NF VALIDATION Validation of alternative analytical methods Application in food microbiology. Summary report

Transcription

1 ACCREDITATION N PORTEE DISPONIBLE SUR 3M Boulevard de l Oise F CERGY PONTOISE NF VALIDATION Validation of alternative analytical methods Application in food microbiology Summary report EN ISO study validation renewal of 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) Plate Quantitative method This document includes 43 pages, with 5 appendixes. Only copies including the totality of this document are authorized. Competences of the laboratory are certified by COFRAC accreditation for the analyses marked with symbol. Version 0 December 3, 2014 ADRIA DEVELOPPEMENT Creac h Gwen - F QUIMPER Cedex - Tél. (33) Fax (33) adria.developpement@adria.tm.fr - Site web : ASSOCIATION LOI DE N SIRET N EXISTENCE N TVA FR

2 Summary 1 INTRODUCTION Dates of the validation studies Alternative method Reference method 4 2 INITIAL VALIDATION AND RENEWAL STUDY RESULTS Method comparison study Linearity Relative accuracy Detection limit (LOD) and quantification limit (LOQ) Relative sensitivity Specificity/Selectivity Practicability Inter-laboratory study Study organisation Control of experimental parameters Logistic conditions Results Statistical interpretations and calculations Conclusion 33 Appendix 1 Flow diagrams of the reference methods 34 Appendix 2 Linearity: raw data 35 Appendix 3 Specificity / Selectivity 37 Appendix 4 Mandel s graphics 39 Appendix 5 Statistical results 41 ADRIA Développement 2/43 December 3, 2014

3 Before comment Quality assurance documents related to this study can be consulted upon request from 3M. The technical protocol and the result interpretation were realized according to the EN ISO and the AFNOR technical rules. Company : 3M Boulevard de l Oise F CERGY PONTOISE cedex Expert Laboratory : ADRIA Développement ZA Creac h Gwen F QUIMPER Cedex Studied method : 3M TM Petrifilm TM High Sensivity Coliform Count (HSCC) plate Validation standard: ISO (October 2003): Food microbiology Protocol for the validation of alternative methods Reference methods : ISO 4831 (August 2006): Horizontal method for the detection and enumeration of coliforms. Most probable number technique ISO 4832 (February 2006): Horizontal method for the enumeration of coliforms - Colony Count Technique Scope: All human food products Certification organism: AFNOR Certification Analyses performed according to the COFRAC accreditation ADRIA Développement 3/43 December 3, 2014

4 1 INTRODUCTION 1.1 Dates of the validation studies The 3M TM Petrifilm TM High Sensivity Coliform Count (HSCC) plate was validated in March 2003 (certificate number 3M 01/07 03/99), for all human foods products. Renewals were obtained in 2003, 2007, 2011 and Alternative method The 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plate is a ready to use culture medium system, which contains modified Violet Red Bile (VRB) nutrients, a cold-water-soluble gelling agent, and a tetrazolium indicator that facilitates colony enumeration. Petrifilm HSCC plates are useful for the enumeration of low levels of coliform bacteria in the food industries. A 5-mL sample is plated onto one plate. This allows a 1-coliform-per-gram sensitivity if a 1:5 dilution of product is made. A 2-coliform-per-gram sensitivity is achieved if a 1:10 dilution is used. The protocol is given below: Sampling and decimal dilution in peptone-salt 5 ml per Petrifilm test (1 Petrifilm test / dilution) Incubation at 30 C 1 C, 24 h 2 h Characteristic (red and gas producers) colonies enumeration (maximum 150 colonies / test) 1.3 Reference method The standard methods (See Appendix 1) are: - ISO 4831 (August 2006): Horizontal method for the detection and enumeration of coliforms. Most probable number technique - ISO 4832 (February 2006): Horizontal method for the enumeration of coliforms - Colony Count Technique ADRIA Développement 4/43 December 3, 2014

5 2 INITIAL VALIDATION AND RENEWAL STUDY RESULTS 2.1 Method comparison study Linearity Linearity is the ability of the method when used with a given matrix to give results that are in proportion to the amount of analyte present in the sample, that is an increase in analyte corresponds to a linear or proportional increase in results. The reference standard for the gas-producing coliforms enumeration is the ISO 4831, which is based on the MPN (Most Probable Number) principle. The MPN based enumeration results are determined within confidence intervals. 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plate technique is based on colonies enumeration principle. Only artificially contaminated samples were analysed during the relative linearity studies. Since it is more relevant to compare methods that are based on the same principle, the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) was compared to the ISO 4832 standard which is also based on colonies enumeration. Note that samples were also analysed with the ISO Food matrices and protocols Five (matrix/strain) pairs were tested. Five contamination levels were evaluated and two replicates were analyzed by tested sample. 50 analyses were performed by the alternative method and the both ISO methods, i.e. ISO 4831 and ISO The contamination levels, the tested samples and the inoculated strains are presented in the table below: Category Strain Level (CFU/g) Minced beef Enterobacter cloacae Vegetable Escherichia coli Fish Escherichia coli ad Milk Enterobacter sakazakii Egg product Klebsiella pneumoniae Results (See Appendix 2) Samples were tested twice by each method. The bi-dimensional graphs are shown on Figures 1 and 2. ADRIA Développement 5/43 December 3, 2014

6 log(alternative method) log(alternative method) log(alternative method) log(alternative method) log(alternative method) 3M Figure 1 - Linearity: bi-dimensional graphs ISO 4831 Minced beef (ISO 4831) Green pepper (ISO 4831) 7,00 7,00 5,00 5,00 3,00 3,00 1,00 1,00 1,00 3,00 5,00 7,00 log(reference method) 1,00 3,00 5,00 7,00 log(reference method) Cod filet (ISO 4831) Milk (ISO 4831) 7,00 7,00 5,00 5,00 3,00 3,00 1,00 1,00 1,00 3,00 5,00 7,00 log(reference method) 1,00 3,00 5,00 7,00 log(reference method) Egg product (ISO 4831) 7,00 5,00 3,00 1,00 1,00 3,00 5,00 7,00 log(reference method) ADRIA Développement 6/43 December 3, 2014

