a literature and patent review.
The Questionnaire was submitted to 12 pharmaceutical and biotechnology companies. 8 companies of the 12 surveyed completed the questionnaire. The questionnaire was supplemented by site visits to 7 additional companies, see details in section 1.2.
The key findings from the web based survey were as follows:

•75% of respondents indicated that the typical formulation consists of at least 5 materials while 25% indicated that it would be more than 6.
•75% or respondents indicated that they would not quantify excipients as part of a cleaning validation program.
•Process equipment consists primarily of stainless steel, However there are small levels of Glass, Polycarbonate and Teflon.
•75% of respondents indicated that the acceptance criteria they would use during a cleaning validation study is:
Not more than 0.1% of the normal therapeutic of any product to appear in the maximum daily dose of the following product.
•62.5% of respondents indicated that the acceptance criteria that they would use for a cleaning verification study is:
NMT 10ppm of any product to appear in another product.
Acceptance criteria are dependent on product mix within the process train and dosage strength.

•75% of respondents indicated that the acceptance criteria for detergent residue in both the cleaning validation and cleaning verification study would be:
NMT 10ppm/cm²
•50% of respondents indicated that for parenterals they would apply tighter limits for active detection, the other 50% said this did not apply to their facility.

•87.5% of respondents indicated that they do not use or have not considered using any form of in-line or at-line verification process. Mainly because at present there is not a viable solution that does not involve the generation of a swabbing plan.
Only one company has been pursuing the option of an at line verification device. Their work is ongoing and there is an opportunity to review the development work carried out to date.
As this company were the only company to have seriously considered an at-line option they were the only company in a position to indicate a recommended acceptance level for residual API or detergent. During their attempts to design an at-line system for the detection of residual substances on equipment they had targeted an acceptance criteria of 0.1ug/cm². However as their testing progressed they determined that acceptance criteria of 1.0ug/cm² was a more realistic target for detection of residual substances. This was based on the fact that visual detection is typically between 1 and 4ug/cm², although to get down to the 1ug level it is necessary to introduce lighting, which in itself can lead to visual detection difficulties due to reflection within the vessels and pipe-work.
Portability and spectral range would be the principal prerequisites for an analytical device.
Only 12.5% of respondents indicated that real time results would be a very important. In general once results were available within a couple of hours then this would be perfectly acceptable.

•62.5% of respondents required quantitative results on residual levels.

•75% of respondents indicated that they would like to see the device used by the operator, while only 12.5% indicated a technician and the remaining 12.5% indicted an analyst would complete the testing.
The company visits and consultations conducted during this portion of the work package included the following companies.
1. Alkermes: Athlone – Ireland

2. Pfizer: Cork – Ireland

3. Sanofi: Germany

4. Merck Serono Germany

5. Synthon Barcelona

6. Rottendorf Germany

7. GSK UK

The in-depth industry consultations served to re-enforce the results of the web based survey. Both pharmaceutical companies and API manufacturers struggle with the combinations of the complexities of the technologies available and the difficulties posed when trying to validate the cleaning processes and programs. This is especially true in the growing environment of contained systems and continuous manufacturing processes.
In the majority of cases the preference has been for clean in place solutions for cleaning process equipment. This removes the variability associated with manual cleaning programs and reduces the risk of exposure to cleaning detergents and solvents.
However the challenge now is to validate cleaning recipes and to continuously verify the effectiveness of the recipe to keep the process stream free from any potential for cross contamination.
This process involves the need for repeated cleaning programs, the development of swabbing profiles based on the assessment of risk and the development of acceptance criteria that will provide a high degree of confidence that the cleaning programs are eliminating the risk of cross contamination.
Almost without exception the industry relies on a combination of swabbing or the monitoring of rinse waters. The monitoring of rinse waters requires high levels of automation to provide assurance that cleaning requirements are being met. Also there is the risk that the residual materials will not be soluble or may be occluded by the equipment.
Swabbing typically relies on removal of any residual materials from the swab. In the main final analysis is conducted using a HPLC method, as the level of transfer from surface to swab and swab to solution for subsequent testing can be quite minute.
The consequence for the industry is extended levels of downtime firstly while disassembling and reassembling for swab access and second for the wait on final results to confirm that the process is clean enough to proceed.
The industry is in need of a technology that will firstly provide the data in real time and secondly allow for submission of more realistic specifications for process release.
OPTI-CLEAN is to be dedicated to measuring in real time the residual levels of API and detergent. The system will use a combination of NIR spectroscopy and Chemical Imaging. The system consists of an

•Optical Imaging Sensor in the SWIR region.
•Miniature piezo-actuated Fabry-Perot tunable filter technology platform for NIR imaging applications
•Optical illumination system
•Electronics to control the system
•Calibration algorithms for the detection and quantification of contaminants and residual APIs
•An algorithm for removing the effect of the background surfaces
•A mechanical enclosure with high level of portability.
•A simple-to-use user interface for monitoring the data and controlling the system.
The OPTI-CLEAN system User Requirements Specifications determined from the various industry consultations are as follows:
•Wavelength range – 1200nm to 2200nm (Through testing and evaluation of a considerable number of substances, the peaks of interest in the spectra for the vast majority of APIs lie in the 1200 to 2100nm range.
•Field of view – A field of view of 30 x 30mm is being targeted. However the option to revert to a smaller field of view will also be considered if this enhances the performance levels of the portable device. This will be considered based on advice and consultation with the pharmaceutical industry.
•API concentration – down to 1µg/cm2
•Speed of analysis – real-time (5 – 10 secs)
•Surface topography – flat, curved and angled surfaces.
•Surface material – stainless steel, glass and polycarbonate
•Weight – around 2.5-3.5kg
•Wavelength resolution – 8nm to 20nm
•Light source – LED in device (if possible) or fed QTH source
•GUI – Comparative and quantifiable (µg/cm2) with spectra

In summary, the OPTI-CLEAN system will be a portable system that will provide fast quantitative detection of low level concentrations of impurities and APIs and detergent residuals (down to 1µg/cm2).

In work package 2 VTT successfully designed and developed a laboratory test rig to facilitate the evaluation of NIR-CI as a solution for the detection of residual substances on typical pharmaceutical surfaces.
The development of the system was conducted in two stages. This was to ensure that DIT could complete their laboratory based testing using the laboratory test rig. The decision to take a two stage approach was to allow for the inclusion of any Fabry Perot upgrades that would occur during the course of the development of the technology. This is due to the fact that the Piezo Fabry Perot Technology which forms an integral part of the NIR-CI system was being developed as part of a parallel project funded by Tekes in Finland. The initial FPI delivered for integration into the NIR-CI system in conjunction with the MCT sensor and the band pass filters had a resolution capability go 20nm, which it was felt may not be adequate for determining the final limits of detection of the NIR-CI technology. In addition the field of view of the early test rig was limited to 9 x 6mm field of view. While this was ideal for some of the laboratory testing it did not meet the current industrial requirements for an area do 5cm x 5cm. A further stage of development provided the project team with the ability to work with a larger field of view and thus support the generation of the specifications for the portable device.
There was a suite of tests conducted on both versions of the test rig and the final results are detailed in the report for deliverable D2.4.
The system was tested to verify its spectral and spatial resolution capability.

The laboratory test rig consists of a number of key features, which provide for optimising the performance of the test rig during the sample evaluation process. The key components for operational purposes are:
• Infrared sensor
• Spectrometer
• Filters
• Lighting

These key hardware components will create the platform on which to build the algorithms and computational models to determine the capability of NIR-CI to detect residual substances on typical pharmaceutical manufacturing surfaces.
Near infrared chemical imaging is an extremely useful and rugged tool that is now applied widely within the pharmaceutical and related industries. Chemical (hyper spectral) imaging is the acquisition over a larger usually contiguous of narrower spectral bands comparable to the traditional (single point) spectroscopic techniques. This is the fundamental concept of chemical imaging, a rapid analytical method that simultaneously delivers spatial, chemical (analytical), structural and functional information.
In the case of the program for the evaluation of NIR-CI as a tool for the detection of residual levels of substances, a staring imaging technology, which is focal plane array (FPA) based, was utilised for the building of the laboratory test rig.
The advantage of a staring imaging device over a push broom device is in the speed of analysis of the staring imaging option.
Array detection systems generate what is known as a hyper cube, which can generate in excess of 80,000 points of interest. In the case of the sensor specified for the laboratory test rig a hyper cube of 284 x 288 with 124 image planes is generated. This delivers over 10 million points of interest.
The detector chosen is an MCT (mercury callium telluride) (HgCdTe) SWIR (short wave infrared) sensor. The sensor is specified for a spectral wavelength range of 0.8 to 2.5µm, with a normalised spectral response as illustrated by the red trace in the image contained in the report on the performance of the test rig.
The spectrometer choice was a fabry perot interferometer (FPI). The advantages of the fabry perot and the stability it provides for the instrument and the spectral responses are in the way it handles the infrared light. Unlike spectrometers that use a grating system and a mechanism to identify and isolate the channels and wavelengths of interest. Some of the channels of interest can be lost in this pre-processing activity.
With the interferometer all infrared light is guided through the aperture in the interferometer. Reflective surfaces within the interferometer alter the distribution of light through the interferometer and transmit single light wavelengths to the sensor. A data processing application converts the raw data into the desired result, which is the spectral image of the sample being analysed.
The use of the interferometer allows the system to utilise the multiplex advantage, which facilitates the information from all frequencies being generated simultaneously improving both the speed and the signal to noise ratio.

