A swift Quantitative Microbiological Risk Asssment (sQMRA)tool
Eric G.Evers *,Jurgen E.Chardon
National Institute for Public Health and the Environment,P.O.Box 1,3720BA Bilthoven,The Netherlands
a r t i c l e i n f o Article history:
Received 19December 2008
Received in revid form 11June 2009Accepted 16June 2009
Keywords:
Risk asssment QMRA
Relative risk
Mathematical model Campylobacter Salmonella
a b s t r a c t
Classical full-scale QMRA is a valuable method to asss the effects of control measures on the public health risk of a pathogen–food product combination.However,development of the QMRA models is time consuming,data needs are substantial and it requires extensive modeling experti.We therefore developed a simplified QMRA model especially aimed at comparing the risk of pathogen–food product combinations.The swift Quantitative Microbiological Risk Asssment (sQMRA)–tool is implemented in Microsoft Excel.Special attention is given to make the sQMRA tool insightful,for educational purpos.Like in full-scale QMRA,pathogen numbers are followed through the food chain,which in this ca starts at retail and ends with the number of human cas of illness.The model is deterministic and includes cross-contamination and preparation (heating)in the kitchen and a do–respon relationship.The gen-eral tup of the sQMRA tool consists of concutive questions for values of each of the 11parameters,always followed by intermediate model output broken down into categories of contamination,cross-con-tamination and preparation.In a parate sheet,model input and output are summarized and exposure as well as cas are attributed to the distinguished categories.As a relative risk measure,intermediate and final model outputs are always compared with results from a full-scale QMRA of Campylobacter on chicken fillet.Example calculations with the sQMRA-tool were done for all combinations of Campylobac-ter spp.and Salmonella spp.with chicken fillet,filet americain (raw minced beef with mayonnai)and tabl
e eggs.Data availability appeared to be partly poor.The predicted risk was highest for Salmonella spp.in table eggs and Campylobacter spp.in chicken fillet.Results indicate that the sQMRA-tool is uful for quickly obtaining relative risk estimates of pathogen–food combinations.It can thus rve as a guide for lection of combinations for applying full-scale QMRA,or for risk management –by facilitating the translation of the results of trend analysis or of a specific rearch project into terms of risk.
Ó2009Elvier Ltd.All rights rerved.
1.Introduction
A full scale quantitative microbiological risk asssment (QMRA)describes the propagation of a pathogenic micro-organism from farm via slaughter and retail to the consumer’s home,using math-ematical modeling techniques.In the last decade veral studies covering the whole or part of this food chain were performed (e.g.Cassin,Lammerding,Todd,Ross,&McColl,1998;Nauta,Evers,Tak-umi,&Havelaar,2001;Ronquist,Nieln,Sommer,Norrung,&Christenn,2003;Schlundt,2000).The estimated exposure of con-sumers to a pathogen is input for the do–respon relationship which results in an estimate of the number of human cas.More importantly,a QMRA model is capable of
不同桃李混芳尘
estimating the public health effect of interventions measures in terms of the reduction of this number.When economics is taken into account,even a cost-utility analysis can be performed (Havelaar et al.,2007).
A disadvantage of the valuable QMRA technique is that it is time-consuming and expensive.For a number of applications a
simplified model will suffice and be preferred,where it must be noted that the ability to estimate the effect of intervention mea-sures will be reduced.The applications include:
–Pre-lection of a pathogen–food product combination:when considering a range of pathogens or food products,a simplified model can lect the most relevant pathogen–food combination to be analyzed by a full-scale QMRA.
–Many public health related National Authorities have large dat-abas of measurements of pathogens along the food chain,which are constantly supplemented with new measurements.Using a simplified technique,a continuous rough trend analysis in terms of public health risk can be performed.
–The results of a rearch project can quickly be interpreted in terms of public health risk,given that pathogen concentrations are determined.
–Education:a simplified model is better accessible and under-standable for scientists that are new in the QMRA rearch area or are not very mathematically inclined.Also,QMRA under-standing by policy makers that have to decide on the implemen-tation of intervention measures will be enlarged.
