Using Indicator Bacteria and <I>Salmonella</I>

2732

Journal of Food Protection, Vol. 70, No. 12, 2007, Pages 2732–2740 Copyright ᮊ, International Association for Food Protection

Using Indicator Bacteria and Salmonella Test Results from Three Large-Scale Beef Abattoirs over an 18-Month Period To Evaluate Intervention System Efﬁcacy and Plan Carcass Testing for Salmonella

JOHN R. RUBY,1 JUN ZHU,2 AND STEVEN C. INGHAM1*

1 2 Department of Food Science and Department of Statistics, University of Wisconsin–Madison, Madison, Wisconsin 53706, USA Downloaded from http://meridian.allenpress.com/jfp/article-pdf/70/12/2732/1678123/0362-028x-70_12_2732.pdf by guest on 29 September 2021

MS 07-140: Received 16 March 2007/Accepted 22 July 2007

ABSTRACT

To develop a process for predicting the likelihood of Salmonella contamination on beef carcasses, we evaluated the influence of several possible causative factors (i.e., year, abattoir, day of week, month, and intervention system components) on the risk of Salmonella and indicator organism contamination. Hide and carcass sponge samples were collected in 2005 to 2006 in six steps at three abattoirs in the East (A), Midwest (B), and Southwest (C) United States. Each abattoir used the same intervention system. Samples were analyzed for aerobic plate counts (APCs; n ϭ 18,990) and Enterobacteriaceae counts (EBCs; n ϭ 18,989) and the presence or absence of Salmonella (n ϭ 5,355). Our results demonstrated that many factors play a significant role in the level of microbial contamination of beef carcasses. Overall, Salmonella prevalence and EBC levels were significantly higher in 2006 than in 2005. APCs and EBCs were highest in abattoirs A (3.57 log CFU/100 cm2) and B (1.31 log CFU/100 cm2). The odds of detecting a positive Salmonella isolate were greatest in abattoir C and lowest in abattoir A. Across the three abattoirs, the overall intervention process effectively reduced microbiological contamination. Salmonella prevalence fell from 45% (preevisceration) to 0.47% (postchilled–lactic acid), and there were APC and EBC reductions of 5.43 and 5.28 log CFU/100 cm2, respectively, from hide-on to postchilled–lactic acid samples. At each abattoir, composites of three individual EBC-negative carcass samples yielded Salmonella-negative results 97 to 99% of the time. These results suggest the possibility of using indicator test results to accurately predict the absence of Salmonella in a beef carcass sample.

Each year in the United States, there are an estimated tions have been put into place at commercial beef process- 1.4 million salmonellosis cases and 400 deaths caused by ing facilities. Many beef processors now apply multiple in- Salmonella infections (28). Ground beef has been impli- tervention technologies, i.e., water washes, organic acid cated as the vehicle for transmission of Salmonella in mul- washes, carcass trimming, and steam vacuuming (4), be- tiple foodborne outbreaks (9–11, 13). The U.S. Food Safety cause the literature has demonstrated that the application of and Inspection Service has documented a decrease in the multiple interventions has greater potential to reduce mi- prevalence of Salmonella in ground beef, from 7.5% of croorganisms than interventions applied individually (7, 8, samples in 1996 to 1.6% in 2004 (24, 25); however, out- 14, 17, 19). breaks of human Salmonella infections associated with Even though pathogen contamination is typically the ground beef continue to occur. Even though beef processing greatest food safety concern associated with beef slaughter, plant practices adhere to current U.S. Food Safety and In- fabrication, and production of raw products, there are draw- spection Service production guidelines, with the history of backs to the use of pathogen testing for evaluating process multiple ground beef–associated outbreaks (9–11) and the control. These drawbacks include the low prevalence of recent ﬁrst multistate outbreak of multidrug-resistant Sal- pathogenic bacteria cells found on beef carcasses and raw monella Typhimurium phage type DT104 associated with products (6) and the costs associated with pathogen testing. consumption of store-bought ground beef (13), there are Researchers have studied the reduction of indicator organ- increased governmental concerns regarding whether current isms, e.g., aerobic plate counts (APCs), generic E. coli, total critical control points (i.e., preventive measures to control coliforms, and Enterobacteriaceae counts (EBCs), as a food safety hazards) and pathogen reduction strategies are measure of intervention system efﬁcacy (7, 20). Others adequate for Salmonella control (12). have reported that limited sampling data preclude the es- Because Salmonella colonizes the gastrointestinal tablishment of a direct relationship between indicator or- tracts of cattle, it is believed that carcasses can become ganisms and pathogens (3). Indicator bacteria analyses cost contaminated with Salmonella during slaughter operations. less than pathogen testing and may alert processors to sit- To decrease carcass contamination, antimicrobial interven- uations in which fecal contamination is occurring, regard- less of whether pathogens are involved. * Author for correspondence. Tel: 608-265-4801; Fax: 608-262-6872; In this study, on-line carcass sponge samples were col- E-mail: [email protected]. lected from three commercial beef abattoirs located in geo- J. Food Prot., Vol. 70, No. 12 SALMONELLA AND INDICATOR BACTERIA ON BEEF CARCASSES 2733

