Research ArticleBanked Human Milk and Quantitative Risk Assessment ofBacillus cereus Infection in Premature Infants ASimulation Study

Antoine Lewin 12 Gilles Delage1 France Bernier3 and Marc Germain 4

1Medical Affairs and Innovation Hema-Quebec Montreal QC Canada2Department of Obstetrics and Gynecology University of Sherbrooke Sherbrooke QC Canada3Quality and Regulatory Affairs Hema-Quebec Montreal QC Canada4Medical Affairs and Innovation Hema-Quebec Quebec QC Canada

Correspondence should be addressed to Antoine Lewin antoinelewinhema-quebecqcca

Received 5 October 2018 Accepted 31 December 2018 Published 3 February 2019

Academic Editor Bruno Pozzetto

Copyright copy 2019Antoine Lewin et al-is is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Background Banked human milk (BHM) offers potential health benefits to premature babies BHM is pasteurized to mitigateinfectious risks but pasteurization is ineffective against sporulating bacteria such as Bacillus cereus Sepsis related to Bacillus cereusin premature infants is severe and can often be fatal Even if a causal link has never been established BHM has been suggested as apotential source of infection in premature infantsObjective Our aim was to estimate the potential risk of Bacillus cereus infectionin preterm infants caused by the ingestion of contaminated pasteurized BHM using different post-pasteurization release criteria(ie 9 sampling of 100 microliters versus the HMBANA guideline of 1 sampling of 100microliters per pool) Methods In theabsence of scientific evidence regarding the risk of Bacillus cereus infection by the ingestion of BHM in premature infants riskassessment usingMonte Carlo simulation with the exponential dose-response model was performed-ree scenarios of infectiousrisk (annual incidence rate of 001 013 and 02) with 18 variations of the B cereus virulent dose (from 05 CFUml to200 CFUml) were simulated Results -e mean risk differential between the two methods of post-pasteurization bacteriologicalcontrol for realistic infectious doses of 30 to 200CFUml ranges from 0036 to 00054 047 to 0070 and 072 to 011 per millionservings for each of the three scenarios Conclusion Simulation highlights the very small risk of Bacillus cereus infection followingthe ingestion of pasteurized BHM even in the worst case scenarios and suggests that a 100-microliter sample for post-pasteurization culture is sufficient

1 Introduction

Human milk is the preferred feeding for infants because ofits known long-term benefits such as an improvement ofneurocognitive development [1ndash3] reduced risk for obesity[4] and protection of cardiovascular health outcomes [5]-ere are some clinical settings in which human milk isparticularly advantageous -is is the case of prematurechildren who are at risk of necrotizing enterocolitis (NEC) asevere and sometimes fatal complication Formula-fed in-fants are reported to have 6 to 20 times the risk of expe-riencing NEC compared with breast milk-fed infants [6]Clearly premature infants should be fed with their ownmotherrsquos milk whenever possible However banked human

milk (BHM) is the best alternative feeding for preterm in-fants when maternal milk is not available or is insufficient[7 8] One possible exception that has been raised by some isthe risk of systemic infection by Bacillus cereus throughBHM [9 10]

Bacillus cereus is a motile spore-forming Gram-positiverod aerobic or facultative anaerobic bacteria of the familyBacillaceae that is ubiquitous worldwide in the environment(air dust and water) [11] As a human pathogen the or-ganism is known for its role as a mediator of self-limitedfoodborne illness However Bacillus cereus is also known asa potential pathogen associated with severe local and sys-temic infections among immunosuppressed patients [12]Cases in premature infants are quite rare but neonatal sepsis

HindawiCanadian Journal of Infectious Diseases and Medical MicrobiologyVolume 2019 Article ID 6348281 8 pageshttpsdoiorg10115520196348281

related to Bacillus cereus is particularly severe and can lead todeath [13ndash15] Because sporulated form of this bacteriumcan resist the pasteurization process which is the methodused by most milk banks to safeguard against possiblecontamination some researchers have raised the possibilitythat donated milk is a possible source of contamination[9 13] To our knowledge the risk is purely theoretical sincethere has never been a single case of Bacillus cereus infectionin a preterm infant proven to be caused by BHM ingestion

Practices vary between milk banks concerning milkpool volume post-pasteurization routine bacterial assessmentbacteriological detection thresholds and concentration stan-dards for qualification [16] Nevertheless the large majority ofNorth American milk banks assess post-pasteurization bac-teriological control based on the Human Milk Banking As-sociation of North America (HMBANA) guideline (ie onesample of 100microliters per pool) [17] According to thisguideline any bacteriological growth is unacceptable after themilk has been pasteurized In the province of Quebec Canada(total population 8 millions) human milk bank post-pas-teurization microbiological analyses of pooled milk are rou-tinely made using 9 samples of 100 microliters per batchwhich is nine times larger than that recommended by theHMBANA -e pasteurized milk is tested by inoculating100microl of the lot that was pasteurized on 9 blood agar platesPlates used for the bacterial detection are incubated 48 hoursat 35plusmn 2degC Based on the results of this bacteriological as-sessment a batch showing the presence of Bacillus cereusregardless of the bacterial load does not qualify for distri-bution (Figure 1) Pre-pasteurization microbiological testingwas done for B cereus S aureus enterobacteria and totalcount screening for women entering the program Womenhaving the presence of B cereus or S aureus or enterobacteriaat more than 10000CFUml or a total count of more than100000CFUml have their milk rejected and are contacted bya nurse to identify and correct their hygiene deficienciesHistorically between 25 and 35 of the milk batches pro-duced by our milk bank have been disqualified because ofmicrobial contamination unsurprisingly Bacillus cereus wasthe microorganism found in 80ndash90 of the batches that weredisqualified We therefore considered the possibility ofaligning our practice with the HMBANArsquos recommended 100microliters of sample per batch -e question however wasraised as to whether the risk of having a smaller samplingvolume could reduce the sensitivity of our culture method to apoint that it would lead to an unacceptable level of risk of Bcereus infection Hence the risk assessment is presented in thispaper

2 Materials and Methods

In the absence of scientific evidence regarding the risk ofBacillus cereus infection caused by the ingestion of BHM inpremature infants Monte Carlo stochastic simulationreflecting variability and uncertainty of parameters wasused Our risk evaluation followed the quantitative micro-biological risk assessment (QMRA) method and thereforeundertook consecutive steps such as (1) hazard identifica-tion (described in the introduction section) (2) exposure

assessment (3) dose-response model and (4) risk charac-terization [18] Statistical analyses were performed usingSAS 94 software (SAS Institute Inc Cary NC USA)

21 Exposure Assessment -e aim of the exposure assess-ment section was to estimate the contamination concen-tration of Bacillus cereus in pasteurized pooledmilk acceptedafter bacteriological assessment and consumed by preterminfants -e model was based on modular process risk(MPR) methodology [19 20] as this chain divides naturallyinto a series of modules (ie pathways) which can be de-scribed by component processes (contamination mixingcontrol and detection) Figure 1 describes the donor milkchain model at Hema-Quebec

-e assumptions and inputs underlying our exposureassessment model were as follows

(1) Pooled Milk Volume 40000 milk donations weresimulated to yield a total of 10000 milk pools basedon observed internal data where (i) pooled milkvolume production during the year 2017 at Hema-Quebec was standardized with a mean volume of9500ml and SD 400ml and (ii) an average of 4different milk donors are necessary to constitute onemilk pool (given a mean milk volume per donationof 2375ml and SD 250ml)

(2) Pathogen Distribution -e total number of patho-gens per pasteurized milk pool (T) simulated wasbased on the hypothesis that (i) the prevalence ofsome level of contamination in any milk donation is100 given the ubiquity of Bacillus cereus in theenvironment and (ii) pasteurization (ε) has no ef-ficacy on reducing the concentration of Bacilluscereus contamination in a contaminated pool-erefore all simulated post-pasteurized milk poolswere considered contaminated and the rejection ratedid not depend on the presence or absence of Bacilluscereus but on the sensitivity of the bacteriologicalcontrol used Contamination concentration perdonation was simulated using lognormal distribu-tion of mean 42 CFUml and SD 213 -is simu-lation was based on our observed internal data(unpublished) with bacteriological post-pasteuriza-tion control using 9 samples of 100 microliters perpool around 30 of pools with a mean Bacilluscereus concentration of 13CFUml were detectedand rejected Moreover this simulated concentrationis consistent with the rejection rate reported in theliterature where when using 1 sample of100microliters for post-pasteurization bacteriologi-cal control around 10 of pools with Bacillus cereuswere detected and rejected [21]

Tpool 1113944Nc

i1(1minus ε) Vd times Cd( 1113857 (1)

where Tpool represents the total number of patho-gens per pasteurized milk pool Nc corresponds to

2 Canadian Journal of Infectious Diseases and Medical Microbiology

the number of donations contaminated withBacillus cereus per milk pool and Nc is equal tothe total number of contaminated donationsBecause the prevalence of Bacillus cereus in milkdonations was assumed to be 100 all donationswere assumed to be contaminated Vd and Cdrepresent milk volume and B cereus concentra-tion per donation respectively e eect ofpasteurization (ε) was set to zero as explainedearlier

(3) Contamination Detection and Pasteurized Milk PoolRejection Exposition of premature infants to Bacilluscereus depends on the probability of bacterial de-tection given the sensitivity of the bacteriologicalcontrol Assuming that bacteria are randomly dis-tributed in the pool following a Poisson distributionthe probability of detection using 100microliters wasdened as follows

Pdetec 1minus eminusλx (2)

where λ corresponds to the total CFU simulated bypool and x corresponds to the volume of samplingper pool en the number of rejected poolsdepending on the number of post-pasteurizationbacteriological sampling tests (k 1 or 9) follows abinomial distribution

NRejec binomial Pdetec k( ) (3)

(4) Feedings e distribution of donor milk bankserving per premature infant was determined usinginternal reports [22] e mean serving size was thenestimated to be 100ml

22 Dose-Response Model Because of the lack of in-formation regarding the relationship between the dose

Donor

Interviewquestionnairesserology

Authorized donation

Donor milk

No Milk pooling

Pasteurization

Bacteriological testlowast

1ndash3 CFUmax 3 agarB cereusS aereusEntero

gt3 CFUagar or1ndash3 CFUgt3 agar

0 CFU9 times 100microl

No

Human milk distribution Discarded milk

Preterm neonate

Discarded milk

Yes

Educqualif mother Yes

Bacteriological testNumber of total bacteria ge105ml or Entero ge104ml

or presence of Bacillus or S aureus

Maternal milk bank

Figure 1 Human milk process at Hema-Quebec milk bank Bacterial control is based on 9 samples of 100microliters per pasteurized milkpool lowastPost-bacteriological sampling was done plating 100 microl on 9 blood agar plates per pool Results were categorized into three scenariosFirst if there is no observed bacteriological growth in the 9 samples then pool milk is accepted for distribution Second if on a maximum of3 plates bacteriological growth was found with no more than 3CFUs per plate other than B cereus S aureus or enterobacteria then milkpool is accepted for distribution if one of the above mentioned bacteria was found within the three positive plates then pool milk wasdiscarded ird if bacteriological growth was observed on more than 3 plates or if more than 3CFUs were observed on a single platewhatever the bacterial species then milk pool was discarded

Canadian Journal of Infectious Diseases and Medical Microbiology 3

ingested and infection following the ingestion of bankedmilk contaminated by Bacillus cereus a full-risk assessmenthad to be modeled -e probability of infection resultingfrom the consumption of servings contaminated with Ba-cillus cereus (P) was based on an exponential dose-responsemodel because it allows inferring the dose-effect relationshipbased on certain epidemiological hypotheses

P 1minus eminusRND (4)

where ND represents the total number of B cereus in theserving and R is a constant specific to each pathogenic agentthat expresses the probability of being infected by a singlepathogenic microorganism present in the serving dependingon the virulence of the bacteria [23ndash27] In other words theR-value is the probability of interaction between the host andthe microorganism [28] By rearranging equation (4) R canbe expressed as follows

R minus[ln(1minusP)]

ND (5)

-e constant R was estimated using the approach de-scribed by Lindqvist and Westoo [25] in which the validityof the model depends on certain epidemiological assump-tions In the present study because an exponential dose-response model is a nonthreshold model and since thesimulation assumes that all serving are contaminated allpremature infants fed with banked milk are considered at

risk -e total number of preterm children exposed to BHMper year and the mean number of servings per infant wereestimated using internal distribution reports approximately1000 children were fed with milk supplied by the Hema-Quebec human milk bank with a mean of 40 portions peryear

In the absence of evidence regarding the incidence ofBacillus cereus infection in preterm infants fed with bankedmilk and the potential infectious dose of B cereus in pas-teurized milk we cannot use an independent estimate of theconstant R therefore variation in parameters for the riskcharacterization was made

23 Risk Characterization -ree scenarios were built toestimate the annual incidence of illness in preterm infantscaused by the ingestion of B cereus-contaminated bankedmilk As previously mentioned there is no evidence of acausal link between banked milk contamination and Bcereus infection in preterm infants we therefore had to makesome epidemiological assumptions for the various scenariosbeing considered As a factual basis for these scenarios weused data from the largest observed outbreak of Bacilluscereus infections in newborns in Quebec (unpublished) In2013ndash2014 at one hospital in our service area 7 preterminfants became infected An investigation concluded thatthese infections were likely to have been due to airbornecontamination that resulted from construction work near

Table 1 R-value for the exponential dose-response model based on simulated scenarios and variation of the infectious doselowast

Dose1 Total CFU per serving R scenario 1 R scenario 2 R scenario 3200 20000 12E minus 10 162E minus 9 250E minus 9100 10000 25E minus 10 325E minus 9 500E minus 975 7500 33E minus 10 433E minus 9 666E minus 965 6500 38E minus 10 500E minus 9 769E minus 955 5500 45E minus 10 591E minus 9 909E minus 950 5000 50E minus 10 650E minus 9 100E minus 845 4500 55E minus 10 722E minus 9 111E minus 840 4000 62E minus 10 813E minus 9 125E minus 835 3500 71E minus 10 929E minus 9 142E minus 830 3000 83E minus 10 108E minus 8 167E minus 820 2000 125E minus 9 163E minus 8 250E minus 815 1500 167E minus 9 217E minus 8 333E minus 812 1200 208E minus 9 271E minus 8 416E minus 89 900 278E minus 9 361E minus 8 556E minus 86 600 417E minus 9 542E minus 8 833E minus 83 300 833E minus 9 108E minus 7 167E minus 71 100 250E minus 8 325E minus 7 500E minus 705 50 500E minus 8 650E minus 7 100E minus 7lowastAnnual incidence of infection of 001 013 and 02 for scenarios 1 2 and 3 respectively 1Dose in CFUml

Table 2 Average contamination of Bacillus cereus in the accepted and rejected pools

Variables N Volume mean (in l) Concentration mean (SD)Lots not detected 1 sampling of 100 microl 9026 903 8574782 064 (168)Lots not detected 9 samplings of 100 microl 6996 700 6643546 021 (036)Lots detected 1 sampling of 100 microl 974 97 925459 3574 (32911)Lots detected 9 sampling of 100 microl 3004 300 2856695 1301 (18801)

4 Canadian Journal of Infectious Diseases and Medical Microbiology

the neonatal unit of the hospital However for the purpose of thesimulation we postulated that all these infections had beenacquired by uncultured contaminated humanmilk (a worst-casesituation) It should be noted that only one of these seven infantshad been exposed to banked human milk

Moreover to build the three scenarios we considered thefollowing denominator data (1) the total number of births inQuebec during that year (n 88867) (2) the total number ofbirths less than 25kg in Quebec (n 6464) and (3) the totalnumber of births at the hospital where the B cereus infectionsoccurred (n 3500) Depending on the denominator beingtaken this resulted in annual estimated incidence rates of 001013 and 02 for scenarios 1 2 and 3 respectively For eachscenario in order to estimate the uncertainty of the constant Rand because no information concerning the minimal infectiousdose of Bacillus cereus in human milk is available sensitivityanalyses were performed Eighteen variations of the infectiousdose of Bacillus cereus in human milk were simulated from05CFUml to 200CFUml -erefore R-value ranged from12E minus 10 to 500E minus 8 162E minus 9 to 650E minus 7 and 250E minus 9 to100E minus 7 for scenarios 1 2 and 3 respectively (Table 1)

3 Results

Ten thousand pools each constituting of 4 mothers weresimulatedwith amean volume of 95 liters of SD 04 and ameanBacillus cereus contamination concentration of 405CFUmlSD 103 and geometric mean of 016

31 Detection and Rejection Probability Following oursimulation using post-pasteurization bacterial controlsampling of 100microliters per milk pool 9026 of the10000 simulated pools containing Bacillus cereus would notbe detected and would be in the milk bank inventory fordistribution -e average concentration of Bacillus cereuscontamination in the pools qualified for distribution wouldbe 064CFUml In addition a rejection rate of 974 (974pools) would be observed with an average Bacillus cereuscontamination concentration of 3574 CFUml -is re-jection rate is similar to those reported by humanmilk banksthat use a post-pasteurization sample of 100microliters forculture [21 29] If post-pasteurization bacterial control

2

19

18

17

16

15

14

Mea

n ris

k di

ffere

ntia

l per

por

tion

(times10

ndash6)

13

12

11

1

09

08

07

06

05

04

03

02

01

0 20 40 60 80Infectious dose (CFUmL)

100 120 140 160 180 2000

Incidence = 02Incidence = 013Incidence = 001

Figure 2 Mean risk differential per portion using the two different post-pasteurization bacteriological control criteria Horizontal axiscorresponds to the 18 variations of the infectious doses according to the 3 different incidence scenarios identified by the correspondingcolors -e vertical axis corresponds to the mean risk differential per million portions

