ABSTRACT

Environmental hygiene monitoring in the food processing environment has become important in current food safety programs to ensure safe food production. However, conventional monitoring of surface hygiene based on visual inspection and microbial counts is slow, tedious, and thus unable to support the current risk-based management system. Therefore, this study was conducted to assess the performance of a real-time total adenylate assay that detected ATP+ADP+AMP (A3) for food contact surface hygiene in 13 food processing plants and two commercial kitchens in Malaysia. The A3 value was compared with the microbial count (aerobic plate count [APC]) on food contact surfaces. Receiver-operating characteristic (ROC) analysis was performed to assess the reliability of the data and to determine the optimal threshold value for hygiene indication of food contact surfaces. Overall, the A3 value demonstrated a weak positive relationship with APC. However, the A3 value significantly correlated with APC for food processing environments associated with raw meat and raw food ingredients such as fruit that harbor a high microbial load. ROC analysis suggested an optimal threshold for the A3 value of 500 relative light units to balance the sensitivity and specificity at 0.728 and 0.719, respectively. The A3 assay as a hygiene indicator for food contact surfaces had an efficiency of 72.1%, indicating its reliability as a general hygiene indicator.

HIGHLIGHTS
  • AMP+ADP+ATP (A3) assay could be a hygiene indicator in food processing environments.

  • A3 assay demonstrated 0.728 sensitivity, 0.719 specificity, and 72.1% efficiency.

  • A3 correlated with APC on factory surfaces associated with raw meat and ingredients.

Foodborne pathogens, such as Listeria monocytogenes and Salmonella, have been reported to contaminate and persist in food processing environments that eventually cross-contaminate food products (21, 31). Of particular concern is L. monocytogenes, which has caused numerous outbreaks globally (18). Furthermore, there have been incidences in which foodborne outbreaks were found to be associated with food processing environments and pathogens were detected on food contact surfaces (22). Research has shown that foodborne pathogens, such as L. monocytogenes and Salmonella, could form strong and persistent biofilms on food contact surfaces if the cleaning process is not carried out properly in food processing plants (8). As such, monitoring of environmental hygiene and sanitation in food processing facilities is an essential element emphasized in a food safety program (32).

However, the detection of microorganisms on food contact surfaces for cleanliness and hygiene assessment in the food processing environment largely depends on a conventional culturing method, which is slow and cumbersome. The conventional culture-based method is not suitable for routine monitoring to fulfill the requirements for hazard analysis and critical control point (HACCP) system and hazard analysis and risk-based preventive controls (HARPC) under the Food Safety Modernization Act. In addition, many food factories rely on visual inspection to assess the cleanliness of food contact surfaces, a practice that has been proven to be inaccurate (25). Thus, numerous studies have been conducted to develop more rapid, easy-to-use, economical, and accurate indicators or detection methods to fulfill the need for routine environmental monitoring under food safety programs. Among the many tests established, the ATP swabbing assay is the most widely used for cleanliness and hygiene monitoring of food processing environments, because it enables rapid verification of sanitation processes on site for implementation of HACCP and HARPC programs (3). The ATP swabbing assay detects ATP from organic matter, including microorganisms and food debris (14). In other words, ATP is not specific for the detection of microorganisms only. Nonetheless, ATP is widely used to indicate cleanliness and hygiene, because the presence of organic matter on food contact surfaces, regardless of whether microorganisms are present, can provide a great nutrient source for the attachment and growth of microorganisms that lead to food safety risks (26, 35).

However, the conventional ATP assay was found in some controlled cases to be less sensitive, because ATP is hydrolyzed to ADP and AMP by high temperature, acid, and alkaline conditions (3). Bakke and Suzuki (3) demonstrated that both AMP and ADP were present in high concentrations in various types of food, including meat, seafood, and fruits; thus, measuring AMP+ADP+ATP (A3) was expected to be a sensitive and accurate indicator of cleanliness and hygiene on food contact surfaces. Although Bakke and Suzuki (3) carried out a fairly comprehensive study on the performance of the A3 assay as a sanitation indicator, the assessment was performed in a laboratory environment. The utilization of the A3 assay for hygiene monitoring in a food processing environment, which has a more complex background, was not investigated. Therefore, this study was conducted to examine the sensitivity, specificity, and reliability of this A3 assay for routine monitoring of cleanliness in food processing environments. In addition, because the ATP test is commonly used only to indicate overall hygiene, not microbial contamination on surfaces, this study compared and correlated A3 measurements (is relative light units [RLU]) with aerobic plate count (APC) to evaluate the effectiveness of the A3 assay in monitoring microbial contamination on surfaces.

