658

Journalof Food Protection, Vol. 67, No. 4, 2004, Pages 658– 665 Copyright Q ,International Association forFood Protection

Escherichia coli O157 Prevalence andEnumeration of Aerobic Bacteria, Enterobacteriaceae, and Escherichia coli O157 at VariousSteps inCommercial Beef Processing Plants †

TERRANCE M.ARTHUR, 1* JOSEPH M.BOSILEVAC, 1 XIANGWU NOU, 1 STEVEN D.SHACKELFORD, 1 TOMMY L.WHEELER, 1 MATTHEW P.KENT, 1 DIVYA JARONI, 1‡ BRUCE PAULING, 2 DELL M.ALLEN, 2 AND MOHAMMAD KOOHMARAIE 1

1RomanL. HruskaU.S. Meat Animal Research Center,Agricultural Research Service, U.S.Department of Agriculture, Clay Center, Nebraska68933-0166; and 2Excel Corporation,151 North Main Street, Wichita, Kansas 67202, USA

MS03-315:Received 10July 2003/ Accepted 21November 2003

ABSTRACT

Theeffectiveness of currentantimicrobial interventions used in reducingthe prevalence or loadof Escherichiacoli O157 andindicator organisms on cattlehides and carcasses at twocommercial beef processing plants was evaluated. Sponge sampling ofbeefcattle was performed at Žvelocationsfrom the initial entry of theanimals to the slaughter  oorto theexit of carcasses fromthe ‘ ‘hotbox’’ cooler.Foreach sample, E. coli O157prevalence was determined and total aerobic bacteria, Enterobac- teriaceae, and E. coli O157were enumerated. E. coli O157was found on 76% of animal hides coming into the plants, but nocarcasses leaving the cooler were identiŽ ed as contaminated with E. coli O157.A positiverelationship was seen between theincidence of E. coli O157in hide samples and that in preevisceration samples. Aerobic plate counts and Enterobacteriaceae countsaveraged 7.8 and 6.2 log CFU/ 100cm 2,respectively,on hides, and 1.4 and 0.4 log CFU/ 100cm 2,respectively,on chilledcarcasses. Aerobic plate counts and Enterobacteriaceae countson preevisceration carcasses were signiŽ cantly related tothe respective levels on the corresponding hides; the carcasses of animals whose hides carried higher numbers of bacteria weremore likely to carry higher numbers of bacteria. Implementation of the sampling protocol described here would allow processorsto evaluate the efŽ cacy of on-line antimicrobial interventions and allow industrywide benchmarking of hygienic practices.

Escherichiacoli O157:H7has been a pathogenof con- surfacecontamination on beef tissue than are individual cernto themeat processing industry for twodecades. Cases interventionsalone (8, 20, 24). Manycommercial beef pro- ofhemorrhagiccolitis caused by E. coli O157:H7were as- cessingplants presently employ several interventions (i.e., sociatedwith consumption of undercooked ground beef in trimming,steam vacuuming, steam pasteurization, water theearly 1980s (26). Inthe United States during 1992 and washes,and organic acid washes) in combination to achieve 1993,an outbreak of E. coli O157:H7infection associated largereductions in carcass contamination in accordance withconsumption of ground beef caused hundreds of ill- withtheir individual HACCP plans (1). nessesand four deaths (31). Theseevents led the Food In2002, the FSIS requiredall raw beefprocessors to Safetyand Inspection Service (FSIS) todeclare the E. coli reassesstheir HACCP plansto ensurethat their critical con- O157:H7organism an adulterant in ground beef and to re- trolpoints were adequatelyaddressing E. coli O157:H7 quirethat meat processors establish hazard analysis and contamination (13). Inverifying process control, testing for criticalcontrol point (HACCP) plansfor theirplants (12). pathogensis generally not useful because of the low num- Sincethis time, numerous intervention strategies focusing bersof bacterial cells (6, 17). Therefore,indicator organ- onprevention of carcass contamination and decontamina- isms,present in sufŽ ciently high numbers throughout the tionof carcasses have been designed, tested, and put into processingline, are monitored to ensure that intervention practiceat commercial processing plants. systemsare functioning properly. T oadequatelyinterpret Recentstudies have demonstrated that combinations of thesedata, the relationships between the indicator organ- antimicrobialinterventions are more effective at reducing ismsand the pathogen(s) of interest must be established (15). *Authorfor correspondence. Tel: 402-762-4227; Fax: 402-762-4149; E-mail: [email protected]. Inthis study, counts of indicator organisms (aerobic †Names are necessary toreport factually on availabledata; however, the bacteriaand Enterobacteriaceae ) and E. coli O157 and the U.S.Department ofAgriculture neither guarantees nor warrants thestan- prevalenceof E. coli O157were assessedat various steps dardof theproduct, and the use ofthename bytheU.S. Department of inprocessingto identifyrelationships that may be exploited Agricultureimplies no approvalof the product to theexclusion of others tomonitor process control. The objectives of this study thatmay alsobe suitable. ‡Presentaddress: Room 119, Veterinary Diagnostic Center, East Campus, were twofold:(i) todetermine the effectiveness of current Universityof Nebraska, Lincoln,NE 68583-0907,USA. interventionsused in reducing the prevalence or levelof E. J.FoodProt., Vol. 67, No. 4 BACTERIALCONTAMINATION DURING BEEFPROCESSING 659