FIGURE 1. Flow diagram of beef slaughter intervention system. Sampling steps are numbered: (1) hide on, (2) preintervention, (3) prepasteurization–lactic acid spray, (4) postpasteurization–lactic acid, (5) postchill, and (6) postchilled–lactic acid. Downloaded from http://meridian.allenpress.com/jfp/article-pdf/70/12/2732/1678123/0362-028x-70_12_2732.pdf by guest on 29 September 2021

graphically different regions of the United States. Samples vacuuming and trimming. Between sampling steps 3 and 4, the were collected at various locations throughout the beef pro- carcasses were treated with an ambient-temperature water wash duction process over an 18-month period. Sample analyses and then a pasteurization water wash (target water temperature, Ͼ Њ included enumeration of indicator organisms (APCs and 82.2 C) and a 4.0 to 5.0% lactic acid spray. Between sampling EBCs) and qualitative analysis for Salmonella. The overall steps 4 and 5, the carcasses were in a cooler and subjected to intermittent spray chilling for 36 to 48 h. Between sampling steps objective of this project was to develop a predictive tool 5 and 6, 4.0 to 5.0% lactic acid was again sprayed on the carcass for plant-specific or perhaps widespread use that would al- immediately before it was moved to the fabrication room. low operators of large-scale abattoirs to evaluate specific indicator test results and operational parameters in terms of Sample collection procedure. Samples were collected daily the likelihood of Salmonella contamination. The specific over an 18-month period beginning the first week of March 2005 objectives of the study were to (i) determine the relation- and ending mid-September 2006. On each sampling day, at each ships existing between the presence of Salmonella, the level sample step, three carcasses were randomly selected and sampled. of APCs, and the presence and levels of EBCs on beef Each hide was sampled with one Speci-Sponge (Nasco, Fort At- kinson, Wis.) moistened with 25 ml of buffered peptone diluent carcasses, and (ii) determine the influence of several pos- (Difco, Becton Dickinson, Sparks, Md.), and each carcass was sible causative factors (i.e., geographic region, day of week, sampled with two Speci-Sponges, each moistened with 25 ml of month, and intervention treatments) on the risk of Salmo- buffered peptone diluent. Excess diluent was removed from each nella and indicator contamination. sponge inside the Speci-Sponge bag before sampling. Sampling was performed as described by Arthur et al. (3) and consisted of MATERIALS AND METHODS sampling an area approximately 8,000 or 100 cm2 on the carcass Sample collection steps and microbial interventions. Car- or hide, respectively. To sample such a large area on the carcass, cass sponge samples were collected from three commercial beef two sponges were used: sponge 1 sampled the inside and outside abattoirs located in geographically different regions of the United round (ϳ4,000 cm2), and sponge 2 sampled the navel-plate-bris- States: East (A), Midwest (B), and Southwest (C). As shown in ket-shank area (ϳ4,000 cm2). Sample sponges were stored in in- Figure 1, samples were collected at six steps throughout the pro- sulated coolers containing ice packs and were transported to the cess: (1) after the hide was split at the midline but before complete laboratory. Microbiological analysis was done within 24 h of the hide removal ϭ hide on, (2) after complete hide removal but prior samples being obtained at the plant. to any antimicrobial intervention ϭ preintervention, (3) after evis- ceration, on-line carcass trimming, and final inspection but before Microbiological analysis. Samples from all steps (1 through carcass washes ϭ prepasteurization–lactic acid spray, (4) after the 6) were analyzed for APCs and EBCs, but only those samples pasteurization and lactic acid sprays but before carcass chilling ϭ collected from the carcass (steps 2 through 6) were analyzed for postpasteurization–lactic acid, (5) after chilling the carcass for 36 Salmonella. to 48 h postmortem ϭ postchill, and (6) after spraying the chilled Sponges were pummeled by stomaching at normal speed for carcasses with lactic acid before going into fabrication ϭ post- 1 min with a Stomacher 400 (Seward, Fisher Scientific, Itasca, chilled–lactic acid. Ill.). For carcass samples, the two sponges were placed in a single At each of the three abattoirs, the microbial intervention sys- sample bag before stomaching. After stomaching, a 2-ml aliquot tems were the same. Before hide-on sampling (step 1), the hide was removed, with 1 ml for APC and 1 ml for EBC analysis. was opened on the rear shanks and the midline, but no interven- Then, six sponges (two sponges per carcass ϫ three carcasses per tion was applied directly to the hide prior to sampling. Once the step) were placed in a single sample bag and enriched for Sal- hide was completely removed, prior to sampling at step 2, carcass monella detection. trimming occurred, and the hide removal cut pattern lines on the APCs were determined by the Spiral Plate Method as de- carcass were treated with steam vacuuming. Between sampling scribed in the U.S. Food and Drug Administration Bacteriological steps 2 and 3, the preeviscerated carcasses were sprayed with 4.0 Analytical Manual (26). Spiral plating was performed by plating to 5.0% lactic acid, eviscerated, and subjected to further steam 100 ␮l of appropriately diluted samples on plate count agar (Dif- 2734 RUBY ET AL. J. Food Prot., Vol. 70, No. 12 co, Becton Dickinson) with an Autoplate 4000 (Spiral Biotech, statistical analysis. For both APCs and EBCs, minimum detectable Norwood, Mass.). The agar plates were then incubated at 35 Ϯ limits for carcass samples and hide samples were 0.625 and 25 1ЊC for 48 Ϯ 3 h. After incubation, enumeration was performed CFU/100 cm2, respectively. When carcass samples (steps 2 as instructed by the manufacturer. Total EBCs were determined through 6) or hide samples (step 1) did not produce any colonies, by plating appropriate dilutions onto 3M Petrifilm Enterobacteri- values of Ϫ0.51 log/100 cm2 (0.3125 CFU/100 cm2 less than the aceae Count Plates (3M Microbiology, St. Paul, Minn.) and in- detection limit) or 1.09 log/100 cm2 (12.5 CFU/100 cm2 less than cubating at 37 Ϯ 1ЊC for 24 Ϯ 2 h. After the Petrifilm was in- the detectable limit) were used, respectively, for statistical anal- cubated, enumeration was performed by selecting plates with iso- ysis. For evaluating the influence of the factors tested and their lated colonies ranging in numbers from 10 to 150 and counting interactions on the amount of microbial contamination, an analysis all red colonies that produced gas, acid (yellow zone surrounding of variance was done (SAS PROC GLM, SAS Institute, Cary, colony), or both. N.C. (22)). For each factor, except for interactions involving both Salmonella detection was performed by first screening the month and day of week, all pairwise comparisons were conducted samples (each of which consisted of liquid expressed from six by the least significant difference procedure. Odds ratios for Sal- sponges) with the automated VIDAS Salmonella enzyme-linked monella were determined by SAS PROC LOGISTIC (SAS Insti- Downloaded from http://meridian.allenpress.com/jfp/article-pdf/70/12/2732/1678123/0362-028x-70_12_2732.pdf by guest on 29 September 2021 fluorescent assay (bioMe´rieux-USA, Durham, N.C.) (2). Each tute (22)) on the factors tested. For evaluating the relationship sample was enriched with 225 ml of buffered peptone diluent and between the presence of Salmonella and the levels of APCs and incubated at 35 Ϯ 1ЊC for 18 to 24 h. After initial enrichment, EBCs, sample averages for APCs and EBCs were calculated for 0.1 and 1.0 ml were transferred in parallel to 10 ml of Rappaport- each sampling day at each step. To evaluate the relationship be- Vassiliadis broth (Difco, Becton Dickinson) and 10 ml of tetra- tween the presence of Salmonella and the presence of Enterobac- thionate broth supplemented with iodine (Difco, Becton Dickin- teriaceae, the averaged EBC values were converted to binary EBC son), respectively. Inoculated Rappaport-Vassiliadis broth and te- data (positive ϭ detected CFU; negative ϭ no CFU). Sample trathionate broth supplemented with iodine were incubated at 41 steps in which Salmonella results did not correspond with an APC to 42ЊC for 18 to 24 h. After incubation, 1.0 ml of Rappaport- and EBC result were excluded from the analysis (n ϭ 127 of Vassiliadis broth and 1.0 ml of tetrathionate broth supplemented 5,355 Salmonella data points were excluded). The relationships with iodine were each transferred into 9 ml of M-broth (Difco, existing between the presence of Salmonella, the level of APC, Becton Dickinson) and incubated at 41 to 42ЊC for 6 to 8 h. After and the presence and levels of EBC were evaluated and analyzed incubation, 1.0 ml from each M-broth sample was transferred into by SAS PROC LOGISTIC (SAS Institute) (22). one test tube, heated for 15 Ϯ 1 min in a water bath at 95 to 100ЊC, and allowed to cool; then, the automated VIDAS Salmo- RESULTS AND DISCUSSION nella assay was performed. Samples positive on the enzyme- APC and EBC enumeration. Most factors (abattoir linked fluorescent assay screen were confirmed as Salmonella (27) ϭ P, year ϭ Y, month ϭ M, day of week ϭ DW, and by first streaking onto brilliant green sulfa agar supplemented with sample step ϭ SS) that were evaluated, and their interac- 0.1% sodium sulfapyridine (Difco, Becton Dickinson) and xylose tions, had a significant effect on APC and EBC. When com- lysine Tergitol 4 (XLT4) agar (Difco, Becton Dickinson) from the Rappaport-Vassiliadis and tetrathionate broth supplemented with bined data (all three abattoirs) were analyzed, all main ef- Ͻ iodine enrichments and incubating for 35 Ϯ 2ЊC for 18 to 24 h. fects were significant (P 0.05) for both APC and EBC, ϭ Three typical colonies from each of the incubated brilliant green except the Y for APC (P 0.18) and the DW for EBC (P sulfa agar supplemented with 0.1% sodium sulfapyridine and ϭ 0.73). All evaluated pairwise interactions for APC were XLT4 agar plates were selected for biochemical analysis. Typical significant (P Ͻ 0.05), except for Y ϫ DW. Interactions colonies on the brilliant green sulfa agar supplemented with 0.1% that did not significantly affect EBC were P ϫ DW, Y ϫ sodium sulfapyridine agar were pink and opaque with a smooth DW, P ϫ Y ϫ DW, DW ϫ SS, P ϫ DW ϫ SS, Y ϫ DW appearance and the entire edge surrounded by a red color in the ϫ SS, and P ϫ Y ϫ DW ϫ SS. In general, the main effects medium. Crowded brilliant green sulfa agar supplemented with and their interactions had a significant effect on the amount 0.1% sodium sulfapyridine plates will produce typical colonies of APC and EBC contamination of beef carcasses. that appear tan against the green background. Typical colonies When data from all six sampling steps were pooled, selected from the XLT4 agar plates appeared black or red with or significant differences (P Ͻ 0.05) in APCs and EBCs were without black centers with a red rim around the colony. Biochem- ical analysis was performed with triple sugar iron (Difco, Becton found between all three abattoirs. The APC was highest 2 Dickinson) and lysine iron agar (Difco, Becton Dickinson) slants. (3.57 log CFU/100 cm ) in abattoir A and second highest 2 Inoculation of the triple sugar iron and lysine iron agar slants was (3.12 log CFU/100 cm ) in abattoir B (Table 1). In contrast performed in parallel with a single colony by stabbing the butt to APC values, the EBC values were highest in abattoir B and streaking the top of the slant in one operation. Inoculated and second highest in abattoir A. For both APC and EBC, slants were incubated at 35 Ϯ 2ЊC for 24 Ϯ 2 h. Isolates that abattoir C had the lowest levels of APCs and EBCs (Table produced typical biochemical reactions on the triple sugar iron 1). and lysine iron agar slants were subjected to serological analysis When data from all three abattoirs were pooled, no with polyvalent (O) antiserum (serogroups A through I; Difco, statistically significant difference in APC was seen between Becton Dickinson). Isolates were considered confirmed with pos- 2005 and 2006 (Table 2). The mean value of EBCs in 2005 itive agglutination from the polyvalent (O) antiserum. was slightly, but significantly, less than the mean value Statistical analysis. Statistical analysis was performed on from 2006 (Table 2). The relatively narrow APC and EBC combined data collected at steps 1 through 6 from abattoirs A, B, 95% confidence intervals at the level of abattoir (Table 1), and C. Main effects included the following: abattoir ϭ P, year ϭ year, and month for all three abattoirs and sampling steps Y, month ϭ M, day of week ϭ DW, and sample step ϭ SS. The combined (Table 2) and the combination of sampling step APC and EBC data were converted to log CFU/100 cm2 prior to and abattoir (Table 3) suggest that operational practices and J. Food Prot., Vol. 70, No. 12 SALMONELLA AND INDICATOR BACTERIA ON BEEF CARCASSES 2735

