There are a few guidelines for the verification of food qualitative methods, including HPB, NMKL, and AOAC. The HPB-MFO5 guideline introduces a “detection limit (DL)” study with artificial contamination at five different spiking levels and three replicates of uninoculated samples on 18 replicate test samples [25]. In spite of performing the numerous spiking tests and in contrast to ISO 16140-3 Protocol 1, the HPB does not offer a numeric DL value. Similarly, the 2013 Australian National Association of Testing Authorities (NATA) guideline offers one method to indicate the LOD for a quantitative microbiological assay through inoculation with a limited number of challenging microorganisms (not more than 5 CFU per unit), followed by the measurement of recovery and determination of eLOD50 in other aspects, such as method validation and verification of alternative and rapid methods [26]. Therefore, in this study, we applied the ISO 16140-3 protocol to verify three routine microbiologic detection methods by calculating the limit of detection at 50% (eLOD50). Our findings successfully verified standard methods of detecting E. coli, S. aureus, and Salmonella spp. in selected food items, indicating that our laboratory can properly use/implement three qualitative methods. A few studies have been conducted on verifying food microbiological testing methods, mostly focused on Salmonella spp. detection methods with limited food items [8–11]. For instance, the study by Cajas Rios et al. verified the horizontal method for the detection of Salmonella spp. in poultry, garlic, strawberries, and animal feed pellets [9]. Freschi et al. also verified a Salmonella spp. detection method using powdered infant formula [27]. Furthermore, different studies have calculated eLOD50 other performance characteristics using models such as the complementary log-log model and generalized linear mixed models [8, 28].
Considering the lack of published validation studies on LOD50 for E. coli and S. aureus, we considered the lowest possible LOD50 (1 CFU per test portion) as the acceptance criteria. According to Granato et al., a theoretically minimal LOD value of 1 CFU/25g of the analytical test portion is the best that can be expected from a qualitative microbiology method [3]. For Salmonella spp. and S. aureus, we had to repeat the tests for a number of food items (minced meat, hamburger, pasteurized whole liquid eggs, and black pepper) due to unacceptable negative results for the high inoculation level or positive results for the blank test portions. As outlined in the results evaluation step of protocol 1, the high level inoculation (9 × LOD50) should only produce positive results and the blank level should not produce a positive result, so we had to repeated the experiment (re-run verification) for all levels of mentioned food items. Such test results may be attributed to the complexity of the microflora, natural contamination, natural microbiota, heterogeneous properties of the semi-solid matrix (matrix effect and motion effect or Poisson distribution effects), diverse behavior of the microorganism in the food matrix, secondary and accidental contamination of food items, the interaction of food ingredients in complex foods, the heterogeneity of samples, and microbial competition; these factors interfere with the blank test results, which should be negative at all testing stages. In cases where the blank test result was positive, the test was repeated. ISO protocols have not suggested any decontamination processes before running the tests. Therefore, these positive results can be attributed to natural and accidental contamination of the test samples; such contamination cannot easily be removed and are not negligible.
Despite using CRMs (WDCM strains) for inoculum preparation, we found that the adoption of ISO 16140-3 Protocol 3 was impractical in this study due to the intrinsic wide concentration range in bacteria count of the CRMs and the mentioned uncertainty in the certificate of analysis of each product. Therefore, microbial counting should be done simultaneously with test portion inoculation to ensure the accuracy and precision of the count of the final inoculum. Nonetheless, the application of microbiological CRMs was the best choice for culture media quality control and inoculum preparation.
In this study, more than 5 + 1 food items were selected since our laboratory has a broad range of applications. All food items, except animal feed (soybean powder), were within the scope of our laboratory. Other food categories, such as meat products (minced meat and hamburger), dairy products (pasteurized milk), pasteurized whole liquid eggs, and black pepper, were the most challenging food items selected for the verification study, as recommended by the ISO.
The LOD is the lowest number of microorganisms that can be detected under the stated testing conditions. Tim Sandle et al. reported a LOD < 10 CFU/g in the evaluation of raw material using the pour plate method and described some challenges in this respect considering the fact that dilution may lead to more losses caused by the lack of homogeneity and the Poisson distribution of microorganisms [29]. Therefore, microbiologically, the LOD must sometimes be considered theoretically rather than practically. Accordingly, many pharmacopeial tests require the use of a low-level challenge (< 100 CFU), which is normally considered to be sufficient [30]. Because the amount of LOD in qualitative microbiological food testing methods is considered much less than this range, we faced many challenges in determining LOD. The main challenges were preparing reputable counts of the inoculums of each organism and the risks of working with pathogenic Salmonella (Biosafety Level 2, Hazard Group 2 biological agent that can cause human disease and might be a hazard to workers).