The certification of a reference material for the evaluation of methods for the enumeration of Bacillus cereus


  • P. H. in ‘t Veld,

    1. Microbiological Laboratory for Health Protection, National Institute of Public Health and the Environment, Bilthoven, The Netherlands
    Search for more papers by this author
  • A. H. Havelaar,

    1. Microbiological Laboratory for Health Protection, National Institute of Public Health and the Environment, Bilthoven, The Netherlands
    Search for more papers by this author
  • N. G. W. M. Van Strijp-Lockefeer

    1. Microbiological Laboratory for Health Protection, National Institute of Public Health and the Environment, Bilthoven, The Netherlands
    Search for more papers by this author

Paul in ‘t Veld, Microbiological Laboratory for Health Protection, National Institute of Public Health and the Environment, PO Box 1, 3720 BA Bilthoven, The Netherlands (e-mail: nl).


A reference material containing Bacillus cereus was certified by the Community Bureau of Reference (BCR) for its number of colony-forming particles (cfp) in 0·1 ml reconstituted capsule solution. To this end, a batch of approximately 15 000 capsules was produced and tested for its homogeneity and stability. The variation in the number of cfp between capsules (homogeneity) was not found to be significantly different from a Poisson distribution. Stability was tested for extended periods at storage temperature (−20 °C) and at various higher temperatures up to 37 °C for 4 weeks to simulate transport conditions. Only at 37 °C did a small but significant decrease in the number of cfp occur. At −20 °C, no decrease in the number of cfp was observed over a period of about 4 years. For certification, 12 laboratories determined the number of cfp on two agars: Mannitol Egg-Yolk Polymyxin agar (MEYP, incubated at 30 °C) and Polymyxin pyruvate Egg-yolk Bromothymol blue Agar (PEMBA, incubated at 37 °C). The certified geometric mean value on MEYP after 24 h of incubation was 53·4 cfp 0·1 ml−1 of the reconstituted capsule solution (95% confidence interval 51·7–55·2) and on PEMBA, 55·0 (95% confidence interval 52·8–57·4). Based on these certified values, user tables were constructed specifying the 95% confidence limits when testing a smaller number of capsules, as would be done in individual laboratories. Based on the information on homogeneity, stability and the certification study, the BCR decided to certify the material as CRM 528.

Bacillus cereus is a spore-forming organism frequently found in various types of food. Normally, B. cereus is present at low levels but when high levels are present, it is capable of producing two types of toxins (entero- and emetic toxins) that can cause food poisoning. Several enterotoxins are currently known ( Beecher & Wong 1994; Agata et al. 1995 ; Granum et al. 1996 ) and one type of emetic toxin, namely cerelulide ( Agata et al. 1994 ). The best characterized enterotoxin is the HBL-complex, a toxin consisting of three units (B, L1 and L2) all of which are needed for activity. Enterotoxins are mostly produced in protein-rich foods, such as meat, and emetic toxins in starch-rich foods, such as rice and pasta ( Granum 1997). To be able to produce toxins at a level sufficient to cause food poisoning, B. cereus has to be present at high levels (>105 cells g−1). In The Netherlands, about 19% of all food-borne cases of disease with known aetiology are caused by B. cereus ( Notermans & van de Giessen 1993). Recently, the formation of enterotoxin by psychrotrophic strains of B. cereus in milk became a focus of interest ( Christiansson et al. 1989 ; Van Netten et al. 1990 ).

For the enumeration of B. cereus in foods, several selective media are available: MEYP agar as described in ISO 7932 ( Anonymous 1993), PEMBA agar ( Holbrook & Anderson 1980) or KG agar ( Kim & Goepfert 1971). All three media contain egg yolk which is needed to exhibit the lecithinase reaction, a characteristic feature of B. cereus.

