P2-78: Generic Modeling Approach for Quantitative Microbial Risk Assessment

1
P2-78 Generic Modeling Approach for Quantitative
Microbial Risk Assessment Thomas P. Oscar, USDA,
ARS/1890 Center of Excellence in Poultry Food
Safety Research, Room 2111, Center for Food
Science and Technology, University of Maryland
Eastern Shore, Princess Anne, MD 21853
410-651-6062 Thomas.Oscar_at_ars.usda.gov
INTRODUCTION Risk analysis is a holistic approach
to food safety that involves three interactive
processes risk assessment, risk management and
risk communication. Risk assessment modeling is
the foundation of risk analysis and is most often
accomplished using Monte Carlo simulation methods
to combine existing knowledge and data into a
prediction of risk. The prediction of risk is
relative rather than absolute because of
knowledge, data and model uncertainty. However,
through the process of scenario analysis relative
risk can be assessed and used to help inform risk
management decisions.
Process Step   Input Input Input Input Input Input
Process Step   Incidence Incidence Extent Extent Extent Extent
Process Step Hazard Event A B Min Median Max Units
Packaging Contamination 25 10 0 1 4 log D
Distribution Growth 20 40 0.1 1 3 log D
Washing Removal 15 30 -0.1 -1 -3 log D
Cooking Survival 10 10 -0.1 -5 -7 log D
Serving Cross-contamination 15 30 -3 -2 -1 log rate
Fig. 3
Fig. 4
Fig. 1
Class Class Class Input Input Input Input
Hazard Food Host A _B RDmin RD50 RDmax
1 - Normal Normal Normal 70_30 4.0 6.0 8.0
2 - High Normal Normal 6_38 3.0 5.0 7.0
3 - Normal High Normal 2_1 3.5 5.5 7.5
4 - High High Normal 2_1 2.5 4.5 6.5
5 - Normal Normal High 5_5 2.0 4.0 6.0
6 - High Normal High 9_17 1.0 3.0 5.0
7 - Normal High High 3_4 1.5 3.5 5.5
8 - High High High 3_4 0.5 2.5 4.5
    Total A_B    
    Hazard 20_60 High Risk  
    Food 10_10 High Risk  
    Host 20_30 High Risk  
Fig. 5
OBJECTIVE To provide a generic example of how
quantitative microbial risk assessment (QMRA) can
be used to provide a relative assessment of risk
for informing risk management decisions.
MATERIALS AND METHODS Case Study. A food company
has two processing plants located in different
regions of the country but that produce the same
food product. End product testing indicated that
the food product produced by both plants was
contaminated with a single species of microbial
hazard. Hazard incidence was higher for food
from Plant A (i.e. 25) but only food from Plant
B (i.e. 10) had caused an outbreak. These
observations caused the risk managers to pose the
following risk question Why did food from plant
B cause an outbreak when food from plant B has a
lower incidence of hazard than food from plant A?
QMRA Model. As a first step, the food company
determined the initial distribution of the hazard
in single food units at packaging. After hazard
identification, the food company developed a
Monte Carlo simulation model to predict consumer
exposure and response. Fig. 1 shows the input
settings and design for the hazard identification
and exposure assessment module, whereas Fig. 2
shows the input settings and design for the
hazard characterization and risk characterization
module for the Plant A and Plant B
scenarios. Incidence of hazard events was modeled
using a discrete distribution, whereas extent of
hazard events was modeled using a pert
distribution. The model was created in an Excel
spreadsheet and was simulated with _at_Risk using
settings of Latin Hypercube sampling, 105
iterations, 200 simulations and randomly selected
random number generator seeds.
Fig. 2
RESULTS AND DISCUSSION Hazard incidence was lower
in food from Plant B at packaging, during
distribution and after washing but was similar to
food from Plant A after cooking and at serving
(Fig. 3). Total hazard number per 100,000 food
units was lower for food from Plant B than Plant
A from packaging through cooking but because of
greater cross-contamination during serving it was
similar to Plant A at serving (Fig. 4). Because
of a higher incidence of a highly virulent strain
of the hazard and because of a higher incidence
of high risk consumers (Fig. 2), the dose that
elicited an adverse health response from 50 of
consumers was lower for food from Plant B than
food from Plant A (Fig. 5). This resulted in a
higher predicted response rate among consumers of
food from Plant B than from Plant A (Fig. 6).
The median response rate was 3 (range 0-11) per
100,000 food units for Plant A and 7.5
(range 1-14) per 100,000 food units for Plant B.
The primary risk factors were gt 50 cells per food
unit at packaging, hazard growth during
distribution, washing of food before cooking that
resulted in higher risk of cross-contamination
during serving, presence of a highly virulent
strain of the hazard and consumption of the food
by a high risk consumer (Table 1). Thus,
although food from Plant B had a lower incidence
of hazard contamination at packaging, the food
from Plant B posed a higher risk of an outbreak
because of a higher incidence of hazard growth
during distribution, a higher incidence of
cross-contamination during serving, a higher
incidence of the virulent strain of the hazard
and and a higher incidence of high risk
consumers. These results demonstrated to the
food company the extreme importance of
considering post-process risk factors when
assessing safety of food at the processing plant.

Hazard number per food unit Hazard number per food unit Hazard number per food unit Hazard number per food unit Hazard number per food unit
Plant Iteration Packaging Distribution Washing Cooking Serving Class Response Dose
A 9,904 1,451 300,130 36,260 0 5,945 7 3,065
A 64,376 202 4,539 27 0 23 8 22
A 69,075 316 5,714 3,214 0 20 8 7
A 93,039 100 1,191 173 0 28 8 15
B 53,459 52 4,045 509 0 81 6 67
B 65,115 62 5,188 343 0 111 6 111
B 69,075 316 5,714 3,214 0 20 8 7
B 71,656 668 92,965 13,705 0 948 6 347
B 87,865 153 3,446 611 0 110 6 106
B 93,039 100 1,191 173 0 28 8 15
Table 1
Fig. 6