MATERIALS AND METHODS

Food manufacturing plants and commercial kitchens

A total of 13 food manufacturing plants and two commercial kitchens located in peninsular Malaysia were recruited for this study. For each plant (except for FMP12), five (n = 5) food contact surfaces were identified and swabbed for A3 detection and microbial assessment. For FMP12, only four food contact surfaces were swabbed in the first visit and only two food contact surfaces were swabbed in subsequent visits. The contact surfaces were identified based on the HACCP plan or in-house hygiene monitoring plan of each food manufacturing plant. Detailed information on the food manufacturing plants and sampling points in the plants is listed in Table 1.

TABLE 1

Food manufacturing plants and commercial kitchens that participated in this study

Food manufacturing plants and commercial kitchens that participated in this study
Food manufacturing plants and commercial kitchens that participated in this study

Sampling for surface hygiene assessment

Five visits were conducted for each of the participating food manufacturing plants and commercial kitchens from February to June 2019. The identified surfaces were swabbed after the cleaning and sanitizing procedures were performed by factory workers. Approximately 100 cm2 of flat surface was swabbed by making five streaks in four directions with both the A3 swab and the microbial swabs simultaneously. The LuciPac A3 Surface swab (Kikkoman Biochemifa Company, Tokyo, Japan) was immediately inserted into a tube for measurement of A3. The microbial swab (Puritan sampling kit) was immediately immersed in a tube containing 5 mL of Letheen broth (Puritan Medical Products, Guilford, ME) and transported on ice to the laboratory for microbial analysis within 4 h.

A3 measurements

A3 was assayed using a LuciPac A3 Surface/Lumitester PD-30 (Kikkoman Biochemifa). Sample collection and processing were performed according to manufacturer instructions. Before A3 measurements were performed, the LuciPac A3 Surface kit was prewarmed in the incubator at 30°C so that when the A3 measurements were performed, the temperature could be maintained at the range of 23 to 30°C. After the tested surface was swabbed with the sample collection swab, it was inserted into the LuciPac A3 tube. The tube that contained the extraction solution and reagent was then shaken vigorously. The tube was immediately inserted into the Lumitester PD-30 to allow measurement of the resulting luminescence. The measurement output was in RLU, and data were recorded electronically. The temperature in the factory was also measured to ensure that all measurements were performed at 23 to 30°C.

Microbial assay

Standard APC was performed to quantitate the microbial count on food contact surfaces. A series of 10-fold dilutions was conducted by transferring 1 mL of sample into 9 mL of buffered peptone water (Merck KGaA, Darmstadt, Germany). Then, 1 mL of diluent was plated in duplicates onto plate count agar (Merck) and incubated at 35°C for 18 to 24 h. The number of isolates grown on the plates was counted following standard protocol. For statistical analysis, a colony count below 25 colonies was included in the analysis by assuming the estimated number. The unit used in the analysis was CFU per 100 cm2 of contact surface. A threshold level at 250 CFU/100 cm2 was used to determine passing or failing of the microbial test. Surfaces with less than 250 CFU/100 cm2 were classified as clean (i.e., passed the microbial test). The pass–fail threshold of 250 CFU/100 cm2 for APC was based on studies by Annette et al. (2), Cooper et al. (12), and Buckalew (7).