TABLE 1. Aerobic plate counts (APCs) and Enterobacteriaceae steps 2 through 6, the highest mean APC occurred at ab- counts (EBCs) on beef carcasses across all years, months, and attoir A. Abattoir C had the lowest mean APC for sampling sampling steps: (1) hide on, (2) preintervention, (3) prepasteuri- steps 1 through 3, 5, and 6. Abattoir B had an intermediate zation–lactic acid spray, (4) postpasteurization–lactic acid, (5) mean APC for sampling steps 2, 3, 5, and 6. With the ex- postchill, and (6) postchilled–lactic acida ception of abattoirs A and B at sampling step 1, all differ- APCs EBCs ences in mean APC between abattoirs were statistically sig- niﬁcant (P Ͻ 0.05). In terms of mean EBC, abattoir C had Region n Mean 95% CI n Mean 95% CI the lowest value at sampling steps 2, 3, 5, and 6. Abattoirs East 5,001 3.57 Ab 3.51–3.64 5,001 1.14 A 1.09–1.20 A and B had the highest value for sampling steps 2, 4, and Midwest 6,340 3.12 B 3.07–3.18 6,339 1.31 B 1.26–1.36 6 and sampling steps 1, 3, and 5, respectively. Although Southwest 7,649 2.16 C 2.12–2.21 7,649 0.66 C 0.62–0.70 the same intervention system was used at all three abattoirs, a there were clearly important differences between abattoirs.