As the interferometer transmits the information from all frequencies simultaneously a system of filters has been added. These filters are servo motor controlled and allow analysis at selected wavelengths without the distortion or the pre-processing of the infrared light prior to its transmission thought the interferometer. This allows for the isolation of specific areas of the spectral wavelength for analysis of peaks of interest within that wavelength.
The sensor has a pixel pitch of 30 x 30µm and an operating temperature range of 200k, with a cool down period of less than 30seconds.
The camera is specified with a signal to noise ratio of 69dB, given that power SNR ratio of 100 equals 20dB for a voltage SNR, then it can be clearly seen that the signal strength for the MCT sensor far exceed the noise level strength.
The sensor has an array operability specification of >99%. The specification of the sensor was determined for suitability of interface with the spectrometer and to ensure that we had the ability to capture the wavelength range for the wide range of pharmaceutical substances currently available.
The spectrometer choice was a fabry perot interferometer (FPI). The advantages of the fabry perot and the stability it provides for the instrument and the spectral responses are in the way it handles the infrared light. Unlike spectrometers that use a grating system and a mechanism to identify and isolate the channels and wavelengths of interest. Some of the channels of interest can be lost in this pre-processing activity.
With the interferometer all infrared light is guided through the aperture in the interferometer. Reflective surfaces within the interferometer alter the distribution of light through the interferometer and transmit single light wavelengths to the sensor. A data processing application converts the raw data into the desired result, which is the spectral image of the sample being analysed.
The use of the interferometer allows the system to utilise the multiplex advantage, which facilitates the information from all frequencies being generated simultaneously improving both the speed and the signal to noise ratio.

As the interferometer transmits the information from all frequencies simultaneously a system of filters has been added. These filters are servo motor controlled and allow analysis at selected wavelengths without the distortion or the pre-processing of the infrared light prior to its transmission thought the interferometer.
This allows for the isolation of specific areas of the spectral wavelength for analysis of peaks of interest within that wavelength.
Illumination of the sample is achieved through the use of tungsten halogen lamps. The tungsten halogen lamp is the most widely used source for infrared light in NIR spectroscopy. It has broadband pseudo-blackbody emission spectrum with no significant structure. The lamps have a long life, typically in the region of 10,000 hours and are relatively inexpensive.
In order to ensure that the performance of the device is optimised during the analysing process, a robust calibration process has been put in place by the system operators at DIT. In order to remove the instrument response component of the chemical image it is necessary to ratio the data to a back ground reference. For reflective measurements the back ground is a separate data cube typically generated from a uniform reflective standard such as white ceramic.
The dark camera response must also be removed from the image cube. This is achieved through the referencing using no lighting, enhanced by the location of the instrument in a dark room with blackened walls.
This calibration routine helps with the consistency of measurement during sample analysis. The continued application of the calibration routine ensures that any potential variability in results is kept to an absolute minimum.
The laboratory test rig, whose components have been outlined above, was developed to meet the requirements of deliverable 2.1 and has been used to generate the data and results that were detailed in the report on the performance of the test rig.

Report on the performance and stability tests with NIR-CI.
The suitability of NIR-CI as an application for the detection of residual levels of pharmaceutical soils is to be determined through the evaluation experiments designed by DIT and executed on the VTT supplied test rig.
One of the early stages of this process is to evaluate the capability of the test rig to detect and image residual levels of lactose on stainless steel coupons.
The development of the laboratory test rig was split into two phases. This approach was taken in order to avail of the technology improvements for the FPI. The FPI was developed in a parallel project funded by Tekes in Finland. There were a number of FPIs allocated to the OptiClean consortium at different phases in the Tekes project.
The laboratory test rig consists of a number of key features, which provide for optimising the performance of the test rig during the sample evaluation process. The key components for operational purposes are:
• Infrared sensor
• Spectrometer
• Filters
• Lighting

These key hardware components will create the platform on which to build the algorithms and computational models to determine the capability of NIR-CI to detect residual substances on typical pharmaceutical manufacturing surfaces.
Near infrared chemical imaging is an extremely useful and rugged tool that is now applied widely within the pharmaceutical and related industries. Chemical (hyper spectral) imaging is the acquisition over a larger usually contiguous of narrower spectral bands comparable to the traditional (single point) spectroscopic techniques. This is the fundamental concept of chemical imaging, a rapid analytical method that simultaneously delivers spatial, chemical (analytical), structural and functional information.
In the case of the program for the evaluation of NIR-CI as a tool for the detection of residual levels of substances, a staring imaging technology, which is focal plane array (FPA) based, was utilised for the building of the laboratory test rig.
The advantage of a staring imaging device over a push broom device is in the speed of analysis of the staring imaging option. Also the option to draw a hand held portable device directly over the surface of process equipment could introduce contamination issues.
Array detection systems generate what is known as a hyper cube, which can generate in excess of 80,000 points of interest. In the case of the sensor specified for the laboratory test rig a hyper cube of 284 x 288 with 124 image planes is generated. This delivers over 10 million points of interest.
The detector chosen for the laboratory test rig is an MCT (Mercury Callium Telluride) (HgCdTe) SWIR (short wave infrared) sensor. The sensor is specified for a spectral wavelength range of 0.8 to 2.5µm, with a normalised spectral response.
The sensor has a pixel pitch of 30 x 30µm and an operating temperature range of 200k, with a cool down period of less than 30 seconds.

The camera is specified with a signal to noise ratio of 69dB, given that power SNR ratio of 100 equals 20dB for a voltage SNR, then it can be clearly seen that the signal strength for the MCT sensor far exceeds the noise level strength.
The sensor has an array operability specification of >99%. The specification of the sensor was determined for suitability of interface with the spectrometer and to ensure that we had the ability to capture the wavelength range for the wide range of pharmaceutical substances currently available.
The spectrometer choice was a Fabry Pérot interferometer (FPI). The advantages of the Fabry Pérot and the stability it provides for the instrument and the spectral responses are in the way it handles the infrared light. Unlike spectrometers that use a grating system and a mechanism to identify and isolate the channels and wavelengths of interest. Some of the channels of interest can be lost in this pre-processing activity.
With the interferometer all infrared light is guided through the aperture in the interferometer. Reflective surfaces within the interferometer alter the distribution of light through the interferometer and transmit single light wavelengths to the sensor. A data processing application converts the raw data into the desired result, which is the spectral image of the sample being analysed.
The use of the interferometer allows the system to utilise the multiplex advantage, which facilitates the information from all frequencies being generated simultaneously improving both the speed and the signal to noise ratio.
As the interferometer transmits the information from all frequencies simultaneously a system of filters has been added. These filters are servo motor controlled and allow analysis at selected wavelengths without the distortion or the pre-processing of the infrared light prior to its transmission thought the interferometer.
This allows for the isolation of specific areas of the spectral wavelength for analysis of peaks of interest within that wavelength.

Illumination of the sample is achieved through the use of tungsten halogen lamps. The tungsten halogen lamp is the most widely used source for infrared light in NIR spectroscopy. It has broadband pseudo-blackbody emission spectrum with no significant structure. The lamps have a long life, typically in the region of 10,000 hours and are relatively inexpensive.
In order to ensure that the performance of the device is optimised during the analysing process, a robust calibration process has been put in place by the system operators at DIT. In order to remove the instrument response component of the chemical image it is necessary to ratio the data to a back ground reference. For reflective measurements the back ground is a separate data cube typically generated from a uniform reflective standard such as white ceramic.
The dark camera response must also be removed from the image cube. This is achieved through the referencing using no lighting, enhanced by the location of the instrument in a dark room with blackened walls.

This calibration routine helps with the consistency of measurement during sample analysis. The continued application of the calibration routine ensures that any potential variability in results is kept to an absolute minimum.
The laboratory test rig, whose components have been outlined above, was developed to meet the requirements of deliverable 2.1 and has been used to generate the data and results that demonstrate the capability of the system for the purposes of the OptiClean project.
The suitability of NIR-CI as an application for the detection of residual levels of pharmaceutical soils is to be determined through the evaluation experiments designed by DIT and executed on the VTT supplied test rig.
One of the early stages of this process is to evaluate the capability of the test rig to detect and image residual levels of lactose on stainless steel coupons. Initial tests used a stain of 500ug per 10uml droplet. Deliverable report D2.2 displays the images generated at this early phase. A clear lactose spectral response was generated with no unexplained or unexpected noise.
Further evaluation of lactose was completed to evaluate the opportunity to detect lower concentration levels using the NIR-CI test rig, these results are outlined below.