7 log(alternative method) log(alternative method) log(alternative method) log(alternative method) log(alternative method) 3M Figure 2 - Linearity: bi-dimensional graphs ISO 4832 Minced beef (ISO 4832) Green pepper (ISO 4832) 7,00 7,00 5,00 5,00 3,00 3,00 1,00 1,00 1,00 3,00 5,00 7,00 log(reference method) 1,00 3,00 5,00 7,00 log(reference method) Cod filet (ISO 4832) Milk (ISO 4832) 7,00 7,00 5,00 5,00 3,00 3,00 1,00 1,00 1,00 3,00 5,00 7,00 log(reference method) 1,00 3,00 5,00 7,00 log(reference method) Egg product (ISO 4832) 7,00 5,00 3,00 1,00 1,00 3,00 5,00 7,00 log(reference method) ADRIA Développement 7/43 December 3, 2014

8 Statistical interpretations Statistical interpretation results are: Table 1 - Test 3M TM Petrifilm TM High-Sensitivity Coliform Count / ISO 4831 Matrix R Selected regression Rob.F Critical value P% Correlation coefficient Regression equation* Minced beef 0.24 OLS log Ref = log Alt Milk 0.10 OLS log Ref = log Alt Green pepper 0.07 OLS log Ref = log Alt Cod filet 0.18 OLS log Ref = log Alt Egg product 0.33 OLS log Ref = log Alt Table 2 - Test 3M TM Petrifilm TM High-Sensitivity Coliform Count / ISO 4832 Matrix R Selected regression Rob.F Critical value P% Correlation coefficient Regression equation* Minced beef 1.29 GMFR log Alt = log Ref Milk 0.29 OLS log Ref = log Alt Green pepper 1.40 GMFR log Alt = log Ref Cod filet 5.00 OLS log Alt = log Ref Egg product 0.40 OLS log Ref = log Alt * x-axis and y-axis choice depends on the selected regression. Statistical interpretation: P > 5 % : not significant 1 % < P < 5 % : significant 0,1 % < P < 1 % : very significant P < 1 % :highly significant The regression straight lines are given figures 3 and 4. ADRIA Développement 8/43 December 3, 2014

9 Référence Reference Référence Reference Reference 3M Figure 3 - Linearity: Regression straight lines ISO 4831 Minced beef - Regression OLS 2 (ISO 4831) Green pepper - Regression OLS 2 (ISO 4831) 6 5 y = 0,8603x + 0, y = 1,0753x - 0, Alternative Alternative Cod fish - Regression OLS 2 (ISO 4831) Milk- Regression OLS 2 (ISO 4831) 6 5 y = 0,9184x + 0, y = 0,9508x + 0, Alternative Alternative Egg product - Regression OLS 2 (ISO 4831) 6 5 y = 0,9779x - 0, Alternative ADRIA Développement 9/43 December 3, 2014

10 Reference Alternative Reference Alternative Alternative 3M Figure 4 - Linearity: Regression straight lines - ISO 4832 Minced beef- Regression GMFR (ISO 4832) Green pepper - Regression GMFR (ISO 4832) y = 0,9368x + 0, Reference Cod fish - Regression OLS 1 (ISO 4832) y = 0,9382x + 0, Reference Milk - Regression OLS 2 (ISO 4832) 6 5 y = 1,0038x + 0, Reference 6 5 y = 0,9389x - 0, Alternative Egg product - Regression OLS 2 (ISO 4832) 6 5 y = 1,0698x - 0, Alternative ADRIA Développement 10/43 December 3, 2014

11 Table 3 - Repeatability limits Matrix ISO 4831 Standard method ISO 4832 Standard method Alternative method Test 3M TM Petrifilm TM High- Sensitivity Coliform Count Milk Minced beef Green pepper Cod filet Egg product M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates and ISO 4832 repeatability limits are similar, and highly better than the ISO 4831 performances. Conclusion 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates and ISO 4832 comparison shows satisfactory results, all the correlation coefficients are higher than 0,97. The ISO 4831 is characterised by a wide range results dispersion, with repeatability limits varying between 0,176 and 2, Relative accuracy The accuracy is the closeness of agreement between a test result and the accepted reference value. The bias is the difference between the expectation of the test results and an accepted reference value. The ISO calculations were done with the results obtained in 1998 and Since no replicate analysis was run with the ISO 4831 method, a GMFR regression was used. New samples from egg and meat products categories were additionally analysed in duplicata by each method in 2007, i.e. the alternative method, the ISO 4831 and ISO 4832 standards, in order to fulfil with the ISO and the AFNOR technical rules. The analysed categories and types of analysed food are presented table 4. ADRIA Développement 11/43 December 3, 2014

12 Table 4 - Number and nature of the samples Categories Meat products Types Raw meat, raw and cooked delicatessen, RTE foods Analysed samples number Interpreted results number Dairy products Raw milks, cheeses, milk powders Seafood Frozen, salads, RTE foods Egg products Egg product, pastries, RTE foods Vegetables and others Raw, frozen, spices, chocolate, flour All products Artificial contamination samples (2007) 21 samples were artificially contaminated with injured cells. Injury treatments and injury efficiencies, i.e. enumerations differences between selective and non selective media, are presented hereafter (See table 5): Table 5 - Stress Reference Strain Origin Treatment delta log Inoculated sample number E.coli 1 Toulouse sausage HT 30 min 55 C E.coli 12 Turkey meat HT 30 min 55 C E.coli 142 Egg product HT 30 min 55 C Enterobacter cloacae 58 Food -20 C Serratia liquefaciens 5 Poultry meat -20 C E.coli 143 Mayonnaise HT 30 min 55 C E.coli Ad222 Egg product -20 C Hafnia alvei 168 Poultry meat -20 C Hafnia alvei 168 Poultry meat -20 C, + 4 C HT: Heat treatment 15 results were kept for interpretation. ADRIA Développement 12/43 December 3, 2014