Comparisons were made between abattoirs. Values for mean and Downloaded from http://meridian.allenpress.com/jfp/article-pdf/70/12/2732/1678123/0362-028x-70_12_2732.pdf by guest on 29 September 2021 95% CI (confidence interval) are log CFU/100 cm2. These differences are likely not attributable to regional dif- b Values in the same column with the same letter are not signifi- ferences in animal suppliers, because there was consider- cantly different (P Ն 0.05). able sourcing overlap between the three abattoirs. The larg- est differences between abattoirs in mean APC and mean conditions remained consistent (i.e., proper Good Manufac- EBC were at sampling steps 2 and 3, perhaps reflecting turing Practices, Sanitation Standard Operating Procedures, differences in hide removal, trimming, or steam vacuuming. and other standard operating procedures were followed). Figure 2 illustrates the hide and carcass microbial pro- Consistent with the seasonal microbial patterns report- files (data combined for all three abattoirs) for APCs and ed in others’ research (23), APC values, averaged across EBCs at each of the six steps sampled in the process. For abattoirs, were generally higher in the warmer months (June both APCs and EBCs, values were significantly different through October) and lower in the colder months (Novem- between all six processing steps, with the exception of sim- ber through May; Table 2). More specifically, the mean ilar EBC values at postpasteurization–lactic acid and APC for December was the lowest and significantly differ- postchilled–lactic acid (sampling steps 4 and 6; Fig. 2). ent from all other months, with the exception of March, These results show that the overall intervention process, and the September mean APC value was the highest and from hide-on (sampling step 1) to postchilled–lactic acid significantly different from all months, except for January, samples (sampling step 6), effectively reduced microbio- June, and October. The mean EBC was lowest in Novem- logical contamination, as evidenced by average APC and ber, December, and March and highest in June and July EBC decreases of 5.43 and 5.28 log CFU/100 cm2, respec- (Table 2). tively. Contamination of the sterile carcass surface during When data were organized by sampling step and ab- the hide removal process resulted in preintervention (sam- attoir, several trends were apparent (Table 3). For sampling pling step 2) APC and EBC values of 3.10 and 1.04 log

TABLE 2. Aerobic plate counts (APCs) and Enterobacteriaceae counts (EBCs) on beef carcasses across all abattoirs and sampling steps: (1) hide on, (2) preintervention, (3) prepasteurization–lactic acid spray, (4) postpasteurization–lactic acid, (5) postchill, and (6) postchilled–lactic acida APCs EBCs

n Mean 95% CI n Mean 95% CI

Year 2005 10,041 2.85 Ab 2.81–2.90 10,040 0.97 A 0.93–1.10 2006 8,949 2.85 A 2.81–2.90 8,949 1.04 B 1.00–1.10 Month January 954 2.80 A 2.67–2.94 954 0.93 E 0.80–1.05 February 930 2.82 CD 2.68–2.96 930 0.98 CD 0.86–1.11 March 2,215 2.69 EF 2.60–2.78 2,215 0.82 F 0.74–0.90 April 2,045 2.73 E 2.64–2.83 2,045 0.96 DE 0.88–1.05 May 2,207 2.87 C 2.78–2.97 2,207 0.99 CD 0.91–1.08 June 2,167 3.00 AB 2.89–3.09 2,167 1.18 A 1.09–1.26 July 2,027 2.95 B 2.85–3.05 2,027 1.16 A 1.07–1.25 August 1,830 2.83 CD 2.73–2.94 1,829 1.03 C 0.94–1.12 September 1,648 3.04 A 2.94–3.15 1,648 1.10 B 1.00–1.20 October 1,020 2.99 AB 2.85–3.12 1,020 1.00 CD 0.89–1.12 November 1,037 2.75 E 2.62–2.87 1,037 0.85 F 0.74–0.97 December 910 2.68 F 2.54–2.81 910 0.86 F 0.73–0.98 a Comparisons were made between years and months. Values for mean and 95% CI (conﬁdence interval) are log CFU/100 cm2. b Values within a category in the same column with the same letter are not signiﬁcantly different (P Ն 0.05). 2736 RUBY ET AL. J. Food Prot., Vol. 70, No. 12

TABLE 3. Aerobic plate counts (APCs) and Enterobacteriaceae counts (EBCs) on beef carcassesa APCs EBCs Sampling step Abattoir U.S. region n Meanb 95% CIb,c n Meanb 95% CIb,c