To be able to rely on results obtained in different laboratories, quality assurance principles have to be applied by those laboratories. Reference Materials (RMs) and Certified RMs (CRMs) are useful tools in this respect. To be able to produce a (C)RM, the number of organisms present must be distributed homogeneously over the samples and must remain stable over a long period of time. Spores were used for the production of the CRM containing B. cereus as they easily survive adverse environmental conditions such as heat, pH and low water activity. The production process for this CRM is comparable with other CRMs developed by the National Institute of Public Health and the Environment, such as Salmonella ( in 't Veld et al. 1992 ) or Listeria monocytogenes ( in 't Veld et al. 1995 ). The difference between an RM and a CRM is the degree of accuracy of the (certified) values stating the number of organisms present in the material. In the ISO guide 35 ( Anonymous 1989), a CRM is defined as an RM that has been certified for its properties by means of a technically valid procedure and should be accompanied by a certificate issued by a certifying body. The certifying body in this respect is the Community Bureau of Reference (BCR) of the European Union.

Materials and methods

Preparation of the material

Bacillus cereus (ATCC 9139) was streaked for purity on sheep blood agar (SBA) and incubated for 24 h at 37 ± 1 °C. After incubation, a single colony was suspended in 10 ml peptone saline solution (PS) by mixing on a vortex mixer. This suspension was used to inoculate 10 plates (0·3 ml per plate) containing the Polymyxin pyruvate Egg yolk Mannitol Bromothymol blue Agar (PEMBA) of Holbrook & Anderson (1980) prepared from original ingredients. The PEMBA plates were used for rapid sporulation of B. cereus and were incubated at 37 ± 1 °C. After 24 h incubation, 5 ml PS were added to each plate and the cells suspended with the use of a sterile glass spreader. The suspension from each plate was pipetted into a single tube and heated for 10 min in a water-bath at 80 ± 0·5 °C to inactivate vegetative cells.

The heated suspension was added to 3 l sterilized milk (evaporated to a dry mass concentration of 240 g l−1, dry fat mass concentration 40 g l−1), which was then spray-dried using a Niro mobil minor spray dryer (Niro Atomizer, Soeberg, Denmark) operated at an inlet temperature of about 190 °C and an outlet temperature of about 70 °C. The highly contaminated milk powder (HCMP) thus obtained was sealed in a polyethylene bag (0·2 mm thick) and stored at 5 °C until required.

To obtain the desired number of colony-forming particles (cfp) in 0·1 ml of the reconstituted capsule solution, 1 g HCMP was mixed with 7 kg sterile skim milk powder. The milk powder used for mixing was a commercially available skim milk powder sterilized by gamma irradiation with a dose of 10 kGy. The mixing of HCMP with skim milk powder was carried out in 13 steps. For each step, equal amounts of contaminated milk powder were mixed with sterile milk powder, except for the last step in which approximately 4 kg contaminated milk powder were mixed with approximately 3 kg sterile milk powder. The first eight steps were done using a mortar and pestle. Each of these steps consisted of mixing the powder for 15–20 s using the pestle, followed by remodelling of the powder using a paper card; this procedure was repeated three times. The remaining five steps were carried out in a 17 l stainless steel drum with a Turbula type T 10 b mixing apparatus (Bachofen, Basel, Switzerland) for 1 h for each step.

Gelatine capsules were filled with the mixed powder using an aluminium filling apparatus in a laminar air flow cabinet at 19 g powder for 60 capsules (0·317 g for each capsule). Before filling, the empty capsules were sterilized by gamma irradiation with a dose of 10 kGy. The first stage was to fill two sets of 60 capsules with powder obtained from different places in the mixed powder, in order to determine the level of contamination and homogeneity of the powder. Ten capsules from each of these two sets of 60 capsules were analysed in duplicate on MEYP and SBA. Before use, the capsules were reconstituted as follows:

i) filling of test tubes (diameter 26 mm) with 10 ± 0·2 ml peptone saline solution;

ii) pre-warming of the tubes in a water-bath maintained at 38·5 ± 0·5 °C for 30 min;

iii) addition of the capsules to the test tubes and mixing on a Vortex mixer for a few seconds 10, 20 and 30 min after the addition of the capsule, leaving the tubes outside the water-bath for as short a time as possible;

iv) after the last mixing, transferring the tubes from the water-bath to iced water;

v) use of the dissolved capsule solution within 2 h.The plates were incubated at 30 °C for 24 h.