13 Raw results Table 6 Food categories Contamination range (log) Meat products 0.56 to 5.16 Dairy products 0.60 to 3.63 Seafood to 4.38 Egg products to 4.91 Vegetables and others 0.18 to 3.87 The bi-dimensional graphs including ISO 4831 confidence intervals are presented Figure 5. ADRIA Développement 13/43 December 3, 2014

14 ISO 4831 ISO 4831 ISO M Figure 5 - Relative accuracy: bi-dimensional graphs PHS- Meat products 8,00 8,00 - PHS PHS- Dairy products 8,00 8,00 PHS PHS- Seafood 8,00 8,00 - PHS ADRIA Développement 14/43 December 3, 2014

15 Log(Reference method) ISO 4831 ISO M PHS- Egg products 8,00 8,00 - PHS PHS- Vegetables 8,00 8,00 - PHS All products PHS 5,00 3,00 1,00 1,00 3,00 5,00-1,00 log(alternative method) ADRIA Développement 15/43 December 3, 2014

16 Interpretation Categories n Selected regression Table 7 a t(a) b t(b) Critical T Ordinate at 0 Meat products 32 GMFR Dairy products 13 GMFR Seafood 12 GMFR Egg products 22 GMFR Vegetables and others 25 GMFR All products 104 GMFR P% Slope at 1 Statistical Interpretation: P > 5 % : not significant 1 % < P < 5 % : significant P < 1 % : highly significant Table 8 Alternative method Reference method Categories Bias D repeatability limit repeatability limit Meat products / Dairy products / Seafood / Egg products / Vegetables and others / All products * * Calculation performed by using the data obtained in 2007, data related to meat and products analysis. Regression straight lines (graphics and equations) are presented Figure 6. ADRIA Développement 16/43 December 3, 2014

17 Alternative Alternative Alternative Alternative Alternative Alternative 3M Figure 6 - Relative accuracy: regression straight lines Meat Products - Regression GMFR Vegetables - Regression GMFR 5,00 y = 0,908x + 0,5558 5,00 y = 0,9414x + 0,3756 3,00 3,00 1,00 1,00 1,00 3,00 5,00 Reference 1,00 3,00 5,00 Reference Seafood - Regression GMFR Egg products - Regression GMFR 5,00 y = 1,0037x + 0,1652 5,00 y = 1,0225x + 0,1497 3,00 3,00 1,00 1,00 3,00 5,00 Reference 1,00 1,00 3,00 5,00 Reference Dairy products - Regression GMFR Tous produits - Régression GMFR All products - Regression GMFR 5,00 y = 0,937x + 0,1042 5,00 y = 0,9746x + 0,2913 3,00 1,00 1,00 3,00 5,00 Reference 3,00 1,00 1,00 3,00 5,00 Reference Référence ADRIA Développement 17/43 December 3, 2014

18 Discussion For all analysed products, the bias between both methods is equal to 0,255 log CFU/g, the alternative method repeatability limit is equal to 0,323 log CFU/g, and the regression equation is: log (alternative method) = 0,975 log (standard method) + 0,291 85% of 3M TM Petrifilm TM plates results are close to or integrated into the ISO 4831 confidence intervals. Some exceptions are observed for the samples mentioned hereafter. The ISO 4832 standard enumeration results are also given. When the data are available, the API identification results of colonies isolated from the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) and the VRBL medium are provided below: Meat products - sample 40, frozen turkey meat: o 3M TM Petrifilm TM High-Sensitivity plates results: 3.11 and 3.02 log (CUF/g) o ISO 4831 results: 1: 3.97 [ ] log (CFU/g) o ISO 4832 results: 2.68 and 2.80 log (CFU/g) - sample 41, frozen turkey meat: o 3M TM Petrifilm TM High-Sensitivity plates enumerations: 2.44 and 2.66 log (CFU/g) o ISO 4831 results: 3.38 [ ] log (CFU/g) o ISO 4832 results: 2.37 and 2.32 log (CFU/g) - sample 7, Pizza with goat cheese and bacon: o 3M TM Petrifilm TM High-Sensitivity plates results: 1.34 and 1.51 log (CFU/g) o ISO 4831 results: 0.56 [ ] log (CFU/g) o ISO 4832 results: 1.60 and 2.11 log (CFU/g) - sample 12, Beef Burgundy o 3M TM Petrifilm TM High-Sensitivity plates results: 1.71 and 1.64 log (CFU/g) o ISO 4831 results: 0.87 [ ] log (CFU/g) o ISO 4832: results 1.60 and 1.00 log (CFU/g) o 6 isolates from the 3M TM Petrifilm TM High-Sensitivity plates were identified: Rhanella aquatilis: 2 isolates Enterobacter agglomerans: 4 isolates o 2 isolates from the VRBL medium were identified: Enterobacter agglomerans: 2 isolates ADRIA Développement 18/43 December 3, 2014