FIGURE 1. Carcasssampling areas. Shad- edareasrepresent locations where carcass spongesamples were collected.

coli O157and indicator organisms and (ii) to develop a througha steampasteurization cabinet. Sampling point 4 wasin toolthat can be used by beef processors to monitor their thechill cooler, where the carcasses were spray chilled for 29 h processes. beforethe Ž nalsamples were taken. MATERIALS AND METHODS Samplecollection. Allsamples were obtained using Speci- Sponges(Nasco, Fort Atkinson, Wis.) moistened with 20 ml of Sampleswere collected during three separate trips each to bufferedpeptone water (Difco Laboratories, Sparks, Md.). Spong- twocommercial fed-beef processing plants. Samples were col- eswerewrung out in the bag and then removed from the bag and lectedat plantA fromthe last week in September to mid-October usedto swab the hide or carcass. The hide sample was collected andat plant B frommid-October to the Ž rstweek of November froma 100-cm 2 areaover the plate region, using a 100-cm 2 tem- 2002.Forty-eight samples were collected from each of Žvesample plateand  ippingthe sponge over midway through taking the sitesduring each trip, for a totalof 288 samples per site. sample.One sample consisted of approximately Ž vevertical and In-plantsampling locations. Samplingwith wetted sponges Žvehorizontalpasses (up and down or sideto side was considered wasdone at Ž velocations on the processing line: point 1: hide, onepass). For each of the four carcass processing points, two 2 sampledafter hide opening but before hide removal; point 2: 4,000-cm areaswere sampled as a singlesample. Sampling of preevisceration,immediately following dehiding before any anti- sufŽcient area at the preevisceration and postevisceration process microbialapplications; point 3: postevisceration, after eviscera- pointswas complicated by the carcasses moving at high line tion,carcass trimming, and Ž nalinspection before the Žnalcarcass speeds.Therefore, efŽ cient sampling was facilitated by usingtwo washes;point 4: postintervention ,inthe chill cooler after all an- sponges,each moistened with 20 ml of buffered peptone water . timicrobialinterventions; and point 5: chilled carcass, 29 h post- Onesponge was used for the inside and outside round area and mortem.Individual animals and carcasses were tagged and tracked onefor the navel-plate-brisket -foreshankarea (Fig. 1). Later, these throughoutthe process. The same carcass was sampled at hide twosponge samples were combined into one sample bag in the (1),preevisceration (2), and postevisceration(3) processing points. laboratory.For the postinterventio nandchilled samples, a single Thecarcass immediately following the carcass sampled at points spongewas used to sample both areas of the carcass, with one 1,2, and 3 wassampled at the postintervention (4) and chilled sideof the sponge used for the inside and outside round area and (5)processing points. theother side of the sponge used for the navel-plate-briske t-fore- shankarea. All sample bags were transported on ice to the labo- In-plantantimicrobial interventions. Theantimicrobial in- ratory,where they were processed immediately (plant A samples terventionsystems utilized by both plants were ver ysimilarwith within3 handplant B sampleswithin 6 h).The inside and outside regardto the types of interventions employed and their physical roundand navel-plate-brisk et-foreshankareas were selected for placementin the processing line. Prior to sampling point 1, the samplingin this experiment because those areas include or lie patternareas of the hide were subjected to high-pressure water adjacentto points where incisions are made to open the hide (hide rinsesand steam vacuuming. Between sampling points 1 and2, patternlines); therefore these points are thought to behotspotsfor thehide was removed and the pattern lines were steam vacuumed. hide-to-carcasscross contamination. Because contamination is not Betweensampling points 2 and3, thecarcasses were further steam evenlydistributed, large areas were sampled to ensure that the vacuumed,passed through a preeviscerationwash cabinet in samplewas representative at each sampling point. whichthey were washed with cold water and 2 to3% lacticacid, thentrimmed and inspected. Between sampling points 3 and4, Sampleprocessing. Spongebags were massaged thoroughly, thecarcasses passed through two wash cabinets, one using 90 8C andaliquots of 2.5 and 5 ml(12.5% of the diluent volume) were waterand the other using peroxyacetic acid, followed by passage removedfrom the hide, postinterventio n,andchilled samples and 660 ARTHUR ETAL. J.FoodProt., Vol. 67, No. 4

TABLE 1. Percentageof samples positive for E. coli O157for each point by trip and overall Plant Ab,c Plant Bb,c

