Quantitative Microbiological Exposure Assessment – QMEA

Quantitative Microbiological Exposure Assessment – QMEA JM Membré INRAE, Nantes •Introduction •Focus on exposure assessment •Probability density •A simple example •Uncertainty •Sensitivitiy analysis •Example with bakery products Plan 2 Introduction 3 Risk Analysis(WHO) Risk Communication Risk Assessment Risk Management · Hazard Identification · Hazard Characterisation · Exposure Assessment · Risk Characterisation · Risk Evaluation · Option Assessment · Option Implementation · Monitoring & Review Interactive exchange of information and opinions concerning risks Scientificbased Policybased 4 Introduction Microbiological risk assessment The purpose of microbiological risk assessment is to characterize the nature and probability of harm resulting from human exposure to the biological agents present in foodstuffs. Hazard identification: A qualitative process to identify microbial hazards of concern in foods. Microbial hazards can include infectious agents or toxins produced by microorganisms. Exposure Assessment: An estimate of the probability and the level of hazard in a food product portion. By identifying the frequency and quantity of food consumed, this also makes it possible to estimate the exposure of the (sub)population to a microbiological hazard. Hazard Characterization: A description of the harmful effects that may result from the ingestion of a hazard, whether a microorganism or its toxin, and the expression of a dose–response relationship where possible. Risk Characterization: The integration of the previous three steps to obtain a risk estimate, that is an estimate of the probability and severity of harmful effects occurring in a given (sub)population, with the uncertainties linked to the consumption of a food product contaminated by the hazard. 5 Introduction Listeria monocytogenes Salmonella Campylobacter Escherichia coli O157 Vibrio parahaemolyticus Hazard Identification 6 Introduction 7 Introduction Exposure Assessment Exposure is deduced from: Quantity of microorganisms ( Propagation of microorganisms from farm to fork) Quantity eaten (portion size and frequency of consumption) Size of the population 8 Introduction Exposure Assessment At the factory, several steps, for instance: Mixing Adding additives (formulation is changing) Thermal treatment Partitionning Cutting, slicing Packaging with or without modified atmosphere Setting a shelf-life … 9 Introduction Often, it is not only important to estimate the current public health risk, but also to evaluate and compare the expected impact of control measures that are taken along the food pathway. If such control measures are to be taken at the farm, obviously, the farm stage has to be included in the food pathway. If such control measures are not considered, a shorter pathway may be used.  Building a fit-for purpose model  What-if scenarios 16 Exposure assessment Main challenges of exposure assessment Various modeling approaches Inactivation and growth can be described by predictive microbiology models that are specific for the hazard. Mixing and partitioning describe how food units are joined or split up (e.g. or milk form a tank distributed over several milk packages)  Probability function Cross-contamination involves a combination of transfer events that leads to contamination of foods, sometimes even in the case of a single microbe transfer event  Mass transfer modelling Sampling and then removing: refers to the selective removal of food items based on a (visual) inspection of their quality, which may be related to the probability and level of contamination of this food item  Probability theory 17 Exposure assessment The ICMSF equation builds on concentrations expressed as log cfu/g, so the exposure can be expressed as the sum of the effects of growth, inactivation and contamination: 𝐻0+ 𝐺 + 𝐶 − 𝑅 = 𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒 where H0 is the initial concentration in a food product, ΣG is the sum of the effect of growth in the process steps that allow growth, ΣR is the sum of the effect of inactivation (reduction) in the process steps that lead to inactivation, ΣC is the increase due to (re-)contamination. 