Hyper spectral images were taken from lactose stains of 250µg, 125µg, 60µg, 10µg and 1µg respectively. Again there was a clear absence of unexpected or unexplained noise in the spectral images.
The hyper spectral images clearly indicate the presence of the lactose even at the low concentration level of 1µg. A micropipette was used to produce the stains using a water dilution and following the DIT staining protocol as outlined in deliverable D3.1.
Further tests were carried out on a number of pharmaceutical APIs, again the spectral responses were very clear with no unexplained or unexpected noise in the images.
In relation to the performance of the test rig through the two stage development, the spectral linearity has been consistent; there have been some slight deviations in linearity if the test rig operating temperature increases due to the lighting. Updates are made for better robustness and measurement repeatability. The main reason for fluctuations between measurements has been identified to be changes in the halogen spectra. For making sure that the halogen process is stabilized as quickly as possible the thermal capacity of the test-.rig is minimized. In addition, the illumination current is monitored and the user can only start measuring in a stable state. Furthermore, the illumination source has been optimized by measuring different halogens lamp with integrating sphere and choosing the most suitable model.
Originally the spectral resolution was measured to be between 15 and 20 nm. A next generation Fabry-Perot component has been fabricated and tested. The new generation has a reduced aperture, now 15mm from the previous 19mm. There are new materials utilised in the reflective surfaces and the new Fabry-Perot is also tuneable. The new Fabry-Perot is installed in the upgraded sections of the laboratory test rig and a second interferometer has been issued to IRIS for the construction of the portable test rig. These enhancements will result in a much improved resolution.

To achieve this, the new optics will be configured into an improved composition of band pass filters to enable the usage of higher orders of FPI filter response.
The FPI response consists of multiple orders, from which only one is used by applying a wide band pass filter. It was clearly evident that the higher order peaks are narrower, going right to left thus giving a better spectral resolution. The trade-off is that the peaks are located more close to each other demanding more and narrower band pass filters. The updated band pass filter set allows using the 4th order peak, giving a resolution close to 10nm. The updated band pass filter responses along with overall transmission values are clearly illustrated in the report for deliverable D2.2. The rectangular responses are for the newer filter options and the arched responses, also illustrated in D2.2 represent the performance of current filters.
The spatial resolution was assessed using a number of MTF plots with three different wavelength and band pass filter combinations. As could be seen from the results detailed using the MTF plots in D2.4 the spatial resolution is very close to the resolution of the sensor pixels.
The SNR was recorded by recording the variation of individual pixels between consecutive frames compared to pixels mean value. SNR was measured with different exposure times and wavelengths. Results are given in the table below. Compared i.e. to a normal RGB machine vision cameras, the results were good.

Exposure
[ms] Fabry[nm] Signal filterDark filter Dark lights SNR [dB] Notes
8 2000 9 0 on 57,3
8 2000 9 9 on -0,5 SNR= 0,9
8 2000 9 0 off 57,7
18 2350 6 0 on 56,0
18 2350 6 6 on 10,3 SNR= 3,3
18 2350 6 0 off 56,9
The spectral resolution of the system depends on the response of the Fabry-Perot unit. It was characterized using a commercial bench-top NIR spectrometer. The resolution is defined to be the full width half maximum of the peak in the Fabry-Perot unit transmittance spectrum. The measurement range is divided into zones according to the band pass filter characteristics, using a different order of a peak in each zone to maximize spectral performance. The measured resolution and used Fabry-Perot order is given in the table below.
Zone indexλ1 λ1 Order Resolution [nm]
1 1101 1375 3 21
2 1376 1575 4 16
3 1576 1644 5 12
4 1645 1875 4 16
5 1876 1950 5 12
6 1951 2180 5 12
7 2181 2257 5 12
8 2258 2300 5 12
9 2301 2391 5 12
10 2392 2438 5 12
11 2439 2490 5 12

The performance matched the computational simulations made for both the optics and the Fabry-Perot unit.
In keeping with their responsibilities in WP2, VTT also had to provide the specifications for the portable device. The design of the portable chemical imager was based on the following:
The design parameters enable an instrument that remains within the specifications of end-users in terms of portability and performance. The chemical imager itself is designed to weigh approximately 4kg and have maximum dimensions of 36 x 23 x 21 cm, which will allow it to be manually handled and used. A tripod can be used to provide stability and flexibility in positioning. Heavier elements of the system, including power supply and computer will be contained in a separate support pack that can be held on the shoulder or placed on the ground or a worktop. The entire system can be packed into two hard plastic cases for transport from site to site. Knowledge transfer from VTT to IRIS has been completed in line with the requirements details in table 2 above, and will facilitate the effective construction of a portable pre-competitive prototype.
The size of the sampled area on the surface is set at 40mm squared, which closely matches the size used in current swabbing techniques. The resolution of the system in wavelength is 20 nm, and spatially is approximately 0.5mm. With respect to the sensor an extended InGas sensor has been chosen, this was due in the main to the long and unpredictable lead time associated with the delivery of the MCT sensors. The noise level of the camera is similar to that existing in the pre-prototype used for chemical analysis validation at DIT, so the camera will not affect the limit of detection. The most significant difference in instrument parameters is the wavelength resolution, which is approximately 10 -20 nm for the laboratory prototype. However, the portable version may be conveniently refitted to attain the narrower bandwidth by replacing the current FPI module with one of higher maximum air gap. This will allow the use of higher harmonics, such as the 3rd or 4th harmonic. This refitting does not require any mechanical alterations, and can be achieved by programming new settings in the software, with the possible replacement of one of the filters. Any requirement to use the higher harmonics will be determined from the performance of the portable test rig once initial laboratory testing has been completed. In addition it is envisaged that 6 band pass filters will be sufficient to achieve the required performance criteria for the portable device.

In Work Package 3 DIT demonstrated the capability of NIR-CI to detect residual levels of pharmaceutical materials on typical process surfaces. This was done through the simulation of staining using a controlled method to achieve consistency in application of stains of different concentration levels. This method was detailed in a protocol which was submitted as deliverable D3.1.
The suitability of NIR-CI as an application for the detection of residual levels of pharmaceutical soils is to be determined through the evaluation experiments designed by DIT on the VTT supplied test rig.
Generating and analysing controlled soils of single materials.

It is not easy to weight out concentrations via a dry powder mix due to these very low concentration levels. Serial dilution seems a natural manner in which to obtain the desired concentrations. Chemicals such as Lactose, Caffeine and acetylsalicylic acid are water soluble. Other tablet ingredients such as MgSt are soluble in ethanol. These controlled soils were analysed using a combination of NIR-CI and Microscopy.

The images detailed in D3.2 clearly illustrate the system’s ability to measure these soiled coupons.
Initial test utilised stains of approximately 250ug of lactose and 60ug of lactose respectively. A micro-pipette was used to produce these stains via a water dilution. The stain was left to dry out leaving only the lactose behind, the stains can be contiguous or non-contiguous depending on concentration.

The spatial and hyperspectral images in the report for D3.2 clearly illustrate the system’s ability to image stains. There was some shadowing introduced by the camera but it was possible to build a simple rule to pick out the pixels that represented the lactose.

Density plots were utilised to evaluate the two stains and similarly to above there was a small third peak present due to shadows in the camera.

An evaluation of lactose stains using an atomiser to generate the soils was also conducted. A stereoscope image of a Lactose stain produced by an atomiser and a “corresponding" hyper spectral image, again demonstrate the system’s ability to be able to distinguish the lactose on the stainless steel surface.
Further work was carried out implementing the learnings from the experiments conducted above. In this case the stereoscope and hyper spectral images illustrated the system’s capability in this case with different concentrations of caffeine from 125ug down to 1ug.
The next phase of the evaluation process was to apply the leanings from phase 1 to the generation of data using blended materials.
As with other materials stainless steel will have absorbance values. In order to build a classification function to identify aberrant pixels some assumptions are made with regard to the distribution of absorbance values of stainless steel pixels. It is assumed that after the hypercube has been auto-scaled that a blank stainless steel coupon will have pixels, at any given wavelength, that follow a standard normal (Gaussian) distribution. Stainless steel had on average higher absorption values across the entire NIR range (1260 to 2200nm) compare to all APIs tested. Hence we would expect residues to have corresponding auto-scaled pixels lower than that of stainless steel pixels. Thus possible thresholds for detecting residue pixels correspond to values below the lower percentiles.
For example P < 0:01 of the standard normal distribution has a corresponding Threshold = -2:362.