Sample point a Trip1 Trip2 Trip3 Total Trip1 Trip2 Trip3 Total Overall

Hide Mean 83.3 87.5 93.8 88.2 AX 50.0 50.0 89.6 63.2 A 75.7 A 95% CId 81.8–93.0 54.8–63.2 70.3–80.5 Preevisceration Mean 25.0 10.4 25.0 20.1 BX 6.5 2.1 18.8 9.2 B 14.7 B 95% CI 13.9–27.6 5.0–15.1 10.8–19.3 Postevisceration Mean 6.3 2.1 4.2 4.2 CY 0.0 0.0 10.4 3.5 BC 3.8 C 95% CI 1.5–8.8 1.1–7.9 1.9–6.7 Postintervention Mean 0.0 2.1 0.0 0.7 CY 0.0 0.0 0.0 0.0 C 0.3 D 95% CI 0.0–3.8 0.0–2.5 0.0–1.9 Chilled Mean 0.0 0.0 0.0 0.0 CY 0.0 0.0 0.0 0.0 C 0.0 D 95% CI 0.0–2.5 0.0–2.5 0.0–1.3 a Prevalenceswithin a samplepoint either ( X)differsigniŽ cantly between plants ( P , 0.01) or (Y)donot differ between plants ( P . 0.05). b Percentageof samples positive for E. coli O157. c Prevalencesin the same column that do not share a commonletter ( A through E)aresigniŽ cantly different ( P , 0.05). d CI,conŽdence interval. fromthe preevisceration and postevisceration samples, respective- andpostevisceration sample bags, which contained two sponges. ly,prior to the addition of enrichment medium. The sample ali- Allsample bags were incubated, subjected to immunomagnetic quotswere used for enumeration of totalaerobic bacteria, Entero- separation,and plated as previouslydescribed (4). Afterthe plates bacteriaceae, and E. coli O157. wereincubated, up tothree suspect colonies were picked and test- Totalaerobic plate counts (APC) and Enterobacteriaceae edby latex agglutination (DrySpot E. coli O157,Oxoid, Basing- counts(EBC) weremade on a Bactometer(BioMerieux, Hazel- stoke,UK). Ina previousstudy, Barkocy-Gallagher et al. (2) dem- wood,Mo.); for those samples with too few organisms to count onstratedthat .90%of samplesthat were presumptively positive onthe Bactometer ,PetriŽlm Aerobic Count Plates or Enterobac- for E. coli O157based on the above methods were actually pos- teriaceae CountPlates (3M Microbiology, St. Paul, Minn.) were itive for E. coli O157:H7.Therefore, for the purposes of this used.Bacterial counts from PetriŽ lm were used to generate stan- study,any sample that produced characteristic E. coli O157 col- dardcurve data for the Bactometer during calibration to facilitate oniesthat gave positive reactions for the O157 latex agglutination dataanalysis for the two systems. assaywas considered positive for E. coli O157:H7. E. coli O157cells were counted using a three-tubemost- probable-number(MPN) method (2). Triplicate10-fold dilution Statisticalanalysis. Tocomparethe prevalence of E. coli serieswere made by transferring100 mlofthesample aliquot into O157between plants and among sampling sites, continuity-ad- 900 mlofbuffered peptone water plus 50 mg/mlferrioxamine in justedchi-square P-valueswere calculated using the PROC FREQ deep-wellmicrotrays. The microtrays were incubated at 25 8C for procedureof SAS (27). Whenmultiple comparisons were made, 2handthen at 37 8Cfor16 to 18 h. After incubation, the trays thepair-wise P-valueswere adjusted using Hommel’ s modiŽca- werekept at 4 8Cuntilthe results from the E. coli O157prevalence tionof theBonferroni procedure (16) toavoidin ated type I error analysiswere obtained. For any sample that was positive for E. rates.APC andEBC datawere log transformed before analysis coli O157in the prevalence analysis, 350 mlfromthe correspond- ofvariance (ANOVA). Foreach sampling site, a one-wayANO- ingbuffered peptone water plus ferrioxamine MPN dilutionswere VA wasconducted to determine whether counts differed between subculturedinto 4.5 ml of MacConkey broth (Difco). The inoc- plants.For each plant and for the combined data set, a one-way ulatedMacConkey tubes were incubated static at 42 8C for 16 to ANOVA wasconducted to determine whether counts differed 24h. Following incubation, the MacConkey tubes were screened amongsampling sites. Hide and preevisceration observations were forthe presence of E. coli O157using the ImmunoCard Stat! E. groupedon thebasis of APC orEBC, andchi-square analysis was coli O157(Meridian Diagnostics, Cincinnati, Ohio). The combi- conductedto determine whether there was a relationshipbetween nationof positive dilution tubes was used to obtain the MPN per APC orEBC andthe prevalence of E. coli O157:H7. mlby using the three-tube MPN tablefrom the Bacteriological RESULTS AND DISCUSSION AnalyticalManual (http://vm.cfsan.fda.gov/;ebam/bam-a2.html (14)). E. coli O157prevalence and enumeration. E. coli E. coli O157detection. Eightymilliliters of trypticsoy broth O157contamination was foundon 75.7% (218 of 288) of wasadded to the hide, postinterventio n,and chilled sample bags, thecattle hides tested, with a rangeof 50 to 93.8% among and160 ml of tryptic soy broth was added to the preevisceration thesix sampling trips (T able1). On average, the hides of J.FoodProt., Vol. 67, No. 4 BACTERIALCONTAMINATION DURING BEEFPROCESSING 661