Data analysis

All data produced by the Lumitester PD-30 and microbial assays from the laboratory were tabulated and processed using an Excel spreadsheet (Microsoft Corp., Redmond, WA). Summary statistics were calculated for the RLU and CFU levels on surfaces. Discrete data were analyzed by logarithmic average with relevant cross-tabulations, and categorical data were analyzed by frequency with stratification as needed. These statistical analyses were conducted using SPSS package 20 (SPSS, Inc., Chicago, IL). Receiver-operating characteristic (ROC) analysis was performed based on the approach described by Biagini et al. (4). In this case, sensitivity was defined as the rate of contaminated surfaces (APC > 250 CFU/100 cm2) in samples with an A3 value below the cutoff value, specificity was the rate of clean surfaces (APC < 250 CFU/100 cm2) in samples with an A3 value greater than the cutoff value, the positive predictive value described the percentage of surfaces with an A3 value below the cutoff value of all contaminated surfaces, the negative predictive value was the percentage of surfaces with an A3 value greater than the cutoff value of all clean surfaces, and efficiency described the percentage of surfaces correctly classified as clean or contaminated based on the A3 value within a specified cutoff. The calculated true-positive rates (sensitivity) across varying cutoffs were plotted against the false-positive rates (1, specificity) to generate the ROC curve. The threshold with sensitivity equal to specificity was selected as the optimal cutoff. The area under the curve that described the inherent validity and accuracy of the diagnostic tests (4) was calculated. The formulae used were as follows:
\(\def\upalpha{\unicode[Times]{x3B1}}\)\(\def\upbeta{\unicode[Times]{x3B2}}\)\(\def\upgamma{\unicode[Times]{x3B3}}\)\(\def\updelta{\unicode[Times]{x3B4}}\)\(\def\upvarepsilon{\unicode[Times]{x3B5}}\)\(\def\upzeta{\unicode[Times]{x3B6}}\)\(\def\upeta{\unicode[Times]{x3B7}}\)\(\def\uptheta{\unicode[Times]{x3B8}}\)\(\def\upiota{\unicode[Times]{x3B9}}\)\(\def\upkappa{\unicode[Times]{x3BA}}\)\(\def\uplambda{\unicode[Times]{x3BB}}\)\(\def\upmu{\unicode[Times]{x3BC}}\)\(\def\upnu{\unicode[Times]{x3BD}}\)\(\def\upxi{\unicode[Times]{x3BE}}\)\(\def\upomicron{\unicode[Times]{x3BF}}\)\(\def\uppi{\unicode[Times]{x3C0}}\)\(\def\uprho{\unicode[Times]{x3C1}}\)\(\def\upsigma{\unicode[Times]{x3C3}}\)\(\def\uptau{\unicode[Times]{x3C4}}\)\(\def\upupsilon{\unicode[Times]{x3C5}}\)\(\def\upphi{\unicode[Times]{x3C6}}\)\(\def\upchi{\unicode[Times]{x3C7}}\)\(\def\uppsy{\unicode[Times]{x3C8}}\)\(\def\upomega{\unicode[Times]{x3C9}}\)\(\def\bialpha{\boldsymbol{\alpha}}\)\(\def\bibeta{\boldsymbol{\beta}}\)\(\def\bigamma{\boldsymbol{\gamma}}\)\(\def\bidelta{\boldsymbol{\delta}}\)\(\def\bivarepsilon{\boldsymbol{\varepsilon}}\)\(\def\bizeta{\boldsymbol{\zeta}}\)\(\def\bieta{\boldsymbol{\eta}}\)\(\def\bitheta{\boldsymbol{\theta}}\)\(\def\biiota{\\boldsymbol{\iota}}\)\(\def\bikappa{\boldsymbol{\kappa}}\)\(\def\bilambda{\boldsymbol{\lambda}}\)\(\def\\bimu{\boldsymbol{\mu}}\)\(\def\binu{\boldsymbol{\nu}}\)\(\def\bixi{\boldsymbol{\xi}}\)\(\def\biomicron{\boldsymbol{\micron}}\)\(\def\bipi{\boldsymbol{\pi}}\)\(\def\birho{\boldsymbol{\rho}}\)\(\def\bisigma{\boldsymbol{\sigma}}\)\(\def\bitau{\boldsymbol{\\tau}}\)\(\def\biupsilon{\boldsymbol{\upsilon}}\)\(\def\biphi{\boldsymbol{\phi}}\)\(\def\bichi{\boldsymbol{\chi}}\)\(\def\bipsy{\boldsymbol{\psy}}\)\(\def\biomega{\boldsymbol{\omega}}\)\(\def\bupalpha{\bf{\alpha}}\)\(\def\bupbeta{\bf{\beta}}\)\(\def\bupgamma{\bf{\gamma}}\)\(\def\bupdelta{\bf{\delta}}\)\(\def\bupvarepsilon{\bf{\varepsilon}}\)\(\def\bupzeta{\bf{\zeta}}\)\(\def\bupeta{\bf{\eta}}\)\(\def\buptheta{\bf{\theta}}\)\(\def\bupiota{\bf{\iota}}\)\(\def\bupkappa{\bf{\kappa}}\)\(\def\\buplambda{\bf{\lambda}}\)\(\def\bupmu{\bf{\mu}}\)\(\def\bupnu{\bf{\nu}}\)\(\def\bupxi{\bf{\xi}}\)\(\def\bupomicron{\bf{\micron}}\)\(\def\buppi{\bf{\pi}}\)\(\def\buprho{\bf{\rho}}\)\(\def\bupsigma{\bf{\sigma}}\)\(\def\buptau{\bf{\tau}}\)\(\def\bupupsilon{\bf{\upsilon}}\)\(\def\bupphi{\bf{\phi}}\)\(\def\bupchi{\bf{\chi}}\)\(\def\buppsy{\bf{\psy}}\)\(\def\bupomega{\bf{\omega}}\)\(\def\bGamma{\bf{\Gamma}}\)\(\def\bDelta{\bf{\Delta}}\)\(\def\bTheta{\bf{\Theta}}\)\(\def\bLambda{\bf{\Lambda}}\)\(\def\bXi{\bf{\Xi}}\)\(\def\bPi{\bf{\Pi}}\)\(\def\bSigma{\bf{\Sigma}}\)\(\def\bPhi{\bf{\Phi}}\)\(\def\bPsi{\bf{\Psi}}\)\(\def\bOmega{\bf{\Omega}}\)\begin{equation}{\rm{Sensitivity}} = {{{\rm{True}}\;{\rm{positive}}} \over {{\rm{True}}\;{\rm{positive}} + {\rm{False}}\;{\rm{negative}}}}\end{equation}
\begin{equation}{\rm{Specificity}} = {{{\rm{True}}\;{\rm{negative}}} \over {{\rm{False}}\;{\rm{positive}} + {\rm{True}}\;{\rm{negative}}}}\end{equation}
\begin{equation}{\rm{Positive}}\;{\rm{predictive}}\;{\rm{value}} = {{{\rm{True}}\;{\rm{positive}}} \over {{\rm{True}}\;{\rm{positive}} + {\rm{False}}\;{\rm{positive}}}} \times {\rm{100\% }}\end{equation}
\begin{equation}{\rm{Negative}}\;{\rm{predictive}}\;{\rm{value}} = {{{\rm{True}}\;{\rm{negative}}} \over {{\rm{True}}\;{\rm{negative}} + {\rm{False}}\;{\rm{negative}}}} \times {\rm{100\% }}\end{equation}
\begin{equation}{\rm{Efficiency}} = {{{\rm{True}}\;{\rm{positive}} + {\rm{True}}\;{\rm{negative}}} \over {{\rm{True}}\;{\rm{positive}} + {\rm{False}}\;{\rm{positive}} + {\rm{True}}\;{\rm{negative}} + {\rm{False}}\;{\rm{negative}}}} \times {\rm{100\% }}\end{equation}