Canadian Journal of Infectious Diseases and Medical Microbiology 5

sampling consisted of nine sample of 100 microliters permilk pool 6996 pools with a mean Bacillus cereus con-centration of 021 would not be detected and would bequalified for distribution 3004 pools would have beendetected with an average Bacillus cereus concentration of1301CFUml and a rejection rate of 30 (Table 2) -issimulated rejection rate and the concentration in therejected pool are consistent with those of our internalobservations

In summary concentration of Bacillus cereus in poolsput into the inventory would be higher by 043CFUmlaverage when detected using one sample of 100microlitersversus 9 samples of 100 microliters Moreover the rejectedmilk volume difference would be 19312 liters corre-sponding to 203 of the total volume simulated

32 Residual Risk Differential Figure 2 represents the meanrisk differential per million between the two methods ofpost-pasteurization bacterial control (ie 1 sample versus 9samples of 100microliters) by scenario and infectious dose-is figure shows that for the vast majority of scenarios evenamong the most pessimistic the mean risk differential as-sociated with sampling of 1 test rather than 9 tests of100microliters is well below the threshold of 11 millionadditional infections per serving It is only when we supposethat the minimal infectious dose is 20CFUml or less that weobserve a risk difference greater than 11 million

Table 3 presents detailed data concerning the residualrisk scenarios comparing a 100-microliter sample to a 900-microliter sample for post-pasteurization culture whenvarying the infectious dose per serving -e mean riskdifferential for scenario 1 ranges from 215 to 000539 per

million for virulent dose of 05 to 200CFUml For scenarios2 and 3 the mean risk differential ranges from 278 to 0070per million and from 428 to 0108 respectively

4 Discussion

Proven invasive infections due to Bacillus cereus have beenreported among immunosuppressed patients particularlypreterm neonates but given the scarcity of cases thepathophysiology of these infections is not well knownMoreover assessment of the origin of contamination due toorganisms such as B cereus that are widely disseminated inthe environment is often difficult and may not yield anobvious source [11] Nonetheless in outbreak investigationscontamination through respiratory route was often docu-mented [11 15] Several studies have also assessed the pu-tative role of human milk in B cereus infection in prematureinfants however clinical strains of B cereus never matchedthe strains found in milk and therefore the tracking of theorigin of B cereus to contaminated bank milk was to datenever demonstrated [6 9]

Our analysis cannot make any statement regarding theactual risk of Bacillus cereus infection resulting from theingestion of contaminated banked human milk Even if therisk is real which remains unproven yet it is likely to beextremely small based on the various scenarios that weexplored in this simulation -e question of whether there isa real threat posed by Bacillus cereus to the safety of bankedhuman milk therefore remains unresolved

What our simulation shows is that even in very pes-simistic scenarios regarding a causal link between bankedmilk and B cereus infection a more stringent culture-sampling strategy makes very little difference in the risk

Table 3 Mean risk of infection by ingestion of banked maternal milk by serving according to different levels of Bacillus cereus infectiousdose in premature neonateslowast

Infectiousdose1

Scenario 1 Scenario 2 Scenario 3100 microl

sampling9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

200 799E minus 9 260E minus 9 539E minus 9 103E minus 7 334E minus 8 696E minus 8 159E minus 7 514E minus 8 108E minus 7100 160E minus 8 519E minus 9 108E minus 8 206E minus 7 669E minus 8 139E minus 7 317E minus 7 103E minus 7 214E minus 775 213E minus 8 693E minus 9 144E minus 8 275E minus 7 892E minus 8 186E minus 7 423E minus 7 137E minus 7 286E minus 765 246E minus 8 799E minus 9 166E minus 8 317E minus 7 103E minus 7 214E minus 7 488E minus 7 158E minus 7 33 E minus 755 290E minus 8 945E minus 9 195E minus 8 375E minus 7 122E minus 7 253E minus 7 576E minus 7 187E minus 7 389E minus 750 319E minus 8 104E minus 8 215E minus 8 412E minus 7 134E minus 7 278E minus 7 634E minus 7 206E minus 7 428E minus 745 355E minus 8 115E minus 8 240E minus 8 458E minus 7 149E minus 7 309E minus 7 705E minus 7 229E minus 7 476E minus 740 399E minus 8 130E minus 8 269E minus 8 515E minus 7 167E minus 7 348E minus 7 793E minus 7 257E minus 7 536E minus 735 456E minus 8 148E minus 8 308E minus 8 589E minus 7 191E minus 7 398E minus 7 906E minus 7 294E minus 7 612E minus 730 532E minus 8 173E minus 8 359E minus 8 687E minus 7 223E minus 7 464E minus 7 106E minus 6 343E minus 7 717E minus 720 799E minus 8 260E minus 8 539E minus 8 103E minus 6 334E minus 7 696E minus 7 158E minus 6 514E minus 7 107E minus 615 106E minus 7 346E minus 8 714E minus 8 137E minus 6 446E minus 7 924E minus 7 211E minus 6 686E minus 7 142E minus 612 133E minus 7 433E minus 8 897E minus 8 172E minus 6 557E minus 7 116E minus 6 264E minus 6 857E minus 7 178E minus 69 177E minus 7 577E minus 8 119E minus 7 229E minus 6 743E minus 7 155E minus 6 352E minus 6 114E minus 6 238E minus 66 266E minus 7 865E minus 8 179E minus 7 343E minus 6 111E minus 6 232E minus 6 528E minus 6 171E minus 6 357E minus 63 532E minus 7 173E minus 7 359E minus 7 687E minus 6 223E minus 6 464E minus 6 106E minus 5 343E minus 6 717E minus 61 160E minus 6 519E minus 7 108E minus 6 206E minus 5 669E minus 6 131E minus 5 317E minus 5 103E minus 5 214E minus 505 319E minus 6 104E minus 6 215E minus 6 412E minus 5 134E minus 5 278E minus 5 634E minus 5 206E minus 5 428E minus 5lowastSimulation scenario was made after post-pasteurized microbiological control using detection of 100 microliters vs 9times100 microliters per pool and for anannual incidence of infection of 001 013 and 02 corresponding to scenarios 1 2 and 3 respectively 1Infectious concentration in CFUml

6 Canadian Journal of Infectious Diseases and Medical Microbiology

incurred by exposed infants At the same time because ofthe ubiquitous nature of this bacterium increasing thesampling volume and thus the sensitivity of the culturemethod can lead to much higher rates of product dis-qualification More specifically our analysis shows that therisk difference per serving associated with a sampling of 1instead of 9 tests of 100 microliters is well below 11 millioninfections a risk that can be considered negligible In allscenarios it is only when the minimal infectious dose was20 CFUml or less for a 100ml serving that the risk dif-ferential was greater than 1 infection1 million portionssuch minimal infectious doses are clearly unrealistic forBacillus cereus Also average concentrations of Bacilluscereus found in the qualified milk batches would only beslightly increased from 021 to 064 CFUml which re-mains well below the standards applied to formula milk(ie 50 CFUg in Europe and 100 CFUg in the USAAustralia and New Zealand) [30ndash32] By comparison theliterature on human and animal exposure by inhalation todifferent doses of spores of Bacillus anthracis reports aminimal infectious dose of at least 600 CFU [33 34] If weassume the same minimal infectious dose of 600 CFU perserving for B cereus in BHM this would still be ten timeshigher than the average B cereus concentration remainingin the qualified milk pools Furthermore because of themuch lower potential for the pathogenicity of Bacilluscereus compared with B anthracis [33 34] this hypothesisis extremely aggressive Finally the gain in supply resultingfrom a less-stringent culture method and its potentialimpact on product availability for the prevention of nec-rotizing enterocolitis largely exceeds the very low theo-retical differential risk as calculated by these models[32 35 36] In our operational setting changing thesampling volume from 900 to 100 microliters is expected todecrease the rate of disqualified culture-positive poolsfrom 30 to 10 percent -is will represent a major im-provement in our capacity to meet the demand for pre-mature infants

To our knowledge this is the very first study usingMonte Carlo simulation which estimates the potential riskof Bacillus cereus infection in preterm infants fed withbanked human milk -e major limitation of our study isthat we had to estimate many of the parameters used in theabsence of empirical data concerning their value Also thesimulation was based on the premise that any bacterial loadin the milk pool is homogeneously distributed which isprobably the case in the vast majority of pools in our ex-perience However the possibility of this load being dis-tributed unevenly due to an inadequate milk mixing processin rare instances cannot be ruled out

5 Conclusion

Our simulation results suggest that if it exists at all the risk ofBacillus cereus infection following the ingestion of pasteurizedbanked milk is extremely small Our analysis also shows thateven in worst-case scenarios a 100-microliter sample forpost-pasteurization culture is amply sufficient to mitigate thisrisk a larger sampling volume would only lead to a higher rate

of disqualification for this important health-care resourcewithout having any significant positive impact on safety

Nevertheless despite no specific mention about Bacilluscereus contamination in the majority of the guidelines thissimulation highlights the importance of the combined use ofpasteurization and bacteriological investigation before hu-man milk distribution for extremely preterm babies Inaddition future studies investigating the association be-tween Bacillus cereus concentration and infection of pre-mature infants are needed

Data Availability

-e simulation code and the data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

-e authors declare that they have no conflicts of interest

References

[1] J-A B Bier T Oliver A E Ferguson and B R VohrldquoHuman milk improves cognitive and motor development ofpremature infants during infancyrdquo Journal of Human Lac-tation vol 18 no 4 pp 361ndash367 2016

[2] D L OrsquoConnor J Jacobs R Hall et al ldquoGrowth and de-velopment of premature infants fed predominantly humanmilk predominantly premature infant formula or a combi-nation of human milk and premature formulardquo Journal ofPediatric Gastroenterology and Nutrition vol 37 no 4pp 437ndash446 2003

[3] B R Vohr B B Poindexter A M Dusick et al ldquoBeneficialeffects of breast milk in the neonatal intensive care unit on thedevelopmental outcome of extremely low birth weight infantsat 18 months of agerdquo Pediatrics vol 118 no 1 pp e115ndashe1232006

[4] J Mahon L Claxton and H Wood ldquoModelling the cost-effectiveness of human milk and breastfeeding in preterminfants in the United Kingdomrdquo Health Economics Reviewvol 6 no 1 2016

[5] A Singhal T J Cole M Fewtrell and A Lucas ldquoBreastmilkfeeding and lipoprotein profile in adolescents born pretermfollow-up of a prospective randomised studyrdquo e Lancetvol 363 no 9421 pp 1571ndash1578 2004

[6] A M Wendelboe C Smelser C A Lucero andL C McDonald ldquoCluster of necrotizing enterocolitis in aneonatal intensive care unit New Mexico 2007rdquo AmericanJournal of Infection Control vol 38 no 2 pp 144ndash148 2010

[7] S Arslanoglu G E Moro and E E Ziegler ldquo-e WapmWorking Group on N Optimization of human milk fortifi-cation for preterm infants new concepts and recommenda-tionsrdquo Journal of Perinatal Medicine vol 38 no 3pp 233ndash238 2010

[8] L M Gartner J Morton R A Lawrence et al ldquoBreastfeedingand the use of human milkrdquo Pediatrics vol 115 no 2pp 496ndash506 2005

[9] J-W Decousser N Ramarao C Duport et al ldquoBacillus cereusand severe intestinal infections in preterm neonates putativerole of pooled breast milkrdquo American Journal of InfectionControl vol 41 no 10 pp 918ndash921 2013

Canadian Journal of Infectious Diseases and Medical Microbiology 7

[10] S Landers and K Updegrove ldquoBacteriological screening ofdonor human milk before and after Holder pasteurizationrdquoBreastfeeding Medicine vol 5 no 3 pp 117ndash121 2010

[11] N J Hilliard R L Schelonka and K B Waites ldquoBacilluscereus bacteremia in a preterm neonaterdquo Journal of ClinicalMicrobiology vol 41 no 7 pp 3441ndash3444 2003

[12] E J Bottone ldquoBacillus cereus a volatile human pathogenrdquoClinical Microbiology Reviews vol 23 no 2 pp 382ndash3982010

[13] R Lotte A-L Herisse Y Berrouane et al ldquoVirulence analysisof Bacillus cereus isolated after death of preterm neonatesnice France 2013rdquo Emerging Infectious Diseases vol 23no 5 pp 845ndash848 2017

[14] N Ramarao L Belotti S Deboscker et al ldquoTwo unrelatedepisodes of Bacillus cereus bacteremia in a neonatal intensivecare unitrdquo American Journal of Infection Control vol 42no 6 pp 694-695 2014

[15] W C Van Der Zwet G A Parlevliet P H Savelkoul et alldquoOutbreak of Bacillus cereus infections in a neonatal intensivecare unit traced to balloons used in manual ventilationrdquoJournal of Clinical Microbiology vol 38 no 11 pp 4131ndash41362000

[16] Y Almutawif B Hartmann M Lloyd W Erber andD Geddes ldquoA retrospective audit of bacterial culture results ofdonated human milk in Perth Western Australiardquo EarlyHuman Development vol 105 pp 1ndash6 2017

[17] Centre for Clinical Practice at NICE (UK) Donor Breast MilkBankse Operation of Donor Milk Bank Services 20120210ed National Institute for Health and Clinical Excellence (UK)London UK 2010

[18] C Haas J Rose and C Gerba Quantitative Microbial RiskAssessment John Wiley amp Sons Hoboken NJ USA 2ndedition 2014

[19] M J Nauta A Modular Process Risk Model Structure forQuantitative Microbiological Risk Assessment and Its Applica-tion in an Exposure Assessment of Bacillus cereus in a REPFEDNational Institute of Public Health and the EnvironmentAmsterdam Netherlands 2001 Contract No 149106 007

[20] M Nauta ldquoMicrobiological risk assessment models for par-titioning and mixing during food handlingrdquo InternationalJournal of Food Microbiology vol 100 no 1ndash3 pp 311ndash3222005

[21] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination du lait maternel par une floreaerobie evaluation des pertes pour un lactariumrdquo Archives dePediatrie vol 22 no 5 pp 461ndash467 2015

[22] Hema-Quebec Guide drsquoinformation sur le don de laitmaternel Banque publique de lait maternel Hema-QuebecMontreal Canada 2014

[23] C N Haas ldquoEstimation of risk due to low doses of micro-organisms a comparison of alternative methodologiesrdquoAmerican Journal of Epidemiology vol 118 no 4 pp 573ndash582 1983

[24] C N Haas ldquoMicrobial dose response modeling past presentand futurerdquo Environmental Science and Technology vol 49no 3 pp 1245ndash1259 2015

[25] R Lindqvist and A Westoo ldquoQuantitative risk assessment forListeria monocytogenes in smoked or gravad salmon andrainbow trout in Swedenrdquo International Journal of FoodMicrobiology vol 58 no 3 pp 181ndash196 2000

[26] D Vose Risk Analysis A Quantitative Guide John Wiley ampSons New York NY USA 3rd edition 2008

[27] P Acai L Valik and D Liptakova ldquoQuantitative risk as-sessment of Bacillus cereus in pasteurised milk produced in

the Slovak Republicrdquo Czech Journal of Food Sciences vol 32no 2 pp 122ndash131 2014

[28] S Regli J B Rose C N Haas and C P Gerba ldquoModeling therisk from Giardia and viruses in drinking waterrdquo Journal -American Water Works Association vol 83 no 11 pp 76ndash841991

[29] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination of human milk with aerobicflora evaluation of losses for a humanmilk bankrdquo Archives dePediatrie 22 22 pp 461ndash467 2015 in French

[30] US Food and Drug Administration Food CompositionStandards Labeling And Economics Compliance ProgramGuidance Manual Chapter 21ndashFood Composition StandardsLabeling and Economics US Food and Drug AdministrationSilver Spring MD USA 2006 httpswwwfdagovdownloadsFoodComplianceEnforcementUCM073339pdf

[31] FSANZ Bacillus cereus Limits in Infant Formula FSANZCanberra BC Autralia 2004 httpswwwfoodstandardsgovaucodeapplicationsdocumentsA454_B_cereus_FARpdf

[32] Commission of the European Communities CommissionRegulation (EC) no 14412007 Amending Regulation (EC) No20732005 on Microbiological Criteria for Foodstuffs Com-mission of the European Communities Luxembourg 2007

[33] M L Cohen and TWhalen ldquoImplications of low level humanexposure to respirable B Anthracisrdquo Applied Biosafetyvol 12 no 2 pp 109ndash115 2016

[34] M E Coleman B -ran S S Morse M Hugh-Jones andS Massulik ldquoInhalation anthrax dose response and riskanalysisrdquo Biosecurity and Bioterrorism Biodefense StrategyPractice and Science vol 6 no 2 pp 147ndash160 2008

[35] R J Schanler R J Shulman and C Lau ldquoFeeding strategiesfor premature infants beneficial outcomes of feeding fortifiedhuman milk versus preterm formulardquo Pediatrics vol 103no 6 pp 1150ndash1157 1999

[36] P M Sisk C A Lovelady R G Dillard K J Gruber andT M OrsquoShea ldquoEarly human milk feeding is associated with alower risk of necrotizing enterocolitis in very low birth weightinfantsrdquo Journal of Perinatology vol 27 no 7 pp 428ndash4332007