1 A East 840 7.34 Ad 7.29–7.39 840 4.64 B 4.56–4.72 B Midwest 1,028 7.40 A 7.35–7.45 1,028 5.35 A 5.29–5.41 C Southwest 1,271 6.18 B 6.14–6.22 1,271 4.70 B 4.66–4.75 2 A East 841 4.61 C 4.56–4.67 841 1.71 D 1.64–1.78 B Midwest 1,071 3.27 E 3.21–3.33 1,071 1.30 D 1.25–1.36 C Southwest 1,272 1.95 I 1.90–1.99 1,272 0.38 F 0.35–0.42 3 A East 841 3.49 D 3.43–3.55 841 0.94 E 0.88–1.00 B Midwest 1,081 3.25 E 3.19–3.31 1,080 1.34 D 1.29–1.39

C Southwest 1,272 1.50 J 1.46–1.53 1,272 Ϫ0.06 H (0.09)e–(0.03) Downloaded from http://meridian.allenpress.com/jfp/article-pdf/70/12/2732/1678123/0362-028x-70_12_2732.pdf by guest on 29 September 2021 4 A East 840 1.26 K 1.21–1.32 840 Ϫ0.33 J (0.37)–(0.30) B Midwest 1,082 0.92 M 0.89–0.94 1,082 Ϫ0.49 L (0.50)–(0.48) C Southwest 1,272 1.09 L 1.06–1.12 1,272 Ϫ0.33 J (0.36)–(0.31) 5 A East 820 2.65 F 2.56–2.74 820 0.04 G (0.02)–0.10 B Midwest 1,038 2.48 G 2.42–2.54 1,038 0.91 E 0.84–0.97 C Southwest 1,281 1.26 K 1.22–1.30 1,281 Ϫ0.26 I (0.29)–(0.23) 6 A East 819 2.03 H 1.94–2.11 819 Ϫ0.21 I (0.26)–(0.17) B Midwest 1,040 1.54 J 1.49–1.60 1,040 Ϫ0.41 K (0.44)–(0.39) C Southwest 1,281 1.03 L 1.00–1.06 1,281 Ϫ0.47 KL (0.48)–(0.46) a Data were combined across years and months, with comparisons made between abattoirs and sampling steps: (1) hide on, (2) preintervention, (3) prepasteurization–lactic acid spray, (4) postpasteurization–lactic acid, (5) postchill, and (6) postchilled–lactic acid. b Log CFU/100 cm2. c CI, conﬁdence interval. d Values in the same column and the same sampling step with the same letter are not signiﬁcantly different (P Ն 0.05). e Parentheses denote negative numbers.

CFU/100 cm2, respectively. The subsequent interventions to 1.46 log CFU/100 cm2, but the final APC level for the applied to the carcass after the hide was removed (i.e., chilled carcasses was significantly higher than the APC lev- preevisceration wash, pasteurization wash, and lactic acid el on the carcasses immediately prior to chilling (sampling wash) further reduced the APCs and EBCs on the carcass step 4). A similar microbial pattern was seen with EBC immediately prior to the fabrication room (sampling step 6) counts, but the 5.0% lactic acid treatment on the chilled to 1.46 and Ϫ0.38 log CFU/100 cm2, respectively. Samples carcass (sampling step 6) reduced the mean EBC value to collected at postpasteurization–lactic acid (sampling step 4) the same level seen immediately before chilling (sampling produced the lowest APC mean values; however, signifi- step 4). The same trends were also seen within each indi- cantly larger APC values were found after chilling (sam- vidual abattoir (Table 3). pling step 6), which suggested regrowth, additional contam- Consistent with others’ research (16), both APC and ination during the chilling process, or both. The 5.0% lactic EBC counts increased during chilling. The increase in mean acid treatment on the chilled carcass reduced the APC value APC value was greater (0.95 log CFU/100 cm2) than for

FIGURE 2. Average APC (right-most of each pair of columns) and EBC (left-most of each pair of columns) values of beef hides or carcasses from combined data (abattoirs A [East], B [Midwest], and C [Southwest] for March 2005 to September 2006) at sampling steps (1) hide on, (2) preintervention, (3) prepasteurization–lactic acid spray, (4) postpasteurization–lactic acid, (5) postchill, and (6) postchilled– lactic acid. Columns having the same letter for a microbial group are not signiﬁcantly different (P Ն 0.05). J. Food Prot., Vol. 70, No. 12 SALMONELLA AND INDICATOR BACTERIA ON BEEF CARCASSES 2737

TABLE 4. Prevalence of Salmonella and Enterobacteriaceae on beef carcasses and the utility of Enterobacteriaceae-negative results as a negative screen for Salmonellaa Enterobacteriaceae Salmonella Utility as a negative Abattoir Region n % positive Odds ratio nb % positive screen for Salmonellac