After a first check to determine whether or not the results met the set criteria, the entire batch of mixed powder was placed into gelatine capsules. Two capsules were taken from each filling of 60 capsules and used to determine the variation in mass of the capsules (including the capsule itself). The variation in mass of the empty capsules is negligible in comparison with the filled capsules. From these capsules, a further 20 capsules was selected at random to determine the level of contamination of the final batch on MEYP and SBA as described above.

Homogeneity studies

The variation in number of cfp between replicates from one reconstituted capsule (T1) and between replicates from different reconstituted capsules of a single batch (T2) were tested separately ( Heisterkamp et al. 1993 ). The formulae for the T1 and T2 test were published earlier by in 't Veld, et al. 1993 ). In the case of a Poisson distribution, T1 and T2 follow a χ2-distribution with, respectively, I . (J−1) and (I−1) degrees of freedom (I is the number of capsules, J is the number of replicates). In this case, the expected values of T1 and T2 are the same as the number of degrees of freedom. Hence, T1/{I . (J−1)} and T2/(I−1) are expected to be equal to one. For the variation between replicates of different capsules of a single batch, the Poisson distribution is theoretically the smallest possible variation which could be achieved. However, over-dispersion between capsules is expected and T2/(I−1) will usually be higher than 1 ( Heisterkamp et al. 1993 ). The homogeneity of the batch was determined based on the results of the stability test at higher temperatures. Only the results from the capsules stored at –20 °C were used. The five T2 values for each medium at different points in time were summed to obtain the overall dispersion test statistic (Thom) for SBA, MEYP and PEMBA. The number of degrees of freedom of Thom, called N, is the sum of the number of capsules examined minus the number of homogeneity tests carried out (so N = 25–5). The value of Thom/N is expected to be equal to one.

Stability studies

Two types of stability tests were done, a stability test at storage temperature (−20 °C) and a stability test at higher (transport) temperatures simulating transport conditions. For testing the stability of the materials stored at −20 °C, 10 capsules were examined in duplicate at regular time intervals on MEYP and SBA. The stability of the material stored at higher temperatures was determined at four different temperatures. The temperatures tested were −20 (reference), 22, 30 and 37 °C. Once a week, over a period of 4 weeks, five capsules from each storage temperature were examined in duplicate on MEYP and SBA. Analyses on PEMBA were also made in parallel with the stability test at higher temperatures, but only for the capsules stored at −20 °C. The results were used to determine the overall measure of dispersion for PEMBA (see homogeneity studies). The counts obtained for each storage temperature were log10-transformed and analysed using linear regression.

Certification study

For the study, 12 laboratories were selected, all experienced with the enumeration of B. cereus and with the reconstitution procedure for the RMs. Each participating laboratory received, by courier service, a parcel containing two series of eight numbered capsules. The capsules were stored at −20 °C upon receipt. The study was carried out according to the fixed time schedule laid down in the analytical protocol. After reconstitution of each capsule, the number of B. cereus cfp in 0·1 ml capsule solution was determined for each capsule in duplicate on two agars (MEYP and PEMBA) and a third optional agar (Sheep Blood Agar, SBA). The method for inoculation and incubation of the plates was described in a Standard Operating Procedure. The procedure for incubation and confirmation of B. cereus using MEYP was based on ISO 7932 ( Anonymous 1993) and that for PEMBA, on method no. L 00.00–25 ( Anonymous 1992) of the German Federal Food Law with confirmation according to ISO 7932. Laboratories incubated SBA according to their own procedures.

The MEYP plates were incubated for 48 h at 30 °C and the number of colonies counted after 24 and 48 h. The PEMBA plates were incubated for 48 h at 37 °C and counted after 24 and 48 h. The plates were re-labelled with random numbers before the first count (after 24 h incubation) was made. The colonies were marked on the base of the plates and after 48 h incubation, only the additional colonies were counted and reported (the same random numbers were used for counting after both 24 and 48 h incubation). Each laboratory selected at random a number of colonies from each medium for confirmation. Confirmation tests were done as described in the ISO 7932.