Several images were generated to show the typical spectra for a number of common APIs versus the raw absorbance spectrum for stainless steel. The resultant spectral responses correlated clearly to the individual APIs utilised in the tests.
Before model building occurs hyper spectral data often under goes transformations via Standard Normal Variate (SNV) or Multi-Scatter Correction (MSC). Derivative algorithms such as the Savitzky-Golay smoothing filters are often applied as well. These methods are applied to each pixel's spectrum and are used to deal with random effects with regard to the intercepts and slopes.
Having demonstrated that it is possible to detect residual levels of blends or single materials on clean stainless steel surfaces, using a transformation of the spectrum to identify substances, the challenge then is to quantify the residual materials on the stainless steel coupons.
Quantification will be done by using the spatial information to estimate the concentration levels per area via a pixel count. Another option is to use pixel spectrum in concentration estimation.
For the quantification of residual substance on stainless steel an design of experiment was created suing various concentrations of caffeine, the design of experiment was then repeated using a common pharmaceutical API.
Experimental Design: Stains of Pure Caffeine were made at 6 levels of weights (1ug, 10ug, 25ug, 50ug, 75ug, 100ug) and repeat twice times at each level. Thus a total of 6*2=12 stains where place on stainless steel coupon. Each stain was scanned twice.

This sweet of tests produced a linear regression on pixel count by stain weight, while this result only had an R2 of 62%, further development of the algorithm achieved a liner regression with an R2 of 96%.
Similarly for detergents the tests for those supplied by the industry demonstrated a linear regression with an R2 of 60%. Further work will need to be completed on detergent samples.
Ina addition it was noted that the majority of detergents are sodium carbonate based, while these don’t have a spectral response with unique peaks of interest they do generate a spectral reaction and can be imaged, so quantification is possible.

The algorithms developed for the purposes of identifying and quantifying residual soils are supported by a number of transformation models such as Standard Normal Variate (SNV) and Multi Scatter Correction (MSC). In addition derivative algorithms such as the Savitzky-Golay smoothing filter and a multivariate Gaussian distribution are also used.
A Hyper spectral Image, denoted x, could be viewed as a multivariate distributed random variable. For simplicity, it will be assumed that these images are independent and identical distribution (I.I.D) and that these variables can be appropriated by a multivariate Gaussian distribution. Since a Hyper spectral Image can consist of an average number of multiple scans, assuming scans are I.I.D the Central Limit Theorem can be utilised.

π(x)=2π^((-k)/2) |〖Σ|〗^((-1)/2) ∫▒〖e^((x-μ)^('Σ^(-1) ) (x-μ)) dx〗

Where k = 125 _ 384 _ 288 = 36; 110; 592 and _ is a covariance matrix of size 36,110,592*36,110,592. Some further assumptions have to be made in order to make the analysis feasible.
Dr Laura
In order to build a classification function to identify aberrant pixels some assumptions are made with regard to the distribution of absorbance values of stainless steel pixels. It is assumed that after the hypercube has been auto scaled that a blank stainless steel coupon will have pixels, at any given wavelength, that follow a standard normal (Gaussian) distribution. Stainless steel had on average higher absorption values across the entire NIR range
(1260-2200nm) compare to all APIs tested. Hence we would expect residues to have corresponding auto scaled pixels lower than that of stainless steel pixels. Thus possible thresholds for detecting residue pixels correspond to values below the lower percentiles.
For example P < 0:01 of the standard normal distribution has a corresponding Threshold = - 2:362.
Currently the Hyper spectral Imaging system is being used over the range 1260 - 2170nm with images taken at every 10nm. Thus there are n = 92 possible wavelengths to build a threshold. For simplicity a method was first developed to choose just one suitable wavelength. The wavelength that results in the smallest standard error associated with a calibration curve is chosen. This method was extended to work with two wavelengths. Unfortunately this procedure has a high computational burden thus generalisation to multiple wavelength combinations may only be possible via Simulated Anneling rather than an exhaustive search.

Example of 2D parameter space for a threshold function
The arrow indicates the combination of two wavelengths that minimises the standard error associated with a calibration curve via thresholding. Note there does not necessarily have to be a unique minimum. There may be many local minimums whose locations will be connected to the absorption spectra of the stainless steel and the residues.
Once the thresholding and modelling has been defined, quantification of the residual substances is achieved through pixel count.

Experimental Design: Stains of Sulfathiazole sodium salt were made at 6 levels of weights (1ug, 10ug, 25ug, 50ug, 75ug, 100ug) and repeat four times at each level. Thus a total of 6*4=24 stains where place on stainless steel coupon. Each stain was scanned twice.

A pixel count was then obtained per stain concentration level (100-1 µg/50mm2) after applying the mask by simply adding those pixels classified as stain after the thresholding. Pixel count data per stain concentration level (100-1 µg/50mm2) was then fitted with a linear model to obtain the different model parameters including intercept, R2, Standard Error and Limit of Detection.
The algorithm and modelling approached developed above on the stainless steel coupons were also applied in the evaluation of stains on glass and polycarbonate materials (Perspex). The algorithm was able to image and quantify the soils on these materials much as it did for stainless, so there is no significant re-modelling required for these surface areas.
As a result of the experimental work carried out to date it is not envisaged that curved or angled surfaces within vessels or typical processes will present an issue, however to determine the opportunities for success in being able to verify the cleanliness of pipe work and valve further activity will be planned utilising the portable test rig. This is necessary as the viewing window of the laboratory test rig is smaller and therefore not influenced by curvature.
The final part of work package 3 was to evaluate the results achieved on the laboratory test rig against a standard industry process for residual level determination and to determine the limits of detection for the system.
In the design of experiment to compare the residual quantification capability of NIR-CI against HPLC, a number of differing concentrations of caffeine and a typical pharmaceutical active were carried out. As per the normal procedure for HPLC recovery a HPLC calibration was also performed to ensure that there was no bias in the results achieved.
HPLC calibration curve for caffeine.

Area under the curve from HPLC chromatogram (273 nm) versus concentration (ug/ml) for the swab extracts of the caffeine stains 1ug to 100ug. Recoveries (calculated as measured*100/theoretical using an external standard calibration curve) were above 85% except for 1ug stains.
The experiment was then taken to the next level where the area under the curve for HPLC of caffeine was compared to the pixel count.

Experimental Design: Stains of Pure Caffeine were made at 6 levels of weights (1ug, 10ug, 25ug, 50ug, 75ug, 100ug) and repeat twice at each level. Thus a total of 6*2=12 stains where place on stainless steel coupon. Each stain was scanned twice.
There was a loss of linearity at the 1ug level versus the HPLC data, this phenomenon was not ignored but as will be seen from the second group of experiments using a typical active pharmaceutical ingredient there was no evidence in the loss of linearity. It is most likely the result obtained above resulted from the manner in which the caffeine stain dried following application on the stainless steel plate

The images contained in the report for D3.4 illustrate the calibration and the HPLC area under the curve for a typical pharmaceutical active.

Experimental Design: Stains of Sulfathiazole sodium salt were made at 6 levels of weights (1ug, 10ug, 25ug, 50ug, 75ug, 100ug) and repeated four times at each level. Thus a total of 6*4=24 stains where place on stainless steel coupon. Each stain was scanned twice.

Area under the curve from HPLC chromatogram (260 nm) versus concentration (ug/ml) for the swab extracts of the Sulfathiazole sodium salt stains 1ug to 100 ug. Recoveries (calculated as measured*100/theoretical using an external standard calibration curve) were above 75% except for 1 ug stains.

In the stains using the typical pharmaceutical active ingredients, in a comparison of the average pixel count versus the weighted average of the stain there was no loss in linearity at the 1ug level. The linear regression R2 value for this set of experiments was 96%.

The data generated in these experiments demonstrated that we could draw comparisons from area under the curve analysis. That using NIR-CI for cleaning verification we have the potential to run comparative results against typical swabbing methodology.

It should be noted at this point that even with HPLC which is the most comment method for cleaning evaluation and validation within the drug product and drug substance there is a significant reduction in recovery levels as you get to the 1ug level.
With regards to the limits of detection this was calculated in the following manner.

• A pixel count versus stain weight calibration curve was built in the range 100 to 10 µg (100, 75, 50, 25, 10 and 1 µg). This calibration curve was prepared using four replicate stains per stain weight. The 4 data points obtained per stain weight were used to build the calibration curve, as opposed to building the calibration curve using one point per stain weight which would correspond to the average of the 4 replicate stains.

• The Limit of Detection was calculated according to the International Cooperation on Harmonisation (ICH) [22] which defines the detection limit of an analytical system as the concentration or amount corresponding to a measurement level 3.3 times the standard error above the value for zero analytes (equation (1)).

Limit of Detection (LOD) = y + 3.3 × σ (1)

Where y is the absorbance for a blank sample and σ is the standard error.

• Equation parameters (intercept and slope), R2 and Standard Error (SE) values of the linear curves were obtained for the different APIs.

• Limit of detection in absorbance units was calculated as the value corresponding to that of the intercept of the curve plus three times the Standard Error of the curve. The limit of detection in weight units (µg) is then calculated from the calibration curve as the x value (µg) that corresponds to the calculated limit of detection in absorbance units.

