Creating a Culture of Data Integrity Using an Automated Detection and Enumeration Method Novartis, Schaffhauserstrasse, 4332 Stein AGStein, Switzerland Sanofi Pasteur, Marcy l Etoile, France Leo Pharma, 39 Route de Chartres, 28501 Vernouillet, France IBSA, Via del Tritone, 169, 00187 Roma, Italy biomérieux Industry Culture Media Group, Chemin de l Orme, 69280 Marcy l Etoile, France Abstract: In the pharmaceutical industry, the ability to understand, follow and predict the microbiology surrounding drugs manufacturing plays a key role in the quality and the safety of the products. The more common way to enumerate microbiological contamination today is the use of solid culture media. The manual enumeration by an analyst of microbial colonies after incubation is still the standard to assess microbiological results. This task is manual, based on human interpretation and can be a source of errors during transcription in LIMS. New rapid testing technologies allow to improve the traceability of the information but generally fail to prove the equivalence of performance compared to a growth based method. EviSight Compact is an innovative system that incubates standard media dishes detects growing microbial colonies. This new technology brings to the user the full traceability of the data, the reliability of the results and the possibility to go back in time to observe early events that occur during microbial colony growth and this without modifying the already validated compendial methods. Keywords Automated microbial method, data integrity, traceability Microbiological Monitoring of water, environment and intermediate products as well as microbiological testing of final products are routine tests in the Pharmaceutical industry. In addition growth promotion tests are routinely performed to evaluate the fertility of growth media before use. The industry performs thousands of tests using agar culture media and in the majority of the cases, Petri dishes are enumerated by a visual count performed by an operator. In this article, we present an innovative system, EviSight Compact, an automated incubator and plate reader. The performance of the technology was challenged in term of detection and enumeration on three main type of applications: 90mm plates reading, Contact plates reading and filters reading by four different and independent pharmaceutical industries. The summary of the statistical analysis performed on the data generated by the Beta sites allowed to evaluate the following criteria: specificity, accuracy, precision and equivalence of the methods. based on traditional (90mm) or Contact Petri dishes according to the applications. Fig. 1: Presentation of the full capacity system The EviSight Compact system (Fig.1) is designed with one computer running the analysis software. With the first version up to four readers can be connected to 1 PC. Material and Methods EviSight Compact presentation The EviSight Compact is a smart incubator system that automatically reads standard commercial plates for detection of microbial contaminants. It is designed to be compatible with filterable products (water-like products) and direct agar reading (environmental control, product control ) Fig. 2: Incubation module and a tray with Petri dishes Each incubation module is independent from the other and can be run individually. The Petri dishes are introduced in the module thanks to a dedicated tray adapted to the type of Petri dish to analyze: 6 wells for 90 mm formats (Fig. 2) and 12 wells for 55mm or contact formats. The modules can be set for an incubation between 20 and 35 C depending on the
applications. Within one incubation circle different temperatures can be chosen. Test microorganisms Test microorganisms included in the study were sourced from culture collection and commercial calibrated strains: BioBall Multishot (biomérieux): Aspergillus brasiliensis ATCC 16404, ref. 56011 Bacillus subtilis ATCC 6633, ref. 56012 Candida albicans ATCC 10231, ref. 56013 Pseudomonas aeruginosa ATCC 9027, ref.56017 Staphylococcus aureus ATCC 6538, ref. 56019 Culture collection: Methylobacterium extorquens NBRC 15911 Pseudomonas fluorescens NBRC 15842 Additional in-house microorganisms were also tested during the study and were sourced from the different pharma industries collection. The recommended dilutions of each microorganism were 5, 25 and 50 CFU/inoculum. Five replicates per microorganism of each concentration were used to perform the study. Culture media and filters The table below describes the different consumables used to perform the study: Product Reference Supplier Culture media 90mm R2A 43551 biomérieux TSA 43011 biomérieux TSA 3P w. neutralizers 43811 biomérieux Sabouraud 3P 43814 biomérieux Culture media 55mm R2A AEB523487 biomérieux Culture media Contact plates Count-Tact 3P 43691 biomérieux Count-Tact SDA 3P 43812 biomérieux White plain filters Cellulose acetate filters 11106-47-ACN Sartorius PVDF filters HVWP047S6 Millipore MicroFunnel filters 516-8963 Pall Table 1: Consumables used during the study Filtration application was tested in combination with R2A and the filters on the panel of microorganisms with the 3 different concentration. The other culture media were tested by direct inoculation of the 3 concentrations. Incubation conditions The tests were performed according to the incubation times and temperatures presented in the table 2. Agar R2A CT 3P TSA 3P CT SDA 3P SDA 3P Microorganisms P. aeruginosa ATCC 9027 B. subtilis ATCC 6633 M. extorquens NBRC 15911 P. fluorescens NBRC 15842 S. aureus ATCC 6538 P. aeruginosa ATCC 9027 B. subtilis ATCC 6633 C. albicans ATCC 10231 A. brasiliensis ATCC 16404 C. albicans ATCC 10231 A. brasiliensis ATCC 16404 Table 2: Incubation times and temperatures Trueness evaluation Incubation time and temperature with EviSight 36h 32,5 C 36h 32,5 C 72h 32,5 C 72h 22,5 C 52h 32,5 C 52h 22,5 C 52h 22,5 C The trueness is defined as the closeness of agreement between the average enumeration results obtained from a large series of test results by the EviSight Compact and average value of manual reading (reference value). The accuracy was evaluated at each targeted concentration (5, 25 and 50 CFU/plate) independently. Precision The precision is defined as the degree of agreement among individual count values obtained by the EviSight system when the procedure is applied repeatedly to multiple samplings of suspensions of laboratory microorganisms across the range of the method. Precision is usually expressed as a standard deviation, variance or %CV for the repeatability of the method. Repeatability precision was estimated using a nested analysis of factors (different series of 5 replicates). Comparison of repeatability variances obtained for EviSight and manual plate reading was performed using the Fisher test (with a risk alpha=5%). Comparison of EviSight Compact enumeration and manual plate counting The comparison of EviSight Compact enumeration and manual plate counts was performed as follows: i. With a linear regression of EviSight enumeration results versus manual plate counts. Linear regression estimates (slope and intercept) were estimated using the least squares method. The R² associated to the linear regression shall be 0.9