RESULTS AND DISCUSSION

A3 assay versus microbial counts

The A3 values and microbial counts collected from 362 food contact surfaces from 15 food processing plants and commercial kitchens in Malaysia showed a weak positive correlation (R2 = 0.157; Fig. 1). Although some studies have reported a relatively better correlation between ATP and microbial count (1, 16, 17), the weak correlation observed in this study is not surprising, because the food contact surfaces tested in this study were from food processing plants, with open and complex environments, rather than a simulated study conducted in a laboratory under a controlled environment. Despite the weak correlation obtained in this study, it was evident that when the A3 value increased, it was more likely to detect a higher count of microorganisms on the food contact surfaces in food processing plants (Fig. 1). However, it was also observable from the scattered plot of A3 values versus APCs (Fig. 1) that low APCs were also detected over a range of A3 values from low to high. Occasionally, some data points with a low A3 value but a high microbial load (>10,000 CFU/100 cm2) were detected (Fig. 1). These contributed to a weaker correlation between overall A3 values and APCs, as well as the efficiency of the A3 assay.

FIGURE 1

Scatterplot of the aerobic plate count (in CFU per 100 cm2) versus the A3 value (in RLU) of 362 food contact surfaces in 15 food processing plants. The horizontal dotted line is the pass–fail threshold of the microbial count for cleanliness of food contact surfaces at 250 CFU/100 cm2.

FIGURE 1

Scatterplot of the aerobic plate count (in CFU per 100 cm2) versus the A3 value (in RLU) of 362 food contact surfaces in 15 food processing plants. The horizontal dotted line is the pass–fail threshold of the microbial count for cleanliness of food contact surfaces at 250 CFU/100 cm2.

The occasionally detected low A3 values with a high microbial load could result from the low concentration of food debris and organic matter on surfaces but a high number of microbial cells that adhered strongly to the food contact surfaces and were resistant to the sanitizing chemicals use in the cleaning and sanitizing regime. Studies have shown that bacteria could rapidly (in about 1 min) form reversible attachments via hydrodynamic and electrostatic interactions upon contacting surfaces (5, 6); given enough time (several hours), the bacteria would strengthen the attachment via van der Waals interactions to form irreversible attachments to the contact surfaces (30). Therefore, if the sanitizing agent failed to inactivate the attached microbial cells, these cells would continue to grow into biofilms that could only be removed mechanically.

However, high A3 values with the low or zero APC observed in this study could be due to two possibilities: (i) the surface had high amounts of food residue but no microbial cells, or (ii) the APC method failed to detect bacteria, which could be viable but nonculturable (VBNC), on surfaces. Increasing scientific evidence shows that bacteria enter a VBNC state under stress and in challenging environments, such as in food processing environments (e.g., heat treatment, low pH, freezing, preservation, and exposure to antimicrobial agents) (9, 10, 15, 27, 33). Furthermore, research found that many foodborne pathogens, such as pathogenic E. coli (27, 33), Salmonella (20, 23), Vibrio (11), L. monocytogenes (20), and Campylobacter (9), could transform into a VBNC state and hence escape detection by conventional culturing methods. These VBNC pathogens could later resuscitate under suitable conditions to pose a public health risk (11, 19, 28). In addition, surfaces with a high amount of food residue but a limited number or absence of microbes could still serve as hotspots with sufficient nutrients to support active growth of microorganisms and pose a risk of contamination to the final food products. Therefore, a high A3 value despite a low microbial count indicates a food safety risk that must not be neglected.