0956-7135/$-e front matter Ó2009Elvier Ltd.All rights rerved.doi:10.1016/j.foodcont.2009.06.013
属金的颜色*Corresponding author.Tel.:+31302744149;fax:+31302744434.E-mail address:eric.evers@rivm.nl (E.G.Evers).Food Control 21(2010)
319–330
Contents lists available at ScienceDirect
Food Control
journal homepage:www.elvi e r.c o m /l o c a
te/foodcont
It must be stresd that when applying a simplified model,the resulting public health risk in terms of numbers of human cas must be interpreted in a relative n,comparing it with a refer-ence study or with other simplified pathogen–product calculations. The absolute values are not very reliable given that this is usually even not the ca for full-scale QMRAs.
Previously,a number of simplified QMRA methods have been developed by others.A good approach
was prented by Ross and Sumner(2002),who implemented their model in an accessible Excel spread sheet,however,they ud categories with inevitable arbitrary weighting factors and a somewhat artificial risk ranking scale.McNab(2003)ud data on exposure and on its impact from farm to consumer in a MS Access application.Both did not include concentrations or numbers of pathogens,which is a necessary in-put of the do–respon relationship and also,numbers of patho-gens are more important for public health risk than prevalences (Nauta&Havelaar,2008;Ronquist et al.,2003).Van Gerwen, Te Giffel,Van’t Riet,Beumer and Zwietering(2000)describe a framework with a stepwi approach including more detail when a transmission route proves to be significant.No predefined model equations are ud;the are chon ad hoc.Teunis,Medema,Kru-idenier,and Havelaar(1997)describe a model for exposure to pathogens via drinking water and the risk of infection,which can be relatively simple as the drinking water chain is more uniform than the food chain.
The objective of this paper is to prent a simplified QMRA method which is implemented in Excel,the swift Quantitative Microbiological Risk Asssment(sQMRA)-tool.The sQMRA tool in-cludes numbers of pathogens and us predefined equations that take the inherent variation in processing in the food chain into ac-count.Relative risk(compared with a reference value)is given for topic是什么意思
five model outputs,including the number of human cas.Special attention is given to make the sQMRA tool insightful,for educa-tional purpos.The sQMRA method builds on previous work by Evers,van der Fels-Klerx,Nauta,Schijven,and Havelaar(2004, 2008)in which human exposure for Campylobacter is estimated for31transmission routes,including food,direct contact and water. Extensions include explicit modeling of food portions,cross-con-tamination and preparation,and the addition of effect modeling.
The sQMRA method is an example of comparative risk asss-ment,a new approach in the t of methods ud for human illness attribution,which include microbial subtyping,epidemiological approaches(study of sporadic infections and outbreaks),interven-tion studies and expert elicitation approaches(Pires et al.,2009).In order to gain insight in the ufulness of the sQMRA method,cal-culations were done for Campylobacter and Salmonella in chicken fillet,filet americain and table eggs.The results are prented here to illustrate the method.
2.Materials and methods
2.1.General
The sQMRA model is a simple risk asssment model in which the propagation of a pathogen is follo
wed in the exposure asss-ment part,which starts at the retail pha.Process considered are cross-contamination and preparation(=heating,in the n of cooking,frying,etc.of the product)in the kitchen,which leads to ven categories of portions.The resulting exposure is input for the effect modeling part,which has risk in terms of infection and illness,per portion as well as at the population level,as output. The model is deterministic and does not include uncertainty and variability.
The sQMRA-tool is implemented in a spread sheet application and is divided into a model sheet and a results sheet.In the model sheet parameter values are inrted,and calculation results are prented at all intermediate points and at the end point.The in-clude the resulting risk and intermediate results such as the num-ber of portions and number of cfu(colony forming units)per portion.In the results sheet,we prent(1)a list of input parame-ter values,(2)attribution of exposure and of cas to the different transmission routes and(3)the relative risk at veral intermedi-ate points and the end point(no.of human cas)in the calculation.
The sQMRA model and tool will be described stepwi below, using screenshots of the tool showing afictitious example together with the corresponding mathematical(left side of page)and Excel equations(right side).From the preparation step onwards,only one example Excel equation will be given for each t of mathe-matical equations.Thereafter,example calculations will be given on Cam
pylobacter and Salmonella in/on chickenfillet,filet americ-ain and table eggs.Filet americain is a bread spread which contains raw minced beef as the main ingredient(70%),together with a mayonnai-bad sauce and spices.It is sometimes called steak tartare.