8 Canadian Journal of Infectious Diseases and Medical Microbiology

the number of donations contaminated withBacillus cereus per milk pool and Nc is equal tothe total number of contaminated donationsBecause the prevalence of Bacillus cereus in milkdonations was assumed to be 100 all donationswere assumed to be contaminated Vd and Cdrepresent milk volume and B cereus concentra-tion per donation respectively e eect ofpasteurization (ε) was set to zero as explainedearlier

(3) Contamination Detection and Pasteurized Milk PoolRejection Exposition of premature infants to Bacilluscereus depends on the probability of bacterial de-tection given the sensitivity of the bacteriologicalcontrol Assuming that bacteria are randomly dis-tributed in the pool following a Poisson distributionthe probability of detection using 100microliters wasdened as follows

Pdetec 1minus eminusλx (2)

where λ corresponds to the total CFU simulated bypool and x corresponds to the volume of samplingper pool en the number of rejected poolsdepending on the number of post-pasteurizationbacteriological sampling tests (k 1 or 9) follows abinomial distribution

NRejec binomial Pdetec k( ) (3)

(4) Feedings e distribution of donor milk bankserving per premature infant was determined usinginternal reports [22] e mean serving size was thenestimated to be 100ml

22 Dose-Response Model Because of the lack of in-formation regarding the relationship between the dose

Donor

Interviewquestionnairesserology

Authorized donation

Donor milk

No Milk pooling

Pasteurization

Bacteriological testlowast

1ndash3 CFUmax 3 agarB cereusS aereusEntero

gt3 CFUagar or1ndash3 CFUgt3 agar

0 CFU9 times 100microl

No

Human milk distribution Discarded milk

Preterm neonate

Discarded milk

Yes

Educqualif mother Yes

Bacteriological testNumber of total bacteria ge105ml or Entero ge104ml

or presence of Bacillus or S aureus

Maternal milk bank

Figure 1 Human milk process at Hema-Quebec milk bank Bacterial control is based on 9 samples of 100microliters per pasteurized milkpool lowastPost-bacteriological sampling was done plating 100 microl on 9 blood agar plates per pool Results were categorized into three scenariosFirst if there is no observed bacteriological growth in the 9 samples then pool milk is accepted for distribution Second if on a maximum of3 plates bacteriological growth was found with no more than 3CFUs per plate other than B cereus S aureus or enterobacteria then milkpool is accepted for distribution if one of the above mentioned bacteria was found within the three positive plates then pool milk wasdiscarded ird if bacteriological growth was observed on more than 3 plates or if more than 3CFUs were observed on a single platewhatever the bacterial species then milk pool was discarded

Canadian Journal of Infectious Diseases and Medical Microbiology 3

ingested and infection following the ingestion of bankedmilk contaminated by Bacillus cereus a full-risk assessmenthad to be modeled -e probability of infection resultingfrom the consumption of servings contaminated with Ba-cillus cereus (P) was based on an exponential dose-responsemodel because it allows inferring the dose-effect relationshipbased on certain epidemiological hypotheses

P 1minus eminusRND (4)

where ND represents the total number of B cereus in theserving and R is a constant specific to each pathogenic agentthat expresses the probability of being infected by a singlepathogenic microorganism present in the serving dependingon the virulence of the bacteria [23ndash27] In other words theR-value is the probability of interaction between the host andthe microorganism [28] By rearranging equation (4) R canbe expressed as follows

R minus[ln(1minusP)]

ND (5)

-e constant R was estimated using the approach de-scribed by Lindqvist and Westoo [25] in which the validityof the model depends on certain epidemiological assump-tions In the present study because an exponential dose-response model is a nonthreshold model and since thesimulation assumes that all serving are contaminated allpremature infants fed with banked milk are considered at

risk -e total number of preterm children exposed to BHMper year and the mean number of servings per infant wereestimated using internal distribution reports approximately1000 children were fed with milk supplied by the Hema-Quebec human milk bank with a mean of 40 portions peryear

In the absence of evidence regarding the incidence ofBacillus cereus infection in preterm infants fed with bankedmilk and the potential infectious dose of B cereus in pas-teurized milk we cannot use an independent estimate of theconstant R therefore variation in parameters for the riskcharacterization was made

23 Risk Characterization -ree scenarios were built toestimate the annual incidence of illness in preterm infantscaused by the ingestion of B cereus-contaminated bankedmilk As previously mentioned there is no evidence of acausal link between banked milk contamination and Bcereus infection in preterm infants we therefore had to makesome epidemiological assumptions for the various scenariosbeing considered As a factual basis for these scenarios weused data from the largest observed outbreak of Bacilluscereus infections in newborns in Quebec (unpublished) In2013ndash2014 at one hospital in our service area 7 preterminfants became infected An investigation concluded thatthese infections were likely to have been due to airbornecontamination that resulted from construction work near

Table 1 R-value for the exponential dose-response model based on simulated scenarios and variation of the infectious doselowast

Dose1 Total CFU per serving R scenario 1 R scenario 2 R scenario 3200 20000 12E minus 10 162E minus 9 250E minus 9100 10000 25E minus 10 325E minus 9 500E minus 975 7500 33E minus 10 433E minus 9 666E minus 965 6500 38E minus 10 500E minus 9 769E minus 955 5500 45E minus 10 591E minus 9 909E minus 950 5000 50E minus 10 650E minus 9 100E minus 845 4500 55E minus 10 722E minus 9 111E minus 840 4000 62E minus 10 813E minus 9 125E minus 835 3500 71E minus 10 929E minus 9 142E minus 830 3000 83E minus 10 108E minus 8 167E minus 820 2000 125E minus 9 163E minus 8 250E minus 815 1500 167E minus 9 217E minus 8 333E minus 812 1200 208E minus 9 271E minus 8 416E minus 89 900 278E minus 9 361E minus 8 556E minus 86 600 417E minus 9 542E minus 8 833E minus 83 300 833E minus 9 108E minus 7 167E minus 71 100 250E minus 8 325E minus 7 500E minus 705 50 500E minus 8 650E minus 7 100E minus 7lowastAnnual incidence of infection of 001 013 and 02 for scenarios 1 2 and 3 respectively 1Dose in CFUml

Table 2 Average contamination of Bacillus cereus in the accepted and rejected pools

Variables N Volume mean (in l) Concentration mean (SD)Lots not detected 1 sampling of 100 microl 9026 903 8574782 064 (168)Lots not detected 9 samplings of 100 microl 6996 700 6643546 021 (036)Lots detected 1 sampling of 100 microl 974 97 925459 3574 (32911)Lots detected 9 sampling of 100 microl 3004 300 2856695 1301 (18801)

4 Canadian Journal of Infectious Diseases and Medical Microbiology

the neonatal unit of the hospital However for the purpose of thesimulation we postulated that all these infections had beenacquired by uncultured contaminated humanmilk (a worst-casesituation) It should be noted that only one of these seven infantshad been exposed to banked human milk

Moreover to build the three scenarios we considered thefollowing denominator data (1) the total number of births inQuebec during that year (n 88867) (2) the total number ofbirths less than 25kg in Quebec (n 6464) and (3) the totalnumber of births at the hospital where the B cereus infectionsoccurred (n 3500) Depending on the denominator beingtaken this resulted in annual estimated incidence rates of 001013 and 02 for scenarios 1 2 and 3 respectively For eachscenario in order to estimate the uncertainty of the constant Rand because no information concerning the minimal infectiousdose of Bacillus cereus in human milk is available sensitivityanalyses were performed Eighteen variations of the infectiousdose of Bacillus cereus in human milk were simulated from05CFUml to 200CFUml -erefore R-value ranged from12E minus 10 to 500E minus 8 162E minus 9 to 650E minus 7 and 250E minus 9 to100E minus 7 for scenarios 1 2 and 3 respectively (Table 1)

3 Results

Ten thousand pools each constituting of 4 mothers weresimulatedwith amean volume of 95 liters of SD 04 and ameanBacillus cereus contamination concentration of 405CFUmlSD 103 and geometric mean of 016

31 Detection and Rejection Probability Following oursimulation using post-pasteurization bacterial controlsampling of 100microliters per milk pool 9026 of the10000 simulated pools containing Bacillus cereus would notbe detected and would be in the milk bank inventory fordistribution -e average concentration of Bacillus cereuscontamination in the pools qualified for distribution wouldbe 064CFUml In addition a rejection rate of 974 (974pools) would be observed with an average Bacillus cereuscontamination concentration of 3574 CFUml -is re-jection rate is similar to those reported by humanmilk banksthat use a post-pasteurization sample of 100microliters forculture [21 29] If post-pasteurization bacterial control

2

19

18

17

16

15

14

Mea

n ris

k di

ffere

ntia

l per

por

tion

(times10

ndash6)

13

12

11

1

09

08

07

06

05

04

03

02

01

0 20 40 60 80Infectious dose (CFUmL)

100 120 140 160 180 2000

Incidence = 02Incidence = 013Incidence = 001

Figure 2 Mean risk differential per portion using the two different post-pasteurization bacteriological control criteria Horizontal axiscorresponds to the 18 variations of the infectious doses according to the 3 different incidence scenarios identified by the correspondingcolors -e vertical axis corresponds to the mean risk differential per million portions

Canadian Journal of Infectious Diseases and Medical Microbiology 5

sampling consisted of nine sample of 100 microliters permilk pool 6996 pools with a mean Bacillus cereus con-centration of 021 would not be detected and would bequalified for distribution 3004 pools would have beendetected with an average Bacillus cereus concentration of1301CFUml and a rejection rate of 30 (Table 2) -issimulated rejection rate and the concentration in therejected pool are consistent with those of our internalobservations

In summary concentration of Bacillus cereus in poolsput into the inventory would be higher by 043CFUmlaverage when detected using one sample of 100microlitersversus 9 samples of 100 microliters Moreover the rejectedmilk volume difference would be 19312 liters corre-sponding to 203 of the total volume simulated

32 Residual Risk Differential Figure 2 represents the meanrisk differential per million between the two methods ofpost-pasteurization bacterial control (ie 1 sample versus 9samples of 100microliters) by scenario and infectious dose-is figure shows that for the vast majority of scenarios evenamong the most pessimistic the mean risk differential as-sociated with sampling of 1 test rather than 9 tests of100microliters is well below the threshold of 11 millionadditional infections per serving It is only when we supposethat the minimal infectious dose is 20CFUml or less that weobserve a risk difference greater than 11 million

Table 3 presents detailed data concerning the residualrisk scenarios comparing a 100-microliter sample to a 900-microliter sample for post-pasteurization culture whenvarying the infectious dose per serving -e mean riskdifferential for scenario 1 ranges from 215 to 000539 per

million for virulent dose of 05 to 200CFUml For scenarios2 and 3 the mean risk differential ranges from 278 to 0070per million and from 428 to 0108 respectively

4 Discussion

Proven invasive infections due to Bacillus cereus have beenreported among immunosuppressed patients particularlypreterm neonates but given the scarcity of cases thepathophysiology of these infections is not well knownMoreover assessment of the origin of contamination due toorganisms such as B cereus that are widely disseminated inthe environment is often difficult and may not yield anobvious source [11] Nonetheless in outbreak investigationscontamination through respiratory route was often docu-mented [11 15] Several studies have also assessed the pu-tative role of human milk in B cereus infection in prematureinfants however clinical strains of B cereus never matchedthe strains found in milk and therefore the tracking of theorigin of B cereus to contaminated bank milk was to datenever demonstrated [6 9]

Our analysis cannot make any statement regarding theactual risk of Bacillus cereus infection resulting from theingestion of contaminated banked human milk Even if therisk is real which remains unproven yet it is likely to beextremely small based on the various scenarios that weexplored in this simulation -e question of whether there isa real threat posed by Bacillus cereus to the safety of bankedhuman milk therefore remains unresolved

What our simulation shows is that even in very pes-simistic scenarios regarding a causal link between bankedmilk and B cereus infection a more stringent culture-sampling strategy makes very little difference in the risk

Table 3 Mean risk of infection by ingestion of banked maternal milk by serving according to different levels of Bacillus cereus infectiousdose in premature neonateslowast

Infectiousdose1

Scenario 1 Scenario 2 Scenario 3100 microl

sampling9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

200 799E minus 9 260E minus 9 539E minus 9 103E minus 7 334E minus 8 696E minus 8 159E minus 7 514E minus 8 108E minus 7100 160E minus 8 519E minus 9 108E minus 8 206E minus 7 669E minus 8 139E minus 7 317E minus 7 103E minus 7 214E minus 775 213E minus 8 693E minus 9 144E minus 8 275E minus 7 892E minus 8 186E minus 7 423E minus 7 137E minus 7 286E minus 765 246E minus 8 799E minus 9 166E minus 8 317E minus 7 103E minus 7 214E minus 7 488E minus 7 158E minus 7 33 E minus 755 290E minus 8 945E minus 9 195E minus 8 375E minus 7 122E minus 7 253E minus 7 576E minus 7 187E minus 7 389E minus 750 319E minus 8 104E minus 8 215E minus 8 412E minus 7 134E minus 7 278E minus 7 634E minus 7 206E minus 7 428E minus 745 355E minus 8 115E minus 8 240E minus 8 458E minus 7 149E minus 7 309E minus 7 705E minus 7 229E minus 7 476E minus 740 399E minus 8 130E minus 8 269E minus 8 515E minus 7 167E minus 7 348E minus 7 793E minus 7 257E minus 7 536E minus 735 456E minus 8 148E minus 8 308E minus 8 589E minus 7 191E minus 7 398E minus 7 906E minus 7 294E minus 7 612E minus 730 532E minus 8 173E minus 8 359E minus 8 687E minus 7 223E minus 7 464E minus 7 106E minus 6 343E minus 7 717E minus 720 799E minus 8 260E minus 8 539E minus 8 103E minus 6 334E minus 7 696E minus 7 158E minus 6 514E minus 7 107E minus 615 106E minus 7 346E minus 8 714E minus 8 137E minus 6 446E minus 7 924E minus 7 211E minus 6 686E minus 7 142E minus 612 133E minus 7 433E minus 8 897E minus 8 172E minus 6 557E minus 7 116E minus 6 264E minus 6 857E minus 7 178E minus 69 177E minus 7 577E minus 8 119E minus 7 229E minus 6 743E minus 7 155E minus 6 352E minus 6 114E minus 6 238E minus 66 266E minus 7 865E minus 8 179E minus 7 343E minus 6 111E minus 6 232E minus 6 528E minus 6 171E minus 6 357E minus 63 532E minus 7 173E minus 7 359E minus 7 687E minus 6 223E minus 6 464E minus 6 106E minus 5 343E minus 6 717E minus 61 160E minus 6 519E minus 7 108E minus 6 206E minus 5 669E minus 6 131E minus 5 317E minus 5 103E minus 5 214E minus 505 319E minus 6 104E minus 6 215E minus 6 412E minus 5 134E minus 5 278E minus 5 634E minus 5 206E minus 5 428E minus 5lowastSimulation scenario was made after post-pasteurized microbiological control using detection of 100 microliters vs 9times100 microliters per pool and for anannual incidence of infection of 001 013 and 02 corresponding to scenarios 1 2 and 3 respectively 1Infectious concentration in CFUml

6 Canadian Journal of Infectious Diseases and Medical Microbiology

incurred by exposed infants At the same time because ofthe ubiquitous nature of this bacterium increasing thesampling volume and thus the sensitivity of the culturemethod can lead to much higher rates of product dis-qualification More specifically our analysis shows that therisk difference per serving associated with a sampling of 1instead of 9 tests of 100 microliters is well below 11 millioninfections a risk that can be considered negligible In allscenarios it is only when the minimal infectious dose was20 CFUml or less for a 100ml serving that the risk dif-ferential was greater than 1 infection1 million portionssuch minimal infectious doses are clearly unrealistic forBacillus cereus Also average concentrations of Bacilluscereus found in the qualified milk batches would only beslightly increased from 021 to 064 CFUml which re-mains well below the standards applied to formula milk(ie 50 CFUg in Europe and 100 CFUg in the USAAustralia and New Zealand) [30ndash32] By comparison theliterature on human and animal exposure by inhalation todifferent doses of spores of Bacillus anthracis reports aminimal infectious dose of at least 600 CFU [33 34] If weassume the same minimal infectious dose of 600 CFU perserving for B cereus in BHM this would still be ten timeshigher than the average B cereus concentration remainingin the qualified milk pools Furthermore because of themuch lower potential for the pathogenicity of Bacilluscereus compared with B anthracis [33 34] this hypothesisis extremely aggressive Finally the gain in supply resultingfrom a less-stringent culture method and its potentialimpact on product availability for the prevention of nec-rotizing enterocolitis largely exceeds the very low theo-retical differential risk as calculated by these models[32 35 36] In our operational setting changing thesampling volume from 900 to 100 microliters is expected todecrease the rate of disqualified culture-positive poolsfrom 30 to 10 percent -is will represent a major im-provement in our capacity to meet the demand for pre-mature infants

To our knowledge this is the very first study usingMonte Carlo simulation which estimates the potential riskof Bacillus cereus infection in preterm infants fed withbanked human milk -e major limitation of our study isthat we had to estimate many of the parameters used in theabsence of empirical data concerning their value Also thesimulation was based on the premise that any bacterial loadin the milk pool is homogeneously distributed which isprobably the case in the vast majority of pools in our ex-perience However the possibility of this load being dis-tributed unevenly due to an inadequate milk mixing processin rare instances cannot be ruled out

5 Conclusion

Our simulation results suggest that if it exists at all the risk ofBacillus cereus infection following the ingestion of pasteurizedbanked milk is extremely small Our analysis also shows thateven in worst-case scenarios a 100-microliter sample forpost-pasteurization culture is amply sufficient to mitigate thisrisk a larger sampling volume would only lead to a higher rate

of disqualification for this important health-care resourcewithout having any significant positive impact on safety