The results of the analyses were reported on a test reporting form. Each laboratory was free to analyse the capsules in one or two series as long as the time for inoculation of all dishes in a series was not longer than 2 h. Each series of capsules contained one (blind) control sample consisting of a gelatine capsule filled with sterile milk powder.

The relationship between reconstitution time and the number of cfp 0·1 ml−1 sample on MEYP and PEMBA was tested, in addition to the certification study, by the organizing laboratory. For this purpose, 10 capsules were examined in two series of five capsules on separate days using separate batches of media. For each capsule, portions of 1 ml were taken after 30, 40, 50, 60, 75 and 90 min of reconstitution. The portions were put directly into iced water before inoculation of the plates. The incubation conditions were identical to those used for the certification study. The plates were counted after 24 h incubation.

Statistical analysis

The results from the blank capsules were excluded from the statistical analysis. The statistical analysis performed on the data is described in detail in a separate report ( Heisterkamp et al. 1993 ). A summary of the analysis carried out is presented below.

First, the variation between the duplicate counts was calculated per laboratory, per medium and per incubation period by means of the T1-test. It was expected that replicate counts would follow a Poisson distribution, in which case the T1 test statistic would follow a χ2-distribution with I (= the number of capsules examined) degrees of freedom (α= 0·05). If the T1 result of a laboratory was significantly different from a χ2-distribution, the T1 results of duplicates per capsule were considered further. Problems related to single capsules were assumed if the T1-value per capsule was significantly different from a χ2-distribution with one degree of freedom and at α= 0·05 divided by I, where I is the total number of capsules examined for one medium (maximum 14) (Bonferroni’s rule).

The variation between capsules was calculated per laboratory, per medium and per incubation period by means of the T2-test. The T2 value was divided by (I−1) (number of degrees of freedom for the T2-test). For each laboratory, the value of T2/(I−1) was compared with the overall measure of dispersion (Thom/N) for that medium. For each laboratory, the value of T2/(I−1) was compared with the Thom/N for that medium by means of a one-sided F-test ( Heisterkamp et al. 1993 ) with I−1 and N = 20 degrees of freedom, with a probability of 95%. The critical threshold was calculated as FI−1,N . ΣThom/N for each laboratory separately, depending on the number of capsules examined.

The laboratory means were compared for MEYP and PEMBA separately and per incubation period using an analysis of variance ( Snedecor & Cochran 1967) on the log10-transformed results. The Grubbs’ test ( Anonymous 1988) was applied to indicate outliers. For SBA, no analysis of variance was carried out because of the limited number of laboratories using this medium.

From the analysis of variance, the variance components were calculated per medium and per incubation period. These variance components were used to calculate the certified values and 95% confidence limits for MEYP and PEMBA. The 95% confidence limits for SBA were calculated as ±2 times the standard deviation (based on log10-transformed counts). The mean values and confidence limits were back-transformed to the original scale.

Technical discussion of analytical procedures

Before the results of the statistical analysis were presented to the participants, a discussion on technical details of the analytical procedure (as laid down in the analytical protocol) was held. This discussion was used to determine whether deviations from the analytical protocol would influence the counts reported. If an influence was expected or could not be excluded, the count(s) were not used in the final analysis of the data and thus, in the calculation of the certified value(s).


Preparation of the material, homogeneity and stability

The highly contaminated milk powder (HCMP) contained approximately 107 cfp g−1 powder. After mixing, the level of contamination and homogeneity was checked against predefined criteria (see Table 1). The results of the first check indicated that all requirements were met, so the entire batch of mixed milk powder was filled into capsules. In total, 15 600 capsules were produced. After the entire batch had been filled into capsules, a check was made to test the variation in weight of the filled capsules. The criterion for the variation in mass was: the standard deviation in mass divided by the average mass of a filled capsule should be less than 0·03. Two sets of 210 and 310 capsules were weighed. The variation in mass was 0·017 and 0·015, respectively. From these capsules, a further 20 capsules was selected at random to determine the level of the final batch on MEYP and SBA as described above. The results of this analysis are also presented in Table 1, together with the criteria for acceptance of the final batch.