TABLE 2.E. coli O157MPN results of E. coli O157(one sample had .220,000MPN/ 100cm 2). Sample point MPNa Plant Ab Plant Bb As awhole,the E. coli O157MPN levelsmay be under- estimated.Certain MPN combinations,speciŽ cally those in- Hide NDc 18 53 dicativeof growth suppression at low dilutions, appear ,60 54 67 moreoften than would be expected by chance. This phe- 60–99 25 14 nomenonhas been seen previously when examining bac- 100–999 28 5 teriain various food matrices (5, 14). Lahti et al. (21) used 1,000–9,999 12 5 anMPNmethodto enumerate E. coli O157in fecalsamples 10,000–99,999 6 0 frombullsan ddocumenteda largerange( ,0.2 to 100,000–999,999 1 0 .160,000MPN/ goffeces) inthe E. coli O157populations Preevisceration ND 115 131 associatedwith individual animals. ,1.5 24 11 Thenumbers of E. coli O157organisms present on 1.5–9 4 2 mostof the positive carcasses (83%) were belowthe de- 10–99 0 0 100–999 1 0 tectionlimit of the MPN methodused ( ,1.5 MPN/100 cm2),butsix carcasses had O157 loads of approximately 2 a LogCFU/ 100cm 2. MPN/100 cm2 andone carcass had an O157 load of 550 b Numberof samples for each class of MPN values. MPN/100 cm2.Barkocy-Gallagheret al. (2) reportedsim- c ND, MPN notdone because O157 was not detected by culture ilar E. coli O157MPN levelsfrom preeviscerationsamples method. collectedin the fall of 2001.The E. coli O157populations for 69%of the positive carcass samples from thatstudy were belowthe MPN detectionlimit, but one carcass car- cattleprocessed at plant A carried E. coli O157more fre- ried an E. coli O157load of 334MPN/ 100cm 2. All of the quently (P , 0.0001)than did those of cattle processed at posteviscerationand postintervention carcasses that har- plantB (88.2%versus 63.2%), although the Ž nalsampling bored E. coli O157did so at levels below the detection periodat plant B hadthe second highest E. coli O157 hide limitof the MPN assayused here ( ,1.5MPN/ 100cm 2, prevalenceof the six sampling trips. Immediatelyfollowing hide removal, the mean E. coli datanot shown). O157prevalence on carcasses was 14.7%(range of 2.1 to Enumerationof aerobic bacteria and Enterobacte- 25%)for thevarious sampling trips. Carcasses sampled pri- riaceae. Otherresearchers have stated that bacterial counts, orto evisceration at plant A were twiceas likely ( P , suchas APC andEBC, arenot correlated with levels of 0.0001)to have detectable E. coli O157contamination as pathogensand cannot be used with index organisms to were thoseat plant B (20.1%versus 9.2%). gaugethe presence or absence of speciŽ c pathogens (15, Followingpreevisceration sampling, the carcasses were 19). Instead,the bacteria enumerated in these tests have washedwith a lacticacid solution, eviscerated, split, beenused as indicator organisms to assess the hygienic sta- trimmed,inspected according to governmental regulations, tusof processing plants and the performance of antimicro- andthen sampled again. The mean E. coli O157prevalence bialinterventions (15).Enterobacteriaceae havebeen con- atthe postevisceration sampling point had decreased to sideredsuitable indicators for theeffectiveness of decon- 3.8%(range of 0.0to 10.4%)among the six sampling trips, taminationprocedures because they are slightly more tol- andthere was nosigniŽcant difference ( P . 0.05)between erantto lactic acid than are several enteric pathogens (29). plants. If theindicator organisms react to antimicrobial interven- E. coli O157was detectedon only 1 of288 (0.3%) tionsin a mannersimilar to that of the pathogen of interest, carcassessampled after the application of all antimicrobial in this case E. coli O157:H7,then these organisms can be interventions.None of the chilled carcass samples were usedto monitor process control. positivefor E. coli O157. Theindicator organism counts for hideswere higher Theprevalence of E. coli O157described herein is for plantA thanfor plantB (Table3), similar to the trend analogousto thatreported for beefcarcasses sampled in the for E. coli O157.The average values for APConcattle (2). fall of 2001 Inthat report, the hide, preevisceration, hidesranged from 8.0to 9.0 log CFU/ 100cm 2 for plant A E. coli andpostinterventio nprevalencelevels of O157:H7 and7.2 to 7.4 log CFU/ 100cm 2 for plantB, andthe EBC were 67.2,27.3, and 1.0%, respectively. Sampling and de- tripaverages ranged from 6.7to 8.0 log CFU/ 100cm 2 for tectionmethodologies used in that study were similarto plantA and4.9 to 5.8 log CFU/ 100cm 2 for plant B. thoseused here. Duringhide removal, bacterial deposition onto the ster- E. coli O157enumeration. Consistentwith the E. coli ilesurface of the carcass resulted in levels of 3.5 and 1.4 O157prevalence data, MPN levelsgenerally were higher logCFU/ 100cm 2 for APC andEBC, respectively.Given for plantA thanfor plantB (Table2). Enumeration of E. thereduction seen in E. coli O157prevalence between the coli O157cells revealed that although most E. coli O157- preeviscerationand postevisceration sampling points, we positivehides (56%) were contaminatedat lowlevels, with assumedthat the corresponding EBC andAPC levelsalso O157levels below the MPN detectionlimit ( ,60 MPN/ woulddecrease. This was notthe case; the APC andEBC 100 cm2), manyhides (41%) had levels between 60 and resultsshowed a slightincrease, resulting in posteviscera- 9,999MPN/ 100cm 2 anda few (3%) haddense populations tionlevels of 3.7 and 1.7 log CFU/ 100cm 2,respectively. 662 ARTHUR ETAL. J.FoodProt., Vol. 67, No. 4

