ParticleScope™ – System Background

Patented Non-Destructive Particle Detection and Measurement System used for Parental (SVI) Products.

 

Phoenix Imaging Ltd. ParticleScope™ Logo

The ParticleScope™ instrument is unique in its approach for the isolation of suspect contamination particles in the solution.  The instrument was developed to aid in studies of human and machine inspection capabilities.

According to cGMP practice, the validation demonstration of any alternative method or mechanism used on a USP approved product must show functionality at least as effective as the preceding method or mechanism. In the case of visible contaminating particles, a demonstration of at least equivalent performance to the single container clinical inspection for visible particles is required. The Knapp-Abramson methodology has achieved de facto status as the standard methodology for this purpose since its 1980 PDA Journal publication.

The present USP Attribute Sampling Inspection methodology, a descendent of MIL SPEC 105D, is in current use as an injectable product batch release assay. This attribute sampling inspection bases results of its analysis on a sample of 2 to 2,000 containers randomly extracted from a production batch. The sample size and the non-conforming number of containers in the test group for acceptance of the production batch are dictated by the AQL required and the size of the production batch. To satisfy the prerequisites for accurate analyses from the Attribute Sampling Tables, the data on which the analysis is based must be, to quote from the introduction to ANSI/ASQC Z1.4-1993, “must be simply repeatable”. Another requirement is that the data considered should be normally distributed. It must be stressed that Sampled Inspection conclusions are incompatible with PAT continuous batch quality improvement perspectives. This incompatibility traces to the crudeness of the Attribute Sampling Decision and it’s planned to restriction to data with “simply repeatable errors” describable by a normal distribution.

This incompatibility traces to four factors: 1) Visible Particle data is probabilistic and is therefore not “simply repeatable”. 2) The random nature of visible particle data cannot be described with a normal distribution curve. 3) The crudeness of the Sampling Inspection rejection set point (it is set at twice the Accept Quality Level, the AQL). 4) The asymptotic approach to zero acceptance of the Operating Characteristic, the OC that can result in the acceptance of a multiply flawed batch as typified in the Detroit “lemon car”.

From Knapp’s seminal publication in the PDA Journal in 1980, the prime parameter for the analysis of visible particle contamination in injectable products has been demonstrated to be the probability of detection for the contaminating particle. Visible particle detection data has been shown to be both randomly sourced and randomly located in the containers of a batch. Consequently, the use of the Attribute Sampling Tables with incompatible probabilistic raw visible particle inspection data has been shown to result in the rejection of good batches and the acceptance of batches with sub- standard quality.

An alternative batch evaluation methodology based on the reject rate of a validated 100% inspection which uses the size of the contaminating particle is proposed and has been demonstrated. The probabilistic nature of the data results in improved accuracy of the determination when made with the entire data of a batch compared to the smaller number of data points available in the sampled data procedure.

Conversion of the manual raw visible inspection data into equivalent particle size measurements is accomplished with a calibration curve. The calibration curve uses a standard test set of containers seeded with a range of single, durable, micro sized spheres whose dimensions are traceable to NIST Dimensional Standards to define the relationship of (particle size) to (particle detectability) in accurately specified test conditions.

The accurately specified test conditions include: the quantity and quality of the light at the inspection point and the contrast of the background. The selected test conditions provide for a ½% overlap between the visible and sub-visible particle regions. The inspection rate and the fatigue of the inspector performing the inspection are also critical test conditions. Particle movement during the inspection period is essential to distinguish the low ratio of contaminating particles from the bulk of the extraneous visible information.

From the previously cited Knapp publication in the PDA Journal in 1980, the prime parameter for visible particle inspection data has been demonstrated to be the probability of detection for the contaminating particle. The direct use of raw visible inspection data has been shown to result in the rejection of good batches and the acceptance of sub-standard quality batches. Conversion of the raw, probabilistic inspection data into equivalent particle size measurements is accomplished with a calibration curve. The calibration curve uses a standard test set of containers with a set of durable range of micro sized spheres, whose dimensions are traceable to NIST Dimensional Standards, to define the relationship of (particle size) to (particle detectability) in accurately specified test conditions. The accurately specified test conditions include: light quality and intensity at the inspection point, the contrast of the background, and particle movement during inspection to distinguish the low ratio of particle signals from the bulk of the extraneous visible data, inspection rate and the fatigue of the inspector performing the inspection.

Each batch is tested to determine the quantity of non-conforming containers in the test group against the batch acceptance limit determined from production records.. An alert level is set at 120% of the standard acceptance limit and an action level is set at 150% of the targeted reject rate.

Note: the alert and action levels are respectively 5 times and twice as sensitive as the hoped for results from the sampling inspection.