The tool is implemented in Microsoft Excel XP.Only built-in functionality and no VBA or macro code is ud.Excel cells for in-put parameters have automatic data validation:only quantitative, positive values are accepted and percentages have to be between 0%and100%.All other cells in the tool are write protected.The sQMRA tool can be obtained by nding an e-mail to the corre-sponding author.
2.2.Model
2.2.1.Symbols
A list of input parameters of the model can be found in Fig.7, which depicts the results sheet of the sQMRA-tool.Question num-ber in the tool,parameter symbol,description and value are gi-ven.Most ud symbols are N for number of portions,S for subdivision of the numbers into fractions of categories of por-tions,D for the do(no.of cfu)of a pathogen and F for the frac-tion of the numbers that cross-contaminate or survive.Note that(for educational purpos)fractions are prented as percent-ages in the tool,whereas all actual calculations are done with fractions.饮用水水质标准
Typically the subscripts of symbols consists of two parts pa-rated by a‘/’.Thefirst part describes the pha or process of the risk asssment:r for retail,cc for cross-contamination,pr for prep-aration,ing for ingestion,inf for infection,and ill for illness.The c-ond part describes portions categories:contamination at retail(+) or not(À),occurrence of cross-contamination(cc+)or not(ccÀ)or not specified(ccx)and the preparation method:prd is done,prh is halfdone,prr is raw and pry is not specified.For example, N pr/+,ccÀ,prd stands for the number of portions after preparation that were contaminated at retail,did not cau cross-contamination and were prepared done.In particular for S,the cond part of the subscript can be a pha or process as well,in which ca S stands for a subdivision within a previous pha/process.For example, e.g.S cc/r stands for the percentage of portions that cross-contaminate the environment given that they are contami-nated in retail.
2.2.2.Title and ca definition(Fig.1)
Characteristics of a risk asssment have to be entered in the ca definition box.The data are not ud in the sQMRA calcula-tions,but have an administrative purpo:they are repeated in the results sheet as identification of the calculation results prented there.
320 E.G.Evers,J.E.Chardon/Food Control21(2010)319–330
根特大学
2.2.
3.Consumption data(Fig.2)
Consumption data necessary for the risk calculations are the number of portions N(Question1)given the ca definition (Fig.1)and the average size of the portion M in grams(Question 2).Ideally the variables are obtained from a national food con-sumption survey.
2.2.4.Retail(Fig.2)
The number of contaminated N r/+(cell I35)and uncontaminated portions N r/À(cell I36)in retail equals
N r=þ¼NS r=þI35¼G22ÃH35
N r=À¼Nð1ÀS r=þÞ¼NS r=ÀI36¼G22ÃH36
where S r/+(Question3;cell H35)and S r/À=1ÀS r/+(cell H36)are the fractions of contaminated and uncontaminated portions at retail, respectively.
It must be noted that S r/+includes the fal negatives(non de-tects),so in general it will be higher than the value that b-tained from national monitoring programs.Not including a correction for fal negatives will lead to an underestimation of the risk,however,the calculation results will still be uful when one is aware of this and taking into account the fact that the ne-glected concentrations will be relatively low.The unit‘portion’is ud at this pha,although it can be a virtual concept when sale size in retail differs from portion size at the moment of consumption.
The do per portion(no.of cfu)for contaminated and uncon-taminated portions,D r/+(cell J46)and D r/À(Cell J47),equals
D r=þ¼C r=þM J46¼G40ÃG24
D r=À¼0J47¼0
where C r/+(Question4)is the average concentration of the pathogen in contaminated portions(cfu/g).Again it must be noted that in principle this average includes the concentrations in fal negative portions.From here on,we will concentrate on contaminated portions.