TABLE 3. MeanAPC andEBC foreach sampling point by trip and overall Plant Aa,b Plant Ba,b Indicator organismSample point Trip1 Trip2 Trip3 Total Trip1 Trip2 Trip3 TotalOverall

APCc Hide 9.0 8.1 8 8.3 AX 7.2 7.3 7.4 7.3 A 7.8 A Preevisceration 4.2 3.6 3.7 3.8 BX 3.0 3.2 3.3 3.2 B 3.5 C Postevisceration 3.8 3.4 3.9 3.7 CY 3.3 4.1 4.0 3.8 C 3.7 B Postintervention 1.4 1.4 1.0 1.2 EY 0.9 1.4 1.5 1.3 D 1.3 E Chilled 1.6 0.9 1.8 1.4 DY 2.0 1.1 1.1 1.4 D 1.4 D

EBCc Hide 8.0 6.6 6.7 7.1 AX 5.5 4.9 5.8 5.4 A 6.2 A Preevisceration 2.5 1.4 1.4 1.8 BX 1.2 0.7 1.0 1.0 C 1.4 C Postevisceration 1.9 1.3 1.3 1.5 CX 1.8 1.7 2.1 1.9 B 1.7 B Postintervention 0.4 0.1 0.0 0.2 DY 0.1 0.1 0.2 0.1 E 0.2 E Chilled 0.3 0.2 0.6 0.3 DY 0.7 0.5 0.3 0.5 D 0.4 D a LogCFU/ 100cm 2. b Meansin the same column for a bacterialtype that do not share a commonletter ( A through E)aredifferent ( P , 0.05). c Meanswithin a samplesite either ( X)differbetween plants ( P , 0.0001) or (Y)donot differ between plants ( P . 0.05).