Table 1. Results and criteria of testing the batch of milk powder directly after mixing (first check) and after filling of all capsules
Medium Arithmetic
mean count *
  • *Criterion for number of cfp: between 40 and 70 cfp 0·1 ml −1 reconstituted capsule.

  • †Criterion for T 1: not significantly different from χ2(limits footnotes and §).

  • 95% Confidence interval: 3·25–20·5.

  • §

    95% Confidence interval: 9·59–34·2.

  • ¶Criterion T 2/(I−1): ≤2.

MEYPfirst check:   
set 157·08·78 0·97
set 259·316·0 2·19
both58·124·8 §1·56
SBAfirst check:   
set 160·36·14 0·83
set 260·513·5 0·56
both60·419·7 §0·66
MEYPall capsules51·615·0 §1·24
SBAall capsules52·513·5 §0·77

Homogeneity studies

The T2 values needed to determine Thom/N were calculated for SBA, MEYP and PEMBA separately. Only three of 15 T1 values were significantly different from the χ2-distribution. This means that the variation between analytical portions of one reconstituted capsule followed a Poisson distribution.

Two out of 15 T2/(I−1) values were above the limit of 2. The factor Thom/N was found to be 0·71 on MEYP, 1·14 on PEMBA and 1·27 on SBA. The mean dispersion values for each medium were less than 2 and none were significantly different from the χ2-distribution. It was therefore concluded that the variation between samples of different capsules followed a Poisson distribution.

Stability studies

The results of the analyses from January 1994 to November 1997 on MEYP are presented in Fig. 1. The results for SBA (t-value = 0·50) did not indicate a significant (α= 0·05) change in number of cfp over the period tested when using linear regression. However, on MEYP (t-value 2·52), a small but significant (P= 0·012) increase in the number of cfp was found. The rate of increase observed was 0·00015 log10 units per week, corresponding to an average increase of 1 cfp year−1.

Figure 1.

Results of stability tests at storage temperature (−20 °C) on MEYP over a period of 200 weeks

Figure 2 presents the results of the stability test at higher temperatures using MEYP. Only at 37 °C was a significant decrease in the mean number of cfp, of 0·34% d−1, observed. Based on these results, it is concluded that shipment of the materials at ambient temperature will have no effect on number of cfp.

Figure 2.

Results of stability tests at various temperatures: (•), −20 °C; (▵), 22 °C; (○), 30 °C; (□), 37 °C over a period of

Relation between reconstitution time and number of cfp

The results of the tests to determine the relation between reconstitution time and the number of cfp 0·1 ml−1 sample are presented in Fig. 3. The results are averaged for the two sets of five capsules each.

Figure 3.

Relationship between reconstitution time at 38 °C and the arithmetic mean of the number of cfp 0·1 ml−1 sample on MEYP (○) and PEMBA (▵)

Certification study

Technical discussion.

Before use, the capsules needed to be reconstituted according to the protocol. The period for reconstitution specified in the protocol was 30 min. The laboratories took 30–40 min for reconstitution, except for one laboratory which took about 67 min. The results from the latter laboratory were not used for certification because the period was too long and multiplication of the organism could not be excluded (see Fig. 3). No other deviations from the protocol were identified that could have influenced the results.

The results reported for the two blank capsules did not show any false positive counts. Confirmation tests on selected colonies according to the ISO 7932 standard showed that all colonies tested were B. cereus. Laboratories tested a total of 120 colonies from both MEYP and PEMBA, and 40 colonies from SBA.

Statistical analysis.