The total do Dtot r/+(no.of cfu;cell K46)stands for the sum of the number of cfu over all portions,which is related to the ca definition(Fig.1).It equals:
Dtot r=þ¼D r=þN r=þK46¼J46ÃI46
2.2.5.Cross-contamination(Fig.3)
It is assumed that a certain part of the portions caus cross-contamination in the kitchen prior to heating and the other part does not.The fractions S cc/+,cc+(cell H64)and numbers N cc/+,cc+(cell I64)of contaminated portions causing cross-contamination can be calculated as follows:
S cc=þ;ccþ¼S r=þS cc=r H64¼H35ÃG54
N cc=þ;ccþ¼N r=þS cc=r I64¼I35ÃG54
鸡飞蛋打的意思where S cc/r(Question5)is the fraction of the portions contaminating the environment given that the portion is contaminated.Analo-gously,the fractions S cc/+,ccÀ(cell H65)and numbers N cc/+,ccÀ(cell I65)of contaminated portions not causing cross-contamination are S cc=þ;ccÀ¼S r=þð1ÀS cc=rÞH65¼H35Ãð1ÀG54Þ
N cc=þ;ccÀ¼N r=þð1ÀS cc=rÞI65¼I35Ãð1ÀG54Þ
Further,it is assumed that cross-contamination is a two-step process.First,prior to food preparation,a
certain part of the patho-gens on the portion F cc(Question6)transfers to the environment. This environment be kitchen equipment or the hands of the person preparing the meal.When cross-contamination
occurs, Fig.1.Ca definition part of the Model sheet of the sQMRA
tool.
Fig.2.Consumption data and Retail part of the Model sheet of the sQMRA tool.
E.G.Evers,J.E.Chardon/Food Control21(2010)319–330321
the do that remains in the portion D cc/+,cc +(no.of cfu;cell J64)and the do that ends up in the environment De cc/+,cc +(no.of cfu;cell K64)are:
D cc =þ;cc þ¼D r =þð1ÀF cc ÞJ64¼J46Ãð1ÀG59ÞDe cc =þ;cc þ¼D r =þF cc
K64¼J46ÃG59
When there is no cross-contamination,the do in the portion D cc/+,cc À(no.of cfu;cell J65)does not change and the environment is not contaminated (cell K65).
D cc =þ;cc À¼D r =þJ65¼J46De cc =þ;cc À¼0K65¼0
Second,a part (F ei ;Question 7)of the pathogens in the environ-ment transfers back to the portion or another part of the meal after heating,if any,and will be ingested by the consumer.The size of this do Dei cc /+,cc +(no.of cfu;cell K76)is
Dei cc =þ;cc þ¼De cc =þ;cc þF ei K76¼K64ÃG71
The rest of the do De cc/+,cc +,Del cc/+,cc +(no.of cfu;cell L76),remains in the environment:
Del cc =þ;cc þ¼De cc =þ;cc þð1ÀF ei ÞL76¼K64Ãð1ÀG71Þ
2.2.6.Preparation (Fig.4)
We distinguish three categories of heating of the portion:done,half-done and raw (=no heating).The fractions S pr/+,ccx,pry (cells H100–H105)and numbers N pr/+,ccx,pry (cells I100–I105)of contami-nated portions that do (x =+)or do not (x =À)cau cross-contam-ination and are prepared done (y =d ),half-done (y =h )or raw (y =r )are calculated as follows:
S pr =þ;ccx ;pry ¼S cc =þ;ccx S pry =cc e :g :H100¼H64ÃG86N pr =þ;ccx ;pry ¼N cc =þ;ccx S pry =cc e :g :I100¼I64ÃG86
where S pry/cc (Question 8)is the fraction of the portions prepared done,half-done or raw given that the portion is contaminated and has caud cross-contamination or not.
The three heating categories are characterized by different frac-tions survival of the pathogen:F prd ,
F prh and F prr ,respectively (Question 9).The do that remains in the portion after heating D pr/+,ccx,pry (no.of cfu/portion;cells J100–J105)equals
D pr =þ;ccx ;pry ¼D cc =þ;ccx F pry e :g :J100¼J64ÃG94
In ca of cross-contamination,the do in the portion that is eventually ingested by the consumer D ing/+,cc +,pry (no.of cfu/por-tion;cells L100–L102)consists of the do remaining after heating and the do from cross-contamination and equals
D ing =þ;cc þ;pry ¼D pr =þ;cc þ;pry þDei cc =þ;cc þe :g :L100¼J100þK100
When there is no cross-contamination,the do ingested by the consumer D ing /+,cc À,pry (no.of cfu/portion;cells L103–L105)is sim-ply equal to the do remaining after heating:
D ing =þ;cc À;pry ¼D pr =þ;cc À;pry e :g :L103¼J103
The total do at the moment of ingestion Dtot ing/+,ccx,pry (no.of cfu;cells M100–M105)corresponding to the ca definition becomes
Dtot ing =þ;ccx ;pry ¼D ing =þ;ccx ;pry N pr =þ;ccx ;pry e :g :M100¼L100ÃI100
and the sum of Dtot ing/+,ccx,pry (no.of cfu;cell M107)for the six con-taminated categories is the total number of pathogens ingested by the considered population.