The detection limits calculated according to the International Cooperation on Harmonisation (ICH) [22] for caffeine and sulfathiazole sodium salt soils on stainless steel coupons were 48.42 and 18.20 µg per 50 mm2 respectively. In the case of caffeine soils on borosilicate glass and Perspex coupons, the calculated limits of detection were 33.46 and 27.24 µg per 50 mm2 respectively. A visual evaluation of the images displayed

in Figure 1 however suggest that the potential exists to lower the limit of detection of API residues to values closer to 1 µg/50 mm2. It is important to note that the limit of detection was calculated based on the standard error of the linear model built using an average of 4 replicates per concentration level. The calculated limits of detection values reflect the standard error associated with the linear model (Equation (1)) fitted to the experimental data and is a result of the marked variability in pixel count values between replicate API residues of the same concentration. Lower limits of detection may be thus achieved using this technology by improving the uniformity and thus replicate precision of the soils. Different staining protocols may be employed to this end such as the use of an atomiser or aerosol as well as testing various drying conditions such as the use of infrared heating or other standardised drying methods. Moreover, in this study only pixel count values as determined by the classification function were used to fit a linear regression model to quantify concentration of API residue on background surfaces. Future work will involve the use of further classification algorithms as well as algorithms to extract information from pixel absorbance values and pixel spectra to obtain additional information from the API soils such as soil thickness and thus develop robust and accurate models with possibly lower limits of detection for the quantification of API residue.

The work carried out above demonstrated two things for the project; firstly can the results achieved using the NIR-CI test rig can be used to show correlations between this approach and the traditional swabbing and HPLC recovery method. This will assist greatly in proving the effectiveness of the portable test rig for ongoing real time cleaning verification. The results achieved do demonstrate that we will be able to prove correlation to the existing swabbing and HPLC solutions. As demonstrated even HPLC % recoveries drop significantly at the lower concentration levels.
Secondly in our numerous discussions with the pharmaceutical and medical device industry the proposed limit of detection for a continuous process would be approaching 1.0ug per cm2 and that this is quite acceptable. The success realised so far in being able to image and quantify soils with a solution of just 1ug demonstrates that we can achieve results approaching this detection level. In addition the work completed to date determines the limits of detection in line with current ICH guidelines.
The efforts From WP1, WP2 and WP3 fed into the design and building of the portable prototype.

The basic functionality of the design and some of the components are based on the successful bench-top prototype delivered during the first half of the project by VTT, for which laboratory tests and the development of chemical analysis algorithms are reported by DIT in the Deliverables D3.2 Efficacy of NIR-CI to identify and quantify low amounts of pharmaceutical residues on surfaces, D3.3 Algorithms for the detection of contaminants with the NIR-CI device, and D3.4 Report on limits of detection for NIR chemical imaging. Accurate measurement of API mass on stainless steel was achieved from 1 to 100 g/cm2. The prototype weighs over 20kg, and so is not portable for inline measurements.
Starting from this point, the design was developed to reduce its weight and size, while maintaining ruggedness and stability, by using lightweight materials for the construction of the mechanical framework. The complexity of some features, such as the optics and electronic control was reduced, while at the same time the performance of critical components such as the camera and Fabry-Perot interferometer was maintained so that the portable device has the ability to reach the same limits of detection as the bench top version.
The principle of operation is as a spectral imaging camera in the staring imager configuration, which is to generate a series of narrow wavelength-band images of the same location. By combining the images together in a data cube, it is possible to measure the infrared spectrum of any pixel in the image or to measure the spatial distribution of a single wavelength band.
The components of the chemical imager are shown in the block diagram in Figure 2. Halogen lamps are used to illuminate the sample surface. Light from the surface is collected by a front lens and passed through a filter and a Fabry-Perot interferometer. These filter out all light except a narrow bandwidth, which is imaged on a short-wave infrared (SWIR) camera. After a single image is captured, the filter/Fabry-Perot system settings are adjusted to allow transmission of another wavelength band, until the desired set of narrow waveband images is collected and stored as a single datacube.
Maximum dimensions of the imager are 36 cm x 23 cm x 21 cm, and the overall weight is estimated to be approximately 4kg.

The chemical imager is controlled by computer, which is carried in a separate portable support pack, along with the power supply, which converts the mains voltage to 12 Volts DC. A microcontroller contained in the chemical imager optical head distributes power to the individual components and controls the acquisition routine. Parameters of the acquisition routine are set from the computer.
Halogen lamps are used as a broadband source emitting strongly from 300 nm to 3000 nm (Figure 3a). The selected lamps are BLV-Licht Reflekto, which are small in size – diameter 35 mm and pins GU4, and have an aluminium parabolic reflector which reflects over 90% of the light generated into a beam of 12 degree divergence. The spot size at the 30cm from the chemical imager is approximately 80mm, which makes efficient use of the light (Figure 3c). The region of interest that is imaged by the lenses is 44 mm x 44 mm, which is within the illuminated spot. Each lamp has 20W of power and 6 are available for use. It will be possible to select the number of lamps to illuminate, so that the light intensity and the heat at the sample surface can be controlled.
The optical design uses two lenses with the object and image planes at the focal distances to either lens (Figure 4). A front collimating lens collects the light from the sample surface and a camera lens images it on the sensor. The light between the lenses is collimated for passing through the Fabry-Perot interferometer and filter. The size of the image relative to the object size is demagnified in the same ratio as the focal lengths.
The collimating lens is an achromatic doublet with a focal length of 300 mm, and the camera lens is a compound lens designed for CCTV applications with a focal length of 50 mm. The sensor chip has area of 9.6 x 7.7 mm. A square of distance 7.7mm on the sensor contains the image of a 44 mm square on the object surface.

The collimating lens is an achromatic doublet optimised for use over the range 1050 – 1620 nm, diameter 1 inch, supplied by Thorlabs (Figure 5(a)). The camera lens is supplied by Kowa and is optimised for the shortwave infrared, with transmission ranging from 100 – 65 % over the range 1200 – 2200 nm. (Figure 5b). The f-number is 1.4 and it has a C-mount that allows direct attachment to the camera. Its focus is set to infinity to form a clear image of the collimated light from the Fabry-Perot.

The collimating lens is held in an adjustable holder that enables positioning to be changed along the optical axis up to 25 mm. This allows the fine focus of the camera to be adjusted after positioning. Rough focus will be achieved by positioning the front of the camera at a distance of 285 mm from the sample surface. A pair of diode laser pointers with overlapping beams at the correct distance will be used to estimate rough positioning.

Six filters are used to cover the wavelength range from 1200 - 2200 nm. Filtering is in two stages. The filters transmit a wide band of light, over approximately 250 nm. The interferometer selects a narrow band from this, of about 25 nm FWHM. This is imaged on the camera. The FPI is set to transmit the desired light as the 2nd harmonic of the cavity (Airgap/3). This narrows the transmitted bandwidth (the fundamental FWHM is about 40 nm), and the filters are used to block out the adjacent harmonics.
The filters are held in a wheel and rotated into position sequentially into the optical path. The filter wheel has been made in IRIS from PVC. Eight slots are included, which contain the six filters, plus an opaque metal shutter to enable dark reference measurement, and a blank slot to allow broadband transmission. Short ½” lengths of lens tube are used to hold the filters in place. The wheel is driven by a servo motor which is connected at the centre (Figure 6). Outer diameter of the wheel is 126mm.

An electronically adjustable, piezo-actuated Fabry-Perot interferometer is used to select narrow passbands from the light collected from the sample surface (Figure 7). The aperture of the module is 18 mm, and the maximum airgap is 7112 nm. The aperture sets the maximum width of the transmitted collimated beam in the current optical configuration. The control box of the FPI is located in the design above the SWIR camera.
The air gap enables selection of the second harmonic up to the top of the instrument wavelength range of 2200 nm (6600 nm airgap). This sets the passband width at approximately 25 nm, which is therefore the wavelength resolution of the system.
The camera is supplied by Photonic Science Ltd., and uses an extended InGaAs sensor covering the wavelength range from 1200 – 2200 nm. The chip is cooled by a single Peltier thermoelectric stage to reduce the temperature by 50 degrees below ambient, and the heat sink is cooled by a fan blowing through the body of the camera. Data output and camera control is via GigaBit Ethernet cable, with a frame rate of 60 fps. Power supply is 12V DC. The sensor has 320 x 256 pixels of 30 x 30 um size. The raw image from the sensor is processed by white pixel correction and flat-fielding to apply median values to defective pixels. Dark current is 5pA, which is less than 25% of full well capacity for a 0.5ms exposure in high gain mode. It has a C-mount for lenses and weighs less than 1 kg.

