Food Control 18 (2007) 789–799

www.elsevier.com/locate/foodcont

Risk assessment for listeriosis in consumers of Parma

and San Daniele hams
a,*
Armando Giovannini , Giacomo Migliorati a, Vincenza Prencipe a, Davide Calderone b,
Carlo Zuccolo c, Paolo Cozzolino d
a
Istituto Zooprofilattico Sperimentale dell’Abruzzo e del Molise ‘‘G. Caporale’’, Via Campo Boario, 64100 Teramo, Italy
b
Consorzio del Prosciutto di Parma, Parma, Italy
c
Consorzio del Prosciutto di San Daniele, San Daniele, Udine, Italy
d
Dipartimento di Sanità Pubblica, Azienda Unità Sanitaria Locale, Parma, Italy

Received 24 September 2005; received in revised form 6 March 2006; accepted 6 March 2006

Abstract

Following the detection of Listeria monocytogenes in samples of dry cured ham imported in the USA from Italy, the Italian Ministry
of Health planned a risk assessment for listeriosis in ham consumers. The risk assessment was performed according to international
guidelines set by FAO and WHO. Expected incidence of illness was 4.7 · 1010 cases per serving in the case of normal adult population,
6.1 · 107 in the case of organ transplanted patients, the most susceptible risk sub-population. Due to the low value of water activity
even in the deeper parts of the ham, bacterial growth had a very slight effect on the probability of illness.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Listeria monocytogenes; Dry cured ham; Risk assessment

1. Introduction • It can grow slowly in many foods even during refriger-

Outbreaks of illness associated with the consumption of
coleslaw, pasteurized milk, and fresh soft cheese in the
early 1980s in the USA led, to the recognition of Listeria
monocytogenes as a foodborne pathogen (CFSAN/FSIS,
2003). This pathogen can cause death and is particularly
difficult to control in ready-to-eat (RTE) foods (i.e., prod-
ucts that may be consumed without any further cooking or
reheating) because (CFSAN/FSIS, 2003):
• It is commonly found in the environment, including
food processing, distribution, and retail environments,
in foods, and in the home.
• It primarily affects a small segment of the population
with a heightened susceptibility.
*
Corresponding author. Tel.: +39 086 133 2281; fax: +39 086 133 2251.
E-mail address: a.giovannini@izs.it (A. Giovannini).
ated storage.
• Unlike most bacteria it is particularly resistant to the
conditions and treatments used to control foodborne
pathogens.
The risk of severe listeriosis is well documented in the
case of soft cheese (Altekruse, Timbo, Mowbray, Bean, &
Potter, 1998; Bemrah, Sanaa, Cassin, Griffiths, & Cerf,
1998; Dalton et al., 1997; Farber, Ross, & Harwig, 1996;
Gilot et al., 1997), and smoked fish (Brett, Short, &
McLaughlin, 1998; CFSAN/FSIS, 2003; Ericsson et al.,
1997; FAO, 1999; Gombas, Chen, Clavero, & Scott, 2003;
Miettinen et al., 1999). It is also evident that different path-
ogen/commodity combinations play different roles in the
epidemiology of human listeriosis: estimates of the expected
number of cases per million servings range from 4.5 · 109
cases per million servings in the case of hard cheese to 0.065
cases per million servings in the case of Frankfurters not
reheated before consumption (CFSAN/FSIS, 2003).