The results of the T1-tests, T2-tests and the geometric means are presented in Table 2 for MEYP and PEMBA agar and in Table 3 for SBA. Results are based on the examination of 14 capsules per laboratory. The results presented were obtained after 24 h incubation; results obtained after 48 h incubation were similar (results not presented).

Table 2. Results per laboratory on MEYP and PEMBA counted after 24 h incubation

T1T2/(I−1) Geometric
T1T2/(I−1) Geometric
  • *Significantly different from χ 2 -distribution (critical value 23·7 at I  = 14 and α = 0·05).

  • †Significantly different from T hom /N (critical value Thom/N . F(I−1),N = 1·59 at I = 14 and α = 0·05).

  • ‡Significantly different from T hom /N (critical value Thom/N . F(I−1),N = 2·57 at I = 14 and α = 0·05).

  • §

    Results based on the examination of 13 capsules.

15·371·83 57·56·501·6962·2
222·11·73 53·911·41·1455·8
37·201·0950·914·0 §2·16 §52·4 §
625·0 *1·0953·815·00·3753·1
710·32·78 53·94·981·9454·3
1031·8 *1·5956·513·81·9958·8
1118·31·98 51·521·83·11 51·8
Table 3. Results of the laboratories using SBA
LabcodeT1T2/(I−1) T hom /N . F(I−1),NGeometric

For the various laboratories, it was shown that the T1-values were significantly different from the χ2-distribution. For laboratory 3, the result of the T1-value of one capsule on PEMBA showed a problem; the results of this capsule were excluded for certification due to difficulties with pipetting the capsule solution (formation of foam in the tube). For the other laboratories with a significant T1-value, no capsule could be identified as causing problems. For these laboratories, the results were therefore considered as unexplained deviations and retained for analysis.

The results of T2/(I−1) on MEYP and PEMBA are presented in Fig. 4. For MEYP, three laboratories found a higher value for T2/(I−1) than the critical threshold. These results were related to the exceptionally low value for Thom/N for this medium. For PEMBA, one laboratory found a higher value for T2/(I−1) than the critical threshold; this result might be explained by random variation of T2/(I−1).

Figure 4.

Relationship between dispersion test statistic (T2/(I−1)) and arithmetic mean number of cfp 0·1 ml−1 sample found on MEYP (▵) and PEMBA (•) after 24 h incubation

Significant differences between laboratory means existed for MEYP and PEMBA after both 24 and 48 h incubation. However, no outlier was indicated in the group of laboratories. Thus, it was concluded that the results were obtained from a sufficiently homogeneous group of laboratories.

The certified values are presented in Table 4. The indicative values on SBA resulted in a geometric mean number of cfp of 50·1 after 24 h incubation (95% confidence interval 48·4–51·7) based on the examination of 56 capsules in four laboratories.

Table 4. Certified values of CRM 528 on MEYP and PEMBA after 24 and 48 h incubation
95% Confidence limits
ProcedureCertified value *Lower limitUpper limitSets of
  • *

    This value is the geometric mean of 11 accepted sets of data, independently obtained by 11 laboratories.

  • Comprising the results of 154 capsules.

  • Comprising the results of 153 capsules.

  • §

    German Federal Food Law method number.

MEYP (ISO 7932) after 24 h incubation53·451·755·211
MEYP (ISO 7932) after 48 h incubation53·752·155·411
PEMBA (L 00·00–25) § after 24 h incubation55·052·857·411
PEMBA (L 00·00–25) § after 48 h incubation55·853·658·011


Certification results

The batch of B. cereus RM fulfilled the requirements of a CRM to a high degree. The homogeneity test of the material showed no significant difference from a Poisson distribution, irrespective of the type of medium used. However, in the certification study, a number of laboratories found a significantly higher dispersion than expected from the results of the producing laboratory, especially on MEYP. In Table 2, it can be seen that the values for T2/(I−1) are, for most laboratories, higher than 1 for a Poisson distribution. The average value for T2/(I−1) on MEYP was 1·4 and on PEMBA, 1·7. However, only one out of the 11 laboratories found a value for T2/(I−1) significantly higher than 1 on MEYP, and one laboratory on PEMBA. Therefore, the homogeneity of the material is good, which can also be derived from the narrow confidence interval of the certified values. Also, the variance components indicate that the variation resulting from differences between capsules is small compared with the variation due to replicates. The variation between laboratories is also small and comparable with the variation between capsules. This is reflected in the small differences between the mean counts per laboratory (see also Table 2).

