Estimation of the POD Function and the LOD of a Binary Microbiological Measurement Method from an Interlaboratory Experiment.

Background: We deal with interlaboratory experiments (collaborative studies) in which k participating laboratories, selected randomly from a population of laboratories, use samples from one and the same material or matrix. They perform binary microbiological measurements for which the measurement results are either "0" (target microorganisms not detected) or "1" (target microorganisms detected). The performance of such a measurement method is described by its probability of detection (POD) function, i.e., the POD as a function of the contamination of the sample (CFU per gram or CFU per milliliter), or by the level of detection (LODp), i.e., the contamination level of the sample that is detected (measurement result "1") with a specified probability p. Objective: We derive an approximate statistical analysis that is simple enough to be implemented in a spreadsheet application. Methods: Under the assumption of a Poisson distribution of the number of CFU in the samples, we estimate the mean POD function of the laboratories and the SD of the laboratory effect based on a complementary log-log model, a special case of the Generalized Linear Model in the special situation in which the contamination level is known by means other than the POD. The estimates are obtained by maximization of the Laplace approximation of the likelihood function. By simulation, a bias correction factor for the estimate of the SD is obtained. With the estimated POD function, LODs can be estimated. The model can also be used to evaluate the relative LOD of an alternative method with repect to a reference method. Results: The EXCEL program PODLOD-interlab_ver1.xls for this method of statistical analysis can be downloaded from http://www.wiwiss.fu-berlin.de/fachbereich/vwl/iso/ehemalige/wilrich. Highlights: A simple approximate statistical method for the estimation of the POD and LOD is derived. The method also allows the estimation of the RLOD of an alternative method with respect to reference method. The method is implemented in an EXCEL program that can be downloaded from http://www.wiwiss.fu-berlin.de/fachbereich/vwl/iso/ehemalige/wilrich.

Trends in Chemometrics: Food Authentication, Microbiology, and Effects of Processing.

In the last decade, the use of multivariate statistical techniques developed for analytical chemistry has been adopted widely in food science and technology. Usually, chemometrics is applied when there is a large and complex dataset, in terms of sample numbers, types, and responses. The results are used for authentication of geographical origin, farming systems, or even to trace adulteration of high value-added commodities. In this article, we provide an extensive practical and pragmatic overview on the use of the main chemometrics tools in food science studies, focusing on the effects of process variables on chemical composition and on the authentication of foods based on chemical markers. Pattern recognition methods, such as principal component analysis and cluster analysis, have been used to associate the level of bioactive components with in vitro functional properties, although supervised multivariate statistical methods have been used for authentication purposes. Overall, chemometrics is a useful aid when extensive, multiple, and complex real-life problems need to be addressed in a multifactorial and holistic context. Undoubtedly, chemometrics should be used by governmental bodies and industries that need to monitor the quality of foods, raw materials, and processes when high-dimensional data are available. We have focused on practical examples and listed the pros and cons of the most used chemometric tools to help the user choose the most appropriate statistical approach for analysis of complex and multivariate data.

The contribution of sampling uncertainty to total measurement uncertainty in the enumeration of microorganisms in foods.

Random samples of each of several food products were obtained from defined lots during processing or from retail outlets. The foods included raw milk (sampled on farm and from a bulk-milk tanker), sprouted seeds, raw minced meat, frozen de-shelled raw prawns, neck-flaps from raw chicken carcasses and ready-to-eat sandwiches. Duplicate sub-samples, generally of 100 g, were examined for aerobic colony counts; some were examined also for counts of presumptive Enterobacteriaceae and campylobacters. After log(10)-transformation, all sets of colony count data were evaluated for conformity with the normal distribution (ND) and analysed by standard ANOVA and a robust ANOVA to determine the relative contributions of the variance between and within samples to the overall variance. Sampling variance accounted for >50% of the reproducibility variance for the majority of foods examined; in many cases it exceeded 85%. We also used an iterative procedure of re-sampling without replacement to determine the effects of sample size (i.e. the number of samples) on the precision of the estimate of variance for one of the larger data sets. The variance of the repeatability and reproducibility variances depended on the number of replicate samples tested (n) in a manner that was characteristic of the underlying distribution. The results are discussed in relation to the use of measurement uncertainty in assessing compliance of results with microbiological criteria for foods.