Nevertheless despite no specific mention about Bacilluscereus contamination in the majority of the guidelines thissimulation highlights the importance of the combined use ofpasteurization and bacteriological investigation before hu-man milk distribution for extremely preterm babies Inaddition future studies investigating the association be-tween Bacillus cereus concentration and infection of pre-mature infants are needed

Data Availability

-e simulation code and the data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

-e authors declare that they have no conflicts of interest

References

[1] J-A B Bier T Oliver A E Ferguson and B R VohrldquoHuman milk improves cognitive and motor development ofpremature infants during infancyrdquo Journal of Human Lac-tation vol 18 no 4 pp 361ndash367 2016

[2] D L OrsquoConnor J Jacobs R Hall et al ldquoGrowth and de-velopment of premature infants fed predominantly humanmilk predominantly premature infant formula or a combi-nation of human milk and premature formulardquo Journal ofPediatric Gastroenterology and Nutrition vol 37 no 4pp 437ndash446 2003

[3] B R Vohr B B Poindexter A M Dusick et al ldquoBeneficialeffects of breast milk in the neonatal intensive care unit on thedevelopmental outcome of extremely low birth weight infantsat 18 months of agerdquo Pediatrics vol 118 no 1 pp e115ndashe1232006

[4] J Mahon L Claxton and H Wood ldquoModelling the cost-effectiveness of human milk and breastfeeding in preterminfants in the United Kingdomrdquo Health Economics Reviewvol 6 no 1 2016

[5] A Singhal T J Cole M Fewtrell and A Lucas ldquoBreastmilkfeeding and lipoprotein profile in adolescents born pretermfollow-up of a prospective randomised studyrdquo e Lancetvol 363 no 9421 pp 1571ndash1578 2004

[6] A M Wendelboe C Smelser C A Lucero andL C McDonald ldquoCluster of necrotizing enterocolitis in aneonatal intensive care unit New Mexico 2007rdquo AmericanJournal of Infection Control vol 38 no 2 pp 144ndash148 2010

[7] S Arslanoglu G E Moro and E E Ziegler ldquo-e WapmWorking Group on N Optimization of human milk fortifi-cation for preterm infants new concepts and recommenda-tionsrdquo Journal of Perinatal Medicine vol 38 no 3pp 233ndash238 2010

[8] L M Gartner J Morton R A Lawrence et al ldquoBreastfeedingand the use of human milkrdquo Pediatrics vol 115 no 2pp 496ndash506 2005

[9] J-W Decousser N Ramarao C Duport et al ldquoBacillus cereusand severe intestinal infections in preterm neonates putativerole of pooled breast milkrdquo American Journal of InfectionControl vol 41 no 10 pp 918ndash921 2013

Canadian Journal of Infectious Diseases and Medical Microbiology 7

[10] S Landers and K Updegrove ldquoBacteriological screening ofdonor human milk before and after Holder pasteurizationrdquoBreastfeeding Medicine vol 5 no 3 pp 117ndash121 2010

[11] N J Hilliard R L Schelonka and K B Waites ldquoBacilluscereus bacteremia in a preterm neonaterdquo Journal of ClinicalMicrobiology vol 41 no 7 pp 3441ndash3444 2003

[12] E J Bottone ldquoBacillus cereus a volatile human pathogenrdquoClinical Microbiology Reviews vol 23 no 2 pp 382ndash3982010

[13] R Lotte A-L Herisse Y Berrouane et al ldquoVirulence analysisof Bacillus cereus isolated after death of preterm neonatesnice France 2013rdquo Emerging Infectious Diseases vol 23no 5 pp 845ndash848 2017

[14] N Ramarao L Belotti S Deboscker et al ldquoTwo unrelatedepisodes of Bacillus cereus bacteremia in a neonatal intensivecare unitrdquo American Journal of Infection Control vol 42no 6 pp 694-695 2014

[15] W C Van Der Zwet G A Parlevliet P H Savelkoul et alldquoOutbreak of Bacillus cereus infections in a neonatal intensivecare unit traced to balloons used in manual ventilationrdquoJournal of Clinical Microbiology vol 38 no 11 pp 4131ndash41362000

[16] Y Almutawif B Hartmann M Lloyd W Erber andD Geddes ldquoA retrospective audit of bacterial culture results ofdonated human milk in Perth Western Australiardquo EarlyHuman Development vol 105 pp 1ndash6 2017

[17] Centre for Clinical Practice at NICE (UK) Donor Breast MilkBankse Operation of Donor Milk Bank Services 20120210ed National Institute for Health and Clinical Excellence (UK)London UK 2010

[18] C Haas J Rose and C Gerba Quantitative Microbial RiskAssessment John Wiley amp Sons Hoboken NJ USA 2ndedition 2014

[19] M J Nauta A Modular Process Risk Model Structure forQuantitative Microbiological Risk Assessment and Its Applica-tion in an Exposure Assessment of Bacillus cereus in a REPFEDNational Institute of Public Health and the EnvironmentAmsterdam Netherlands 2001 Contract No 149106 007

[20] M Nauta ldquoMicrobiological risk assessment models for par-titioning and mixing during food handlingrdquo InternationalJournal of Food Microbiology vol 100 no 1ndash3 pp 311ndash3222005

[21] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination du lait maternel par une floreaerobie evaluation des pertes pour un lactariumrdquo Archives dePediatrie vol 22 no 5 pp 461ndash467 2015

[22] Hema-Quebec Guide drsquoinformation sur le don de laitmaternel Banque publique de lait maternel Hema-QuebecMontreal Canada 2014

[23] C N Haas ldquoEstimation of risk due to low doses of micro-organisms a comparison of alternative methodologiesrdquoAmerican Journal of Epidemiology vol 118 no 4 pp 573ndash582 1983

[24] C N Haas ldquoMicrobial dose response modeling past presentand futurerdquo Environmental Science and Technology vol 49no 3 pp 1245ndash1259 2015

[25] R Lindqvist and A Westoo ldquoQuantitative risk assessment forListeria monocytogenes in smoked or gravad salmon andrainbow trout in Swedenrdquo International Journal of FoodMicrobiology vol 58 no 3 pp 181ndash196 2000

[26] D Vose Risk Analysis A Quantitative Guide John Wiley ampSons New York NY USA 3rd edition 2008

[27] P Acai L Valik and D Liptakova ldquoQuantitative risk as-sessment of Bacillus cereus in pasteurised milk produced in

the Slovak Republicrdquo Czech Journal of Food Sciences vol 32no 2 pp 122ndash131 2014

[28] S Regli J B Rose C N Haas and C P Gerba ldquoModeling therisk from Giardia and viruses in drinking waterrdquo Journal -American Water Works Association vol 83 no 11 pp 76ndash841991

[29] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination of human milk with aerobicflora evaluation of losses for a humanmilk bankrdquo Archives dePediatrie 22 22 pp 461ndash467 2015 in French

[30] US Food and Drug Administration Food CompositionStandards Labeling And Economics Compliance ProgramGuidance Manual Chapter 21ndashFood Composition StandardsLabeling and Economics US Food and Drug AdministrationSilver Spring MD USA 2006 httpswwwfdagovdownloadsFoodComplianceEnforcementUCM073339pdf

[31] FSANZ Bacillus cereus Limits in Infant Formula FSANZCanberra BC Autralia 2004 httpswwwfoodstandardsgovaucodeapplicationsdocumentsA454_B_cereus_FARpdf

[32] Commission of the European Communities CommissionRegulation (EC) no 14412007 Amending Regulation (EC) No20732005 on Microbiological Criteria for Foodstuffs Com-mission of the European Communities Luxembourg 2007

[33] M L Cohen and TWhalen ldquoImplications of low level humanexposure to respirable B Anthracisrdquo Applied Biosafetyvol 12 no 2 pp 109ndash115 2016

[34] M E Coleman B -ran S S Morse M Hugh-Jones andS Massulik ldquoInhalation anthrax dose response and riskanalysisrdquo Biosecurity and Bioterrorism Biodefense StrategyPractice and Science vol 6 no 2 pp 147ndash160 2008

[35] R J Schanler R J Shulman and C Lau ldquoFeeding strategiesfor premature infants beneficial outcomes of feeding fortifiedhuman milk versus preterm formulardquo Pediatrics vol 103no 6 pp 1150ndash1157 1999

[36] P M Sisk C A Lovelady R G Dillard K J Gruber andT M OrsquoShea ldquoEarly human milk feeding is associated with alower risk of necrotizing enterocolitis in very low birth weightinfantsrdquo Journal of Perinatology vol 27 no 7 pp 428ndash4332007

8 Canadian Journal of Infectious Diseases and Medical Microbiology

ingested and infection following the ingestion of bankedmilk contaminated by Bacillus cereus a full-risk assessmenthad to be modeled -e probability of infection resultingfrom the consumption of servings contaminated with Ba-cillus cereus (P) was based on an exponential dose-responsemodel because it allows inferring the dose-effect relationshipbased on certain epidemiological hypotheses

P 1minus eminusRND (4)

where ND represents the total number of B cereus in theserving and R is a constant specific to each pathogenic agentthat expresses the probability of being infected by a singlepathogenic microorganism present in the serving dependingon the virulence of the bacteria [23ndash27] In other words theR-value is the probability of interaction between the host andthe microorganism [28] By rearranging equation (4) R canbe expressed as follows

R minus[ln(1minusP)]

ND (5)

-e constant R was estimated using the approach de-scribed by Lindqvist and Westoo [25] in which the validityof the model depends on certain epidemiological assump-tions In the present study because an exponential dose-response model is a nonthreshold model and since thesimulation assumes that all serving are contaminated allpremature infants fed with banked milk are considered at

risk -e total number of preterm children exposed to BHMper year and the mean number of servings per infant wereestimated using internal distribution reports approximately1000 children were fed with milk supplied by the Hema-Quebec human milk bank with a mean of 40 portions peryear

In the absence of evidence regarding the incidence ofBacillus cereus infection in preterm infants fed with bankedmilk and the potential infectious dose of B cereus in pas-teurized milk we cannot use an independent estimate of theconstant R therefore variation in parameters for the riskcharacterization was made

23 Risk Characterization -ree scenarios were built toestimate the annual incidence of illness in preterm infantscaused by the ingestion of B cereus-contaminated bankedmilk As previously mentioned there is no evidence of acausal link between banked milk contamination and Bcereus infection in preterm infants we therefore had to makesome epidemiological assumptions for the various scenariosbeing considered As a factual basis for these scenarios weused data from the largest observed outbreak of Bacilluscereus infections in newborns in Quebec (unpublished) In2013ndash2014 at one hospital in our service area 7 preterminfants became infected An investigation concluded thatthese infections were likely to have been due to airbornecontamination that resulted from construction work near

Table 1 R-value for the exponential dose-response model based on simulated scenarios and variation of the infectious doselowast

Dose1 Total CFU per serving R scenario 1 R scenario 2 R scenario 3200 20000 12E minus 10 162E minus 9 250E minus 9100 10000 25E minus 10 325E minus 9 500E minus 975 7500 33E minus 10 433E minus 9 666E minus 965 6500 38E minus 10 500E minus 9 769E minus 955 5500 45E minus 10 591E minus 9 909E minus 950 5000 50E minus 10 650E minus 9 100E minus 845 4500 55E minus 10 722E minus 9 111E minus 840 4000 62E minus 10 813E minus 9 125E minus 835 3500 71E minus 10 929E minus 9 142E minus 830 3000 83E minus 10 108E minus 8 167E minus 820 2000 125E minus 9 163E minus 8 250E minus 815 1500 167E minus 9 217E minus 8 333E minus 812 1200 208E minus 9 271E minus 8 416E minus 89 900 278E minus 9 361E minus 8 556E minus 86 600 417E minus 9 542E minus 8 833E minus 83 300 833E minus 9 108E minus 7 167E minus 71 100 250E minus 8 325E minus 7 500E minus 705 50 500E minus 8 650E minus 7 100E minus 7lowastAnnual incidence of infection of 001 013 and 02 for scenarios 1 2 and 3 respectively 1Dose in CFUml

Table 2 Average contamination of Bacillus cereus in the accepted and rejected pools

Variables N Volume mean (in l) Concentration mean (SD)Lots not detected 1 sampling of 100 microl 9026 903 8574782 064 (168)Lots not detected 9 samplings of 100 microl 6996 700 6643546 021 (036)Lots detected 1 sampling of 100 microl 974 97 925459 3574 (32911)Lots detected 9 sampling of 100 microl 3004 300 2856695 1301 (18801)

4 Canadian Journal of Infectious Diseases and Medical Microbiology

the neonatal unit of the hospital However for the purpose of thesimulation we postulated that all these infections had beenacquired by uncultured contaminated humanmilk (a worst-casesituation) It should be noted that only one of these seven infantshad been exposed to banked human milk

Moreover to build the three scenarios we considered thefollowing denominator data (1) the total number of births inQuebec during that year (n 88867) (2) the total number ofbirths less than 25kg in Quebec (n 6464) and (3) the totalnumber of births at the hospital where the B cereus infectionsoccurred (n 3500) Depending on the denominator beingtaken this resulted in annual estimated incidence rates of 001013 and 02 for scenarios 1 2 and 3 respectively For eachscenario in order to estimate the uncertainty of the constant Rand because no information concerning the minimal infectiousdose of Bacillus cereus in human milk is available sensitivityanalyses were performed Eighteen variations of the infectiousdose of Bacillus cereus in human milk were simulated from05CFUml to 200CFUml -erefore R-value ranged from12E minus 10 to 500E minus 8 162E minus 9 to 650E minus 7 and 250E minus 9 to100E minus 7 for scenarios 1 2 and 3 respectively (Table 1)

3 Results

Ten thousand pools each constituting of 4 mothers weresimulatedwith amean volume of 95 liters of SD 04 and ameanBacillus cereus contamination concentration of 405CFUmlSD 103 and geometric mean of 016

31 Detection and Rejection Probability Following oursimulation using post-pasteurization bacterial controlsampling of 100microliters per milk pool 9026 of the10000 simulated pools containing Bacillus cereus would notbe detected and would be in the milk bank inventory fordistribution -e average concentration of Bacillus cereuscontamination in the pools qualified for distribution wouldbe 064CFUml In addition a rejection rate of 974 (974pools) would be observed with an average Bacillus cereuscontamination concentration of 3574 CFUml -is re-jection rate is similar to those reported by humanmilk banksthat use a post-pasteurization sample of 100microliters forculture [21 29] If post-pasteurization bacterial control

2

19

18

17

16

15

14

Mea

n ris

k di

ffere

ntia

l per

por

tion

(times10

ndash6)

13

12

11

1

09

08

07

06

05

04

03

02

01

0 20 40 60 80Infectious dose (CFUmL)

100 120 140 160 180 2000

Incidence = 02Incidence = 013Incidence = 001

Figure 2 Mean risk differential per portion using the two different post-pasteurization bacteriological control criteria Horizontal axiscorresponds to the 18 variations of the infectious doses according to the 3 different incidence scenarios identified by the correspondingcolors -e vertical axis corresponds to the mean risk differential per million portions

Canadian Journal of Infectious Diseases and Medical Microbiology 5

sampling consisted of nine sample of 100 microliters permilk pool 6996 pools with a mean Bacillus cereus con-centration of 021 would not be detected and would bequalified for distribution 3004 pools would have beendetected with an average Bacillus cereus concentration of1301CFUml and a rejection rate of 30 (Table 2) -issimulated rejection rate and the concentration in therejected pool are consistent with those of our internalobservations

In summary concentration of Bacillus cereus in poolsput into the inventory would be higher by 043CFUmlaverage when detected using one sample of 100microlitersversus 9 samples of 100 microliters Moreover the rejectedmilk volume difference would be 19312 liters corre-sponding to 203 of the total volume simulated

32 Residual Risk Differential Figure 2 represents the meanrisk differential per million between the two methods ofpost-pasteurization bacterial control (ie 1 sample versus 9samples of 100microliters) by scenario and infectious dose-is figure shows that for the vast majority of scenarios evenamong the most pessimistic the mean risk differential as-sociated with sampling of 1 test rather than 9 tests of100microliters is well below the threshold of 11 millionadditional infections per serving It is only when we supposethat the minimal infectious dose is 20CFUml or less that weobserve a risk difference greater than 11 million

Table 3 presents detailed data concerning the residualrisk scenarios comparing a 100-microliter sample to a 900-microliter sample for post-pasteurization culture whenvarying the infectious dose per serving -e mean riskdifferential for scenario 1 ranges from 215 to 000539 per

million for virulent dose of 05 to 200CFUml For scenarios2 and 3 the mean risk differential ranges from 278 to 0070per million and from 428 to 0108 respectively

4 Discussion

Proven invasive infections due to Bacillus cereus have beenreported among immunosuppressed patients particularlypreterm neonates but given the scarcity of cases thepathophysiology of these infections is not well knownMoreover assessment of the origin of contamination due toorganisms such as B cereus that are widely disseminated inthe environment is often difficult and may not yield anobvious source [11] Nonetheless in outbreak investigationscontamination through respiratory route was often docu-mented [11 15] Several studies have also assessed the pu-tative role of human milk in B cereus infection in prematureinfants however clinical strains of B cereus never matchedthe strains found in milk and therefore the tracking of theorigin of B cereus to contaminated bank milk was to datenever demonstrated [6 9]