19 - sample 13, Beef tongue with Madeira wine sauce o 3M TM Petrifilm TM High-Sensitivity plates results: 1.60 and 1.49 log (CFU/g) o ISO 4831 results: 0.56 [ ] log (CFU/g) o ISO 4832 results: 1.70 et 1.30 log (CFU/g) - sample 31, Pizza with ham and mushrooms o 3M TM Petrifilm TM High-Sensitivity plates results: 3.15 and 3.13 log (CFU/g) o ISO 4831 results: 2.18 [ ] log (CFU/g) o ISO 4832 results: 2.63 and 2.65 log (CFU/g) o 10 isolates from the 3M TM Petrifilm TM High-Sensitivity plates were identified: Serratia liquefaciens: 4 isolates Enterobacter agglomerans: 2 isolates Hafnia alvei: 3 isolates o 5 isolates from the VRBL medium were identified: Serratia liquefaciens: 1 isolat Enterobacter agglomerans: 4 isolates Hafnia alvei: 1 isolat - sample 210, Pâté : o 3M TM Petrifilm TM High-Sensitivity plates results: 2.30 and 1.15 log (CFU/g) o ISO 4831 results: 1.36 [ ] log (CFU/g) 2.15 [ ] log (CFU/g) o ISO 4832 results: 1.26 and 1.51 log (CFU/g) - sample 299, Smoked ham: o 3M TM Petrifilm TM High-Sensitivity plates results: 3.15 and 3.15 log (CFU/g) o ISO 4831 results: 0.95 [ ] log (CFU/g) 0.60 [ ] log (CFU/g) o ISO 4832 results: 2.95 and 2.86 log (CFU/g) Dairy products - sample 93, semi-hard cheese (Pyrénées) o 3M TM Petrifilm TM High-Sensitivity plates results: 2.72 and 2.67 log (CFU/g) o ISO 4831 results: 1.96 [ ] log (CFU/g) o ISO 4832 results: 2.26 and 2.15 log (CFU/g) Seafood - sample 55, Raw frozen coaldfish filet o 3M TM Petrifilm TM High-Sensitivity plates results: 1.08 and 1.08 log (CFU/g) o ISO 4831 results: 0.32 [ ] log (CFU/g) o ISO 4832 results: <1 and <1 log (CFU/g) ADRIA Développement 19/43 December 3, 2014

20 Egg products - sample 24, Pastry (Religieuse): o o o 3M TM Petrifilm TM High-Sensitivity plates results: 1.60 and 1.69 log (CFU/g) ISO 4831 results: 0.46 [ ] log (CFU/g) ISO 4832 results: <1 and 1.30 log (CFU/g) - sample 103, Pastry (Religieuse): o o o 3M TM Petrifilm TM High-Sensitivity plates results: 1.78 and 1.45 log (CFU/g) ISO 4831 results: [ ] log (CFU/g) ISO 4832 results: <1 and 1.78 log (CFU/g) - sample 144, Pastry (Eclair): o o o 3M TM Petrifilm TM High-Sensitivity plates results: 4.35 and 4.36 log (CFU/g) ISO 4831 results: 3.18 [ ] log (CFU/g) ISO 4832 results: 3.98 and 4.07 log (CFU/g) - sample 213, Omelette: o 3M TM Petrifilm TM High-Sensitivity plates results: 2.38 and 2.20 log (CFU/g) o ISO 4831 results: 1.97 [ ] log (CFU/g) 1.36 [ ] log (CFU/g) o 4832 results: 1.18 and < 1 log (CFU/g) - sample 377, Pastry (Mille-feuilles): o 3M TM Petrifilm TM High-Sensitivity plates results: 0.30 and 0.78 log (CFU/g) o ISO 4831 results: 1.36 [ ] log (CFU/g) 0.60 [ ] log (CFU/g) o ISO 4832 results: 1.18 and 1.00 log (CFU/g) Vegetables and miscellaneous - sample 77, Frozen mixed vegetables: o o o 3M TM Petrifilm TM High-Sensitivity plates results: 1.85 and 1.69 log (CFU/g) ISO 4831 method results: 0.63 [ ] log (CFU/g) ISO 4832 results: <1 and <1 log (CFU/g) o 5 isolates from 3M TM Petrifilm TM High-Sensitivity plates results were identified to Enterobacter sakazakii species The observed repeatability and specificity clearly confirm the performances the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates, while the ISO 4831 repeatability limits enhance a result dispersion. Note as well that the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) and the ISO 4832 method provide equivalent enumeration results, confirming that the observed characteristic colonies with the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates correspond to coliforms. All the colonies identified with biochemical galleries were classified in the coliform group. Finally, note that the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates favour the recovery of stressed cells, in comparison to the ISO 4832 standard. According to these results, the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates show satisfactory relative accuracy. ADRIA Développement 20/43 December 3, 2014

21 2.1.3 Detection limit (LOD) and quantification limit (LOQ) The critical level is the defined as the smallest amount which can be detected (not null), but not quantified as an exact value. Below this value, it cannot be sure that the true value is not null. The detection limit is defined as being higher than the critical level because it involves a power, the probability 1-, which has to be well over 50 %, for example 95 %. The quantification limit is defined as the smallest amount of analyte (that is the lowest actual number of organisms) which can be measured and quantified with defined precision and accuracy under the experimental conditions by the method under validation. Protocol The detection limit of the alternative method was done with pure cultures. Three different levels of inoculation were tested, with six replicates per level, i.e. a total of 18 analyses by the alternative method. One E. coli strain was used for the limit of detection of the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates. Quantification limit was calculated for six independent blank samples. Results Data are intrinsic to the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates and are presented in the following tables: Table 9 Level Positive samples nb Standard deviation So Bias 0 0/6 / / 0,25 4/ ,5 6/ / / Formulas Obtained Values (CFU/ml) LC 1.65 So + Xo 0.8 LOD 3.3 So + Xo 1.6 LOQ 10 So + Xo 4.9 ADRIA Développement 21/43 December 3, 2014