The effect of the number of sample units tested on the precision of microbial colony counts.

Studies on the precision of chemical methods of analysis, and the associated 'sampling uncertainty', suggest that analysis of eight replicate sample units (the sample size) is required to ensure adequate analytical precision. The primary purpose of this work was to assess whether these findings are equally applicable in microbiological examination of foods. We examined the effect of sample size on the analytical precision of microbiological data by iteratively 're-sampling without replacement' (SNR). Using both theoretical data sets and colony counts from foods we demonstrate that SNR provides an effective and efficient guide to (a) choosing the number of samples to be examined in order to optimise precision and (b) deciding whether logarithmic transformation of the raw data is appropriate. We also discuss theoretical aspects of the procedure and their impact on the results obtained.

The inhibitory effect of natural microflora of food on growth of Listeria monocytogenes in enrichment broths.

The aims of this study were to (i) compare the inhibitory effects of the natural microflora of different foods on the growth of Listeria monocytogenes during enrichment in selective and non-selective broths; (ii) to isolate and identify components of the microflora of the most inhibitory food; and (iii) to determine which of these components was most inhibitory to growth of L. monocytogenes in co-culture studies. Growth of an antibiotic-resistant marker strain of L. monocytogenes was examined during enrichment of a range of different foods in Tryptone Soya Broth (TSB), Half Fraser Broth (HFB) and Oxoid Novel Enrichment (ONE) Broth. Inhibition of L. monocytogenes was greatest in the presence of minced beef, salami and soft cheese and least with prepared fresh salad and chicken pâté. For any particular food the numbers of L. monocytogenes present after 24h enrichment in different broths increased in the order: TSB, HFB and ONE Broth. Numbers of L. monocytogenes recovered after enrichment in TSB were inversely related to the initial aerobic plate count (APC) in the food but with only a moderate coefficient of determination (R(2)) of 0.51 implying that microbial numbers and the composition of the microflora both influenced the degree of inhibition of L. monocytogenes. In HFB and ONE Broth the relationship between APC and final L. monocytogenes counts was weaker. The microflora of TSB after 24h enrichment of minced beef consisted of lactic acid bacteria, Brochothrix thermosphacta, Pseudomonas spp., Enterobacteriaceae, and enterococci. In co-culture studies of L. monocytogenes with different components of the microflora in TSB, the lactic acid bacteria were the most inhibitory followed by the Enterobacteriaceae. The least inhibitory organisms were Pseudomonas sp., enterococci and B. thermosphacta. In HFB and ONE Broth the growth of Gram-negative organisms was inhibited but lactic acid bacteria still reached high numbers after 24h. A more detailed study of the growth of low numbers of L. monocytogenes during enrichment of minced beef in TSB revealed that growth of L. monocytogenes ceased at a cell concentration of about 10(2)cfu/ml when lactic acid bacteria entered stationary phase. However in ONE Broth growth of lactic acid bacteria was slower than in TSB with a longer lag time allowing L. monocytogenes to achieve much higher numbers before lactic acid bacteria reached stationary phase. This work has identified the relative inhibitory effects of different components of a natural food microflora and shown that the ability of low numbers of L. monocytogenes to achieve high cell concentrations is highly dependent on the extent to which enrichment media are able to inhibit or delay growth of the more effective competitors.

Minimising the between-sample variance in colony counts on foods.

The objective of this study was to assess whether it is possible to minimise the variance in colony counts on replicate target samples of foods by aseptic compositing of the samples or by increasing the quantity of sample examined. The results show that compositing reduces the overall variance, and hence the standard deviation, to very low levels, although in some cases the overall variance remains relatively high, reflecting the heterogeneous distribution of microorganisms in the foods. Increasing the weight of target sample examined (e.g. from 10 g to 100g) had a pronounced effect on the mean log(10) colony count and significantly reduced the variance of the mean. The results are discussed in relation to the quantity of sample that is recommended for examination in international and other standards.

Measurement uncertainty of the EU methods for microbiological examination of red meat.