18 Exposure assessment Main challenges of exposure assessment Probability distribution 19 • Let’s consider a practical example •to go to work everyday, someone needs to take successively a bike, a train and a bus •the time cycling (depends on his fitness) is about 10 min, maybe only 5 and sometimes 15 min •the time on the train is 50 min, never below and unlikely 55 min •the time on the bus depends on the traffic, likely value is 20 min, but on a busy day, it could be 30 min, while sometimes it takes only 15 min •The person starts working at 9h •At what time does he need to leave his place ? Decision 20 Probabilistic Risk Assessment -> Decision At what time does he need to leave his place ? •15+55+30 = 100 min. He leaves his place everyday at 7h20, “just in case” (worst case scenario) •if he has a very serious meeting starting at 9h, he leaves his place at 7h20, but generally he leaves at 7h40 (10+50+20= 80 min) (pragmatic decision, realistic...?) •he wants to analyse the problem more in detail before making his decision and he wants to be able to justify (transparency and traceability) his decision 21 Probabilistic Risk Assessment -> Decision + + = The time cycling (depends on his fitness) is about 10 min, maybe only 5 and sometimes 15 min The time on the train is 50 min, never below and unlikely 55 min The time on the bus depends on the traffic, likely value is 20 min, but the busy day, it could be 30 min, sometimes it takes only 15 min 22 The journey takes 100 min once per year (1/354) Probabilistic Risk Assessment -> Decision Lessons Learnt •If he chooses 100 min, he will be always in advance (except once a year) •If he chooses the median value (83 min), he will be late one day in 2 •if he chooses another time, he can determine the risk to be late... and make the “right” decision (management) •Probabilistic approach helps in quantifying the risk and making decision •PRA makes sense when: •Inputs/Factors have definitively more than 1 value (range of values) •Problem is complex, justification of decisions difficult (“fit for purpose”) 23 Probabilistic Risk Assessment -> Decision Percentage of people in a given population eating packed salmon? To answer this question, organize a survey Problem of time and money: although the survey is representative of the population, there is a limited number of people interviewed: 100 people interviewed in total (N) 10 of them eat packed salmon (S)  Prevalence = 10% (based upon Lindqvist et al. 2000) 24 Uncertainty 100 people interviewed Prevalence = 10% Confidence interval around this value: 6-18% To reduce this confidence, what do I need to do? Riskbeta(s+1, n-s+1), s=nbr positive response, n=number of people 25 Uncertainty + + = The time cycling (depends on his fitness) is about 10 min, maybe only 5 and sometimes 15 min The time on the train is 50 min, never below and unlikely 55 min The time on the bus depends on the traffic, likely value is 20 min, but the busy day, it could be 30 min, sometimes it takes only 15 min 32 Sensitivity analysis + + = The time cycling (depends on his fitness) is about 10 min, maybe only 5 and sometimes 15 min The time on the train is 50 min, never below and unlikely 55 min The time on the bus depends on the traffic, likely value is 20 min, but the busy day, it could be 30 min, sometimes it takes only 15 min 33 Sensitivity analysis 34 Sensitivity analysis Example with bakery products 35 Brioche-type Product •Best-by-date around 21 days •Water activity (aw) around 0.86 •No preservative •Slightly acidic (pH ca 5.2) •In-factory mould contamination: possible •Key formulation factor: aw  Combination of aw and shelf-life to avoid spoilage? Example with bakery products 36 Predictive models for SL determination •Time (primary model): •the storage time varies with consumer’s habits. •Formulation and environmental factor (secondary models): •the formulation, and particularly aw, does not vary for a given product  set at ≠ values (what-if scenario). •the storage temperature varies with region & season 37 Example with bakery products 0.27 15.72 90.0% 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0 5 10 15 20 25 30 Probability density Storage time (days) Shelf-life Best before date (BBD) and Shelf-life (= storage time): • Storage time varies with consumer’s habits (BBD is fixed) •A general rule for modelling has been recently suggested: Time = Exp(Best-Before-Date / 4) Roccato et al. 2017. 10.1016/j.foodres.2017.02.017 Ex BBD 21 days 38 Example with bakery products Individual spore variability 39 •Individual spore variability has to be taken into account •Lag time: Indλ ~ Normal(Indmean λ, Indsd λ) 0% 20% 40% 60% 80% 100% 050 100 150 200 250 300 350 400 450 500 550 Cumulated individual lag times Time (h) Increase of stress (due to aw, temper.) Increase of Indmean Increase of Indsd Adapted from Dagnas et al. 2015 (10.1016/j.ijfoodmicro.2015.07.008) Example with bakery products Probability of spoilage •Spoilage rate = Pr Indλ≤Storage time 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0 5 10 15 20 25 30 Probability density Storage time (days) 1Ind𝑚𝑒𝑎𝑛 = 1λ𝑜𝑝𝑡 x γ(aw) x γ(T) Aw: 0.85; 0.86; 0.87….. 0.90 0.00 0.05 0.10 0.15 0.20 0.25 16 18 20 22 24 26 28 30 Probability density Temperature (°C) BBD: 7, 14…. 35 days •Spoilage: proba to achieve visible growth before shelf − life 40 Example with bakery products Brioche-type Product 0.80 0.82 0.84 0.86 0.88 0.90 714 21 28 Water activity "Best-Before-Date" (duration expressed in days) RR 0.7 RR 1 RR 1.1 RR 1.2 Combinations of aw and BBD having same spoilage rate •Iso-risk curve, expressed in Relative Risk (RR) •RR=1 for current BBD and formulation 41 Example with bakery products