22 CV(<x(Y)>) CV(<X(Y)>) 3M Relative sensitivity The relative sensitivity is defined as the ability of the alternative method to detect two different amounts of analyte measured by the reference method within a given matrix, at a specified average value, or over the whole measurement range; that is, it is the minimal quantity variation (increase of the analyte concentration x) which gives a significant variation of the measured signal (response y). Data are intrinsic to the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates and correspond to the relative linearity study results interpretation. The accuracy patterns are presented in figure 7. Figure 7 - Accuracy patterns PHS / ISO ,0% 1800,0% 1600,0% 1400,0% 1200,0% 1000,0% 800,0% 600,0% 400,0% 200,0% 0,0% Reference method CFU/g (X) Milk Egg product Cod filet Green pepper Minced beef PHS / ISO ,0% 50,0% 40,0% 30,0% 20,0% 10,0% 0,0% Méthode de référence ufc/g (X) Milk Egg product Cod filet Green pepper Minced beef ADRIA Développement 22/43 December 3, 2014

23 2.1.5 Specificity/Selectivity The specificity is defined as the degree to which a method is affected (or not) by the other components present in a multi-component sample. That is the ability of a method to measure exactly a given analyte, or its amount, within the sample without interference from non-target components such as a matrix effect, or background noise. The selectivity is defined as a measure of the degree of non-interference in the presence of non-target analytes. A method is selective if it can be used to detect the analyte under examination, and that a guarantee can be provided that the detected signal can only be a product by that specific analyte. 33 positive strains and 31 negative strains were tested according to the ISO standard requirement. The raw data are given in Appendix 3. Similar results were observed with the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates and the ISO 4831 method, except for one target strain (Enterobacter cloacae 48), which showed a gas producer phenotype on 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates, while a negative result was observed in BLBVB media. This strain was used to inoculate three samples during the relative accuracy study; for the three inoculated samples, all the colonies observed on 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates were gas producers, while only one sample was positive by the ISO 4831 standard. ADRIA Développement 23/43 December 3, 2014

24 2.2 Practicability Time to results is decreased with the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates, in comparison to the ISO 4831 standard: - 24 h 2 h are required by the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates for positive and negative results, - 24 to 48 h are required by the ISO standard for negative results, - 96 h are required by the ISO standard for positive results. The use of 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates method reduces the handling time. The workflow of the both techniques according to the number of samples is presented Table 10. Table 10 Manipulation time 15 samples Steps Test 3M TM Petrifilm TM High-Sensitivity ISO 4831 Sampling 26 min 26 min Dilution and stomach 23 min 23 min Analysis 30 min 225 min Sub-culturing / 75 min Reading 25 min 15 min Total 104 min 364 min Time/sample 6.9 min 24.3 min No specific maintenance or traceability procedures are required. The volume of required materials and wastes is significantly reduced with the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC), compared to the NF ISO 4831 standard. ADRIA Développement 24/43 December 3, 2014

25 2.3 Inter-laboratory study Since it is more relevant to compare methods that are based on the same principle, the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) was compared to the ISO 4832 standard. Both methods are based on colonies enumeration, while it is not the case for the ISO Study organisation 12 laboratories were involved in this study. Detailed instructions were sent to the collaborators by the expert laboratory. Pasteurised half-skimmed milk was inoculated with Escherichia coli 94, isolated from dairy product. Inoculation levels targeted were: - 0 CFU/ml, CFU/ml, CFU/ml, CFU/ml. Each laboratory received eight flasks of 25 ml milk, i.e. two flasks per inoculation level. One non-inoculated sample was added to the package for total viable count microflora (NF ISO 4833 method). Blind coded samples were placed in isothermal boxes, containing cooling blocks, and express-shipped to the different laboratories. A temperature control flask with a temperature probe was added to the package in order to register temperature profile during transport and at reception. Samples were shipped in 24 h to laboratories. Collaborative study laboratories and the expert laboratory carried out the analyses with the alternative and reference methods. ADRIA Développement 25/43 December 3, 2014

26 2.3.2 Control of experimental parameters Strain stability during transport In order to evaluate the Escherichia coli 94 strain stability during transport, bacterial count of all inoculated flasks were checked at different time, i.e. inoculation time, after 24 h and 48 h of storage at 2-8 C. Results are reported in Table 11. Table 11 - Escherichia coli 94 count with ISO 4832 method (CFU/ml) Repetition 1 Repetition 2 Repetition 3 Day Day Day No evolution of the strain was observed after 48 h storage at 2-8 C in the isothermal box. Results obtained with both methods The results obtained by the expert laboratory for Escherichia coli count for the both methods are given table 12. Table 12 - Expert laboratory results (in log CFU/g) Targeted rate ISO 4832 Alternative method (log CFU/g) Duplicate 1 Duplicate 2 Duplicate 1 Duplicate 2 < 1 < 1 < 1 < 0.2 < to to to Contamination levels targeted were reached. ADRIA Développement 26/43 December 3, 2014

27 2.3.3 Logistic conditions Sample temperature at receipt Measured temperatures on receipt are listed table 13 Table 13 - Sample temperature on receipt Laboratories Measured temperature Receipt samples: Temperature measured by on receipt ( C) day and hour the temperature probe ( C) A /04/07 10h B /04/07 09h C /04/07 08h D /04/07 10h E /04/07 09h F /04/07 10h G /04/07 11h H Empty flask 17/04/07 12h I /04/07 12h J /04/07 11h K /04/07 14h L /04/07 07h Sample temperature during delivery No problem was observed during transport, i.e. the measured temperatures were between -0.5 and 2.5 C Results 12 laboratories participated to the study. Since lab G has performed the tests incubation at 44 C, the results were discarded from the calculations. The results of 11 labs were thus used for the ISO calculations. The results are summarised in table 14. ADRIA Développement 27/43 December 3, 2014