A3 values and microbial counts on food contact surfaces in different food processing plants and commercial kitchens

Although the overall A3 value showed a weak positive correlation with the microbial load (Fig. 1), the trend was inconsistent across the 15 food processing plants and commercial kitchens studied (Fig. 2). Among the 15 food processing plants and commercial kitchens, a strong and significant positive correlation between the A3 value and the microbial count on food contact surfaces was observed in FMP02 (meat and chicken processing plant; R2 = 0.582), FMP04 (paratha processing plant; R2 = 0.686), FMP06 (durian products processing plant; R2 = 0.500), and FMP14 (commercial kitchen producing ready-to-eat [RTE] sushi; R2 = 0.579). In contrast, no significant relationship between the A3 value and the microbial count could be observed for FMP08 (dry spice processing plant), FMP09 (condensed milk processing plant), FMP10 (dairy processing plant), FMP12 (instant coffee processing plant), FMP13 (commercial kitchen producing RTE salad), and FMP15 (commercial kitchen producing RTE meal boxes) (Fig. 2). The results suggested that the A3 assay was probably most useful as a hygiene indicator for food contact surfaces in food processing plants handling raw foods such as meat, seafood, and fruit that are prone to carrying a higher number of microorganisms. In contrast, the A3 value did not perform well as a microbial indicator for food processing plants or food contact surfaces dealing with low-risk ingredients with an expected low number of microorganisms. In such cases, the A3 value contributed by the microbial cells was insignificant relative to the concentration of the organic matter.

FIGURE 2

Scatterplot of the aerobic plate count (in CFU per 100 cm2) versus the A3 value (in RLU) of 362 food contact surfaces for each of the 15 food processing plants. The horizontal dotted line is the pass–fail threshold of the microbial count for cleanliness of food contact surfaces at 250 CFU/100 cm2.

FIGURE 2

Scatterplot of the aerobic plate count (in CFU per 100 cm2) versus the A3 value (in RLU) of 362 food contact surfaces for each of the 15 food processing plants. The horizontal dotted line is the pass–fail threshold of the microbial count for cleanliness of food contact surfaces at 250 CFU/100 cm2.

The types of sanitizing chemical used in each plant or commercial kitchen did not seem to affect the A3 values and microbial count (Fig. 2). Alcohol (n = 4) and quaternary ammonium (n = 5) were the most commonly applied sanitizing chemicals among the 15 food processing plants and commercial kitchens included in this study. Other types of sanitizers used by these plants and kitchens included Oxonia (hydrogen peroxide, acetic acid, and para-acetic acid), Vortexx (acetic acid, hydrogen peroxide, para-acetic acid, secondary alkane sulphonates, and octanoic acid), chlorine dioxide, and hot water (Table 1).

ROC analysis

For the A3 assay to be useful as a microbial indicator, it is imperative to determine the A3 threshold value for the detection of microbial-contaminated food contact surfaces (i.e., APC > 250 CFU/cm2). In the sanitation monitoring experiment conducted by Bakke and Suzuki (3), the threshold value of the A3 assay as a hygiene indicator of surfaces was determined via a cleaning simulation experiment with meat-exposed stainless steel plates. They found that the A3 value of thoroughly cleansed and disinfected stainless steel was below 200 RLU and thus suggested 200 RLU as the benchmark value. However, when we calculated the sensitivity, specificity, and efficiency of the A3 assay based on 200 RLU as the benchmark value, the specificity (0.516) and efficiency (58.56%) were found to be relatively low and thus offset the high sensitivity (0.827) of the A3 assay as a reliable indicator. Hence, in this study, ROC analysis, which has recently gained popularity in clinical diagnostic test assessments (13, 24, 29, 34), was adopted to determine the optimal threshold of the A3 assay as a hygiene indicator in a food processing environment. Determining and selecting the optimal cutoff or threshold value of a particular test assay involves optimizing between sensitivity and specificity. There is no standard way to decide how a threshold value is selected. Some clinical applications might emphasize high sensitivity (i.e., ability to correctly detect positive cases) over specificity (i.e., ability to correctly identify negative cases), whereas others might emphasize high specificity (4). In this study, we assumed that the cost of a false-positive result was equal to the cost of a false-negative result (i.e., sensitivity was equal to specificity); therefore, 500 RLU was selected as the optimum threshold value through ROC analysis (Fig. 3). In this case, false-positive results suggested that food residue was high on the surface when the microbial count was no more than 250 CFU/100 cm2, whereas false-negative results suggested that food residue was low on a surface when the microbial count was more than 250 CFU/100 cm2. Although it could seem like the cost of false negatives was higher, surfaces with a high amount of food residue but a limited number or absence of microbes could serve as hotspots with sufficient nutrients to support active growth of microorganisms and hence pose a risk of contamination to the final food product. Thus, both false-positive and false-negative cases were assumed to have equal cost in this study. Shifting the threshold value from 200 to 500 RLU significantly improved the efficiency of the A3 assay, from 58.56 to 72.1%. Based on the A3 threshold value at 500 RLU, 72.8% (sensitivity) of food contact surfaces with APC > 250 CFU/100 cm2 would have A3 > 500 RLU and were correctly classified as microbial-contaminated surfaces, whereas 71.9% (specificity) of food contact surfaces with APC ≤ 250 CFU/100 cm2 would have A3 < 500 RLU and were correctly categorized as clean surfaces (Table 2).