A East 1,432 7.89 1.00 Ad 1,412 66.6 99.2 B Midwest 1,797 17.08 2.88 B 1,690 57.5 97.9 C Southwest 2,126 19.61 3.88 C 2,126 52.1 97.3 a Data are combined across years, months, and sampling steps: (2) preintervention, (3) prepasteurization–lactic acid spray, (4) postpasteurization–lactic acid, (5) postchill, and (6) postchilled–lactic acid. Comparisons were made between abattoirs. b Number of EBC results with a corresponding Salmonella result. Downloaded from http://meridian.allenpress.com/jfp/article-pdf/70/12/2732/1678123/0362-028x-70_12_2732.pdf by guest on 29 September 2021 c Percentage of Enterobacteriaceae-negative samples that were Salmonella negative. d Values in the same column with same letter are not significantly different (P Ն 0.05). the mean EBC value (0.58 log CFU/100 cm2). The differ- (Fig. 3). Abattoirs A and B had prevalence primarily in the ence seen in increases in levels of EBC and APC is most summer; however, samples collected in November and Jan- likely attributed to either non-EBC bacteria recovering from uary produced unexpectedly high prevalence (Fig. 3), in- any injury or growing at a faster rate than the EBC bacteria dicating both summer and winter seasons can be a problem. or the carcasses being contaminated with greater levels of The causes of seasonal variation in the incidence of food- non-EBC bacteria during chilling. Additional studies would borne pathogens are poorly understood, although other re- have to be performed to effectively evaluate whether researchers (5, 21, 23) have reported seasonal Salmonella growth, contamination, or combinations of both caused the prevalence patterns. Further investigations into climate var- increased bacterial levels. iation and beef production operation practices (i.e., farm Salmonella prevalence. All main effects had a statis- and feedlot hygienic practices and transportation conditions tically significant (P Ͻ 0.05) influence on the prevalence and distances) would be required to better understand the of Salmonella, except for the day of the week (P ϭ 0.079). reasons for the occurrence of higher than expected inci- Salmonella prevalence for abattoirs A, B, and C was 7.89, dences of Salmonella in the winter months at abattoirs A 17.08, and 19.61%, respectively (Table 4). Odds ratio anal- and B. ysis showed that a sample from abattoir C was 3.88 times Figure 4 shows the Salmonella incidence rates for beef and a sample from abattoir B was 2.88 times more likely carcasses from combined data (all three abattoirs) at each to be Salmonella positive than a sample from abattoir A of the sampled process steps. The intervention system, (Table 4). Others have reported a higher incidence of Sal- which included a preevisceration wash, a pasteurization monella in fecal samples on the pen floor from feedlots wash, and lactic acid washes, proved very effective at re- located in the Southwest than from feedlots located in the ducing the prevalence of Salmonella. Carcasses sampled Midwest (15), suggesting an important correlation between before the intervention system had a Salmonella incidence feedlot environment and postharvest carcass contamination rate of 45.18%, whereas only 0.47% of the carcasses sam- with Salmonella. pled after the intervention system were Salmonella positive Individual abattoir data suggested that abattoir C ex- (Fig. 4). As previously described, carcasses sampled at perienced higher Salmonella incidence in the spring and fall postchilled–lactic acid (sampling step 6) had the lowest

FIGURE 3. Monthly Salmonella incidence rates for combined beef carcass samples collected at (2) preintervention, (3) prepasteurization–lactic acid spray, (4) postpasteurization–lactic acid, (5) postchill, and (6) postchilled–lactic acid sampling steps from March 2005 to September 2006 in abattoirs A (East, left column), B (Mid- west, center column), and C (Southwest, right column). There were no Salmonella- positive samples in February at abattoir A. 2738 RUBY ET AL. J. Food Prot., Vol. 70, No. 12

FIGURE 4. Incidence of Salmonella on beef carcasses. Data are combined across abattoirs A (East), B (Midwest), and C (Southwest) between March 2005 and Sep- tember 2006 at sampling steps (2) preintervention, (3) prepasteurization–lactic acid spray, (4) postpasteurization–lactic acid, (5) postchill, and (6) postchilled–lactic acid. Columns having the same letter are not signiﬁcantly different (P Ն 0.05). Downloaded from http://meridian.allenpress.com/jfp/article-pdf/70/12/2732/1678123/0362-028x-70_12_2732.pdf by guest on 29 September 2021

probability of testing positive for Salmonella. The odds ra- ues were significantly related (APC, P Ͻ 0.0001; EBC, P tios of Salmonella detection increased in reverse order with Ͻ 0.0001) to the presence of Salmonella. Furthermore, the the applied microbial interventions: postchill (sampling step odds ratio indicated that, across the three abattoirs, for ev- 5), postpasteurization–lactic acid (sampling step 4), prepas- ery log increase of APCs and EBCs, there was a 51 and teurization–lactic acid (sampling step 3), and preinterven- 56% greater probability of a Salmonella-positive result at tion (sampling step 2). Sampling at these steps was 2.22, that sampling step, respectively. This relationship could be 5.13, 96.06, and 226.30 times more likely to result in a used to establish a protocol for basing prefabrication Salmonella-positive carcass than sampling at postchilled– (postchilling) carcass Salmonella testing frequency on the lactic acid (sampling step 6). APC or EBC results obtained for the same group of car- Unlike the APC and EBC carcass microbial profiles, casses before chilling. there was no indication of regrowth or additional contam- The relationship of binary EBC (positive ϭ detected ination during the chilling process for Salmonella. This sug- CFU; negative ϭ no CFU) data with the presence or ab- gests that although general microbial regrowth or additional sence of Salmonella was also evaluated. In particular, we contamination of the carcass may occur during chilling, focused on the relationship between the absence of Entero- Salmonella cells are not likely to repair or multiply during bacteriaceae and the absence of Salmonella. For each ab- a normal 48-h chill period. attoir, with data pooled for all sample steps, the occurrence of three carcass samples at a given sampling step being Relationship between Salmonella, APC, and EBC EBC negative corresponded with the composite sample mean values. In an effort to develop tools for choosing testing Salmonella negative Ͼ97% of the time (Table 4). A pathogen testing frequency, the data were evaluated to de- similarly strong link between EBC- and Salmonella-nega- termine the relationship between levels of indicator organ- tive results was seen when combined data from the three isms (APCs and EBCs) and the presence or absence of abattoirs were grouped by year or month (Table 5). When Salmonella. It has been suggested that organism levels, data were analyzed for each sampling step across the three such as APCs and EBCs, do not correlate with pathogen abattoirs (Table 5), EBC-negative results for sampling steps prevalence and therefore cannot be used to gauge the pres- 4, 5, and 6 were linked to Salmonella-negative results 98 ence or absence of specific pathogens (18). Other research- to 99% of the time. However, EBC-negative results at sam- ers have taken caution when interpreting the relationship pling steps 2 and 3 corresponded to Salmonella-negative between indicator organisms and pathogen presence or ab- results only about 72 and 79% of the time, respectively. sence because of limited numbers of samples (3). It has Therefore, our results suggest that for these three abattoirs, been suggested that levels of indicator organisms may be EBC results from sampling steps 4, 5, and 6 can be used used successfully if sufficient data exist to establish a re- as an accurate predictor of the absence of Salmonella in a lationship between the presence and absence of a pathogen sample. The EBC results for prechill carcasses (sampling and the indicator organisms (1). The data set collected and step 4) could therefore be used to set the Salmonella testing reported in this study is unique because of its large size frequency for that step and possibly also for chilled car- (5,228 data points each for APC, EBC, and Salmonella iso- casses (sampling steps 5 and 6). Reducing the frequency of late) and temporal breadth (March 2005 to September Salmonella testing would result in cost savings for the pro- 2006). Thus, evaluation of the relationship between levels cessor. Further studies will investigate the ramifications of of indicator organisms (i.e., APCs and EBCs) and the pres- using EBC results at a given sampling step to set the Sal- ence or absence of Salmonella is appropriate. The logistic monella testing frequency at the corresponding step or later regression analysis determined both APC and EBC log val- sampling steps. J. Food Prot., Vol. 70, No. 12 SALMONELLA AND INDICATOR BACTERIA ON BEEF CARCASSES 2739