Thereare numerous possible explanations for thisdiscrep- respectivelevels on the corresponding carcasses (Fig. 2). ancy.It is possiblethat the preevisceration lactic acid rinse Whengrouped by sampling trip, groups with larger bacte- hada greatereffect on E. coli O157:H7than it had on the rialloads and higher incidence rates on the hide subse- indicatororganism population. In vitro studies have shown quentlyhad larger bacterial loads or higher incidence rates thatgram-negative bacteria are more susceptible to lactic onthe carcass. This association was strongerfor aerobic acidthan are gram-positive bacteria (32). Thisfact could bacteriaand Enterobacteriaceae than for E. coli O157. In explainthe lack of reduction seen in the APC. However, previousstudies, antimicrobial interventions focused on the basedon the results from 20years of research, Smulders hide,such as dehairing, were translatedinto lower overall and Greer (30) statedthat one should expect an average bacterialloads and lower incidences of pathogens on car- reductionof 1.5 log in APC from lacticacid treatment of casses (7, 23). Nou et al. (23) demonstratedthat hide de- carcasses.Alternatively, based on previous in-plant reports hairing,which in effectsanitizes the animal hide, can great- indicatingthat preevisceration lactic acid rinses result in a lyreduce carcass bacterial loads and eliminate E. coli reductionof APC, coliforms,and generic E.coli(9, 10), a O157.Barkocy-Gallagher et al. (3) showed that .66% of secondscenario is also possible. The total bacterial popu- the E. coli O157:H7isolates from beefcarcasses identiŽ ed lationcould have been reduced by thepreevisceration lactic latein processingcould be trackedby pulsed-Želd gel elec- acidrinse, and carcasses could have become subsequently trophoresisto the same carcass early in processing. This recontaminatedwith bacteria other than E. coli O157:H7 Žndingindicates that the E. coli O157population associated duringevisceration, splitting, trimming, and/ orinspection. withthe animal or carcasspersisted throughout processing. Theseactivities may have led to thedeposition of additional Thesefacts indicate that interventions focused on reducing bacteriaonto the carcass, masking any reduction due to the thenumber of bacteria and pathogens on the hide can have organicacid wash. W eareaware ofonly one other study alargeimpact on the levels of pathogens on the carcass. inwhich carcasses in commercial facilities were sampled Relationshipsbetween the levelsof APC andEBC bothbefore preevisceration rinsing and at a similarpostev- andthe incidenceof E. coli O157on carcasses. When iscerationpoint (before Ž nalinterventions) (11). In that theAPC andEBC levelswere groupedinto classes, sig- study,the prevalence of E. coli O157:H7decreased be- niŽcant associations were seenbetween the APC orEBC tweenthese points, as was thecase in the present study, classand the prevalence of E. coli O157for carcasssam- butlevels of otherbacteria were notstudied. Further studies ples (P , 0.05).The samples from higherAPC andEBC areneeded to identify which scenario is correct. classeswere morelikely to be positive for E. coli O157 Followingpostevisceration sampling, the carcasses (Table4). Caution should be taken when interpreting these were subjectedto washes with hot water and peroxyacetic resultsbecause the data set was limitedto 288samples from acidand to steam pasteurization. These interventions re- onlytwo different plants in one season. This relationship ducedAPC andEBC to1.3 and 0.2 log CFU/ 100cm 2, maybe the result primarily of differences among trips or respectively.Cooling the carcasses via spray chilling and betweenplants rather than a directrelationship between storageat refrigerated temperatures held the bacterial pop- APC andEBC andprevalence of E. coli O157.Therefore, ulationsessentially constant with minimal growth, leading moredata are needed to conŽ rm theseŽ ndings.A similar tochilledsample numbers of 1.4and 0.4 log CFU/ 100cm 2 analysiswas doneby Siragusa et al. (28), whereAPC val- for APC andEBC, respectively. uesfrom preevisceration,postinterventio n,and chilled sam- Relationshipsbetween hideand carcass contami- pleswere groupedinto classes and relationships with the nation. Levelsof APC, EBC,andto a lesserextent E. coli incidenceof E. coli biotype1, anindicatorof fecalcontam- O157prevalence on hides were signiŽcantly related to the ination,were assessed.In that study, a strongassociation J.FoodProt., Vol. 67, No. 4 BACTERIALCONTAMINATION DURING BEEFPROCESSING 663

FIGURE 2. Correlationsbetween hide and preeviscerationbacterial levels. (A) APC levelsfor the preevisceration and hide samplesare plotted for each sampling trip (n 5 48per trip). (B) EBC levelsfor the preeviscerationand hide samples are plot- tedfor each sampling trip ( n 5 48 per trip). (C) E. coli O157prevalence for the preeviscerationand hide samples is plotted foreach sampling trip ( n 5 48per trip).

was identiŽed between APC classand the incidence of an nation.By modifying intervention schemes to maintain E. coli–positivesample for theoverall data set. However , APC andEBC levelsbelow maximum target values, pro- nosuch relationship was seenwhen the analysis was re- cessorsare likely to reduce the prevalence of E. coli O157 strictedto the preevisceration samples. Although indicator oncarcasses. organismlevels cannot be used for directpresence or ab- senceanalysis of E. coli O157,they may be useful as a EfŽcacy of interventions. Bothplants employed the guidelinefor theminimization of E. coli O157contami- conceptof multiple sequential interventions using very 664 ARTHUR ETAL. J.FoodProt., Vol. 67, No. 4