Our analysis cannot make any statement regarding theactual risk of Bacillus cereus infection resulting from theingestion of contaminated banked human milk Even if therisk is real which remains unproven yet it is likely to beextremely small based on the various scenarios that weexplored in this simulation -e question of whether there isa real threat posed by Bacillus cereus to the safety of bankedhuman milk therefore remains unresolved

What our simulation shows is that even in very pes-simistic scenarios regarding a causal link between bankedmilk and B cereus infection a more stringent culture-sampling strategy makes very little difference in the risk

Table 3 Mean risk of infection by ingestion of banked maternal milk by serving according to different levels of Bacillus cereus infectiousdose in premature neonateslowast

Infectiousdose1

Scenario 1 Scenario 2 Scenario 3100 microl

sampling9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

200 799E minus 9 260E minus 9 539E minus 9 103E minus 7 334E minus 8 696E minus 8 159E minus 7 514E minus 8 108E minus 7100 160E minus 8 519E minus 9 108E minus 8 206E minus 7 669E minus 8 139E minus 7 317E minus 7 103E minus 7 214E minus 775 213E minus 8 693E minus 9 144E minus 8 275E minus 7 892E minus 8 186E minus 7 423E minus 7 137E minus 7 286E minus 765 246E minus 8 799E minus 9 166E minus 8 317E minus 7 103E minus 7 214E minus 7 488E minus 7 158E minus 7 33 E minus 755 290E minus 8 945E minus 9 195E minus 8 375E minus 7 122E minus 7 253E minus 7 576E minus 7 187E minus 7 389E minus 750 319E minus 8 104E minus 8 215E minus 8 412E minus 7 134E minus 7 278E minus 7 634E minus 7 206E minus 7 428E minus 745 355E minus 8 115E minus 8 240E minus 8 458E minus 7 149E minus 7 309E minus 7 705E minus 7 229E minus 7 476E minus 740 399E minus 8 130E minus 8 269E minus 8 515E minus 7 167E minus 7 348E minus 7 793E minus 7 257E minus 7 536E minus 735 456E minus 8 148E minus 8 308E minus 8 589E minus 7 191E minus 7 398E minus 7 906E minus 7 294E minus 7 612E minus 730 532E minus 8 173E minus 8 359E minus 8 687E minus 7 223E minus 7 464E minus 7 106E minus 6 343E minus 7 717E minus 720 799E minus 8 260E minus 8 539E minus 8 103E minus 6 334E minus 7 696E minus 7 158E minus 6 514E minus 7 107E minus 615 106E minus 7 346E minus 8 714E minus 8 137E minus 6 446E minus 7 924E minus 7 211E minus 6 686E minus 7 142E minus 612 133E minus 7 433E minus 8 897E minus 8 172E minus 6 557E minus 7 116E minus 6 264E minus 6 857E minus 7 178E minus 69 177E minus 7 577E minus 8 119E minus 7 229E minus 6 743E minus 7 155E minus 6 352E minus 6 114E minus 6 238E minus 66 266E minus 7 865E minus 8 179E minus 7 343E minus 6 111E minus 6 232E minus 6 528E minus 6 171E minus 6 357E minus 63 532E minus 7 173E minus 7 359E minus 7 687E minus 6 223E minus 6 464E minus 6 106E minus 5 343E minus 6 717E minus 61 160E minus 6 519E minus 7 108E minus 6 206E minus 5 669E minus 6 131E minus 5 317E minus 5 103E minus 5 214E minus 505 319E minus 6 104E minus 6 215E minus 6 412E minus 5 134E minus 5 278E minus 5 634E minus 5 206E minus 5 428E minus 5lowastSimulation scenario was made after post-pasteurized microbiological control using detection of 100 microliters vs 9times100 microliters per pool and for anannual incidence of infection of 001 013 and 02 corresponding to scenarios 1 2 and 3 respectively 1Infectious concentration in CFUml

6 Canadian Journal of Infectious Diseases and Medical Microbiology

incurred by exposed infants At the same time because ofthe ubiquitous nature of this bacterium increasing thesampling volume and thus the sensitivity of the culturemethod can lead to much higher rates of product dis-qualification More specifically our analysis shows that therisk difference per serving associated with a sampling of 1instead of 9 tests of 100 microliters is well below 11 millioninfections a risk that can be considered negligible In allscenarios it is only when the minimal infectious dose was20 CFUml or less for a 100ml serving that the risk dif-ferential was greater than 1 infection1 million portionssuch minimal infectious doses are clearly unrealistic forBacillus cereus Also average concentrations of Bacilluscereus found in the qualified milk batches would only beslightly increased from 021 to 064 CFUml which re-mains well below the standards applied to formula milk(ie 50 CFUg in Europe and 100 CFUg in the USAAustralia and New Zealand) [30ndash32] By comparison theliterature on human and animal exposure by inhalation todifferent doses of spores of Bacillus anthracis reports aminimal infectious dose of at least 600 CFU [33 34] If weassume the same minimal infectious dose of 600 CFU perserving for B cereus in BHM this would still be ten timeshigher than the average B cereus concentration remainingin the qualified milk pools Furthermore because of themuch lower potential for the pathogenicity of Bacilluscereus compared with B anthracis [33 34] this hypothesisis extremely aggressive Finally the gain in supply resultingfrom a less-stringent culture method and its potentialimpact on product availability for the prevention of nec-rotizing enterocolitis largely exceeds the very low theo-retical differential risk as calculated by these models[32 35 36] In our operational setting changing thesampling volume from 900 to 100 microliters is expected todecrease the rate of disqualified culture-positive poolsfrom 30 to 10 percent -is will represent a major im-provement in our capacity to meet the demand for pre-mature infants

To our knowledge this is the very first study usingMonte Carlo simulation which estimates the potential riskof Bacillus cereus infection in preterm infants fed withbanked human milk -e major limitation of our study isthat we had to estimate many of the parameters used in theabsence of empirical data concerning their value Also thesimulation was based on the premise that any bacterial loadin the milk pool is homogeneously distributed which isprobably the case in the vast majority of pools in our ex-perience However the possibility of this load being dis-tributed unevenly due to an inadequate milk mixing processin rare instances cannot be ruled out

5 Conclusion

Our simulation results suggest that if it exists at all the risk ofBacillus cereus infection following the ingestion of pasteurizedbanked milk is extremely small Our analysis also shows thateven in worst-case scenarios a 100-microliter sample forpost-pasteurization culture is amply sufficient to mitigate thisrisk a larger sampling volume would only lead to a higher rate

of disqualification for this important health-care resourcewithout having any significant positive impact on safety

Nevertheless despite no specific mention about Bacilluscereus contamination in the majority of the guidelines thissimulation highlights the importance of the combined use ofpasteurization and bacteriological investigation before hu-man milk distribution for extremely preterm babies Inaddition future studies investigating the association be-tween Bacillus cereus concentration and infection of pre-mature infants are needed

Data Availability

-e simulation code and the data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

-e authors declare that they have no conflicts of interest

References

[1] J-A B Bier T Oliver A E Ferguson and B R VohrldquoHuman milk improves cognitive and motor development ofpremature infants during infancyrdquo Journal of Human Lac-tation vol 18 no 4 pp 361ndash367 2016

[2] D L OrsquoConnor J Jacobs R Hall et al ldquoGrowth and de-velopment of premature infants fed predominantly humanmilk predominantly premature infant formula or a combi-nation of human milk and premature formulardquo Journal ofPediatric Gastroenterology and Nutrition vol 37 no 4pp 437ndash446 2003

[3] B R Vohr B B Poindexter A M Dusick et al ldquoBeneficialeffects of breast milk in the neonatal intensive care unit on thedevelopmental outcome of extremely low birth weight infantsat 18 months of agerdquo Pediatrics vol 118 no 1 pp e115ndashe1232006

[4] J Mahon L Claxton and H Wood ldquoModelling the cost-effectiveness of human milk and breastfeeding in preterminfants in the United Kingdomrdquo Health Economics Reviewvol 6 no 1 2016

[5] A Singhal T J Cole M Fewtrell and A Lucas ldquoBreastmilkfeeding and lipoprotein profile in adolescents born pretermfollow-up of a prospective randomised studyrdquo e Lancetvol 363 no 9421 pp 1571ndash1578 2004

[6] A M Wendelboe C Smelser C A Lucero andL C McDonald ldquoCluster of necrotizing enterocolitis in aneonatal intensive care unit New Mexico 2007rdquo AmericanJournal of Infection Control vol 38 no 2 pp 144ndash148 2010

[7] S Arslanoglu G E Moro and E E Ziegler ldquo-e WapmWorking Group on N Optimization of human milk fortifi-cation for preterm infants new concepts and recommenda-tionsrdquo Journal of Perinatal Medicine vol 38 no 3pp 233ndash238 2010

[8] L M Gartner J Morton R A Lawrence et al ldquoBreastfeedingand the use of human milkrdquo Pediatrics vol 115 no 2pp 496ndash506 2005

[9] J-W Decousser N Ramarao C Duport et al ldquoBacillus cereusand severe intestinal infections in preterm neonates putativerole of pooled breast milkrdquo American Journal of InfectionControl vol 41 no 10 pp 918ndash921 2013

Canadian Journal of Infectious Diseases and Medical Microbiology 7

[10] S Landers and K Updegrove ldquoBacteriological screening ofdonor human milk before and after Holder pasteurizationrdquoBreastfeeding Medicine vol 5 no 3 pp 117ndash121 2010

[11] N J Hilliard R L Schelonka and K B Waites ldquoBacilluscereus bacteremia in a preterm neonaterdquo Journal of ClinicalMicrobiology vol 41 no 7 pp 3441ndash3444 2003

[12] E J Bottone ldquoBacillus cereus a volatile human pathogenrdquoClinical Microbiology Reviews vol 23 no 2 pp 382ndash3982010

[13] R Lotte A-L Herisse Y Berrouane et al ldquoVirulence analysisof Bacillus cereus isolated after death of preterm neonatesnice France 2013rdquo Emerging Infectious Diseases vol 23no 5 pp 845ndash848 2017

[14] N Ramarao L Belotti S Deboscker et al ldquoTwo unrelatedepisodes of Bacillus cereus bacteremia in a neonatal intensivecare unitrdquo American Journal of Infection Control vol 42no 6 pp 694-695 2014

[15] W C Van Der Zwet G A Parlevliet P H Savelkoul et alldquoOutbreak of Bacillus cereus infections in a neonatal intensivecare unit traced to balloons used in manual ventilationrdquoJournal of Clinical Microbiology vol 38 no 11 pp 4131ndash41362000

[16] Y Almutawif B Hartmann M Lloyd W Erber andD Geddes ldquoA retrospective audit of bacterial culture results ofdonated human milk in Perth Western Australiardquo EarlyHuman Development vol 105 pp 1ndash6 2017

[17] Centre for Clinical Practice at NICE (UK) Donor Breast MilkBankse Operation of Donor Milk Bank Services 20120210ed National Institute for Health and Clinical Excellence (UK)London UK 2010

[18] C Haas J Rose and C Gerba Quantitative Microbial RiskAssessment John Wiley amp Sons Hoboken NJ USA 2ndedition 2014

[19] M J Nauta A Modular Process Risk Model Structure forQuantitative Microbiological Risk Assessment and Its Applica-tion in an Exposure Assessment of Bacillus cereus in a REPFEDNational Institute of Public Health and the EnvironmentAmsterdam Netherlands 2001 Contract No 149106 007

[20] M Nauta ldquoMicrobiological risk assessment models for par-titioning and mixing during food handlingrdquo InternationalJournal of Food Microbiology vol 100 no 1ndash3 pp 311ndash3222005

[21] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination du lait maternel par une floreaerobie evaluation des pertes pour un lactariumrdquo Archives dePediatrie vol 22 no 5 pp 461ndash467 2015

[22] Hema-Quebec Guide drsquoinformation sur le don de laitmaternel Banque publique de lait maternel Hema-QuebecMontreal Canada 2014

[23] C N Haas ldquoEstimation of risk due to low doses of micro-organisms a comparison of alternative methodologiesrdquoAmerican Journal of Epidemiology vol 118 no 4 pp 573ndash582 1983

[24] C N Haas ldquoMicrobial dose response modeling past presentand futurerdquo Environmental Science and Technology vol 49no 3 pp 1245ndash1259 2015

[25] R Lindqvist and A Westoo ldquoQuantitative risk assessment forListeria monocytogenes in smoked or gravad salmon andrainbow trout in Swedenrdquo International Journal of FoodMicrobiology vol 58 no 3 pp 181ndash196 2000

[26] D Vose Risk Analysis A Quantitative Guide John Wiley ampSons New York NY USA 3rd edition 2008

[27] P Acai L Valik and D Liptakova ldquoQuantitative risk as-sessment of Bacillus cereus in pasteurised milk produced in

the Slovak Republicrdquo Czech Journal of Food Sciences vol 32no 2 pp 122ndash131 2014

[28] S Regli J B Rose C N Haas and C P Gerba ldquoModeling therisk from Giardia and viruses in drinking waterrdquo Journal -American Water Works Association vol 83 no 11 pp 76ndash841991

[29] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination of human milk with aerobicflora evaluation of losses for a humanmilk bankrdquo Archives dePediatrie 22 22 pp 461ndash467 2015 in French

[30] US Food and Drug Administration Food CompositionStandards Labeling And Economics Compliance ProgramGuidance Manual Chapter 21ndashFood Composition StandardsLabeling and Economics US Food and Drug AdministrationSilver Spring MD USA 2006 httpswwwfdagovdownloadsFoodComplianceEnforcementUCM073339pdf

[31] FSANZ Bacillus cereus Limits in Infant Formula FSANZCanberra BC Autralia 2004 httpswwwfoodstandardsgovaucodeapplicationsdocumentsA454_B_cereus_FARpdf

[32] Commission of the European Communities CommissionRegulation (EC) no 14412007 Amending Regulation (EC) No20732005 on Microbiological Criteria for Foodstuffs Com-mission of the European Communities Luxembourg 2007

[33] M L Cohen and TWhalen ldquoImplications of low level humanexposure to respirable B Anthracisrdquo Applied Biosafetyvol 12 no 2 pp 109ndash115 2016

[34] M E Coleman B -ran S S Morse M Hugh-Jones andS Massulik ldquoInhalation anthrax dose response and riskanalysisrdquo Biosecurity and Bioterrorism Biodefense StrategyPractice and Science vol 6 no 2 pp 147ndash160 2008

[35] R J Schanler R J Shulman and C Lau ldquoFeeding strategiesfor premature infants beneficial outcomes of feeding fortifiedhuman milk versus preterm formulardquo Pediatrics vol 103no 6 pp 1150ndash1157 1999

[36] P M Sisk C A Lovelady R G Dillard K J Gruber andT M OrsquoShea ldquoEarly human milk feeding is associated with alower risk of necrotizing enterocolitis in very low birth weightinfantsrdquo Journal of Perinatology vol 27 no 7 pp 428ndash4332007

8 Canadian Journal of Infectious Diseases and Medical Microbiology

the neonatal unit of the hospital However for the purpose of thesimulation we postulated that all these infections had beenacquired by uncultured contaminated humanmilk (a worst-casesituation) It should be noted that only one of these seven infantshad been exposed to banked human milk

Moreover to build the three scenarios we considered thefollowing denominator data (1) the total number of births inQuebec during that year (n 88867) (2) the total number ofbirths less than 25kg in Quebec (n 6464) and (3) the totalnumber of births at the hospital where the B cereus infectionsoccurred (n 3500) Depending on the denominator beingtaken this resulted in annual estimated incidence rates of 001013 and 02 for scenarios 1 2 and 3 respectively For eachscenario in order to estimate the uncertainty of the constant Rand because no information concerning the minimal infectiousdose of Bacillus cereus in human milk is available sensitivityanalyses were performed Eighteen variations of the infectiousdose of Bacillus cereus in human milk were simulated from05CFUml to 200CFUml -erefore R-value ranged from12E minus 10 to 500E minus 8 162E minus 9 to 650E minus 7 and 250E minus 9 to100E minus 7 for scenarios 1 2 and 3 respectively (Table 1)

3 Results

Ten thousand pools each constituting of 4 mothers weresimulatedwith amean volume of 95 liters of SD 04 and ameanBacillus cereus contamination concentration of 405CFUmlSD 103 and geometric mean of 016

31 Detection and Rejection Probability Following oursimulation using post-pasteurization bacterial controlsampling of 100microliters per milk pool 9026 of the10000 simulated pools containing Bacillus cereus would notbe detected and would be in the milk bank inventory fordistribution -e average concentration of Bacillus cereuscontamination in the pools qualified for distribution wouldbe 064CFUml In addition a rejection rate of 974 (974pools) would be observed with an average Bacillus cereuscontamination concentration of 3574 CFUml -is re-jection rate is similar to those reported by humanmilk banksthat use a post-pasteurization sample of 100microliters forculture [21 29] If post-pasteurization bacterial control

2

19

18

17

16

15

14

Mea

n ris

k di

ffere

ntia

l per

por

tion

(times10

ndash6)

13

12

11

1

09

08

07

06

05

04

03

02

01

0 20 40 60 80Infectious dose (CFUmL)

100 120 140 160 180 2000

Incidence = 02Incidence = 013Incidence = 001

Figure 2 Mean risk differential per portion using the two different post-pasteurization bacteriological control criteria Horizontal axiscorresponds to the 18 variations of the infectious doses according to the 3 different incidence scenarios identified by the correspondingcolors -e vertical axis corresponds to the mean risk differential per million portions