FIGURE 3

ROC curve based on the A3 value obtained in the analysis of 362 food contact surfaces in 15 food processing plants for indication of hygiene (microbial load threshold at 250 CFU/100 cm2). The area under the curve is 0.759. Threshold value of 398 RLU (sensitivity, specificity) = (0.753, 0.633); threshold value of 631 RLU (sensitivity, specificity) = (0.654, 0.751).

FIGURE 3

ROC curve based on the A3 value obtained in the analysis of 362 food contact surfaces in 15 food processing plants for indication of hygiene (microbial load threshold at 250 CFU/100 cm2). The area under the curve is 0.759. Threshold value of 398 RLU (sensitivity, specificity) = (0.753, 0.633); threshold value of 631 RLU (sensitivity, specificity) = (0.654, 0.751).

TABLE 2

Comparison of conventional hygiene indicators based on microbial count (aerobic plate count), A3 assay with manufacturer's recommended threshold at 200 RLU, and optimal threshold based on ROC analysis in this study at 500 RLU in classifying food contact surfaces as clean or contaminated

Comparison of conventional hygiene indicators based on microbial count (aerobic plate count), A3 assay with manufacturer's recommended threshold at 200 RLU, and optimal threshold based on ROC analysis in this study at 500 RLU in classifying food contact surfaces as clean or contaminated
Comparison of conventional hygiene indicators based on microbial count (aerobic plate count), A3 assay with manufacturer's recommended threshold at 200 RLU, and optimal threshold based on ROC analysis in this study at 500 RLU in classifying food contact surfaces as clean or contaminated

However, 72.8% sensitivity also meant that 27.2% of the surfaces with APC > 250 CFU/100 cm2 would be wrongly identified by the A3 assay as clean, and this is unacceptable if the A3 assay is to be used to detect contaminated surfaces. Further analysis showed that of the 81 contaminated surfaces (APC > 250 CFU/100 cm2), 22 surfaces (27.2%) were wrongly identified as clean by the A3 assay (≤500 RLU). The frequency of false negatives was found to be high in FMP01, FMP07, and FMP09 (buffer tank) (Fig. 4). Except for FMP09, most false negatives could be reduced by assigning a lower A3 threshold value that was specific to the sampling area. Therefore, based on the results and observations in this study, although a threshold of 500 RLU could be applied to most surface areas for hygiene monitoring, preliminary screening to compare the A3 reading with the APC for each site is strongly recommended to increase the reliability of the A3 assay as a microbial indicator for food contact surfaces.

FIGURE 4

Scatterplot of the aerobic plate count (in CFU per 100 cm2) versus the A3 value (RLU) of food contact surfaces with false-negative results.

FIGURE 4

Scatterplot of the aerobic plate count (in CFU per 100 cm2) versus the A3 value (RLU) of food contact surfaces with false-negative results.

In conclusion, the A3 assay could be a potential alternative to the conventional culture-based approach for hygiene monitoring of food contact surfaces to better manage food safety risks in food processing environments, particularly those that involve the processing of raw meat and raw food ingredients such as fruit that harbor high microbial loads. Based on the ROC analysis, 500 RLU was set as an optimal threshold for the A3 value to balance between sensitivity and specificity at 0.728 and 0.719, respectively. The A3 assay as a hygiene indicator for food contact surfaces has an efficiency of 72.1%. However, precautions need to be taken when A3 is used as a microbial indicator in food processing plants, because certain sites or surfaces could have very low sensitivity due to the types of food surfaces, food matrices, and microorganisms. Hence, it is suggested that a preliminary study be conducted before the adoption of the A3 assay as an alternative to the conventional culture-based approach for hygiene monitoring of food contact surfaces.

ACKNOWLEDGMENTS

This study was a collaboration between University of Malaya and Everest 23 Professional Solutions to evaluate the performance of the A3 assay for applications in food industries in Malaysia. Everest 23 is a food safety training company. The LuciPac A3 Surface kits and microbial swabs were sponsored by Kikkoman Biochemifa. The Lumitester PD-30 was owned by Everest 23 and was lent to the research team for the purpose of this study. The microbial analysis was conducted and supported by the IPPP Infra Microbiology Laboratory of the University of Malaya.