TABLE 4. Relationshipbetween APC andEBC ofpreeviscera- recoveredfrom beef cattle andcarcasses at processingplants in the tioncarcasses and prevalence of E. coli O157on preevisceration midwestern states ofthe United States. Appl.Environ. Microbiol. carcasses 67:3810–3818. 4.Barkocy-Gallagher, G. A.,E. D.Berry,M. Rivera-Betancourt,T .M. No. (%) of E. coli Arthur,X.Nou,and M. Koohmaraie.2002. Development of methods Indicator Log CFU/ O157–positive forthe recovery of Escherichiacoli O157:H7 and Salmonella from organism 100 cm2 n samplesa beef carcass spongesamples andbovine fecal andhide samples. J. Food Prot. 65:1527–1534. APC .4 59 14 (24) A 5.Blodgett, R. J.,and W .E.Garthright.1998. Several MPNmodels ,4 227 28 (12) B forserial dilutionswith suppressed growth at lowdilutions. Food EBC .2 57 14 (25) A Microbiol. 15:91–99. ,2 228 27 (12) B 6.Brown, M. H.,C. O.Gill,J. Hollingsworth,R. NickelsonII, S. Seward,J. J.Sheridan,T .Stevenson,J. L.Sumner,D. M.Theno,W . a Percentageswithin an indicator organism that do not share a R.Usborne,and D. Zink.2000. The role of microbiological testing commonletter ( A, B)aresigniŽ cantly different ( P , 0.05). insystems forassuring the safety ofbeef. Int.J. FoodMicrobiol. 62:7–16. 7.Castillo, A., J. S.Dickson,R. P.Clayton,L. M.Lucia,and G. R. similarantimicrobial applications. By minimizing deposi- Acuff.1998. Chemical dehairingof bovine skin to reduce pathogenic tionof bacteria onto the carcass and using subsequent ef- bacteria andbacteria offecal origin. J. FoodProt. 61:623–625. fectivedecontamination, the processors were ableto main- 8.Castillo, A., L. M.Lucia,K. J.Goodson,J. W.Savell,and G. R. tain E. coli O157populations at less than detectable levels Acuff.1999. Decontamination of beef carcass surface tissueby steam vacuumingalone and combined with hot water andlactic acid onall of the carcasses tested after chilling. sprays. J.FoodProt. 62:146–151. HACCP testing. Inlight of the requirement that beef 9.Castillo, A., L. M.Lucia,I. Mercado,and G. R.Acuff.2001. In- plantevaluation of a lactic acid treatment forreduction of bacteria processingplants reassess their HAACP plansregarding the onchilled beef carcasses. J.FoodProt. 64:738–740. efŽcacy of the antimicrobial interventions used to control 10.Dormedy, E. S.,M.M.Brashears, C.N.Cutter,and D. E.Burson. E. coli O157:H7,we offer thesampling method detailed 2000.Validation of acid washes as critical controlpoints in hazard herefor useby the industry as a toolto evaluate process analysisand critical controlpoint systems. J.FoodProt. 63:1676– controland validate critical control points. This method of- 1680. 11.Elder ,R.O.,J.E.Keen, G.R.Siragusa,G. A.Barkocy-Gallagher, fers clearadvantages over those previously reported for M.Koohmaraie,and W .W.Laegreid.2000. Correlation of entero- HAACP monitoring (18). First,the technique is noninva- hemorrhagic Escherichiacoli O157prevalence infeces, hides,and siveand does not result in any product loss. Second, the carcasses ofbeef cattle duringprocessing. Proc.Natl. Acad. Sci. procedurecan be performed efŽ ciently and adequately on USA 97:2999–3003. theline at operational chain speeds. Third, the microbio- 12.Federal Register.1996.Pathogen reduction; hazard analysisand crit- logichygiene of a plantis assessed at pointsencompassing ical controlpoint (HACCP) systems, Žnalrule. Fed.Regist. 61: 38805–38989. thecarcasses’ entrance to and exit from theslaughter  oor. 13.Federal Register.2002. E. coli O157:H7contamination of beef prod- Severalresearchers have demonstrated that the hide is pos- ucts. Fed.Regist. 67:62325–62334. siblythe major source of beefcarcass contamination (2, 22, 14.Food and Drug Administration. 1998. Bacteriological analytical 23, 25). Therefore,processors must know the bacterial lev- manual,8th ed., revision A. AOAC International,Gaithersburg, Md. elson animalscoming into the plants to accurately evaluate 15.Food Safety and Inspection Service. 2002. Guidance for minimizing the risk of Escherichiacoli O157:H7 and Salmonella inbeef slaugh- antimicrobialintervention performance at various points in ter operations.Availableat: http://www.fsis.usda.gov/oa/topics/ theprocessing line. An industrywide standard protocol for o157.htm.Accessed 4April2003. microbiologictesting would allow individual plants to 16.Hommel, G.1988.A stagewise rejective multipletest procedure benchmarktheir hygienic standard and evaluate the efŽ cacy basedon a modiŽed Bonferroni test. Biometrika 75:383–386. ofcurrent on-line antimicrobial interventions. 17.Jay, J. M.(ed.).2000. Modern food microbiology. Aspen Publishers, Gaithersburg,Md. ACKNOWLEDGMENTS 18.Jericho, K. W.F.,G.C.Kozub,J. A.Bradley,V .P.J.Gannon,E. J. Golsteyn-Thomas,M. Gierus,B. J.Nishiyama,R. K.King,E. E. We thankJulie Dyer ,SueHauver, Bruce Jasch, Dee Kucera, Frank Tanaka,S. D’Souza,and J. M.Dixon-MacDougall.1996. Microbi- Reno,and Greg Smithfor technical support and Carol Grummert forsec- ologicalveriŽ cation of thecontrol of theprocesses ofdressing,cool- retarial support. ingand processing of beef carcasses at ahighline-speed abattoir . FoodMicrobiol. 13:291–301. REFERENCES 19.Johnson, J. 1996.Predictive microorganisms as anindication of 1.Bacon, R. T.,K.E.Belk,J. N.Sofos,R. P.Clayton,J. O.Reagan, pathogencontamination. Proc.Recip. Meat Conf. 49:138–143. andG. C.Smith.2000. Microbial populations on animal hidesand 20.Kang, D. H.,M. Koohmaraie,and G. R.Siragusa.2001. Application beef carcasses at differentstages ofslaughter in plants employing ofmultiple antimicrobial interventions for microbial decontamina- multiple-sequentialinterventions for decontamination. J.FoodProt. tionof commercial beef trim. J.FoodProt. 64:168–171. 63:1080–1086. 21.Lahti, E., O. Ruoho,L. Rantala,M. L.Hanninen,and T .Honkanen- 2.Barkocy-Gallagher, G. A.,T .M.Arthur,M.Rivera-Betancourt,X. Buzalski.2003. Longitudinal study of Escherichiacoli O157 in a Nou,S. D.Shackelford,T .L.Wheeler,and M. Koohmaraie.2003. cattle Žnishingunit. Appl.Environ. Microbiol. 69:554–561. Seasonalprevalence ofShiga toxin– producing Escherichiacoli , in- 22.McEvoy, J. M.,A.M.Doherty,M. Finnerty,J. J.Sheridan,L. cludingO157:H7 and non-O157 serotypes, and Salmonella in com- McGuire,I. S.Blair,D.A.McDowell,and D. Harrington.2000. The mercial beef processingplants. J. FoodProt. 66:1978–1986. relationshipbetween hidecleanliness andbacterial numberson beef 3.Barkocy-Gallagher, G. A.,T .M.Arthur,G. R.Siragusa,J. E.Keen, carcasses at acommercial abattoir. Lett.Appl. Microbiol. 30:390– R.O.Elder,W .W.Laegreid,and M. Koohmaraie.2001. Genotypic 395. analyses of Escherichiacoli O157:H7and O157 nonmotile isolates 23.Nou, X., M. Rivera-Betancourt,J. M.Bosilevac,T .L.Wheeler,S. J.FoodProt., Vol. 67, No. 4 BACTERIALCONTAMINATION DURING BEEFPROCESSING 665