Canadian Journal of Infectious Diseases and Medical Microbiology 5

sampling consisted of nine sample of 100 microliters permilk pool 6996 pools with a mean Bacillus cereus con-centration of 021 would not be detected and would bequalified for distribution 3004 pools would have beendetected with an average Bacillus cereus concentration of1301CFUml and a rejection rate of 30 (Table 2) -issimulated rejection rate and the concentration in therejected pool are consistent with those of our internalobservations

In summary concentration of Bacillus cereus in poolsput into the inventory would be higher by 043CFUmlaverage when detected using one sample of 100microlitersversus 9 samples of 100 microliters Moreover the rejectedmilk volume difference would be 19312 liters corre-sponding to 203 of the total volume simulated

32 Residual Risk Differential Figure 2 represents the meanrisk differential per million between the two methods ofpost-pasteurization bacterial control (ie 1 sample versus 9samples of 100microliters) by scenario and infectious dose-is figure shows that for the vast majority of scenarios evenamong the most pessimistic the mean risk differential as-sociated with sampling of 1 test rather than 9 tests of100microliters is well below the threshold of 11 millionadditional infections per serving It is only when we supposethat the minimal infectious dose is 20CFUml or less that weobserve a risk difference greater than 11 million

Table 3 presents detailed data concerning the residualrisk scenarios comparing a 100-microliter sample to a 900-microliter sample for post-pasteurization culture whenvarying the infectious dose per serving -e mean riskdifferential for scenario 1 ranges from 215 to 000539 per

million for virulent dose of 05 to 200CFUml For scenarios2 and 3 the mean risk differential ranges from 278 to 0070per million and from 428 to 0108 respectively

4 Discussion

Proven invasive infections due to Bacillus cereus have beenreported among immunosuppressed patients particularlypreterm neonates but given the scarcity of cases thepathophysiology of these infections is not well knownMoreover assessment of the origin of contamination due toorganisms such as B cereus that are widely disseminated inthe environment is often difficult and may not yield anobvious source [11] Nonetheless in outbreak investigationscontamination through respiratory route was often docu-mented [11 15] Several studies have also assessed the pu-tative role of human milk in B cereus infection in prematureinfants however clinical strains of B cereus never matchedthe strains found in milk and therefore the tracking of theorigin of B cereus to contaminated bank milk was to datenever demonstrated [6 9]

Our analysis cannot make any statement regarding theactual risk of Bacillus cereus infection resulting from theingestion of contaminated banked human milk Even if therisk is real which remains unproven yet it is likely to beextremely small based on the various scenarios that weexplored in this simulation -e question of whether there isa real threat posed by Bacillus cereus to the safety of bankedhuman milk therefore remains unresolved

What our simulation shows is that even in very pes-simistic scenarios regarding a causal link between bankedmilk and B cereus infection a more stringent culture-sampling strategy makes very little difference in the risk

Table 3 Mean risk of infection by ingestion of banked maternal milk by serving according to different levels of Bacillus cereus infectiousdose in premature neonateslowast

Infectiousdose1

Scenario 1 Scenario 2 Scenario 3100 microl

sampling9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

200 799E minus 9 260E minus 9 539E minus 9 103E minus 7 334E minus 8 696E minus 8 159E minus 7 514E minus 8 108E minus 7100 160E minus 8 519E minus 9 108E minus 8 206E minus 7 669E minus 8 139E minus 7 317E minus 7 103E minus 7 214E minus 775 213E minus 8 693E minus 9 144E minus 8 275E minus 7 892E minus 8 186E minus 7 423E minus 7 137E minus 7 286E minus 765 246E minus 8 799E minus 9 166E minus 8 317E minus 7 103E minus 7 214E minus 7 488E minus 7 158E minus 7 33 E minus 755 290E minus 8 945E minus 9 195E minus 8 375E minus 7 122E minus 7 253E minus 7 576E minus 7 187E minus 7 389E minus 750 319E minus 8 104E minus 8 215E minus 8 412E minus 7 134E minus 7 278E minus 7 634E minus 7 206E minus 7 428E minus 745 355E minus 8 115E minus 8 240E minus 8 458E minus 7 149E minus 7 309E minus 7 705E minus 7 229E minus 7 476E minus 740 399E minus 8 130E minus 8 269E minus 8 515E minus 7 167E minus 7 348E minus 7 793E minus 7 257E minus 7 536E minus 735 456E minus 8 148E minus 8 308E minus 8 589E minus 7 191E minus 7 398E minus 7 906E minus 7 294E minus 7 612E minus 730 532E minus 8 173E minus 8 359E minus 8 687E minus 7 223E minus 7 464E minus 7 106E minus 6 343E minus 7 717E minus 720 799E minus 8 260E minus 8 539E minus 8 103E minus 6 334E minus 7 696E minus 7 158E minus 6 514E minus 7 107E minus 615 106E minus 7 346E minus 8 714E minus 8 137E minus 6 446E minus 7 924E minus 7 211E minus 6 686E minus 7 142E minus 612 133E minus 7 433E minus 8 897E minus 8 172E minus 6 557E minus 7 116E minus 6 264E minus 6 857E minus 7 178E minus 69 177E minus 7 577E minus 8 119E minus 7 229E minus 6 743E minus 7 155E minus 6 352E minus 6 114E minus 6 238E minus 66 266E minus 7 865E minus 8 179E minus 7 343E minus 6 111E minus 6 232E minus 6 528E minus 6 171E minus 6 357E minus 63 532E minus 7 173E minus 7 359E minus 7 687E minus 6 223E minus 6 464E minus 6 106E minus 5 343E minus 6 717E minus 61 160E minus 6 519E minus 7 108E minus 6 206E minus 5 669E minus 6 131E minus 5 317E minus 5 103E minus 5 214E minus 505 319E minus 6 104E minus 6 215E minus 6 412E minus 5 134E minus 5 278E minus 5 634E minus 5 206E minus 5 428E minus 5lowastSimulation scenario was made after post-pasteurized microbiological control using detection of 100 microliters vs 9times100 microliters per pool and for anannual incidence of infection of 001 013 and 02 corresponding to scenarios 1 2 and 3 respectively 1Infectious concentration in CFUml

6 Canadian Journal of Infectious Diseases and Medical Microbiology

incurred by exposed infants At the same time because ofthe ubiquitous nature of this bacterium increasing thesampling volume and thus the sensitivity of the culturemethod can lead to much higher rates of product dis-qualification More specifically our analysis shows that therisk difference per serving associated with a sampling of 1instead of 9 tests of 100 microliters is well below 11 millioninfections a risk that can be considered negligible In allscenarios it is only when the minimal infectious dose was20 CFUml or less for a 100ml serving that the risk dif-ferential was greater than 1 infection1 million portionssuch minimal infectious doses are clearly unrealistic forBacillus cereus Also average concentrations of Bacilluscereus found in the qualified milk batches would only beslightly increased from 021 to 064 CFUml which re-mains well below the standards applied to formula milk(ie 50 CFUg in Europe and 100 CFUg in the USAAustralia and New Zealand) [30ndash32] By comparison theliterature on human and animal exposure by inhalation todifferent doses of spores of Bacillus anthracis reports aminimal infectious dose of at least 600 CFU [33 34] If weassume the same minimal infectious dose of 600 CFU perserving for B cereus in BHM this would still be ten timeshigher than the average B cereus concentration remainingin the qualified milk pools Furthermore because of themuch lower potential for the pathogenicity of Bacilluscereus compared with B anthracis [33 34] this hypothesisis extremely aggressive Finally the gain in supply resultingfrom a less-stringent culture method and its potentialimpact on product availability for the prevention of nec-rotizing enterocolitis largely exceeds the very low theo-retical differential risk as calculated by these models[32 35 36] In our operational setting changing thesampling volume from 900 to 100 microliters is expected todecrease the rate of disqualified culture-positive poolsfrom 30 to 10 percent -is will represent a major im-provement in our capacity to meet the demand for pre-mature infants

To our knowledge this is the very first study usingMonte Carlo simulation which estimates the potential riskof Bacillus cereus infection in preterm infants fed withbanked human milk -e major limitation of our study isthat we had to estimate many of the parameters used in theabsence of empirical data concerning their value Also thesimulation was based on the premise that any bacterial loadin the milk pool is homogeneously distributed which isprobably the case in the vast majority of pools in our ex-perience However the possibility of this load being dis-tributed unevenly due to an inadequate milk mixing processin rare instances cannot be ruled out

5 Conclusion

Our simulation results suggest that if it exists at all the risk ofBacillus cereus infection following the ingestion of pasteurizedbanked milk is extremely small Our analysis also shows thateven in worst-case scenarios a 100-microliter sample forpost-pasteurization culture is amply sufficient to mitigate thisrisk a larger sampling volume would only lead to a higher rate

of disqualification for this important health-care resourcewithout having any significant positive impact on safety

Nevertheless despite no specific mention about Bacilluscereus contamination in the majority of the guidelines thissimulation highlights the importance of the combined use ofpasteurization and bacteriological investigation before hu-man milk distribution for extremely preterm babies Inaddition future studies investigating the association be-tween Bacillus cereus concentration and infection of pre-mature infants are needed

Data Availability

-e simulation code and the data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

-e authors declare that they have no conflicts of interest

References

[1] J-A B Bier T Oliver A E Ferguson and B R VohrldquoHuman milk improves cognitive and motor development ofpremature infants during infancyrdquo Journal of Human Lac-tation vol 18 no 4 pp 361ndash367 2016

[2] D L OrsquoConnor J Jacobs R Hall et al ldquoGrowth and de-velopment of premature infants fed predominantly humanmilk predominantly premature infant formula or a combi-nation of human milk and premature formulardquo Journal ofPediatric Gastroenterology and Nutrition vol 37 no 4pp 437ndash446 2003

[3] B R Vohr B B Poindexter A M Dusick et al ldquoBeneficialeffects of breast milk in the neonatal intensive care unit on thedevelopmental outcome of extremely low birth weight infantsat 18 months of agerdquo Pediatrics vol 118 no 1 pp e115ndashe1232006

[4] J Mahon L Claxton and H Wood ldquoModelling the cost-effectiveness of human milk and breastfeeding in preterminfants in the United Kingdomrdquo Health Economics Reviewvol 6 no 1 2016

[5] A Singhal T J Cole M Fewtrell and A Lucas ldquoBreastmilkfeeding and lipoprotein profile in adolescents born pretermfollow-up of a prospective randomised studyrdquo e Lancetvol 363 no 9421 pp 1571ndash1578 2004

[6] A M Wendelboe C Smelser C A Lucero andL C McDonald ldquoCluster of necrotizing enterocolitis in aneonatal intensive care unit New Mexico 2007rdquo AmericanJournal of Infection Control vol 38 no 2 pp 144ndash148 2010

[7] S Arslanoglu G E Moro and E E Ziegler ldquo-e WapmWorking Group on N Optimization of human milk fortifi-cation for preterm infants new concepts and recommenda-tionsrdquo Journal of Perinatal Medicine vol 38 no 3pp 233ndash238 2010

[8] L M Gartner J Morton R A Lawrence et al ldquoBreastfeedingand the use of human milkrdquo Pediatrics vol 115 no 2pp 496ndash506 2005

[9] J-W Decousser N Ramarao C Duport et al ldquoBacillus cereusand severe intestinal infections in preterm neonates putativerole of pooled breast milkrdquo American Journal of InfectionControl vol 41 no 10 pp 918ndash921 2013

Canadian Journal of Infectious Diseases and Medical Microbiology 7

[10] S Landers and K Updegrove ldquoBacteriological screening ofdonor human milk before and after Holder pasteurizationrdquoBreastfeeding Medicine vol 5 no 3 pp 117ndash121 2010

[11] N J Hilliard R L Schelonka and K B Waites ldquoBacilluscereus bacteremia in a preterm neonaterdquo Journal of ClinicalMicrobiology vol 41 no 7 pp 3441ndash3444 2003

[12] E J Bottone ldquoBacillus cereus a volatile human pathogenrdquoClinical Microbiology Reviews vol 23 no 2 pp 382ndash3982010

[13] R Lotte A-L Herisse Y Berrouane et al ldquoVirulence analysisof Bacillus cereus isolated after death of preterm neonatesnice France 2013rdquo Emerging Infectious Diseases vol 23no 5 pp 845ndash848 2017

[14] N Ramarao L Belotti S Deboscker et al ldquoTwo unrelatedepisodes of Bacillus cereus bacteremia in a neonatal intensivecare unitrdquo American Journal of Infection Control vol 42no 6 pp 694-695 2014

[15] W C Van Der Zwet G A Parlevliet P H Savelkoul et alldquoOutbreak of Bacillus cereus infections in a neonatal intensivecare unit traced to balloons used in manual ventilationrdquoJournal of Clinical Microbiology vol 38 no 11 pp 4131ndash41362000

[16] Y Almutawif B Hartmann M Lloyd W Erber andD Geddes ldquoA retrospective audit of bacterial culture results ofdonated human milk in Perth Western Australiardquo EarlyHuman Development vol 105 pp 1ndash6 2017

[17] Centre for Clinical Practice at NICE (UK) Donor Breast MilkBankse Operation of Donor Milk Bank Services 20120210ed National Institute for Health and Clinical Excellence (UK)London UK 2010

[18] C Haas J Rose and C Gerba Quantitative Microbial RiskAssessment John Wiley amp Sons Hoboken NJ USA 2ndedition 2014

[19] M J Nauta A Modular Process Risk Model Structure forQuantitative Microbiological Risk Assessment and Its Applica-tion in an Exposure Assessment of Bacillus cereus in a REPFEDNational Institute of Public Health and the EnvironmentAmsterdam Netherlands 2001 Contract No 149106 007

[20] M Nauta ldquoMicrobiological risk assessment models for par-titioning and mixing during food handlingrdquo InternationalJournal of Food Microbiology vol 100 no 1ndash3 pp 311ndash3222005

[21] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination du lait maternel par une floreaerobie evaluation des pertes pour un lactariumrdquo Archives dePediatrie vol 22 no 5 pp 461ndash467 2015

[22] Hema-Quebec Guide drsquoinformation sur le don de laitmaternel Banque publique de lait maternel Hema-QuebecMontreal Canada 2014

[23] C N Haas ldquoEstimation of risk due to low doses of micro-organisms a comparison of alternative methodologiesrdquoAmerican Journal of Epidemiology vol 118 no 4 pp 573ndash582 1983

[24] C N Haas ldquoMicrobial dose response modeling past presentand futurerdquo Environmental Science and Technology vol 49no 3 pp 1245ndash1259 2015

[25] R Lindqvist and A Westoo ldquoQuantitative risk assessment forListeria monocytogenes in smoked or gravad salmon andrainbow trout in Swedenrdquo International Journal of FoodMicrobiology vol 58 no 3 pp 181ndash196 2000

[26] D Vose Risk Analysis A Quantitative Guide John Wiley ampSons New York NY USA 3rd edition 2008

[27] P Acai L Valik and D Liptakova ldquoQuantitative risk as-sessment of Bacillus cereus in pasteurised milk produced in

the Slovak Republicrdquo Czech Journal of Food Sciences vol 32no 2 pp 122ndash131 2014

[28] S Regli J B Rose C N Haas and C P Gerba ldquoModeling therisk from Giardia and viruses in drinking waterrdquo Journal -American Water Works Association vol 83 no 11 pp 76ndash841991

[29] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination of human milk with aerobicflora evaluation of losses for a humanmilk bankrdquo Archives dePediatrie 22 22 pp 461ndash467 2015 in French

[30] US Food and Drug Administration Food CompositionStandards Labeling And Economics Compliance ProgramGuidance Manual Chapter 21ndashFood Composition StandardsLabeling and Economics US Food and Drug AdministrationSilver Spring MD USA 2006 httpswwwfdagovdownloadsFoodComplianceEnforcementUCM073339pdf

[31] FSANZ Bacillus cereus Limits in Infant Formula FSANZCanberra BC Autralia 2004 httpswwwfoodstandardsgovaucodeapplicationsdocumentsA454_B_cereus_FARpdf

[32] Commission of the European Communities CommissionRegulation (EC) no 14412007 Amending Regulation (EC) No20732005 on Microbiological Criteria for Foodstuffs Com-mission of the European Communities Luxembourg 2007

[33] M L Cohen and TWhalen ldquoImplications of low level humanexposure to respirable B Anthracisrdquo Applied Biosafetyvol 12 no 2 pp 109ndash115 2016

[34] M E Coleman B -ran S S Morse M Hugh-Jones andS Massulik ldquoInhalation anthrax dose response and riskanalysisrdquo Biosecurity and Bioterrorism Biodefense StrategyPractice and Science vol 6 no 2 pp 147ndash160 2008

[35] R J Schanler R J Shulman and C Lau ldquoFeeding strategiesfor premature infants beneficial outcomes of feeding fortifiedhuman milk versus preterm formulardquo Pediatrics vol 103no 6 pp 1150ndash1157 1999

[36] P M Sisk C A Lovelady R G Dillard K J Gruber andT M OrsquoShea ldquoEarly human milk feeding is associated with alower risk of necrotizing enterocolitis in very low birth weightinfantsrdquo Journal of Perinatology vol 27 no 7 pp 428ndash4332007

8 Canadian Journal of Infectious Diseases and Medical Microbiology

sampling consisted of nine sample of 100 microliters permilk pool 6996 pools with a mean Bacillus cereus con-centration of 021 would not be detected and would bequalified for distribution 3004 pools would have beendetected with an average Bacillus cereus concentration of1301CFUml and a rejection rate of 30 (Table 2) -issimulated rejection rate and the concentration in therejected pool are consistent with those of our internalobservations

In summary concentration of Bacillus cereus in poolsput into the inventory would be higher by 043CFUmlaverage when detected using one sample of 100microlitersversus 9 samples of 100 microliters Moreover the rejectedmilk volume difference would be 19312 liters corre-sponding to 203 of the total volume simulated