28 Table 14 Summary results of the ISO 4832 method and 3M TM Petrifilm TM High-Sensitivity count plates (CFU/ml) Level 0 Level 1 Level 2 Level 3 Laboratories Standard Alternative Standard Alternative Standard Alternative Standard Alternative method method method method method method method method A <1 <1 <0.2 < B <1 <1 <0.2 < C <1 <1 <0.2 < D <1 <1 <0.2 < E <1 <1 <0.2 < F <1 <1 <0.2 < H <1 <1 <0.2 < I <1 <1 <0.2 < J <1 <1 <0.2 < K <1 <1 <0.2 < L <1 <1 <0.2 < Statistical interpretations and calculations The calculations were realised according to the amendment number 1 of the ISO standard (2011): Accuracy: closeness of agreement between a measurement result and the accepted reference value. Note: Accuracy refers to a combination of trueness and precision Trueness: closeness of agreement between the expectation of a measurement result and the accepted reference value. Note: the measure of trueness is usually expressed in terms of bias Precision: closeness of agreement between independent measurement results obtained under stipulated conditions. Note: quantitative measures of precision depend critically on the stipulated conditions. Repeatability conditions and reproducibility conditions are particular sets of extreme stipulated conditions Repeatability: precision under repeatability conditions ADRIA Développement 28/43 December 3, 2014

29 Repeatability conditions: measurement conditions where independent measurement results are obtained with the same method on identical measurement items in the same laboratory by the same operator using the same equipment within short interval of time. Repeatability standard deviation: standard deviation of measurement results obtained under repeatability conditions Repeatability limit (r): value less than or equal to which the absolute difference between two measurement results obtained under repeatability conditions is expected to be with a probability of 95% Reproducibility: precision under reproducibility conditions Reproducibility conditions: measurement conditions where measurement results are obtained with the same method on identical measurement items in different laboratories with different operators using different equipment Reproducibility standard deviation: standard deviation of measurement results obtained under reproducibility conditions Reproducibility limit (R): value less than or equal to which the absolute difference between two measurement results obtained under reproducibility conditions is expected to be with a probability of 95% ADRIA Développement 29/43 December 3, 2014

30 Scrutiny of the measurement results for consistency In order to identify other measurement results or laboratories that could be inconsistent, two graphical consistency techniques were realized: the robust Mandel s h-and k-statistics. These graphics are given in Appendix 4. All the statistical results are given in Appendix 5. The number of values above the indicator for h and k are given below: Mandel s values Number of observed values above the indicator Reference method Alternative method h > 1 % Lab F Level 1 / h > 5 % Lab F Level 1 Lab F Level 3 Lab J Level 2 Lab L Levels 1 and 2 k> 1 % / / k > 5 % Lab C Level 3 Lab E Level 3 Lab D Level 2 Lab F Level 3 Lab L Level 1 No inconsistent result was observed. All the results were used for the statistical analysis. Comparison of the trueness and precision characteristics of the reference method and alternative methods The statistical values are summarized hereafter: Reference method Alternative method Level Repeatability Reproducibility Repeatability Reproducibility Median Median standard deviation standard deviation standard deviation standard deviation ADRIA Développement 30/43 December 3, 2014

31 Bias of the alternative method In order to estimate the bias of the alternative method with respect to the reference method for each level, Dij and t are calculated as described below: Dij Y ij, Alt Y ij, Ref t median i ( Dij) / (2 p) Diff If t is larger than 2, the alternative method is significantly biased with respect to the reference method. The values are given in table 15. Table 15 Bias values obtained for each level Level Bias (D) t Interpretation Bias not significant Bias not significant Bias not significant All the calculated biases are not significant. Comparison of the repeatability standard deviations If the ratio Srj, Alt / Srj, Ref. of the repeatability standard deviations of the alternative method and the reference method is larger than 2, the precision under repeatability conditions of the alternative method is considered to be lower than that of the reference method. If this ratio is smaller than 0.5, the precision under repeatability conditions of the alternative method is considered to be greater than that of the reference method. ADRIA Développement 31/43 December 3, 2014

32 The ratio values are given in table 16. Table 16 Repeatability Reference method Alternative method Level Ratio Alternative / Repeatability Repeatability Repeatability Repeatability Reference method standard deviation standard deviation The repeatability performances of the compared methods are equivalent. Comparison of the reproducibility standard deviations If the ratio Srj, Alt / Srj, Ref. of the reproducibility standard deviations of the alternative method and the reference method is larger than 2, the precision under reproducibility conditions of the alternative method is considered to be lower than that of the reference method. If this ratio is smaller than 0,5, the precision under reproducibility conditions of the alternative method is considered to be greater than that of the reference method. The ratio values are given in table 17. Table 17 Reproducibility Level Reference method Reproducibility Reproducibility standard deviation Alternative method Reproducibility Reproducibility standard deviation Ratio Alternative / Reference methods The reproducibility performances of the compared methods are equivalent. ADRIA Développement 32/43 December 3, 2014

33 2.4 Conclusion The conclusions of the method comparison study are the following: 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates relative linearity, relative accuracy, inclusivity and exclusivity performances are satisfactory. The alternative method repeatability limit is highly better than ISO 4831 one. 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates offers important economic savings by: o standardizing the analysis o minimizing the training time, the volume of wastes and the manipulations. The workflow study demonstrates a reduction of the handling time and time to result with the 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates, in comparison to the ISO standard. The inter-laboratory was done by comparing the performances of plate count methods, i.e. 3M TM Petrifilm TM High Sensitivity Coliform Count (HSCC) plates and the ISO 4832 standard. The conclusions of the interlaboratory, study are the following: The biases between the compared methods show satisfactory results, as well as the alternative method repeatability and reproducibility performances. The 3M TM Petrifilm TM High-Sensitivity Coliform Count (HSCC) plates repeatability limits vary from to log CFU/g, reproducibility limits from to log CFU/g. ADRIA Développement 33/43 December 3, 2014