and slope of the linear regression shall not be diverging more than 20% from 1.0 (within 0.8-1.2) 1. ii. With a Bland Altman plot: the difference in counts is plotted against the manual reading counts in order to see if there is a trend in the difference over the tested range. Agreement limits are calculated as mean of the differences ± 1.96 x standard deviation and plotted on the graph. This interval is supposed to contain 95% of the values. Results Kinetic growth and go back in time feature The EviSight Compact system runs a high magnification technology enabling a detection of micro-colonies at an early stage of their development when they have a size of 30-50µm. The EviSight Compact takes a picture of the incubated plates each 30 minutes and powerful algorithms analyze the development of microorganisms that are present on the culture media. The system follows then the kinetic of growth of the different microorganisms (see figure 3). Figure 3: Count evolution of a pure culture of C. albicans according to incubation time As the system is able to maintain the pictures taken during the incubation, it allows to look at the microbial growth at different times of interest (before swarming of colonies, pigmentation ). The figure 4 illustrates the possibility to go back in time and to analyze at different stages of the growth on the media. Figure 4: A. brasiliensis growth observed by the EviSight Compact system at different incubation times on a contact plate Evaluation of the EviSight Compact System The evaluation was performed at four different locations in pharmaceutical industries. The objective was to assess the performance in term of detection and enumeration of three main applications (90mm and contact plates direct reading and filters reading) and compare the results obtained with the traditional human eye reading. Accuracy evaluation The enumeration results from both EviSight Compact and manual reading were compared using a boxplot graphical representation. This allows to visualize characteristics of the distribution of a quantitative variable.
Comparison of EviSight Compact enumeration and manual plate counting The figure 5 shows the linear regression obtained for each application using all data results from all concentrations together. The equation of the linear regression is displayed on the graph with the associated R 2. Figure 4: Boxplots of enumerations obtained for EviSight Compact (left side of the pair) and manual counting (right side of the pair) Both EviSight Compact and manual count enumerations are comparable. Precision The precision was estimated for each application, at each concentration level and expressed in terms of variance, standard deviation and %CV. Table 3: Precision estimates for EviSight and for manual reading Statistical comparison of repeatability variances obtained with both techniques was performed using the Fisher test (alpha=5%) by application for each concentration level. Figure 5: Linear regression of EviSight Compact enumeration results versus mean of manual plate counts for each application Results are very satisfactory : - Intercept estimates are very close to 0, - Slope estimates are all comprised within the acceptance range of 0.8-1.2, - R 2 values are all higher than the acceptance criteria of 0.9. The following Bland-Altman plot figures show for each strain and media the difference in enumerations results between EviSight Compact and manual reading (EviSight manual reading) versus the manual reading counts. Also displayed on the graphs : - the plain black horizontal line corresponds to the mean difference, - the dashed black horizontal lines correspond to the agreement limits, - the grey horizontal line corresponds to the null difference. Table 4: Results of statistical comparison of EviSight Compact and manual reading repeatability variances All repeatability variances were shown to be statistically comparable between EviSight Compact and manual plate reading. Figure 6: Brand-Altman plot for 90 mm plates applications (per microorganism and culture media)
but even effective, this method can introduce a natural variability between individuals. To counter this variability, a cumbersome four eye verification may be used but, in some cases of swarming, provide a result is sometimes impossible. The automated kinetic reader EviSight Compact is a solution to obtain traceable, reproducible and standardized enumerations of colonies on regular Petri dishes with the same level of accuracy than visual reading. References 1. PDA Technical Report 33 Figure 7: Brand-Altman plot for contact plates applications (per microorganism and culture media) Figure 8: Brand-Altman plot for filtration applications (per microorganism and culture media) All the figures show no statistical differences of enumeration between both methods. However, on the figure 7, a higher variability seems to appears on SDA with A. brasiliensis. Aspergillus is a mold, and grown on a specific media such as Sabouraud Dextrose Agar, this microorganism exhibits a morphology and a sporulation challenges the enumeration. This is even more true when the plate is a contact plate with a small surface of agar. In such a case the go back in time feature of the system allows to look at the mold development at an early stage and provide an accurate enumeration (Fig. 4). Conclusion Microbiological results are well known to be highly variable. In particular, microorganisms display an infinite number of growth behaviors (speed, colony size, shape and color etc.) challenging the enumeration. Operators visual microbiological analysis is still the gold standard,