Lessons from the organization of a proficiency testing program in food microbiology by interlaboratory comparison: analytical methods in use, impact of methods on bacterial counts and measurement uncertainty of bacterial counts.

The proficiency testing program in food microbiology RAEMA (Réseau d'Analyses et d'Echanges en Microbiologie des Aliments), created in 1988, currently includes 450 participating laboratories. This interlaboratory comparison establishes proficiency in detection of Salmonella and Listeria monocytogenes, as well as enumeration of aerobic micro-organisms, Enterobacteriaceae, coliforms, beta-glucuronidase-positive Escherichia coli, anaerobic sulfito-reducing bacteria, Clostridium perfringens, coagulase-positive staphylococci, and L. monocytogenes. Twice a year, five units samples are sent to participants to assess their precision and trueness for enumeration and detection of micro-organisms. Most of participating laboratories use standard or validated alternative methods, they were 50-70% in 1994 and, for 5 years, they are 95%. An increasing use of alternative methods was also observed. This phenomenon is all the more significant as standard methods are laborious and time consuming; thus, 50% of the laboratories use alternative methods for the detection of Salmonella and L. monocytogenes. More and more laboratories use ready-to-use media and although the percentage is variable according to the microflora, we can consider that, today, 50-60% of the laboratories participating to the proficiency program only use ready-to-use media. The internal quality assurance programs lead also to an increasing use of media quality controls. The impact of analytical methods on bacterial counts was assessed by grouping together the results obtained by participating laboratories during the 10 last testing schemes from 1999 to 2003. The identified significant factors influencing enumeration results are variable from one microflora to another. Some of them significantly influence many microflora: the plating method (spiral plating or not) is influential for aerobic micro-organisms, Enterobacteriaceae, coliforms, and staphylococci, the type of culture medium and the medium manufacturer is influential for aerobic micro-organisms, Enterobacteriaceae, coliforms, E. coli, anaerobic sulfito-reducing bacteria, staphylococci, and L. monocytogenes. Others are specific of some micro-organisms: the resuscitation broth for L. monocytogenes, the mode of medium preparation for staphylococci and the incubation temperature for C. perfringens. These effects lead generally to small differences of about 0.1 log10 cfu g(-1), except for the enumeration of anaerobic sulfito-reducing bacteria, where the difference reaches 0.7 log10 cfu g(-1). These results, although difficult to extrapolate to all actual situations, which associate numerous food constituents and physiological states of bacteria to detect or numerate, allow nevertheless the quantification of interlaboratory variations linked to the methods in use. The analysis of bacterial counts obtained by the laboratories participating to the RAEMA proficiency testing program allowed also to validate a formula to calculate the repeatability of bacterial counts and to estimate the between-laboratory uncertainties for the majority of micro-organisms enumerated in food microbiology. The repeatability uncertainty is only indirectly affected by the method in use but depends essentially on the number of counted colonies. On the other hand, the between-laboratory uncertainty varies with the enumeration method in use, this variability is relatively small for the enumerations calling for methods without colony confirmation, i.e. for the enumeration of aerobic micro-organisms, Enterobacteriaceae, 'total' and thermotolerant coliforms, beta-glucuronidase-positive E. coli and coagulase-positive staphylococci with the technique using the rabbit-plasma fibrinogen agar. For these methods, the average between-laboratory standard deviation is 0.17 log10 cfu g(-1). The between-laboratory uncertainty is, on the contrary, larger for more complex techniques. For the enumeration of coagulase-positive staphylococci with the Baird-Parker agar, the between-laboratory standard deviation is equal to 0.23 log10 cfu g(-1), it is equal to 0.28 log10 cfu g(-1) for the enumeration of L. monocytogenes, to 0.34 log10 cfu g(-1) for the enumeration of C. perfringens, and to 0.47 log10 cfu g(-1) for the enumeration of anaerobic sulfito-reducing bacteria.