32 Residual Risk Differential Figure 2 represents the meanrisk differential per million between the two methods ofpost-pasteurization bacterial control (ie 1 sample versus 9samples of 100microliters) by scenario and infectious dose-is figure shows that for the vast majority of scenarios evenamong the most pessimistic the mean risk differential as-sociated with sampling of 1 test rather than 9 tests of100microliters is well below the threshold of 11 millionadditional infections per serving It is only when we supposethat the minimal infectious dose is 20CFUml or less that weobserve a risk difference greater than 11 million

Table 3 presents detailed data concerning the residualrisk scenarios comparing a 100-microliter sample to a 900-microliter sample for post-pasteurization culture whenvarying the infectious dose per serving -e mean riskdifferential for scenario 1 ranges from 215 to 000539 per

million for virulent dose of 05 to 200CFUml For scenarios2 and 3 the mean risk differential ranges from 278 to 0070per million and from 428 to 0108 respectively

4 Discussion

Proven invasive infections due to Bacillus cereus have beenreported among immunosuppressed patients particularlypreterm neonates but given the scarcity of cases thepathophysiology of these infections is not well knownMoreover assessment of the origin of contamination due toorganisms such as B cereus that are widely disseminated inthe environment is often difficult and may not yield anobvious source [11] Nonetheless in outbreak investigationscontamination through respiratory route was often docu-mented [11 15] Several studies have also assessed the pu-tative role of human milk in B cereus infection in prematureinfants however clinical strains of B cereus never matchedthe strains found in milk and therefore the tracking of theorigin of B cereus to contaminated bank milk was to datenever demonstrated [6 9]

Our analysis cannot make any statement regarding theactual risk of Bacillus cereus infection resulting from theingestion of contaminated banked human milk Even if therisk is real which remains unproven yet it is likely to beextremely small based on the various scenarios that weexplored in this simulation -e question of whether there isa real threat posed by Bacillus cereus to the safety of bankedhuman milk therefore remains unresolved

What our simulation shows is that even in very pes-simistic scenarios regarding a causal link between bankedmilk and B cereus infection a more stringent culture-sampling strategy makes very little difference in the risk

Table 3 Mean risk of infection by ingestion of banked maternal milk by serving according to different levels of Bacillus cereus infectiousdose in premature neonateslowast

Infectiousdose1

Scenario 1 Scenario 2 Scenario 3100 microl

sampling9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

100 microlsampling

9times100 microlsampling

Riskdifferential

200 799E minus 9 260E minus 9 539E minus 9 103E minus 7 334E minus 8 696E minus 8 159E minus 7 514E minus 8 108E minus 7100 160E minus 8 519E minus 9 108E minus 8 206E minus 7 669E minus 8 139E minus 7 317E minus 7 103E minus 7 214E minus 775 213E minus 8 693E minus 9 144E minus 8 275E minus 7 892E minus 8 186E minus 7 423E minus 7 137E minus 7 286E minus 765 246E minus 8 799E minus 9 166E minus 8 317E minus 7 103E minus 7 214E minus 7 488E minus 7 158E minus 7 33 E minus 755 290E minus 8 945E minus 9 195E minus 8 375E minus 7 122E minus 7 253E minus 7 576E minus 7 187E minus 7 389E minus 750 319E minus 8 104E minus 8 215E minus 8 412E minus 7 134E minus 7 278E minus 7 634E minus 7 206E minus 7 428E minus 745 355E minus 8 115E minus 8 240E minus 8 458E minus 7 149E minus 7 309E minus 7 705E minus 7 229E minus 7 476E minus 740 399E minus 8 130E minus 8 269E minus 8 515E minus 7 167E minus 7 348E minus 7 793E minus 7 257E minus 7 536E minus 735 456E minus 8 148E minus 8 308E minus 8 589E minus 7 191E minus 7 398E minus 7 906E minus 7 294E minus 7 612E minus 730 532E minus 8 173E minus 8 359E minus 8 687E minus 7 223E minus 7 464E minus 7 106E minus 6 343E minus 7 717E minus 720 799E minus 8 260E minus 8 539E minus 8 103E minus 6 334E minus 7 696E minus 7 158E minus 6 514E minus 7 107E minus 615 106E minus 7 346E minus 8 714E minus 8 137E minus 6 446E minus 7 924E minus 7 211E minus 6 686E minus 7 142E minus 612 133E minus 7 433E minus 8 897E minus 8 172E minus 6 557E minus 7 116E minus 6 264E minus 6 857E minus 7 178E minus 69 177E minus 7 577E minus 8 119E minus 7 229E minus 6 743E minus 7 155E minus 6 352E minus 6 114E minus 6 238E minus 66 266E minus 7 865E minus 8 179E minus 7 343E minus 6 111E minus 6 232E minus 6 528E minus 6 171E minus 6 357E minus 63 532E minus 7 173E minus 7 359E minus 7 687E minus 6 223E minus 6 464E minus 6 106E minus 5 343E minus 6 717E minus 61 160E minus 6 519E minus 7 108E minus 6 206E minus 5 669E minus 6 131E minus 5 317E minus 5 103E minus 5 214E minus 505 319E minus 6 104E minus 6 215E minus 6 412E minus 5 134E minus 5 278E minus 5 634E minus 5 206E minus 5 428E minus 5lowastSimulation scenario was made after post-pasteurized microbiological control using detection of 100 microliters vs 9times100 microliters per pool and for anannual incidence of infection of 001 013 and 02 corresponding to scenarios 1 2 and 3 respectively 1Infectious concentration in CFUml

6 Canadian Journal of Infectious Diseases and Medical Microbiology

incurred by exposed infants At the same time because ofthe ubiquitous nature of this bacterium increasing thesampling volume and thus the sensitivity of the culturemethod can lead to much higher rates of product dis-qualification More specifically our analysis shows that therisk difference per serving associated with a sampling of 1instead of 9 tests of 100 microliters is well below 11 millioninfections a risk that can be considered negligible In allscenarios it is only when the minimal infectious dose was20 CFUml or less for a 100ml serving that the risk dif-ferential was greater than 1 infection1 million portionssuch minimal infectious doses are clearly unrealistic forBacillus cereus Also average concentrations of Bacilluscereus found in the qualified milk batches would only beslightly increased from 021 to 064 CFUml which re-mains well below the standards applied to formula milk(ie 50 CFUg in Europe and 100 CFUg in the USAAustralia and New Zealand) [30ndash32] By comparison theliterature on human and animal exposure by inhalation todifferent doses of spores of Bacillus anthracis reports aminimal infectious dose of at least 600 CFU [33 34] If weassume the same minimal infectious dose of 600 CFU perserving for B cereus in BHM this would still be ten timeshigher than the average B cereus concentration remainingin the qualified milk pools Furthermore because of themuch lower potential for the pathogenicity of Bacilluscereus compared with B anthracis [33 34] this hypothesisis extremely aggressive Finally the gain in supply resultingfrom a less-stringent culture method and its potentialimpact on product availability for the prevention of nec-rotizing enterocolitis largely exceeds the very low theo-retical differential risk as calculated by these models[32 35 36] In our operational setting changing thesampling volume from 900 to 100 microliters is expected todecrease the rate of disqualified culture-positive poolsfrom 30 to 10 percent -is will represent a major im-provement in our capacity to meet the demand for pre-mature infants

To our knowledge this is the very first study usingMonte Carlo simulation which estimates the potential riskof Bacillus cereus infection in preterm infants fed withbanked human milk -e major limitation of our study isthat we had to estimate many of the parameters used in theabsence of empirical data concerning their value Also thesimulation was based on the premise that any bacterial loadin the milk pool is homogeneously distributed which isprobably the case in the vast majority of pools in our ex-perience However the possibility of this load being dis-tributed unevenly due to an inadequate milk mixing processin rare instances cannot be ruled out

5 Conclusion

Our simulation results suggest that if it exists at all the risk ofBacillus cereus infection following the ingestion of pasteurizedbanked milk is extremely small Our analysis also shows thateven in worst-case scenarios a 100-microliter sample forpost-pasteurization culture is amply sufficient to mitigate thisrisk a larger sampling volume would only lead to a higher rate

of disqualification for this important health-care resourcewithout having any significant positive impact on safety

Nevertheless despite no specific mention about Bacilluscereus contamination in the majority of the guidelines thissimulation highlights the importance of the combined use ofpasteurization and bacteriological investigation before hu-man milk distribution for extremely preterm babies Inaddition future studies investigating the association be-tween Bacillus cereus concentration and infection of pre-mature infants are needed

Data Availability

-e simulation code and the data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

-e authors declare that they have no conflicts of interest

References

[1] J-A B Bier T Oliver A E Ferguson and B R VohrldquoHuman milk improves cognitive and motor development ofpremature infants during infancyrdquo Journal of Human Lac-tation vol 18 no 4 pp 361ndash367 2016

[2] D L OrsquoConnor J Jacobs R Hall et al ldquoGrowth and de-velopment of premature infants fed predominantly humanmilk predominantly premature infant formula or a combi-nation of human milk and premature formulardquo Journal ofPediatric Gastroenterology and Nutrition vol 37 no 4pp 437ndash446 2003

[3] B R Vohr B B Poindexter A M Dusick et al ldquoBeneficialeffects of breast milk in the neonatal intensive care unit on thedevelopmental outcome of extremely low birth weight infantsat 18 months of agerdquo Pediatrics vol 118 no 1 pp e115ndashe1232006

[4] J Mahon L Claxton and H Wood ldquoModelling the cost-effectiveness of human milk and breastfeeding in preterminfants in the United Kingdomrdquo Health Economics Reviewvol 6 no 1 2016

[5] A Singhal T J Cole M Fewtrell and A Lucas ldquoBreastmilkfeeding and lipoprotein profile in adolescents born pretermfollow-up of a prospective randomised studyrdquo e Lancetvol 363 no 9421 pp 1571ndash1578 2004

[6] A M Wendelboe C Smelser C A Lucero andL C McDonald ldquoCluster of necrotizing enterocolitis in aneonatal intensive care unit New Mexico 2007rdquo AmericanJournal of Infection Control vol 38 no 2 pp 144ndash148 2010

[7] S Arslanoglu G E Moro and E E Ziegler ldquo-e WapmWorking Group on N Optimization of human milk fortifi-cation for preterm infants new concepts and recommenda-tionsrdquo Journal of Perinatal Medicine vol 38 no 3pp 233ndash238 2010

[8] L M Gartner J Morton R A Lawrence et al ldquoBreastfeedingand the use of human milkrdquo Pediatrics vol 115 no 2pp 496ndash506 2005

[9] J-W Decousser N Ramarao C Duport et al ldquoBacillus cereusand severe intestinal infections in preterm neonates putativerole of pooled breast milkrdquo American Journal of InfectionControl vol 41 no 10 pp 918ndash921 2013

Canadian Journal of Infectious Diseases and Medical Microbiology 7

[10] S Landers and K Updegrove ldquoBacteriological screening ofdonor human milk before and after Holder pasteurizationrdquoBreastfeeding Medicine vol 5 no 3 pp 117ndash121 2010

[11] N J Hilliard R L Schelonka and K B Waites ldquoBacilluscereus bacteremia in a preterm neonaterdquo Journal of ClinicalMicrobiology vol 41 no 7 pp 3441ndash3444 2003

[12] E J Bottone ldquoBacillus cereus a volatile human pathogenrdquoClinical Microbiology Reviews vol 23 no 2 pp 382ndash3982010

[13] R Lotte A-L Herisse Y Berrouane et al ldquoVirulence analysisof Bacillus cereus isolated after death of preterm neonatesnice France 2013rdquo Emerging Infectious Diseases vol 23no 5 pp 845ndash848 2017

[14] N Ramarao L Belotti S Deboscker et al ldquoTwo unrelatedepisodes of Bacillus cereus bacteremia in a neonatal intensivecare unitrdquo American Journal of Infection Control vol 42no 6 pp 694-695 2014

[15] W C Van Der Zwet G A Parlevliet P H Savelkoul et alldquoOutbreak of Bacillus cereus infections in a neonatal intensivecare unit traced to balloons used in manual ventilationrdquoJournal of Clinical Microbiology vol 38 no 11 pp 4131ndash41362000

[16] Y Almutawif B Hartmann M Lloyd W Erber andD Geddes ldquoA retrospective audit of bacterial culture results ofdonated human milk in Perth Western Australiardquo EarlyHuman Development vol 105 pp 1ndash6 2017

[17] Centre for Clinical Practice at NICE (UK) Donor Breast MilkBankse Operation of Donor Milk Bank Services 20120210ed National Institute for Health and Clinical Excellence (UK)London UK 2010

[18] C Haas J Rose and C Gerba Quantitative Microbial RiskAssessment John Wiley amp Sons Hoboken NJ USA 2ndedition 2014

[19] M J Nauta A Modular Process Risk Model Structure forQuantitative Microbiological Risk Assessment and Its Applica-tion in an Exposure Assessment of Bacillus cereus in a REPFEDNational Institute of Public Health and the EnvironmentAmsterdam Netherlands 2001 Contract No 149106 007

[20] M Nauta ldquoMicrobiological risk assessment models for par-titioning and mixing during food handlingrdquo InternationalJournal of Food Microbiology vol 100 no 1ndash3 pp 311ndash3222005

[21] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination du lait maternel par une floreaerobie evaluation des pertes pour un lactariumrdquo Archives dePediatrie vol 22 no 5 pp 461ndash467 2015

[22] Hema-Quebec Guide drsquoinformation sur le don de laitmaternel Banque publique de lait maternel Hema-QuebecMontreal Canada 2014

[23] C N Haas ldquoEstimation of risk due to low doses of micro-organisms a comparison of alternative methodologiesrdquoAmerican Journal of Epidemiology vol 118 no 4 pp 573ndash582 1983

[24] C N Haas ldquoMicrobial dose response modeling past presentand futurerdquo Environmental Science and Technology vol 49no 3 pp 1245ndash1259 2015

[25] R Lindqvist and A Westoo ldquoQuantitative risk assessment forListeria monocytogenes in smoked or gravad salmon andrainbow trout in Swedenrdquo International Journal of FoodMicrobiology vol 58 no 3 pp 181ndash196 2000

[26] D Vose Risk Analysis A Quantitative Guide John Wiley ampSons New York NY USA 3rd edition 2008

[27] P Acai L Valik and D Liptakova ldquoQuantitative risk as-sessment of Bacillus cereus in pasteurised milk produced in

the Slovak Republicrdquo Czech Journal of Food Sciences vol 32no 2 pp 122ndash131 2014

[28] S Regli J B Rose C N Haas and C P Gerba ldquoModeling therisk from Giardia and viruses in drinking waterrdquo Journal -American Water Works Association vol 83 no 11 pp 76ndash841991

[29] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination of human milk with aerobicflora evaluation of losses for a humanmilk bankrdquo Archives dePediatrie 22 22 pp 461ndash467 2015 in French

[30] US Food and Drug Administration Food CompositionStandards Labeling And Economics Compliance ProgramGuidance Manual Chapter 21ndashFood Composition StandardsLabeling and Economics US Food and Drug AdministrationSilver Spring MD USA 2006 httpswwwfdagovdownloadsFoodComplianceEnforcementUCM073339pdf

[31] FSANZ Bacillus cereus Limits in Infant Formula FSANZCanberra BC Autralia 2004 httpswwwfoodstandardsgovaucodeapplicationsdocumentsA454_B_cereus_FARpdf

[32] Commission of the European Communities CommissionRegulation (EC) no 14412007 Amending Regulation (EC) No20732005 on Microbiological Criteria for Foodstuffs Com-mission of the European Communities Luxembourg 2007

[33] M L Cohen and TWhalen ldquoImplications of low level humanexposure to respirable B Anthracisrdquo Applied Biosafetyvol 12 no 2 pp 109ndash115 2016

[34] M E Coleman B -ran S S Morse M Hugh-Jones andS Massulik ldquoInhalation anthrax dose response and riskanalysisrdquo Biosecurity and Bioterrorism Biodefense StrategyPractice and Science vol 6 no 2 pp 147ndash160 2008

[35] R J Schanler R J Shulman and C Lau ldquoFeeding strategiesfor premature infants beneficial outcomes of feeding fortifiedhuman milk versus preterm formulardquo Pediatrics vol 103no 6 pp 1150ndash1157 1999

[36] P M Sisk C A Lovelady R G Dillard K J Gruber andT M OrsquoShea ldquoEarly human milk feeding is associated with alower risk of necrotizing enterocolitis in very low birth weightinfantsrdquo Journal of Perinatology vol 27 no 7 pp 428ndash4332007