D.Shackelford,B. L.Gwartney,J. O.Reagan,and M. Koohmaraie. andM. Koohmaraie.1998. The incidence of Escherichiacoli on beef 2003.Effect ofchemical dehairingon the prevalence of Escherichia carcasses andits association with aerobic mesophilic plate count coli O157:H7and the levels ofaerobic bacteria and Enterobacteri- categories duringthe slaughter process. J. FoodProt. 61:1269–1274. aceae oncarcasses ina commercial beef processingplant. J. Food 29.Smulders, F .J.M.,P.Barendsen,J. G.vanLogtestijin, D. A.A. Prot. 66:2005–2009. Mossel,and G. M.vander Marel. 1986. Lactic acid:considerations 24.Phebus, R. K.,A.L.Nutsch,D. E.Schafer,R.C.Wilson,M. J. infavour of itsacceptance as ameat decontaminant. J.FoodTech- Riemann,J. D.Leising,C. L.Kastner,J.R.Wolf,and R. K.Prasai. nol. 21:419–436. 1997.Comparison of steam pasteurizationand other methods for 30.Smulders, F .J.M.,andG. G.Greer. 1998.Integrating microbial reductionof pathogens on surfaces offreshly slaughtered beef. J. decontaminationwith organic acids inHACCP programmes formus- Food Prot. 60:476–484. cle foods:prospects and controversies. Int.J. FoodMicrobiol. 44: 25.Reid, C.-A., A. Small,S. M.Avery,and S. Buncic.2002. Presence 149–169. offood-bornepathogens on cattle hides. FoodControl 13:411–415. 31.Tuttle, J., T.Gomez, M.P.Doyle,J. G.Wells,T .Zhao,R. V.Tauxe, 26.Riley, L. W.,R.S.Remis, S.D.Helgerson,H. B.McGee, J.G. andP .M.GrifŽn. 1999. Lessons from a large outbreakof Esche- Wells,B. R.Davis,R. J.Hebert,E. S.Olcott,L. M.Johnson,N. T. richia coli O157:H7infections: insights into the infectious dose and Hargrett,P .A.Blake,and M. L.Cohen.1983. Hemorrhagic colitis methodof widespread contamination of hamburger patties. Epide- associated witha rare Escherichiacoli serotype. N.Engl.J. Med. miol.Infect. 122:185–192. 308:681–685. 32.V anNetten, P .,J.H.Huisin’ t Veld,and D. A.A.Mossel.1994. The 27.SAS. 1998. SAS user’s guide:statistics. SASInstitute, Cary, N.C. effect oflactic acid decontaminationon the micro ora on meat. J. 28.Siragusa, G. R.,W .J.Dorsa,C. N.Cutter,G.L.Bennett,J. E.Keen, Food Safety 14:243–257.