TABLE 5. Prevalence of Salmonella and Enterobacteriaceae on beef carcasses and utility of Enterobacteriaceae-negative results as a negative screen for Salmonellaa Enterobacteriaceae Salmonella Utility as a (negative) n % positive Odds ratio nb % positive screen for Salmonellac

Year 2005 2,753 13.95 1.00 Ad 2,735 57.37 98.20 2006 2,602 17.41 1.71 B 2,493 58.28 97.40 Month January 265 16.60 1.32 AB 265 53.58 100.00 February 256 14.06 1.06 AB 256 53.91 98.31 Downloaded from http://meridian.allenpress.com/jfp/article-pdf/70/12/2732/1678123/0362-028x-70_12_2732.pdf by guest on 29 September 2021 March 621 14.49 1.43 AB 619 48.47 95.61 April 571 12.43 1.12 A 567 54.85 95.70 May 625 11.52 1.00 A 613 54.98 100.00 June 601 12.81 1.11 A 597 63.32 99.54 July 568 20.60 2.66 D 568 64.96 97.49 August 583 22.81 3.18 D 514 61.28 96.48 September 484 17.15 2.17 CD 453 64.02 98.16 October 278 11.87 1.50 ABC 273 63.00 97.06 November 253 20.95 3.04 D 252 55.95 99.10 December 250 11.20 1.12 B 251 51.39 99.25 Sampling step 2 1,074 45.18 226.30 A 1,051 96.57 72.22 3 1,080 28.15 96.06 B 1,053 91.45 78.89 4 1,079 2.32 5.13 C 1,054 17.93 98.27 5 1,061 1.04 2.22 D 1,036 63.32 99.74 6 1,061 0.47 1.00 E 1,034 19.25 99.64 a For year and month comparisons, data were combined across abattoirs and sampling steps: (2) preintervention, (3) prepasteurization– lactic acid spray, (4) postpasteurization–lactic acid, (5) postchill, and (6) postchilled–lactic acid. For sampling step comparisons, data were combined across abattoirs, years, and months. b Number of EBC results that corresponded with a Salmonella result. c Percentage of Enterobacteriaceae-negative samples that were Salmonella negative. d Values within each category in the same column with the same letter are not signiﬁcantly different (P Ն 0.05).

The data in this study were obtained from abattoirs REFERENCES with an up-to-date, typical, large-scale beef intervention 1. American Meat Science Association. 1999. The role of microbiolog- process over a longer time frame than previous research ical testing in beef food safety programs. The scientific perspective. has been able to evaluate. Many of our results corroborate Consensus of the 1999 symposium. American Meat Science Asso- previous research (3, 4, 23), showing the effectiveness of ciation, Savoy, Ill. the ‘‘multihurdle approach’’ of a beef intervention system, 2. AOAC International. 2004. Official method 2004.03. Official methods of analysis of AOAC International, 18th ed. AOAC Internation- as well as demonstrating that many factors, including geo- al, Gaithersburg, Md. graphic region, year, month, and intervention treatment, 3. Arthur, T. M., J. M. Bosilevac, X. Nou, S. D. Shackelford, T. L. have an influence on the level of both pathogen (i.e., Sal- Wheeler, M. P. Kent, D. Jaroni, B. Pauling, D. M. Allen, and M. monella) and indicator organism (i.e., APC and EBC) con- Koohmaraie. 2004. Escherichia coli O157 prevalence and enumeration of aerobic bacteria, Enterobacteriaceae, and Escherichia coli tamination on beef carcasses. However, the magnitude of at various steps in commercial beef processing plants. J. Food Prot. the data set is unique, which provided a much better un- 67:658–665. derstanding of the relationships between indicator organ- 4. Bacon, R. T., K. E. Belk, J. N. Sofos, R. P. Clayton, J. O. Reagan, isms (i.e., APCs and EBCs) and the presence of Salmonella. and G. C. Smith. 2000. Microbial populations on animal hides and Individual processors could conduct similar studies to gain beef carcasses at different stages of slaughter in plants employing multiple-sequence interventions for decontamination. J. Food Prot. a better understanding of the relationship between indicator 63:1080–1086. organisms (i.e., APCs and EBCs) and the presence of Sal- 5. Barkocy-Gallagher, G. A., T. M. Arthur, M. Rivera-Betancourt, X. monella in their abattoirs and maximize the cost-effective- Nou, S. D. Shackelford, T. L. Wheeler, and M. Koohmaraie. 2003. ness of qualitatively testing carcasses for Salmonella. Fur- Seasonal prevalence of Shiga toxin–producing Escherichia coli, including O157:H7 and non-O157 serotypes, and Salmonella in com- ther research is planned to determine parameters that must mercial beef processing plants. J. Food Prot. 66:1978–1986. be met by a given abattoir-specific study to adequately de- 6. Brown, M. H., C. O. Gill, J. Hollingsworth, R. Nickelson II, S. scribe the EBC-Salmonella relationship in their operation. Seward, J. J. Sheridan, T. Stevenson, J. L. Sumner, D. M. Theno, W. 2740 RUBY ET AL. J. Food Prot., Vol. 70, No. 12