8 Canadian Journal of Infectious Diseases and Medical Microbiology

incurred by exposed infants At the same time because ofthe ubiquitous nature of this bacterium increasing thesampling volume and thus the sensitivity of the culturemethod can lead to much higher rates of product dis-qualification More specifically our analysis shows that therisk difference per serving associated with a sampling of 1instead of 9 tests of 100 microliters is well below 11 millioninfections a risk that can be considered negligible In allscenarios it is only when the minimal infectious dose was20 CFUml or less for a 100ml serving that the risk dif-ferential was greater than 1 infection1 million portionssuch minimal infectious doses are clearly unrealistic forBacillus cereus Also average concentrations of Bacilluscereus found in the qualified milk batches would only beslightly increased from 021 to 064 CFUml which re-mains well below the standards applied to formula milk(ie 50 CFUg in Europe and 100 CFUg in the USAAustralia and New Zealand) [30ndash32] By comparison theliterature on human and animal exposure by inhalation todifferent doses of spores of Bacillus anthracis reports aminimal infectious dose of at least 600 CFU [33 34] If weassume the same minimal infectious dose of 600 CFU perserving for B cereus in BHM this would still be ten timeshigher than the average B cereus concentration remainingin the qualified milk pools Furthermore because of themuch lower potential for the pathogenicity of Bacilluscereus compared with B anthracis [33 34] this hypothesisis extremely aggressive Finally the gain in supply resultingfrom a less-stringent culture method and its potentialimpact on product availability for the prevention of nec-rotizing enterocolitis largely exceeds the very low theo-retical differential risk as calculated by these models[32 35 36] In our operational setting changing thesampling volume from 900 to 100 microliters is expected todecrease the rate of disqualified culture-positive poolsfrom 30 to 10 percent -is will represent a major im-provement in our capacity to meet the demand for pre-mature infants

To our knowledge this is the very first study usingMonte Carlo simulation which estimates the potential riskof Bacillus cereus infection in preterm infants fed withbanked human milk -e major limitation of our study isthat we had to estimate many of the parameters used in theabsence of empirical data concerning their value Also thesimulation was based on the premise that any bacterial loadin the milk pool is homogeneously distributed which isprobably the case in the vast majority of pools in our ex-perience However the possibility of this load being dis-tributed unevenly due to an inadequate milk mixing processin rare instances cannot be ruled out

5 Conclusion

Our simulation results suggest that if it exists at all the risk ofBacillus cereus infection following the ingestion of pasteurizedbanked milk is extremely small Our analysis also shows thateven in worst-case scenarios a 100-microliter sample forpost-pasteurization culture is amply sufficient to mitigate thisrisk a larger sampling volume would only lead to a higher rate

of disqualification for this important health-care resourcewithout having any significant positive impact on safety

Nevertheless despite no specific mention about Bacilluscereus contamination in the majority of the guidelines thissimulation highlights the importance of the combined use ofpasteurization and bacteriological investigation before hu-man milk distribution for extremely preterm babies Inaddition future studies investigating the association be-tween Bacillus cereus concentration and infection of pre-mature infants are needed

Data Availability

-e simulation code and the data used to support thefindings of this study are available from the correspondingauthor upon request

Conflicts of Interest

-e authors declare that they have no conflicts of interest

References

[1] J-A B Bier T Oliver A E Ferguson and B R VohrldquoHuman milk improves cognitive and motor development ofpremature infants during infancyrdquo Journal of Human Lac-tation vol 18 no 4 pp 361ndash367 2016

[2] D L OrsquoConnor J Jacobs R Hall et al ldquoGrowth and de-velopment of premature infants fed predominantly humanmilk predominantly premature infant formula or a combi-nation of human milk and premature formulardquo Journal ofPediatric Gastroenterology and Nutrition vol 37 no 4pp 437ndash446 2003

[3] B R Vohr B B Poindexter A M Dusick et al ldquoBeneficialeffects of breast milk in the neonatal intensive care unit on thedevelopmental outcome of extremely low birth weight infantsat 18 months of agerdquo Pediatrics vol 118 no 1 pp e115ndashe1232006

[4] J Mahon L Claxton and H Wood ldquoModelling the cost-effectiveness of human milk and breastfeeding in preterminfants in the United Kingdomrdquo Health Economics Reviewvol 6 no 1 2016

[5] A Singhal T J Cole M Fewtrell and A Lucas ldquoBreastmilkfeeding and lipoprotein profile in adolescents born pretermfollow-up of a prospective randomised studyrdquo e Lancetvol 363 no 9421 pp 1571ndash1578 2004

[6] A M Wendelboe C Smelser C A Lucero andL C McDonald ldquoCluster of necrotizing enterocolitis in aneonatal intensive care unit New Mexico 2007rdquo AmericanJournal of Infection Control vol 38 no 2 pp 144ndash148 2010

[7] S Arslanoglu G E Moro and E E Ziegler ldquo-e WapmWorking Group on N Optimization of human milk fortifi-cation for preterm infants new concepts and recommenda-tionsrdquo Journal of Perinatal Medicine vol 38 no 3pp 233ndash238 2010

[8] L M Gartner J Morton R A Lawrence et al ldquoBreastfeedingand the use of human milkrdquo Pediatrics vol 115 no 2pp 496ndash506 2005

[9] J-W Decousser N Ramarao C Duport et al ldquoBacillus cereusand severe intestinal infections in preterm neonates putativerole of pooled breast milkrdquo American Journal of InfectionControl vol 41 no 10 pp 918ndash921 2013

Canadian Journal of Infectious Diseases and Medical Microbiology 7

[10] S Landers and K Updegrove ldquoBacteriological screening ofdonor human milk before and after Holder pasteurizationrdquoBreastfeeding Medicine vol 5 no 3 pp 117ndash121 2010

[11] N J Hilliard R L Schelonka and K B Waites ldquoBacilluscereus bacteremia in a preterm neonaterdquo Journal of ClinicalMicrobiology vol 41 no 7 pp 3441ndash3444 2003

[12] E J Bottone ldquoBacillus cereus a volatile human pathogenrdquoClinical Microbiology Reviews vol 23 no 2 pp 382ndash3982010

[13] R Lotte A-L Herisse Y Berrouane et al ldquoVirulence analysisof Bacillus cereus isolated after death of preterm neonatesnice France 2013rdquo Emerging Infectious Diseases vol 23no 5 pp 845ndash848 2017

[14] N Ramarao L Belotti S Deboscker et al ldquoTwo unrelatedepisodes of Bacillus cereus bacteremia in a neonatal intensivecare unitrdquo American Journal of Infection Control vol 42no 6 pp 694-695 2014

[15] W C Van Der Zwet G A Parlevliet P H Savelkoul et alldquoOutbreak of Bacillus cereus infections in a neonatal intensivecare unit traced to balloons used in manual ventilationrdquoJournal of Clinical Microbiology vol 38 no 11 pp 4131ndash41362000

[16] Y Almutawif B Hartmann M Lloyd W Erber andD Geddes ldquoA retrospective audit of bacterial culture results ofdonated human milk in Perth Western Australiardquo EarlyHuman Development vol 105 pp 1ndash6 2017

[17] Centre for Clinical Practice at NICE (UK) Donor Breast MilkBankse Operation of Donor Milk Bank Services 20120210ed National Institute for Health and Clinical Excellence (UK)London UK 2010

[18] C Haas J Rose and C Gerba Quantitative Microbial RiskAssessment John Wiley amp Sons Hoboken NJ USA 2ndedition 2014

[19] M J Nauta A Modular Process Risk Model Structure forQuantitative Microbiological Risk Assessment and Its Applica-tion in an Exposure Assessment of Bacillus cereus in a REPFEDNational Institute of Public Health and the EnvironmentAmsterdam Netherlands 2001 Contract No 149106 007

[20] M Nauta ldquoMicrobiological risk assessment models for par-titioning and mixing during food handlingrdquo InternationalJournal of Food Microbiology vol 100 no 1ndash3 pp 311ndash3222005

[21] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination du lait maternel par une floreaerobie evaluation des pertes pour un lactariumrdquo Archives dePediatrie vol 22 no 5 pp 461ndash467 2015

[22] Hema-Quebec Guide drsquoinformation sur le don de laitmaternel Banque publique de lait maternel Hema-QuebecMontreal Canada 2014

[23] C N Haas ldquoEstimation of risk due to low doses of micro-organisms a comparison of alternative methodologiesrdquoAmerican Journal of Epidemiology vol 118 no 4 pp 573ndash582 1983

[24] C N Haas ldquoMicrobial dose response modeling past presentand futurerdquo Environmental Science and Technology vol 49no 3 pp 1245ndash1259 2015

[25] R Lindqvist and A Westoo ldquoQuantitative risk assessment forListeria monocytogenes in smoked or gravad salmon andrainbow trout in Swedenrdquo International Journal of FoodMicrobiology vol 58 no 3 pp 181ndash196 2000

[26] D Vose Risk Analysis A Quantitative Guide John Wiley ampSons New York NY USA 3rd edition 2008

[27] P Acai L Valik and D Liptakova ldquoQuantitative risk as-sessment of Bacillus cereus in pasteurised milk produced in

the Slovak Republicrdquo Czech Journal of Food Sciences vol 32no 2 pp 122ndash131 2014

[28] S Regli J B Rose C N Haas and C P Gerba ldquoModeling therisk from Giardia and viruses in drinking waterrdquo Journal -American Water Works Association vol 83 no 11 pp 76ndash841991

[29] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination of human milk with aerobicflora evaluation of losses for a humanmilk bankrdquo Archives dePediatrie 22 22 pp 461ndash467 2015 in French

[30] US Food and Drug Administration Food CompositionStandards Labeling And Economics Compliance ProgramGuidance Manual Chapter 21ndashFood Composition StandardsLabeling and Economics US Food and Drug AdministrationSilver Spring MD USA 2006 httpswwwfdagovdownloadsFoodComplianceEnforcementUCM073339pdf

[31] FSANZ Bacillus cereus Limits in Infant Formula FSANZCanberra BC Autralia 2004 httpswwwfoodstandardsgovaucodeapplicationsdocumentsA454_B_cereus_FARpdf

[32] Commission of the European Communities CommissionRegulation (EC) no 14412007 Amending Regulation (EC) No20732005 on Microbiological Criteria for Foodstuffs Com-mission of the European Communities Luxembourg 2007

[33] M L Cohen and TWhalen ldquoImplications of low level humanexposure to respirable B Anthracisrdquo Applied Biosafetyvol 12 no 2 pp 109ndash115 2016

[34] M E Coleman B -ran S S Morse M Hugh-Jones andS Massulik ldquoInhalation anthrax dose response and riskanalysisrdquo Biosecurity and Bioterrorism Biodefense StrategyPractice and Science vol 6 no 2 pp 147ndash160 2008

[35] R J Schanler R J Shulman and C Lau ldquoFeeding strategiesfor premature infants beneficial outcomes of feeding fortifiedhuman milk versus preterm formulardquo Pediatrics vol 103no 6 pp 1150ndash1157 1999

[36] P M Sisk C A Lovelady R G Dillard K J Gruber andT M OrsquoShea ldquoEarly human milk feeding is associated with alower risk of necrotizing enterocolitis in very low birth weightinfantsrdquo Journal of Perinatology vol 27 no 7 pp 428ndash4332007

8 Canadian Journal of Infectious Diseases and Medical Microbiology

[10] S Landers and K Updegrove ldquoBacteriological screening ofdonor human milk before and after Holder pasteurizationrdquoBreastfeeding Medicine vol 5 no 3 pp 117ndash121 2010

[11] N J Hilliard R L Schelonka and K B Waites ldquoBacilluscereus bacteremia in a preterm neonaterdquo Journal of ClinicalMicrobiology vol 41 no 7 pp 3441ndash3444 2003

[12] E J Bottone ldquoBacillus cereus a volatile human pathogenrdquoClinical Microbiology Reviews vol 23 no 2 pp 382ndash3982010

[13] R Lotte A-L Herisse Y Berrouane et al ldquoVirulence analysisof Bacillus cereus isolated after death of preterm neonatesnice France 2013rdquo Emerging Infectious Diseases vol 23no 5 pp 845ndash848 2017

[14] N Ramarao L Belotti S Deboscker et al ldquoTwo unrelatedepisodes of Bacillus cereus bacteremia in a neonatal intensivecare unitrdquo American Journal of Infection Control vol 42no 6 pp 694-695 2014

[15] W C Van Der Zwet G A Parlevliet P H Savelkoul et alldquoOutbreak of Bacillus cereus infections in a neonatal intensivecare unit traced to balloons used in manual ventilationrdquoJournal of Clinical Microbiology vol 38 no 11 pp 4131ndash41362000

[16] Y Almutawif B Hartmann M Lloyd W Erber andD Geddes ldquoA retrospective audit of bacterial culture results ofdonated human milk in Perth Western Australiardquo EarlyHuman Development vol 105 pp 1ndash6 2017

[17] Centre for Clinical Practice at NICE (UK) Donor Breast MilkBankse Operation of Donor Milk Bank Services 20120210ed National Institute for Health and Clinical Excellence (UK)London UK 2010

[18] C Haas J Rose and C Gerba Quantitative Microbial RiskAssessment John Wiley amp Sons Hoboken NJ USA 2ndedition 2014

[19] M J Nauta A Modular Process Risk Model Structure forQuantitative Microbiological Risk Assessment and Its Applica-tion in an Exposure Assessment of Bacillus cereus in a REPFEDNational Institute of Public Health and the EnvironmentAmsterdam Netherlands 2001 Contract No 149106 007

[20] M Nauta ldquoMicrobiological risk assessment models for par-titioning and mixing during food handlingrdquo InternationalJournal of Food Microbiology vol 100 no 1ndash3 pp 311ndash3222005

[21] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination du lait maternel par une floreaerobie evaluation des pertes pour un lactariumrdquo Archives dePediatrie vol 22 no 5 pp 461ndash467 2015

[22] Hema-Quebec Guide drsquoinformation sur le don de laitmaternel Banque publique de lait maternel Hema-QuebecMontreal Canada 2014

[23] C N Haas ldquoEstimation of risk due to low doses of micro-organisms a comparison of alternative methodologiesrdquoAmerican Journal of Epidemiology vol 118 no 4 pp 573ndash582 1983

[24] C N Haas ldquoMicrobial dose response modeling past presentand futurerdquo Environmental Science and Technology vol 49no 3 pp 1245ndash1259 2015

[25] R Lindqvist and A Westoo ldquoQuantitative risk assessment forListeria monocytogenes in smoked or gravad salmon andrainbow trout in Swedenrdquo International Journal of FoodMicrobiology vol 58 no 3 pp 181ndash196 2000

[26] D Vose Risk Analysis A Quantitative Guide John Wiley ampSons New York NY USA 3rd edition 2008

[27] P Acai L Valik and D Liptakova ldquoQuantitative risk as-sessment of Bacillus cereus in pasteurised milk produced in

the Slovak Republicrdquo Czech Journal of Food Sciences vol 32no 2 pp 122ndash131 2014

[28] S Regli J B Rose C N Haas and C P Gerba ldquoModeling therisk from Giardia and viruses in drinking waterrdquo Journal -American Water Works Association vol 83 no 11 pp 76ndash841991

[29] C Dewitte P Courdent C Charlet D Dumoulin R Courcoland V Pierrat ldquoContamination of human milk with aerobicflora evaluation of losses for a humanmilk bankrdquo Archives dePediatrie 22 22 pp 461ndash467 2015 in French

[30] US Food and Drug Administration Food CompositionStandards Labeling And Economics Compliance ProgramGuidance Manual Chapter 21ndashFood Composition StandardsLabeling and Economics US Food and Drug AdministrationSilver Spring MD USA 2006 httpswwwfdagovdownloadsFoodComplianceEnforcementUCM073339pdf

[31] FSANZ Bacillus cereus Limits in Infant Formula FSANZCanberra BC Autralia 2004 httpswwwfoodstandardsgovaucodeapplicationsdocumentsA454_B_cereus_FARpdf

[32] Commission of the European Communities CommissionRegulation (EC) no 14412007 Amending Regulation (EC) No20732005 on Microbiological Criteria for Foodstuffs Com-mission of the European Communities Luxembourg 2007

[33] M L Cohen and TWhalen ldquoImplications of low level humanexposure to respirable B Anthracisrdquo Applied Biosafetyvol 12 no 2 pp 109ndash115 2016

[34] M E Coleman B -ran S S Morse M Hugh-Jones andS Massulik ldquoInhalation anthrax dose response and riskanalysisrdquo Biosecurity and Bioterrorism Biodefense StrategyPractice and Science vol 6 no 2 pp 147ndash160 2008

[35] R J Schanler R J Shulman and C Lau ldquoFeeding strategiesfor premature infants beneficial outcomes of feeding fortifiedhuman milk versus preterm formulardquo Pediatrics vol 103no 6 pp 1150ndash1157 1999

[36] P M Sisk C A Lovelady R G Dillard K J Gruber andT M OrsquoShea ldquoEarly human milk feeding is associated with alower risk of necrotizing enterocolitis in very low birth weightinfantsrdquo Journal of Perinatology vol 27 no 7 pp 428ndash4332007

8 Canadian Journal of Infectious Diseases and Medical Microbiology

Stem Cells International

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

MEDIATORSINFLAMMATION

of

EndocrinologyInternational Journal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Disease Markers

Hindawiwwwhindawicom Volume 2018

BioMed Research International

OncologyJournal of

Hindawiwwwhindawicom Volume 2013

Hindawiwwwhindawicom Volume 2018

Oxidative Medicine and Cellular Longevity

Hindawiwwwhindawicom Volume 2018

PPAR Research

Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018

Immunology ResearchHindawiwwwhindawicom Volume 2018

Journal of

ObesityJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Computational and Mathematical Methods in Medicine

Hindawiwwwhindawicom Volume 2018

Behavioural Neurology

OphthalmologyJournal of

Hindawiwwwhindawicom Volume 2018

Diabetes ResearchJournal of

Hindawiwwwhindawicom Volume 2018

Hindawiwwwhindawicom Volume 2018

Research and TreatmentAIDS

Hindawiwwwhindawicom Volume 2018

Gastroenterology Research and Practice

Hindawiwwwhindawicom Volume 2018

Parkinsonrsquos Disease

Evidence-Based Complementary andAlternative Medicine

Volume 2018Hindawiwwwhindawicom

Submit your manuscripts atwwwhindawicom