Bacterial counts of the stability test at the storage temperature were stable for almost 4 years. The increase in the number of cfp observed on MEYP is regarded as of no, or minor, importance as the increase is very small. As long as this batch of certified RM (CRM) is available for use, the monitoring of its stability will continue. The stability data at higher temperatures indicate that normal (air)mail without cooling of the RM is possible for shipment to other laboratories if the transport time is limited to 1 week.

The reconstitution procedure is regarded as the main critical factor for obtaining reproducible results. In the certification study, one laboratory took 65–70 min for reconstitution of the capsules whereas the protocol specified 30–40 min. Additional tests indicated that up to 60 min, no increase in the number of cfp occurs but after 75 min, a substantial increase was observed. Hence, the results of this laboratory were excluded from further analysis. These data also indicate that the time limits in the protocol are not that critical and can be exceeded as long as the total time needed for reconstitution is less than 1 h.

Use of the CRM

The certificate states the certified values with their 95% confidence limits. These certified values are obtained from a large number of examinations carried out during the certification study. As the user of a CRM will examine only a small number of capsules, user tables are prepared for the interpretation of results in a single laboratory, presenting the 95% confidence limits for different combinations of capsules and replicates per capsule that are likely to be used in practice. These limits are presented in Fig. 5.

Figure 5.

 The 95% confidence limits of the geometric mean of number of cfp 0·1 ml−1 on MEYP after 24 h incubation for different combinations of capsules and replicates, calculated from the certified value and the variance components of Bacillus cereus CRM 528. (□), LL(1) = lower limit using one replicate; (▵), LL(2) = lower limit using two replicates; (○), UL(1) = upper limit using one replicate; (▵), UL(2) = upper limit using two replicates

The limits are calculated using the certified value and variance components per method on the log10 scale. Back-transformation of the obtained values for the upper and lower limit will give the limits on the normal scale. Due to the back-transformation, the values on the original scale represent geometric mean values. The limit values are rounded to whole counts. For the lower limit, values are rounded to the lowest whole count and for the upper limit, values are rounded to the highest whole count.

Figure 5 shows that the examination of more than a few capsules does not lead to a substantial improvement in the confidence limits. It is therefore recommended that two capsules should be tested in duplicate each time the CRM is used.

From all the results presented, it can be concluded that the enumeration of B. cereus on MEYP or PEMBA using the CRM is highly reproducible or, in other words, is very precise.


The authors wish to thank the following participants of the certification study: Dr R. Betts, Campden and Chorleywood Food Research Association, UK; Mrs M. de Buyser, Laboratoire Central d’Hygiène d’Alimentaire, Ministère de l’Agriculture, France; Mr B.J. Hartog, TNO-Nutrition, The Netherlands; Dr J. Kramer, University of Bonn, Germany; Dr V. Sanguinetti, Universita di Bologna, Italy; Prof. M. Stecchini, Instituto di Tecnologie Alimentari, Italy; Mr H. Stegeman, RIKILT-DLO, The Netherlands; Prof. Dr P. Teufel, Bundesgesundheitsamt Robert Von Ostertag Institute, Germany; Dr I.M. Vicente da Cruz, Depart. de Tecnol. das Indust. Alimentares, Portugal; Mrs G. Vlaemynck, Rijkszuivelstation, Belgium; Dr C. Wray, Central Veterinary Laboratory, Ministry of Agriculture, Fisheries and Food, UK. Mrs N.G.W.M. van Strijp-Lockefeer and Mrs J.A. van Dommelen are thanked for the statistical analysis of the certification study. This work has been supported by the EC, Standards, Measurements and Testing Programme under contract no MAT1-CT92–0001.