R. Usborne, and D. Zink. 2000. The role of microbiological testing 1998. Decontamination of inoculated beef with sequential spraying in systems for assuring the safety of beef. Int. J. Food Microbiol. treatment. J. Food Prot. 63:890–893. 62:7–16. 18. Johnson, J. 1996. Predictive microorganisms as an indication of 7. Castillo, A., L. M. Lucia, K. J. Goodson, J. W. Savell, and G. R. pathogen contamination. Proc. Recip. Meat Conf. 49:138–143. Acuff. 1998. Comparison of water wash, trimming and combined 19. Kang, D. H., M. Koohmaraie, and G. R. Siragusa. 2001. Application hot water and lactic acid treatments for reducing bacteria of fecal of multiple antimicrobial interventions for microbial decontamina- origin on beef carcasses. J. Food Prot. 61:823–828. tion of commercial beef trim. J. Food Prot. 64:168–171. 8. Castillo, A., L. M. Lucia, K. J. Goodson, J. W. Savell, and G. R. 20. Marshall, K. M., S. E. Niebuhr, G. R. Acuff, L. M. Lucia, and J. S. Acuff. 1999. Decontamination of beef carcass surface tissues by Dickson. 2005. Identiﬁcation of Escherichia coli O157:H7 meat pro- steam vacuuming alone and combined with hot water and lactic acid cessing indicators for fresh meat through comparison of the effects J. Food Prot. sprays. J. Food Prot. 62:146–151. of selected antimicrobial interventions. 68:2580–2586. 21. Rivera-Betancourt, M., S. D. Shackelford, T. M. Arthur, K. E. West- 9. Center for Disease Control. 1972. Salmonella Typhimurium—Min- morland, G. Bellinger, M. Rossman, J. O. Reagan, and M. Kooh- nesota, Wisconsin, Michigan. Morb. Mortal. Wkly. Rep. 21:411–416. maraie. 2004. Prevalence of Escherichia coli O157:H7, Listeria 10. Centers for Disease Control and Prevention. 1995. Outbreak of Sal- monocytogenes, and Salmonella in two geographically distant com-

monella serotype Typhimurium infection associated with eating raw Downloaded from http://meridian.allenpress.com/jfp/article-pdf/70/12/2732/1678123/0362-028x-70_12_2732.pdf by guest on 29 September 2021 mercial beef processing plants in the United States. J. Food Prot. ground beef—Wisconsin, 1994. Morb. Mortal. Wkly. Rep. 44:905– 67:295–302. 909. 22. SAS. 2003. SAS user’s guide: version 9.1. SAS Institute, Cary, N.C. 11. Centers for Disease Control and Prevention. 2002. Outbreak of mul- 23. Sofos, J. N., S. H. Kochevar, G. R. Bellinger, D. R. Buege, D. D. tidrug-resistant Salmonella Newport—United States, January–April Hancock, S. C. Ingham, J. B. Morgan, J. O. Reagan, and G. C. 2002. Morb. Mortal. Wkly. Rep. 51:545–548. Smith. 1999. Sources and extent of microbiological contamination 12. Centers for Disease Control and Prevention. 2006. Multistate out- of beef carcasses in seven United States slaughtering plants. J. Food break of Salmonella Typhimurium infections associated with eating Prot. 62:140–145. ground beef—United States, 2004–2006. Morb. Mortal. Wkly. Rep. 24. U.S. Department of Agriculture. 1996. Nationwide federal plant raw 55:180–182. ground beef microbiological survey, August 1993–March 1994. 13. Dechet, A. M., E. Scallan, K. Gensheimer, R. Hoekstra, J. Gunder- Available at: http://www.fsis.usda.gov/OPHS/baseline/rwgrbeef.pdf. man-King, J. Lockett, D. Wrigley, W. Chege, J. Sobel, and the Mul- Accessed 5 February 2007. tistate Working Group. 2006. Outbreak of multidrug-resistant Sal- 25. U.S. Department of Agriculture. 2006. Salmonella testing of raw meat monella enterica Typhimurium deﬁnitive type 104 infection linked & poultry products, 1998–2004: progress report. Available at: http:// ࿞ ࿞ ࿞ ࿞ to commercial ground beef, northeastern United States, 2003–2004. www.fsis.usda.gov/science/Progress Report Salmonella Testing Clin. Infect Dis. 42:747–752. 1998-2004/index.asp. Accessed 5 February 2007. 14. Dorsa, W. J., C. N. Cutter, G. R. Siragusa, and M. Koohmaraie. 1996. 26. U.S. Food and Drug Administration. 1998. Bacteriological analytical manual, 8th ed., revision A. AOAC International, Gaithersburg, Md. Microbial decontamination of beef and sheep carcasses by steam, 27. U.S. Food Safety and Inspection Service. 2004. Microbiological lab- hot water spray washes, and a steam-vacuum sanitizer. J. Food Prot. oratory guidebook. 2004. Isolation and identiﬁcation of Salmonella 59:127–135. from meat, poultry, and egg products. MLG 4.03 revision 3. U.S. 15. Fedorka-Cray, P. J., D. A. Dargatz, L. A. Thomas, and J. T. Gray. Food Safety and Inspection Service, Washington, D.C. 1998. Survey of Salmonella serotypes in feedlot cattle. J. Food Prot. 28. Voetsch, A. C., T. J. Van Gilder, F. J. Angulo, M. M. Farley, S. 61:525–530. Shallow, R. Marcus, P. R. Cieslak, V. C. Deneen, and R. V. Tauxe, 16. Gill, C. O., and C. Landers. 2003. Effects of spray-cooling processes for the Emerging Infections Program FoodNet Working Group. 2004. on the microbiological conditions of decontaminated beef carcasses. FoodNet estimate of the burden of illness caused by nontyphoidal J. Food Prot. 66:1247–1252. Salmonella infections in the United States. Clin. Infect Dis. 38:S127– 17. Graves Delmore, L. R., J. N. Sofos, G. R. Schmidt, and G. C. Smith. 134.