The mechanical structure is based on an aluminium frame with front-walls made of PVC and methacrylate (Figure 10). These carry the weight of the components, which are directly attached to this frame. A handle on top of the imager and a base at the bottom for attachment of a tripod are directly integrated with the load-bearing frame. Encapsulation of the imager is completed by attachment of panels to the sides, top and bottom, which may be either flat sheets or curved surfaces.
The design parameters enable an instrument that remains within the specifications of end-users in terms of portability and performance. The chemical imager itself is designed to weigh approximately 4kg and have maximum dimensions of 36 x 23 x 21 cm, which will allow it to be manually handled and used. A tripod can be used to provide stability and flexibility in positioning. Heavier elements of the system, including power supply and computer will be contained in a separate support pack that can be held on the shoulder or placed on the ground or a worktop. The entire system can be packed into two hard plastic cases for transport from site to site.
The size of the sampled area on the surface is set at 44mm squared, which closely matches the size used in current swabbing techniques. The resolution of the system in wavelength is 25 nm, and spatially is approximately 0.5mm. The noise level of the camera is similar to that existing in the pre-prototype used for chemical analysis validation at DIT, so the camera will not affect the limit of detection. The most significant difference in instrument parameters is the wavelength resolution, which is approximately 10 -20 nm for the pre-prototype. However, the portable version may be conveniently refitted to attain the narrower bandwidth by replacing the FPI module with one of higher maximum airgap. This will allow the use of higher harmonics, such as the 4th or 5th. This refitting does not require any mechanical alterations, and can be achieved by programming new settings in the software, with the possible replacement of one of the filters.
In conjunction with the design and building of the prototype an operation and user manual was also generated to support the use and testing of the industrial prototype.

The OPTI-CLEAN prototype is built in two parts: the handheld device that holds the optical system and the central unit with the main computer and the power supply as explained in detail in D4.1 Designs of the OPTI-CLEAN Prototype.
The handheld device is controlled by the control board that provides the USB communications with the main computer at the central unit and the control of the filters wheel and the lamps. It includes also a USB HUB to allow the main computer to communicate with the Fabry-Perot interferometer.
The main computer controls the operation of the handheld device, the image acquisition from the InGaAs camera and provides the user interface that allows the use of the prototype.
The communication protocols are based on the following.

The protocol used is based on messages built with the following format:
Messages PC  Control board:
P d ? lrc
P d V d d d lrc
? lrc
STX> F d lrc
L d V d lrc
Messages Control board  PC:
P d V d d d lrc
I d d lrc
OPTICLEAN Vd.d lrc
Formatting bytes:
: Start Of Text (ASCII 2)
: End Of Text (ASCII 3)
Lrc: Longitudinal Redundancy Check (less significant byte of the binary sum of all bytes of the message including the STX and ETX)
Characters used:
Uppercase 'P': meaning 'parameter'
Uppercase 'V': meaning 'value'
Uppercase 'F': meaning 'filter' - set the filter position
Uppercase 'L': meaning 'lamp' - switch the lamp on/off
Question mark '?': meaning a requirement
Uppercase 'I': meaning 'incidence'
Lowercase 'd': decimal digit in ASCII

The electronics control board is custom made to serve the power and control requirements of the chemical imaging system, although a direct Gigabit Ethernet link is maintained from the computer to the camera to ensure the fastest possible frame rate and data acquisition. All other parts are powered and accessed through the control board, with communication by USB.

The support pack, including mains to DC transformers and a laptop computer, is housed within a hard plastic case from Peli Products. Another Peli case is used to store and transport the portable module. The entire system is robust, lightweight and portable, and is suitable for use in a wide range of at-line locations.

The building of the prototype optics and mechanics carried out in work package 4 and the integration of the whole system including the electronics developed in work package 5 finished with the assembly of the OPTI-CLEAN prototype.

Once stated the features and functionalities of the software that controls the OPTI-CLEAN prototype, its implementation is defined in the annex I as flow charts whilst the actual implementation of the code will be documented in the report D5.2 OPTI-CLEAN User Interface and database. The user interface that will provide the access to these features and functionalities is detailed in the report D4.3 First version of the OPTI-CLEAN user manual where its installation and operation are also specified.

In work package 6 the testing of the portable prototype was to take place in industry, however despite the construction and configuration of the portable prototype in accordance with the project timeline. Issues arose during the testing. Images were not of a suitable quality to achieve the limits of detection experienced with the laboratory test rig. Also there were some anomalies in the spectral profiles of the test materials.

The issue with the anomalies in the spectral profile was deemed to be an issue with curvature on the reflective surfaces of the Fabry Perot Interferometer. In conjunction with this issue there was also some issues with the calibration protocol for the FPI. The FPI had to be removed from the portable device and returned to VTT for assessment.
It was decided at this point to replace the FPI with a unit form a later stage in the FPI development process. Again this FPI required calibration and driver integration before installation into the portable device. The FPI was installed and some further calibration was required following integration with the OptiClean system. At this point the system underwent a series of tests to determine its performance parameters. During this testing there were still some issues with the quality of the images impacting on the system ability to meet the limits of detection required. Due to the delays encountered when correcting the issues experienced with the Fabry Perot this phase of testing was now taking place in the period when we should have been testing in industry. In consultation with industry, while keeping them updated on our progress, it was industry's preference that we prove the system in an academic environment before returning to industry to conduct in line trials. The ongoing testing at both DIT and IRIS and the modification to the algorithm could not improve the performance of the system with regard to limits of detection. The system is capable of identify products and providing a spectral response based on verification standards. However the spectral response has some smoothening of the peaks and the hyperspectral image while identifying the stain included to much pixilation to be able to identify lower concentrations. Currently the system is achieving a lower limit approaching 25ug/cm2. This is not close to the original capability of the laboratory test rig and certainly not within industry requirements. To this end the images were sent to Photonic Sciences, the supplier of the SWIR sensor. They applied some clean up software algorithms to the images but could not improve on the results. It was then necessary to return the sensor to Photonic Sciences, they tested the sensor and discovered an issue with one of the logic boards. This issue took some time to discover as this type of fault typically results in an inability to generate images at all. The sensor has now been returned to Innopharma and it will be reinstalled in the OptiClean device to complete further testing. The camera was sent to Photonic Sciences in November 2013 and returned to Innopharma in February 2014. Innopharma and the other SMEs are committed to the successful implementation of this technology. It forms a strategic part of the Innopharma PAT portfolio and will be important for Innopharma, Rikola and Kuava with respect to future exploitation.
Images associated with the deliverable reports are appended as an attachment to this document.

Potential Impact:
The SME benefit will come from the fact that pharmaceutical processors are obliged to validate cleaning procedures as a customer/market requirement, whereby it ensures the safety and purity of the product, as a regulator requirement, as well as from an internal quality control and compliance point of view. The introduction of Near Infra Red Chemical Imaging (NIR-CI) technology to rapidly quantify the contaminant levels remaining on product contact equipment following the execution of cleaning activities will remove the current challenges posed by conventional cleaning validation methods. The proposed results of this project will deliver a mobile device capable of rapidly identifying and quantifying active and detergent residue levels. It is envisaged that the results will increase the competitiveness of thousands of companies operating within the bulk and finishing pharmaceutical industries, biopharmaceutical industry and medical device industry across Europe.
In terms of the SME participants, on the one hand, SERVIPLAST, as manufacturer of pharmaceuticals and as such end-users of the results of this project, the uptake of the technology will deliver a number plethora of benefits, from reduced down time for equipment, reduced risk of cross contamination for product shared equipment, greater assurance for regulatory bodies, greater flexibility of manufacturing equipment to support non-routine manufacturing activities, for example, clinical trial manufacture. Previously the validation requirements prior to the introduction of a new active ingredient to a commercial line on a temporary basis would be too high to justify its use. Moreover, they will be able to eliminate swab analysis, thereby eliminating HPLC equipment and resource requirements. They will also realise financial savings through laboratory headcount reduction, elimination of analysis time / instrumentation and reduced manufacturing cycle time. These savings would be greater than €800,000 for a typical pharmaceutical manufacturing facility. They will also be equipped with a technology that will enable them to reduce the risk of cross contamination as a result of continuously verifying the cleaning process.
The SMEs from the NIR-CI-based technology supply chain (INNOPHARMA, RIKOLA, KUAVA) will also improve their competitiveness as a result of this research activity through the commercialisation of the OPTICLEAN technology. They will be able to take the results of this project to market, thereby tapping into a highly attractive business opportunity by supplying the novel OPTI-CLEAN to meet demand within the lucrative bulk and finishing pharmaceutical industries, biopharmaceutical industry and medical device industry across Europe and indeed globally. Discussions with both multinational pharmaceutical companies and detergent suppliers have confirmed the real need for the technology. Given that cleaning validation is conducted in every pharmaceutical company worldwide, and that there is potential for sale of multiple units to each manufacturing site (typically 5 units), the initial revenue potential is in excess of €21million (this is discussed in greater detail in Section B.3.1).
Innovation and the uptake of new technologies will prove of paramount importance and developments such as OPTI-CLEAN will go a long way towards assisting to strengthen Europe’s reputation as a consistent supplier of high quality and safe pharmaceuticals, biopharmaceuticals and medical devices. To this end, the SME participants are committed, via their active participation in this Research for SMEs project, to guiding the RTD performers in the development of a Near Infra Red Chemical Imaging (NIR-CI) technology to rapidly quantify the contaminant levels remaining on product contact equipment following the execution of cleaning activities, that will be capable of overcoming the limitations of existing cleaning validation methods in industry. The SMEs will actively participate in the work plan in terms of mobilising a bottom-up approach whereby they will guide and support the RTD performers in delivering usable results that they will validate and exploit. The SMEs will take on ownership of the beneficial results of this project and the responsibility for their full exploitation at European-wide level.
The results of this project will represent a significant advance of the state-of-the-art in cleaning validation and indeed the developed mobile NIR-CI cleaning validation system will be a worldwide breakthrough for the pharmaceutical, biopharmaceutical and medical device industries. It is envisaged that by overcoming the limitations of the existing state-of-the-art in cleaning validation that a number of key innovations and technological progress will emerge from the research.