REFERENCES

1.
Amodio,
E.,
Cannova
L.,
Villafrate
M. R.,
Merendino
A. M.,
Aprea
L.,
and
Calamusa
G.
2014
.
Analytical performance issues: comparison of ATP bioluminescence and aerobic bacterial count for evaluating surface cleanliness in an Italian hospital
.
J. Occup. Environ. Hyg
.
11
(2)
:
D23
D27
.
2.
Annette,
E. C.,
Rajagopal
R.,
Lauer
J.,
and
Allwood
P.
2011
.
Assessment of hygienic quality of surfaces in retail food service establishments based on microbial counts and real-time detection of ATP
.
J. Food Prot
.
74
:
686
690
.
3.
Bakke,
M.,
and
Suzuki
S.
2018
.
Development of a novel hygiene monitoring system based on the detection of total adenylate (ATP+ADP+AMP)
.
J. Food Prot
.
81
:
729
737
.
4.
Biagini,
R. E.,
Krieg
E. F.,
Pinkerton
L. E.,
and
Hamilton
R.
2001
.
Receiver operating characteristics analyses of Food and Drug Administration–cleared serological assays for natural rubber latex-specific immunoglobulin E antibody
.
Clin. Diagn. Lab. Immunol
.
8
:
1145
1149
.
5.
Boks,
N. P.,
Busscher
H. J.,
van der Mei
H. C.,
and
Norde
W.
2008
.
Bond-strengthening in staphylococcal adhesion to hydrophilic and hydrophobic surfaces using atomic force microscopy
.
Langmuir
24
:
12990
12994
.
6.
Boks,
N. P.,
Kaper
H. J.,
Norde
W.,
Busscher
H. J.,
and
van der Mei
H. C.
2008
.
Residence time dependent desorption of Staphylococcus epidermidis from hydrophobic and hydrophilic substrata
.
Colloids Surf. B Biointerfaces
67
:
276
278
.
7.
Buckalew,
J.
1996
.
Surface sanitation and microbiological food quality of a university foodservice operation
.
J. Food Serv. Syst
.
9
:
25
39
.
8.
Carpentier,
B.,
and
Cerf
O.
2011
.
Review—persistence of Listeria monocytogenes in food industry equipment and premises
.
Int. J. Food Microbiol
.
145
:
1
8
.
9.
Chai,
L. C.,
Robin
T.,
Ragavan
U. M.,
Gunsalam
J. W.,
Bakar
F. A.,
Ghazali
F. M.,
Radu
S.,
and
Kumar
M. P.
2007
.
Thermophilic Campylobacter spp. in salad vegetables in Malaysia
.
Int. J. Food Microbiol
.
117
:
106
111
.
10.
Colwell,
R. R.,
Brayton
P. R.,
Grimes
D. J.,
Roszak
D. R.,
Huq
S. A.,
and
Palmer
L. M.
1985
.
Viable but non-culturable Vibrio cholerae and related pathogens in the environment: implication for release of genetically engineered microorganisms
.
Biotechnology
3
:
817
820
.
11.
Colwell,
R. R.,
Brayton
P. R.,
Herrington
D.,
Tall
B. D.,
Huq
A.,
and
Levine
M. M.
1996
.
Viable but nonculturable Vibrio cholerae O1 revert to a culturable state in human intestine
.
World J. Microb. Biotechnol
.
12
:
28
31
.
12.
Cooper,
R.,
Griffith
C.,
Malik
R.,
Obee
P.,
and
Looker
N.
2007
.
Monitoring the effectiveness of cleaning in four British hospitals
.
Am. J. Infect. Control
35
:
338
341
.
13.
Darmon,
M.,
and
Vincent
J.
2011
.
Diagnostic performance of fractional excretion of urea in the evaluation of critically III patients with acute kidney injury: a multicenter cohort study
.
Crit. Care
15
:
R178
.
14.
Davidson,
C. A.,
Griffith
C. J.,
Peters
A. C.,
and
Fielding
L. M.
1999
.
Evaluation of two methods for monitoring surface cleanliness—ATP bioluminescence and traditional hygiene swabbing
.
Luminescence
14
:
33
38
.
15.
Dukan
S.,
Levi
Y.,
and
Touati
D.
1997
.
Recovery of culturability of an HOCl-stressed population of Escherichia coli after incubation in phosphate buffer: resuscitation or regrowth?
Appl. Environ. Microbiol
.
63
:
4204
4209
.
16.
Eed,
H. R.,
Abdel-Kader
N. S.,
El Tahan
M. H.,
Dai
T.,
and
Amin
R.
2016
.
Bioluminescence-sensing assay for microbial growth recognition
.
J. Sens.
2016
:
1492467
.
17.
Gibbs,
S. G.,
Sayles
H.,
Colbert
E. M.,
Hewlett
A.,
Chaika
O.,
and
Smith
P. W.
2014
.