Three parallel trials were made of EU methods proposed for the microbiological examination of red meat using two analysts in each of seven laboratories within the UK. The methods involved determination of aerobic colony count (ACC) and Enterobacteriaceae colony count (ECC) using simulated methods and a freeze-dried standardised culture preparation. Trial A was based on a simulated swab test, Trial B a simulated meat excision test and Trial C was a reference test on reconstituted inoculum. Statistical analysis (ANOVA) was carried out before and after rejection of outlying data. Expanded uncertainty values (relative standard deviation x2) for repeatability and reproducibility, based on the log10 cfu/ml, on the ACC ranged from +/-2.1% to +/-2.7% and from +/-5.5% to +/-10.5%, respectively, depending upon the test procedure. Similarly for the ECC, expanded uncertainty estimates for repeatability and reproducibility ranged from +/-4.6% to +/-16.9% and from +/-21.6% to +/-23.5%, respectively. The results are discussed in relation to the potential application of the methods.

Assessment of measurement uncertainty for quantitative methods of analysis: comparative assessment of the precision (uncertainty) of bacterial colony counts.

An interlaboratory trial was made, by three analysts in each of the 19 laboratories, of three International Standards Organisation (ISO) colony count methods for aerobic organisms (ACC), Enterobacteriaceae (ECC) and Escherichia coli (EcCC). All tests were done in duplicate and were further replicated by plating both on culture media supplied by the organisers and on each laboratory's own choice of culture media. In order to avoid any influence of food matrix on the results, the inoculum for each test was a freeze-dried ampoule of a standardised mixed culture. After collation of test results, individual data sets were examined for obvious non-consistency, and colony counts for each individual test were determined both as simple and 'weighted' mean values. The derived colony counts were then log(10)-transformed and examined statistically. Estimates of repeatability and reproducibility for each set of results were derived and used to calculate the parameters for the uncertainty of measurement. Estimated values of the uncertainty of reproducibility and repeatability for the ACC ranged, respectively, from 9.3 to 12.1% and 2.0 to 3.9% of the mean log(10) colony count, depending on the specific culture media, the method of deriving the mean count and the statistical procedure employed. Similarly, estimates of the uncertainty of reproducibility and repeatability for the ECC ranged, respectively, from 14.0 to 17.4% and 4.1 to 6.7%. The estimates of uncertainty of reproducibility and repeatability for the EcCC data ranged, respectively, from 21.1 to 30.9% and 6.7 to 15.4%. Whilst the uncertainty estimates for the ACC data conform to long-held views on the repeatability and reproducibility of microbial count data, the estimates for the ECC and EcCC are considerably greater. It was notable that no benefits were seen in the use of the weighted mean as compared to simple mean colony count. The importance of uncertainty estimates of colony count data in the assessment of the microbiological quality of foods is discussed.

A critical review of measurement uncertainty in the enumeration of food micro-organisms.

Derivation of uncertainty provides a way to standardize the expression of variability associated with any analytical procedure. The published information on uncertainty associated with data obtained using microbiological procedures is reviewed to highlight the causes and magnitude of such variability in food microbiology. We also suggest statistical procedures that can be used to assess variability (and hence, uncertainty), within and between laboratories, including procedures that can be used routinely by microbiologists examining foods, and the use of 'robust' methods which allow the retention of 'outlying' data. Although concerned primarily with variability associated with colony count procedures, we discuss also the causes of variability in presence/absence and indirect methods, such as limiting dilution, most probable number and modern instrumental methods of microbiological examination. Recommendations are also made concerning the most important precautions to be taken in order to minimize uncertainty in microbiology. These include strict internal controls at all stages of microbiological testing, as well as validation of methods, trend analysis, use of reference materials and participation in proficiency testing schemes. It is emphasized that the distribution of microbes in foods is inherently heterogeneous, and that this review only addresses uncertainty of measurement with respect to the sample taken, not the lot or consignment of food from which the sample was taken.

Application of "robust" methods to the analysis of collaborative trial data using bacterial colony counts.

We describe techniques for the application of two methods, robust to the presence of "outliers", to the hierarchical analysis of variance of bacterial count data from collaborative trials. The techniques are tested against both artificially-generated data with known distributional parameters and actual trial results containing outliers. The relative merits of the robust methods are discussed in comparison with conventional ANOVA techniques.