In terms of a contribution to the advancement of knowledge, as an emerging technique, Chemical Imaging integrates conventional imaging and spectroscopy to attain both spatial and spectral information from an object. As was discussed above, the combination of the chemical selectivity of vibrational spectroscopy with the power of image visualisation, enables this technique to deliver a more complete description of ingredient concentration and distribution in heterogeneous solids, semi-solids, powders, suspensions and liquids.
Fabry-Perot interferometers using piezo-elements for actuation is a well-known technology. VTT has been developing piezo-actuator –based Fabry-Perot interferometers for spectral measurement applications since 2006. These devices follow conventional techniques at many levels, whereby three piezo actuators are used to move the mirrors and three plate capacitances are used for closed loop feedback control of the device. Such a principle has been presented for example by Rees et al. in the 80’s. VTT’s piezo-actuated Fabry-Perot Interferometers (piezo-FPIs) have been constructed using silica substrates and Ti-Ag-SiO2 mirrors. A complete camera system utilizing the piezo-FPI tuneable filter component has been made for Unmanned Aerial Vehicles for the spectral range of 500 – 900 nm. The complete device weighed less than 350 g including batteries and it was used to identify diseases in pear trees. Another system that has been built is a prototype of a chemical imaging spectrometer operating in the wavelength range of 1000 – 2500 nm. The concentration distribution for caffeine, aspirin and acetaminophen in an Excedrin™ tablet was successfully measured. Figure 2 shows the developed piezo-FPI filter versions. These components enable the development of spectrometers and spectral imagers at much lower cost and size than previously possible.
Most of the effort for developing these piezo-FPI versions has been focused on the 7-mm aperture device, designed for the visible range. The benefits of the new devices compared to for example Acousto-Optic Tuneable filter (AOTF) or Liquid Crystal Tuneable Filter (LCTF) devices are their small size and weight, speed of wavelength tuning, high optical throughput, independence of polarization state of incoming light and the capability to record three wavelengths simultaneously. Recently this novel method for utilising three multiple orders of a Fabry-Perot simultaneously has been developed and demonstrated. Here the key is to use a sensor element with pixels dedicated to certain wavelength ranges.
The project will result in a mobile chemical imaging technology that will rapidly quantify chemical residues from pharmaceutical manufacturing equipment surfaces. It is envisaged that the ownership of the rights to exploit the proposed results will afford the participating SMEs with a real business opportunity with a clear economic impact. To this end, they aim to protect the knowledge generated during the project, in order to enable the generation of income and competitive benefits to be derived through the exploitation of the IPR, which will bring a clear strategic and competitive impact for the participating SMEs.

As cleaning validation is a necessary and time-consuming part of manufacturing pharmaceuticals, biopharmaceuticals and medical device manufacturing, it is envisaged that MERRION and SERVIPLAST as end-users of the technology, will benefit by gaining early and affordable access to a novel device that will enable them to greatly improve on their current cleaning validation practices thereby bringing multiple benefits:
•Rapid turnaround time, reduced down time for equipment, and very high throughput- Current swab testing can take anywhere from 3-5 days if done in-house and up to 2-3 weeks if contracted by an outside lab . Providing a portable at-line device that can monitor directly on solid surfaces for contamination will enable the equipment being cleaned to be returned to production more rapidly and efficiently and with greater reliability. This will not only shorten cleaning validation time from days to hours, it will also increase product throughput in the plant, thereby preventing bottlenecks and giving producers greater flexibility to react to market needs. It is expected that these benefits will equate to millions in cost savings and increased turnover.
•Reduce costs- In view of tighter regulations the cost of validating clean surfaces has escalated over the last 10 years. The uptake of the OPTI-CLEAN technology will eliminate swab analysis, thereby eliminating HPLC equipment and resource requirements. Costs of laboratory tests have been cited at $250 (€190 approx.) for HPLC assays (non-acetonitrile solvent system). Further method validations begin at $8500 (€6500 approx.) for HPLC. 50% can be added to these costs for acetonitrile based solvent systems. To this end, financial savings realised through laboratory headcount reduction, elimination of analysis time / instrumentation and reduced manufacturing cycle time. These savings would be greater than €800,000 for a typical pharmaceutical manufacturing facility.
•Reduced risk of cross contamination for product shared equipment and prevention of rejects- by having access to a knowledge-based technology that rapidly quantifies chemical residues from pharmaceutical manufacturing equipment surfaces, will provide operators with real-time reliable information that will enable them to make rapid production decision, remove the subjectivity and limitations of human error, and thereby prevent cross-contamination between batches and subsequent rejections. Moreover, there will be reduced risk of cross contamination as a result of continuously verifying the cleaning process. Importantly, it will prevent contaminated product reaching the market where it could have devastating effects for patient health, as well as a crippling impact on the company through costly recalls, detrimental bad publicity and potently costly law suits.
•Greater assurance for regulatory bodies- the proposed chemical imaging technology will greatly facilitate conformance with applicable GMP (Good Manufacturing Practice) guidelines , whereby it is a requirement of GMP that manufacturers identify what validation work is needed to prove control of the critical aspects of their particular operations. It is stated that validated analytical methods having sensitivity to detect residues or contaminants should be used and the detection limit for each analytical method should be sufficiently sensitive to detect the established acceptable level of the residue or contaminant. This will be possible with the OPTI-CLEAN technology.
•Reduction of human error - by replacing operator dependent swabbing techniques, the effectiveness of which require significant training and depend greatly on the skill and expertise of the operator, the risk of human error will be greatly decreased.
•Reduce the costs and the time to market for new product developments- By greatly speeding up the cleaning validation process, greater flexibility of manufacturing equipment to support non-routine manufacturing activities will be enabled, for example, clinical trial manufacture. Previously the validation requirements prior to the introduction of a new active ingredient to a commercial line on a temporary basis would be too high to justify its use. This will help to bring down the costs of new product development. This would be of significant benefit in that the cost of researching and developing a new chemical or biological entity was estimated at € 1,059 million in 2006.

All of the above benefits will significantly boost the competitiveness of the participating pharmaceutical, biopharmaceutical and medical device manufacturing SMEs. In turn, these benefits will stimulate market demand for the proposed OPTI-CLEAN system among the European (and indeed global) sectors. Moreover, increasingly stringent regulatory requirements (EU Guide to Good Manufacturing Practice; the U.S. Food and Drug Administration, FDA), increasing schedule and cost pressures, as well as the demand for globally active service providers, will fuel demand for technologies such as OPTI-CLEAN. This opens up an enormous opportunity for the SMEs from the OPTI-CLEAN supply chain (INNOPHARMA, RIKOLA and KUAVA) to tap into this market, thereby benefiting from increased sales and income from the supply of the Fabry Perot Interferometer (RIKOLA) and the provision and maintenance of software (KUAVA), as well as the manufacture and sale of the complete solution (INNOPHARMA). They expect to access new international markets and open up distribution networks across Europe, and indeed beyond, as a result of the commercialisation of the novel OPTI-CLEAN NIR-CI cleaning validation system.
In order to quantify these competitive benefits for each SME, the direct economic impacts have been calculated (Table 6), along with the subsequent new jobs that could be created as a result. These initial estimations are based on an analysis of the type of exploitation that each beneficiary will make of the results in keeping with their positioning on the user and supply chain, their existing turnover and employees and their capacity to exploit the results of the project, and the resulting competitive benefits.
Table 6: Estimated direct economic impact for the SME participants

Short Name Type of exploitation 2015 2016 2017 2018 Total New Jobs

INNOPHARMA Commercialisation of 180980 407205 814410 1176370 2578964 10
Complete system
SERVIPLAST System uptake 106465 273766 577951 760462 1718654 4