34 Appendix 1 Flow diagrams of the reference methods ISO 4831: Horizontal method for the detection and enumeration of coliforms. Most probable number technique Solid sample (1/10 dilution) Liquid sample 10 ml x 3 tubes Lauryl sulfate tryptose broth (LST) Double concentration Incubation 30 C 24 h 2 h 1 ml x 3 tubes Main suspension and decimal dilutions Lauryl sulfate tryptose broth (LST) Simple concentration Incubation at 30 C 24 h 2 h to 48 h 2 h Gas production confirmation Transfer 10 µl of positive tubes into a brilliant green lactose broth (BLBVB) Incubation at 30 C 1 ; 24 h 2 h or 48 h 2 h ISO 4832:Horizontal method for the enumeration of coliforms - Colony Count Technique Suspension and decimal dilutions in peptone-salt 1 ml per Petri dish (2 Petri dishes per dilution) Pour 15 ml VRBA Solidification Add a double layer of VRBA Incubation 24 h 2 h at 30 C 1 Count characteristic colonies (red 0.5 mm sometimes surrounded with red colorant) ADRIA Développement 34/43 December 3, 2014

35 N sample Product Dilution Replicate 1 Petrifilm PHS (gas producer) Replicate 2 cfu/g Replicate 1 cfu/g Replicate 2 log Replicate 1 Appendix 2 Linearity: raw data ISO 4832 Replicate1 Replicate2 cfu/g log Dilution Replicate Replicate2 plate a plate b plate a plate b 1 cfu/g Replicate 2 log cfu/g Replicate 1 log cfu/g Replicate 2 cfu/g Replicate 1 cfu/g Replicate 2 ISO ,67 2, ,36 2, ,38 1, ,79 2, ,60 2, ,04 2, Raw ground ,08 3, ,11 2, ,88 3,38 beef ,74 3, ,43 3, ,04 3, ,11 3, ,97 3, ,32 2, ,04 2, ,93 1,90 9,2 93 0,96 1, Ne ,53 2, ,45 2, ,46 2, Raw green ,90 3, pepper ,00 3, ,43 1, ,64 3, ,62 3, ,63 2, ,08 4, ,10 4, ,38 3, log cfu/g Replicate ,91 1, ,65 1, ,38 1, Ne Ne ,62 2, ,49 2, ,38 2, Raw fish fillet ,11 3, ,89 2, ,97 3, ,53 3, ,34 3, ,66 3, ,15 4, ,88 3, ,63 4, log cfu/g Replicate 2 Analyses performed according to the COFRAC accreditation ADRIA Développement 35/43 December 3, 2014

36 n Sample 93 Product Dilution Replicate Replicate 1 2 Petrifilm PHS (gas producer) cfu/g replicate 1 cfu/g replicate 2 log replicate1 log replicate2 Dilution Replicate1 ISO 4832 Replicate2 plate a plate b plate a plate b cfu/g replicate 1 cfu/g replicate 2 log cfu/g replicate1 log cfu/g replicate2 cfu/g replicate 1 cfu/g replicate 2 ISO 4831 log cfu/g replicate ,08 1, ,90 1, ,97 2, Ne ne ,43 2, ,26 2, ,38 2, Lait ,98 3, ,94 2, ,88 2, ,46 3, ,26 2, ,04 2, ,96 3, ,58 3, ,97 3, ,09 1, ,93 1, ,97 1, Ne ne ,41 2, ,30 2, ,38 1, Egg product ,89 2, ,98 2, ,36 2, ,28 3, ,28 3, ,32 3, ,88 3, ,90 3, ,63 3, log cfu/g replicate2 Analyses performed according to the COFRAC accreditation ADRIA Développement 36/43 December 3, 2014

37 Appendix 3 Specificity / Selectivity Data Positive strains Strains PCA PHS30 C(gas +) BLBVB log UFC/ml log UFC/ml % R a b Citrobacter diversus Adria < Citrobacter diversus cip < Citrobacter freundii Adria < Citrobacter freundii cip E. coli O157:H7 cip E. coli E. coli E. coli E. coli cip Enterobacter aerogenes cip Enterobacter aerogenes cip Enterobacter agglomerans Adria Enterobacter cloacae Adria Enterobacter cloacae Adria Enterobacter sakazakii Adria Enterobacter sakazakii Adria Hafnia alvei < Klebsiella oxytoca Adria Klebsiella oxytoca cip Klebsiella pneumoniae Adria Klebsiella pneumoniae cip Raw data Strain Origin Positive strains PCA (ufc/plate) ADRIA Développement 37/43 December 3, 2014 VRBL (ufc/plate) PHS (ufc/petrifilm) Citrobacter gillenii Ad343 Food (G) +/+ Enterobacter intermedius 60 Vegetable (G) +/+ Enterobacter kobei Ad342 Food (G) +/+ Citrobacter freundii 35 Vegetable (G) +/+ Enterbacter agglomerans 74 Cheese (G) +/+ Klebsiella oxytoca 42 Food (G) +/+ BLBVB (gas+/-) Enterobacter sakazakii 95 Fromage frais (G) +/+ (weak) Escherichia vulneris 11 Egg product (NG) +/+ (weak) Escherichia coli (G) +/+ Escherichia coli (G) +/+ Erwinia carotovora CIP82.83T Toulouse sausage (G) +/+ Enterobacter cloacae 48 Food (G weak) -/-