Evaluation of the relationship between the adenosine triphosphate (ATP) bioluminescence assay and the presence of Bacillus anthracis spores and vegetative cells
.
Int. J. Environ. Res. Public Health
11
:
5708
5719
.
18.
Government of Canada.
2016
.
Listeriosis
.
19.
Gupte,
A. R.,
de Rezende
C. L. E.,
and
Joseph
S. W.
2003
.
Induction and resuscitation of viable but nonculturable Salmonella enterica serovar Typhimurium DT104
.
Appl. Environ. Microbiol
.
69
:
6669
6675
.
20.
Highmore,
C. J.,
Warner
J. C.,
Rothwell
S. D.,
Wilks
S. A.,
and
Keevil
C. W.
2018
.
Viable-but-nonculturable Listeria monocytogenes and Salmonella enterica serovar Thompson induced by chlorine stress remain infectious
.
mBio
9
(2)
:
e00540-18.
21.
Kummar,
R.,
and
Indrayan
A.
2011
.
Receiver operating characteristic (ROC) curve for medical researchers
.
Indian Pediatr
.
48
:
277
289
.
22.
Leong,
D.,
NicAogáin
K.,
Luque-Sastre
L.,
McManamon
O.,
Hunt
K.,
Alvarez-Ordóñez
A.,
Scollard
O.,
Schmalenberger
A.,
Fanning
S.,
O'Byrne
C.,
and
Jordan
K.
2017
.
A 3-year multi-food study of the presence and persistence of Listeria monocytogenes in 54 small food businesses in Ireland
.
Int. J. Food Microbiol
.
249
:
18
26
.
23.
Liao,
H.,
Jiang
L.,
and
Zhang
R.
2018
.
Induction of a viable but non-culturable state in Salmonella Typhimurium by thermosonication and factors affecting resuscitation
.
FEMS Microbiol. Lett.
365
(2)
:
fnx249
.
24.
Lloyd,
C. J.
1998
.
Using smooth receiver operating characteristic curves to summarize and compare diagnostic systems
.
J. Am. Stat. Assoc
.
93
:
1356
1364
.
25.
Moore,
G.,
and
Griffith
C.
2002
.
A comparison of traditional and recently developed methods for monitoring surface hygiene within the food industry: an industry trial
.
Int. J. Environ. Health Res
.
12
:
317
329
.
26.
Nauta,
M.
2010
.
Campylobacter transfer from naturally contaminated chicken thighs to cutting boards is inversely related to initial load,” a comment on: J. Food Prot. 72(9):1836–1840 (2009)
.
J. Food Prot
.
73
:
6
7
.
27.
Oliver,
J. D.
2010
.
Recent findings on the viable but nonculturable state in pathogenic bacteria
.
FEMS Microbiol. Rev
.
34
:
415
425
.
28.
Panntdaporn,
N.,
Kawamoto
K.,
Asakura
H.,
and
Makino
S.-I.
2006
.
Resuscitation of the viable but non-culturable state of Salmonella enterica serovar Oranienburg by recombinant resuscitation-promoting factor derived from Salmonella Typhimurium strain LT2
.
Int. J. Food Microbiol
.
106
:
241
247
.
29.
Reddy,
S.,
Dutta
S.,
and
Narang
A.
2008
.
Evaluation of lactate dehydrogenase, creatine kinase and hepatic enzymes for retrospective diagnosis of perinatal asphyxia among sick neonates
.
Indian Pediatr
.
45
:
144
147
.
30.
Renner,
L. D.,
and
Weibel
D. B.
2011
.
Physicochemical regulation of biofilm formation
.
MRS Bull
.
36
:
347
355
.
31.
Rodrigues,
C.,
de Sá
C.,
and
de Melo
C.
2017
.
An overview of Listeria monocytogenes contamination in ready to eat meat, dairy and fishery foods
.
Cienc. Rural
47
(2)
:
1
8
.
32.
Schmidt,
R.
2012
.
Food equipment hygienic design: an important element of a food safety program
.
Food Safety Magazine
December 2012/January 2013.
33.
Xu,
H.-S.,
Roberts
N.,
Singleton
F. L.,
Attwell
R. W.,
Grimes
D.,
and
Colwell
R. R.
1982
.
Survival and viability of nonculturable Escherichia coli and Vibrio cholerae in the estuarine and marine environment
.
Microb. Ecol
.
8
:
313
323
.
34.
Zhou,
X. H.,
Obuchowski
N. A.,
and
McClish
D. K.
2002
.
Statistical methods in diagnostic medicine
.
Wiley & Sons
,
New York
.
35.
Zottola,
E. A.
1994
.
Microbial biofilms in the food processing industry: should they be a concern?
Int. J. Food Microbiol
.
23
:
125
148
.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/cc-by-nc-nd/4.0/)