KUAVA Software supply 111312 250451 500903 723526 1586192 8

RIKOLA Supply of FPI 90000 189000 342000 531000 1152000 2

Totals 488757 1120422 2235264 3191358 7035801 24

Beyond strengthening the competitiveness of the participating SMEs, this OPTI-CLEAN project will also contribute to improving industrial competitiveness across the European Union through the commercialisation of the system in EU-27. Figures published in 2008 by the European Commission (Eurostat) show that the pharmaceutical industry is the industrial sector which invests most in research & development with 15.3% of total EU private R&D expenditure. Of the 635,000 people it employs in Europe, 117,000 work in R&D . To this end, the research-based pharmaceutical industry is one of Europe’s leading high-technology industrial employers. In fact, recent studies in some countries showed that the research-based pharmaceutical industry generates three to four times more employment indirectly - upstream and downstream - than it does directly, a significant proportion being high value added jobs (e.g. clinical science, universities, etc.)19. The pharmaceutical industry is also the sector with the highest ratio of R&D investment to net sales. It amounts to approximately 3.5% of total EU manufacturing value-added and 19.2% of the total worldwide business R&D expenditure. The research-based pharmaceutical industry is a key asset of the European economy representing about 19.2% of global business R&D investments and about 3.5% of the total EU manufactured exports19.
European demand for pharmaceuticals is expanding due to an ageing population, earlier diagnosis of disease and wider use of pharmaceuticals. However, despite this growing domestic demand, since the early 1990s, Europe has been losing competitiveness with respect to its main competitors, in particular the US. Data for 2007 and preliminary figures for 2008 confirm the vulnerability of Europe’s research-based pharmaceutical industry. Benchmarking and performance indicators show Europe's relative lack of attractiveness for pharmaceutical R&D investments. Between 1990 and 2008, R&D investment in United States grew by 5.6 times whilst in Europe it only grew by 3.5 times. Today there is rapid growth in the research environment in emerging economies such as China and India, resulting in closures of R&D sites in Europe and openings of new sites on the Asian continent.
Additionally, the European medical device manufacturing industry represents a lucrative €55 billion market that has not yet been fully capitalised on by the supply sector. Maintaining the competitiveness of this sector, Moreover, the European medical devices market has, in recent years, seen an increase in the quality of devices and components manufactured in Asia . Medical device manufacturers based in Europe are currently running the risk of losing increasing market share to those based in Asia.
Equipping the European bulk and finishing pharmaceutical industries, biopharmaceutical industry and medical device industry with a technology that will enable them to safeguard product quality and safety, reduce costs, increase production throughput, meet with international good manufacturing practices, and facilitate quicker new product development, will undoubtedly contribute to increasing the competitiveness of these sectors at European level, as well as to safeguarding the futures of the 635,000 employed in the pharmaceutical and biopharmaceutical industries, and hundreds of thousands in the medical device sector19 (in Ireland alone 24000 people are employed in the sector ), as well as their related supply and value chains in Europe. We also envisage the creation of at least 1200-1500 new jobs through increased pharmaceutical, biopharmaceutical production in the EU, as a result of increased product throughput and capacity in existing plants, as well as knock-on jobs along the value chain, and in the manufacture of the OPTI-CLEAN system itself to serve the European (and global) market.

Once OPTI-CLEAN is ready to launch in the market, it is envisaged that the ownership of the rights to exploit the proposed technology will afford the industry partners with a real business opportunity through the uptake and commercialisation of a novel NIR-CI cleaning validation technology with very promising market potential at EU, and indeed global, level.
Cleaning validation in the pharmaceutical and biotechnology industry is a major concern in drug manufacturing. From 1999 to 2006 around 35% of the "Warning Letters" from the FDA contained remarks on cleaning and cleaning validation . The verification that potential contaminant levels are below acceptable levels has driven demand for a number of analytical techniques . It is predicted that, as the need for fast and reliable cleaning validation techniques will grow in the future, techniques that can provide low-level specific analysis with high speed most likely will be the leading techniques on the market . This spells excellent promise for OPTI-CLEAN once it enters the market.
In fact, the worldwide pharmaceutical market is very lucrative for pharmaceutical producers and also for equipment suppliers, service providers and consultants . The world pharmaceutical market more than doubled in size between 1998 and 2006 . In 2009 global pharmaceutical sales were worth approximately 750 billion-760 billion U.S. dollars . The US, however, is the world market leader with a 39.3% share of world production, followed by Europe and Japan. Seven emerging markets- Brazil, China, India, Mexico, Russia, South Korea, and Turkey- contribute nearly 25% of growth worldwide and were expected to grow 12–13% in 2008 to $85–90 billion .
To satisfy demand, a large number of capital projects have been initiated and completed worldwide, creating what appears to be the ideal growth market for companies that specialise in the design and delivery of pharmaceutical plants and equipment, and especially suppliers of specialised equipment25.
The proposed OPTI-CLEAN system will offer a number of Unique Selling Points (USPs) in view of its innovation and advancement of the current state-of-the art, such as its speed, its ease of use, removing the need for swabbing and laboratory analysis, its mobility enabling it to serve multiple lines and equipment in the facility, etc. It is expected that these USPs will afford OPTI-CLEAN with a considerable edge in the marketplace. During Subtask 8.3.1-Unique Selling Points & Market Potential, an in-depth analysis of these USPs of the OPTI-CLEAN technology will be carried out, whereby the benefits, costs and Return-On-Investment (RIO) ratios of the technology will be carefully studied, in order to effectively position the OPTI-CLEAN system in the market, which will contribute to a powerful marketing and communication campaign to support the launch of the system post-project.
As a preliminary estimation of market potential, should the proposed OPTI-CLEAN system be brought to market at the expect unit cost of €60,000, and if it fulfils the technical, regulatory requirements and enable conformance with applicable GMP (Good Manufacturing Practice) guidelines, within 5 years of completion of the project, EU and global cumulative sales could have reached a conservative 300 units for that period, representing €18 M worth of turnover. These figures are deemed as highly realistic and achievable in view of the cutting edge benefits that the technology offers, the willingness of the sector to invest in equipment, technology and R&D (the context of the cost of the OPTI-CLEAN in comparison to the cost of for example, researching and developing a new chemical or biological entity, estimated at € 1,059 million in 200618, is highly accessible, especially in view of the benefits that it will bring), as well as the fact that globally there are well over 10,000 pharmaceutical manufacturing sites that could invest in the OPTI-CLEAN technology, and that there is potential for sale of multiple units to each manufacturing site (typically 5 units).
To this end, the envisaged sale of 300 units in the first 5 years of market operation is indeed pessimistic, and represents a modest 0.6% of the full market potential for the system in the pharmaceutical sector alone. However, potential in the biopharmaceuticals sector looks promising, as new system designs are needed to meet growing market demand, especially as cleaning following the fermentation process is currently creating production bottlenecks25. Moreover, the European medical device manufacturing industry represents a lucrative €55 billion market that has not yet been fully capitalised on by the supply sector . It is considered reasonable to suggest that a further €3 M could be generated over that same period through after-sales services, consumables, not to mention any spin-off developments that may arise due to the wider application of the NIR-CI platform technology for the validation of cleaning in other applications, such as cleaning validation in dietary supplements, nutraceuticals, medical feeds manufacture, etc. The SMEs may also explore the possibility of licensing the technology to other companies inside and outside the EU to heighten the strategic market penetration and acceptance needed to achieve the above economic aims.
The results achieved to date in the project clearly demonstrate that the portable cleaning verification technology is capable of delivering on the potential outlined above both within the European and wider pharmaceutical industry and directly for the participating SMEs.

This level of economic benefit increases the SMEs ability to not only sustain their position but also to realise the potential for growth. In the case of Innopharmalabs their technology department has already increased from 1 to 4 people and with the possibility of 2 additional people in 2014. Some of this additional headcount is directly associated with the ongoing development work on the OptiClean technology.
The increase in the levels of safety of pharmaceutical products has a significant impact for the patient. The supply chain becomes more reliable, the product safer and process more robust. This is especially important as the growth potential for European pharmaceutical companies is the supply of drug product into the emerging markets, such as China, India and Africa. Robust, efficient and effective process will be key to sustaining competitive advantage as the industry moves into these particular markets.
In addition the regulatory landscape has changed significantly since 2011, the FDA has issues their new guidelines with respect to process development and validation and this year 2014 the European pharmaceutical industry has followed suit with the publication of its draft guidelines in January of this year. Process analytical technologies are very much to the fore in the new guidelines and in particular cleaning process and cleaning verification.
The main process for dissemination will be to continue the work with academic partners and expand on the publications for NIR-CI as a cleaning verification tool. The publication of results in white paper form or article sin journals such as pharmaceutical Technology will be key to opening the opportunities to partner with individual companies within the pharmaceutical industry to promote and exploit the result of this project. The dissemination plan will be reviewed once the next phase of testing has been completed in Q2 of this year 2014. Once the result achieved meet the original specifications for the technology a plan will be put in place to formally launch the technology and to provide it as a commercial offering through Innopharma's technology portfolio. The successful outcome of the trials and ongoing verification in academia are key to the successful launch of this technology. The target would be to launch the technology at IFPAC in 2015 and to accompany the launch with presentations on results achieved using the device in commercial scale pharmaceutical processes. there are a number of forums available to the SME to achieve this objective. Some of the SMEs from the OptiClean project are also involved with the Promis project in Finland, the ERC (engineering research centre) in the USA, the PMTC (pharmaceutical manufacturing technology centre) and the SSPC (solid dose pharmaceutical cluster) in Ireland, these forum provide a natural environment to facilitate the testing, generation of articles and publication of data form the testing process.

List of Websites:

www.opticlean-fp7.eu
Mr Ian Jones
jonesi@innopharmalabs.com

Final Report Summary - OPTI-CLEAN (Effective and reliable optical system for cleaning validation in pharmaceuticals manufacture)

Udostępnij tę stronę

Pobierz