The conversion of raw visible inspection data to particle size can be accomplished by repeated re-inspection of each container to achieve 0.05 significance level rejection probability as described in Knapp’s publication. The mean rejection probability determination is converted to particle size with a calibration curve. An alternative, more economically desirable method is to use a calibrated low power stereo microscope to directly measure the particle size to determine the accept/reject status of the container. From present data, the edge of the Reject Zone is set at contaminating particles sizes equal to or larger than 95µm (average rejection probability = 0.7071).

The use of a calibration curve based on NIST traceable particle size measurement converts Knapp’s methodology from a local validation method whose results were difficult to communicate to a securely transportable quality determination derived from NIST dimensional standards.

Following the quality control mishaps in Japan in 1995 in which insect bodies, hair and lint were found in accepted containers of injectable products, the USP then reaffirmed its reliance on the single container inspection for contaminating particles at the injection site.

Therefore any other manual, semi-automated, or automated particle inspection procedure must be shown to be at least equivalent to the manual benchmark performance before it can be used on any USP listed product.

It has been demonstrated that a well-defined inspection environment is required to obtain consistent results from manual human inspection procedures. The environment must provide uniform illumination within a large enough volume of inspection chamber to allow for position deviations while being inspected. The amount of lighting (intensity at the subject) is critical to achieve repeatable results. It has been found that 550 foot-candles at the point of inspection will allow the detection of a 95 µm diameter object to be detected 70.7% of the time by trained inspectors with 20/20 vision at a distance of approximately 400 mm. Small changes in the lighting conditions can have a large influence in the apparent size of the contaminating particle.

Optimized Manual Inspection Environments

The art of visual inspection requires a consistent environment for the acquisition and analysis of all the information in the sharp image and its blur surround. Phoenix Imaging has developed a well defined inspection environment to support the requirements of human inspection procedures. The environment provides very uniform illumination within a 1 cubic foot volume of inspection chamber. Special inspection chambers such as the Phoenix Imaging MIB-50 & MIB-100 have been constructed to yield a standardized inspection environment with a large useful inspection volume. The large inspection volume of the MIB-100 provides a 550 foot-candle illumination source anywhere within the volume, minimizing the effects of variations in container placement from sample to sample. The MIB series of inspection environments implement a patented dual-source illumination approach to produce consistent results.

MIB-100 with Side Shelves

MIB-100 in Standard Configuration with Optional Production Side Shelves added.

The designs of all the Dual-Sided MIB products (i.e. MIB-50™, MIB-80™, MIB-90™ and MIB-100™), strive to minimize inspector fatigue. The MIB models 80-100 allow the inspector to adjust arm rest height, in-feed and acceptable product tray height, tray angle, and tray tilt without the use of any tools.  Each inspector can adjust the unit for maximum ergonomic use and comfort. The MIB-90™ & MIB-100™ provides a PLC based timer for inspection pace and inspection count. The MIB-50™ is a bench top inspection system designed to fit in to laboratories where space is a concern. The MIB-50™ provides about 7.5 Liters of inspection volume with the illumination within ±10% of center value.

If any factor is compromised, the reliability of the instrument will be questionable and possibly jeopardize the ability to validate the inspection process. None more so than the standard sample set used for the training and qualification of inspectors.

An essential lack in visible particle measurements has been that of an accurate transferable standard. This lack is now being supplied by Phoenix Imaging as a standard stable micro-particle set whose measurements are referred to the national dimensional standards maintained by the NIST. This standard set when inspected for 5 seconds per container against the recommended Kodak 18% Gray background results in a procedure with twice the output of white/black background inspection at higher accuracy. The higher accuracy is due to the fifty percent reduction of movements required by the inspector for the container inspection.

The strobe-free light quality and 550 foot-candle intensity provided by the MIB is essential for the accuracy of the particle size detectability data required for an accurate transferable standard.

The Phoenix Imaging RLPS™ Standard Particle Calibration Set provides a unique set of single seeded NIST traceable particles in stabilized water. The set includes glass, stainless steel and polystyrene spheres that range in diameter from 40µm to 1,000µm, fibers of nylon and cellulose that range in length of 0.5 mm to 2 mm, aluminum and glass shards of measured length. The sample set also includes “Clean” or uncontaminated containers. The entire set is comprised of 50 containers each coded with a unique serial number to identify the contents. The range of detectability for this set is from poorly detected (<10%) to a size that is securely detected in every inspection.

The standard particle set is used in conjunction with the MIB Standard Lighting Environment to train human inspectors the art of manual inspection and to determine the inspection reliability limit of the individual inspectors. The beginning of a national standard calibration curve has been published at the 2004 PDA Annual Meeting. All data generated in the same defined environment will added by Phoenix Imaging to improve the precision of the national standard curve.