38 Data Negative strains Strains PCA PHS30 C(gas +) BLBVB log UFC/ml log UFC/ml % R b Aeromonas hydrophila cip Bacillus subtilis ATCC < Edwarsiella tarda cip < Enterococcus faecalis ATCC < Erwinia carotovora cip < Lactobacillus plantarum cip A < Proteus vulgaris Adria < Pseudomonas fluorescens cip < Salmonella enteritidis cip < Shigella flexneri cip < Staphylococcus aureus cip < Yersinia enterocolitica cip < Citrobacter diversus Adria < Hafnia alvei Adria < Data Negative strains Strain Origin PCA VRBL PHS (ufc/plate) (ufc/plate) (ufc/petrifilm) BLBVB(gas +/-) Aeromonas hydrophila CIP 7430 / /- Aeromonas hydrophila CIP 5750 / /- Citrobacter koseri 71 Frozen vegetables (NG) -/- Hafnia alvei A00C067 Cockerel (NG) -/- Proteus vulgaris 43 Ham (NG) -/- Providencia rettgeri 112 Egg product (NG) -/- Yersinia enterocolitica AC066 Cockerel (NG) -/- Citrobacter freundii 23 Toulouse sausage (NG) -/- Citrobacter freundii 53 Vegetables (NG) -/- Citrobacter diversus 100 Liver (NG) -/- Enterobacter amnigenus 126 Ground beef (NG) -/- Enterobacter fergusonii 2876 Environmental sample (NG) -/- Serratia marcescens Ad447 Raw milk (NG) -/- Escherichia hermanii (NG) -/- Escherichia hermanii (NG) -/- Citrobacter koseri cip (NG) -/- Serratia liquefaciens 5 Egg (NG) -/- ADRIA Développement 38/43 December 3, 2014

39 Mandel's robust h values Mandel's robust h values 3M Appendix 4 Mandel s graphics Laboratory h Ref level 1 h Ref level 2 h Ref level 3 h Alt level 1 h Alt level 2 h Alt level 3 h5% h1% h5% h1% A B C D E F H I J K L Reference method robust analysis 3,00 1,00-1, ,00 - A B C D E F H I J K L Laboratory h 1%=3.08 h 5%=2,04 h Ref level 1 h Ref level 2 h Ref level 3 Alternative method robust analysis 3,00 1,00-1, ,00 - A B C D E F H I J K L Laboratory h 1%=3.08 h 5%=2,04 h Alt level 1 h Alt level 2 h Alt level 3 ADRIA Développement 39/43 December 3, 2014

40 Mandel's robust k values Mandel's robust k values 3M Laboratory k Ref level 1 k Ref level 2 k Ref level 3 k5% K1% k Alt level 1 k Alt level 2 k Alt level 3 A B C D E F H I J K L Reference method robust analysis 3,50 3,00 2,50 1,50 1,00 0,50 A B C D E F H I J K L Laboratory k 1%=2,59 k 5%=1,82 k Ref level 1 k Ref level 2 k Ref level 3 Alternative method robust analysis 3,00 2,50 k 1%=2,59 1,50 1,00 0,50 A B C D E F H I J K L Laboratory k 5%=1,82 k Alt level 1 k Alt level 2 k Alt level 3 ADRIA Développement 40/43 December 3, 2014

41 Appendix 5 Statistical results Level 1 Laboratories number (p) 11 Laboratoires Level 1 Level 1 Mean si Deviation h values k values Laboratoires Ref Alt Mean si Deviation h values k values D A 1,978 2,041 2,010 0,05-0,032 0,032 0,564 1,039 A 1,863 1,924 1,894 0,04-0,030 0,030-0,043 0,473-0,116 B 1,978 1,892 1,935 0,06 0,043-0,043-0,324 1,398 B 1,949 2,079 2,014 0,09-0,065 0,065 0,759 1,007 0,079 C 1,881 1,919 1,900 0,03-0,019 0,019-0,740 0,625 C 1,806 1,968 1,887 0,11-0,081 0,081-0,086 1,259-0,013 D 1, ,962 0,05-0,038 0, ,236 D 1,881 1,763 1,822 0,08 0,059-0,059-0,521 0,911-0,140 E 0 2,041 2,021 0,03-0,021 0,021 0,696 0,676 E 1,949 1,851 1,900 0,07 0,049-0, ,761-0,120 F 1,792 1,839 1,816 0,03-0,023 0,023-1,743 0,758 F 1,690 1,748 1,719 0,04-0,029 0,029-1,206 0,450-0,096 H 2, ,021 0,03 0,021-0,021 0,696 0,676 H 1,978 1,949 1,964 0,02 0,014-0,014 0,421 0,220-0,057 I 2,079 2,041 2,060 0,03 0,019-0,019 1,167 0,617 I 1,602 1,826 1,714 0,16-0,112 0,112-1,240 1,738-0,346 J 1,851 1,863 1,857 0, ,247 0,197 J 2,204 2,079 2,142 0,09 0,062-0,062 1,606 0,969 0,284 K 1,929 1,987 1,958 0,04-0,029 0,029-0,048 0,936 K 1,959 1,929 1,944 0,02 0,015-0,015 0,292 0,230-0,014 L 2,114 1,987 2,050 0,09 0,064-0,064 1,049 2,076 L 2,342 2,255 2,299 0,06 0,044-0,044 2,653 0,676 0,248 m 1, m 1,90032 Accuracy median D -0,0571 Cn(2p) 2,089 Cn(2p) 2,089 Cp 1,9710 Cn(p) 1,9710 Cn(p) 1,9710 Qn(D) 0, n=2*p n 22 n=2*p n 22 Qdiff 0, f 12 f 12 t 0,93 l 60 l 60 n=p n 11 n=p n 11 f 6 f 6 l 15 l 15 Qn(2p) 0, Qn(2p) 0, Qn(p) 0, Qn(p) 0, Qwithin 0,03063 Qwithin 0,06446 Qbetween 0,08408 Qbetween 0,15024 Sr(repeatability standard deviation) 0,0433 Sr 0,0912 RSDr(variation coefficient of repeatability) 2,21% RSDr 4,80% r(repeatability limit)) 0,121 r 0,255 SL 2 (between laboratories standard deviation 613 SL 2 0,01842 SR(reproductibility standard deviation) 0,0895 SR 0,1635 RSDR(variation coefficient of reproductibility 0,046 RSDR 0,086 R(reproductibility limit)) 0, R 0, ADRIA Développement 41/43 December 3, 2014