2.2.7.Infection and illness (Fig.5)
The probability of infection for an ingested do of pathogens,P inf/+,ccx,pry (cells I122–I127),is calculated with the exponential do–respon relationship:
P inf =þ;ccx ;pry ¼1Àe ÀrD ing =þ;ccx ;pry e :g :I122¼1Àe Àr ÃH122
where r is the probability of infection for one single pathogen (Haas,1983).It can simply be derived that
r ¼
ln 2ID 50
r ¼
ln 2G112
where the ID50value (Question 10)is the do per portion that caus infection in half of the expod population (P inf =0.5).In the sQMRA tool the ID50value is ud as input,as this is a value that is better understandable for non-risk asssors and comes clo to the often ud concept of the infectious do.
It is assumed that the probability to get ill given infection (P ill/inf ;Question 11)is a fixed value,independent of the do.So the prob-ability of illness per portion (P ill/+,ccx,pry ;cells J122–J127)becomes:
P ill =þ;ccx ;pry ¼P inf =þ;ccx ;pry P ill =inf e :g :J122¼I122ÃG117
The corresponding numbers of infections (I inf/+,ccx,pry ;cells L122–L127)and cas (I ill/+,ccx,pry ;cells M122–M127)given the ca defini-tion
are
Fig.3.Cross-contamination part of the Model sheet of the sQMRA tool.
322 E.G.Evers,J.E.Chardon /Food Control 21(2010)319–330
I inf =þ;ccx ;pry ¼P inf =þ;ccx ;pry N pr =þ;ccx ;pry e :g :L122¼I122ÃK122I ill =þ;ccx ;pry ¼P ill =þ;ccx ;pry N pr =þ;ccx ;pry
e :g :M122¼J122ÃK122
The sum of the numbers over the categories gives the total num-ber of infections and cas (cells L130and M130).2.2.8.Model overview
A model overview is given in Fig.6.
2.3.Model input and output (RESULTS –sheet;Fig.7)
The intention of the RESULTS sheet is to prent the most esn-tial model input (input parameters)and output (exposure,effect,relative risk)in a conci format.
2.3.1.Input parameters
The ction ‘input parameters’starts with repeating the ca definition (cells D6–G10;e also Fig.1).
Then,all input parameter values from the MODEL sheet are prented in a list,including question number,parameter symbol,a shortened version of the question and the inrted value (cells D11–G28).The time point of calculation is captured with a time stamp (cell D30–F30).2.3.2.Attribution of exposure
Attribution of exposure is defined here as the relative impor-tance of the end point of exposure for the different transmission
routes at the population level.The calculation of absolute expo-sures (no.of cfu)for the different transmission routes is prented in Table 1.It must be noted that the cross-contamination route in-cludes all cfu’s that are ingested via cross-contamination (i.e.,the route via the environment and back),regardless of the way of prep-aration,whereas the three preparation routes refer to ingestion of cfu’s that stayed in the portion and survived the heating (if any).The attribution of exposure (the relative importance)is calculated by dividing the absolute exposures (Table 1)by the total exposure (=the sum of the exposures given in Table 1,or cell M107in the MODEL sheet).The calculations are located in a parate ction of the sheet (AB50–AC56)(not shown).The result is given in cells K17–M22and in the corresponding pie chart (cells K6–M16).2.3.3.Attribution of effect
For attribution of cas a concept is ud which is analogous to the Population Attributable Risk (PAR)in public health epidemiol-ogy studies.We estimate the number of cas when a transmission route is switched off (I ill/attr )and determine the decrea in number of cas relative to the ba scenario with best estimates for all 11input parameters (I ill/ba ;cell M130in the MODEL sheet).In formula:
母亲的诗歌attribution of cas ¼
I ill =ba ÀI ill =attr
I ill =
ba
Fig.4.Preparation part of the Model sheet of the sQMRA
tool.
Fig.5.Infection and illness part of the Model sheet of the sQMRA tool.
E.G.Evers,J.E.Chardon /Food Control 21(2010)319–330323待命是什么意思