0956-7135/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodcont.2006.03.014
790 A. Giovannini et al. / Food Control 18 (2007) 789–799
Control strategies adopted in the various countries differ
greatly. Concerning RTE, for example, the USA and the
European Commission have adopted a ‘‘zero-tolerance’’
policy. Other countries, such as Canada have ‘‘zero-toler-
ance’’ or tolerance policies depending on the type of RTE
food. Nevertheless, even countries that adopted a general
‘‘zero-tolerance’’ policy have chosen, in practice, different
appropriate levels of protection (ALOPs) on the basis of
different sampling plans, ranging from the absence of L.
monocytogenes in 1 g to the absence of L. monocytogenes
in 10 samples of 25 g each. It is noteworthy that for the
same RTE foods, i.e. cooked RTE crustaceans and smoked
salmon, some countries maintain a ‘‘zero-tolerance’’ policy,
while Australia and New Zealand decided to permit a level
of up to 100 CFU in one of five samples of 25 g. The reason
for the difference is due to the small amounts of these RTE
foods consumed in the two countries, and which would
result in 1 case every 1600 years in the case of cooked
RTE crustaceans. Therefore, the most objective way to
decide the ALOP of a country seems to perform specific
risk assessments for each relevant RTE food, giving the
expected incidence of infection per million servings con-
sumed and to estimate the yearly expected number of cases
on the basis of the amount of that specific food consumed
in a country. The present research was initiated following
the detection by the Food Safety Inspection Service (FSIS)
in the USA in June 2002 and its subsequent notification in
July of the presence of L. monocytogenes in dry cured hams
imported from Italy. The FSIS authorities and representa-
tives of the Italian Ministry of Health met in Washington,
DC and defined the strategy to be adopted in Italy to
ensure the control of L. monocytogenes in exported ham.
This strategy included, among other activities, the sampling
programs, the standardization of laboratory methods to eval-
uate prevalence of L. monocytogenes in ham, and a risk
assessment for L. monocytogenes and Salmonella in cured
pork ham. The aim of this study is to perform a risk assess-
ment on dry cured hams produced by the Parma and the
San Daniele Consortia and to estimate the expected inci-
dence of listeriosis in the normal adult populations and
in some at risk sub-populations.
2. Materials and methods
2.1. Collection and analysis of samples
2.1.1. Samples
At each of the two Consortia, a separate sampling was
planned according to the criteria given in the UNI ISO
norms 2859:1993, part 1 and part 2. The sampling was
designed to guarantee the consumer that any prevalence
of L. monocytogenes contamination >5% would be detected
with a 90% probability, and to guarantee the producer that
any prevalence of L. monocytogenes contamination <2.5%
would be recognized as a low prevalence with no attendant
regulatory consequences (i.e. no systemic corrective action
would be necessary). A random sample of 306 cured hams
was to be selected at the end of the production chain. The
sample was stratified by producer and by type of product
(bone-in, de-boned and sliced hams) and proportionate
to the amount of product (weight) exported to the USA
in 2001. Due to (i) the low number of samples collected
of bone-in hams, (ii) the different criteria for specimens col-
lection in de-boned and sliced hams and (iii) the lower sen-
sitivity of the analysis of sliced hams – due to the lower
contamination level – only the results obtained from de-
boned hams were used in the exposure assessment.
2.1.2. Collection of specimens
In the case of bone-in and de-boned hams, from the por-
tion not covered by the skin of each sampled ham, one or
more surface slices (3–5 mm thick) were cut, to collect a
total sample of 100 g (Fig. 1). In the case of sliced hams,
a packet of vacuum packed sliced ham has been considered
as a single specimen.
2.1.3. Laboratory analysis of samples
Each specimen was divided into two equal aliquots of
50 g. Twenty-five grams of the first aliquot was tested qual-
itatively for L. monocytogenes (FSIS, 2002) with the
remaining 25 g tested quantitatively using a MPN tech-
nique (FSIS, 2002). Samples positive to the qualitative
technique and negative to the MPN technique have been
considered to be at a contamination level between the
two published detection thresholds for the two techniques,
i.e. between 0.04 and 0.3 organisms per gram (FSIS, 2002).
2.2. Exposure assessment
2.2.1. Estimation of the, frequency distribution of surface
contamination
According to the literature (Gill, McGinnis, Rahn, &
Houde, 1996; Kilsby & Pugh, 1981), the contaminated
samples are detectable contaminations arising from a con-
tinuous log normal distribution of contamination, and the
minimum detectable level from the presence/absence tests
is typically 1 organism in 25 g or 0.04 organisms per gram.
A low percentage of samples are contaminated at or above
this level with the remaining samples having non-detectable
levels (i.e., <0.04 organisms per gram) (Fig. 2).
To estimate the mean and the standard deviation of the
lognormal distribution, wide ranges of values for the mean
and for the standard deviation have been considered and
the best fitting pair of values has been chosen as that
minimizing the D statistics in the one-sample Kolmogo-
rov–Smirnov test for goodness-of-fit (Siegel & Castellan,
1988):
Z xi 
X observedi 

MaxðDlj ;rj Þ ¼ Max  Normalðxi ; lj ; rj Þ  
N 
1
where D is the Kolmogorov–Smirnov statistics; xi is each
measured contamination level; lj the hypothesized mean;
rj the hypothesized standard deviation; Observedi the num-
 A. Giovannini et al. / Food Control 18 (2007) 789–799 791

Fig. 1. Technique used to collect specimens and indication of the surface potentially contaminated (‘tested specimen’ and ‘serving’).

Fig. 2. Example of a log-normal frequency distribution of bacterial contamination. Adapted from CFSAN/FSIS, 2003.

ber of samples showing that specified contamination level,
and N the total number of samples.
The uncertainty of the mean and the standard deviation
has been estimated using the non-parametric bootstrap
(Vose, 2000), and employing the same estimation tech-
nique. To obtain convergence, 1000 iterations of the boot-
strap sampling were performed using @Risk (Palisade
Corporation, version 3.5.2) and the pairs of values for
the mean and the standard deviation recorded.
The mean and the standard deviation of the recorded
values and the Pearson’s correlation coefficient for the
pairs of parameters simulated by bootstrap were calculated
also, to be used in further simulations.
The goodness-of-fit of the estimated frequency distribu-
tion of the contamination levels was evaluated graphically
by comparing the cumulative distribution of observed data
against the cumulative theoretical distribution.
2.2.2. Conversion of surface contamination levels into
contamination per serving
Since contamination by L. monocytogenes is located on
the surface of the ham, the levels of contamination per
serving were calculated from the distribution of surface
contamination (Fig. 1). Based on the roughly elliptical
shape and thickness (3–5 mm) of the slices and the spe-
cific weight of ham (weighted average of fat and muscle
792 A. Giovannini et al. / Food Control 18 (2007) 789–799
components = 0.96 g/cm3), the total volume, and the sur-
face area of a typical slice, were calculated. Finally, a mul-
tiplying factor to be used to convert the level of surface
contamination measured in the samples down to the level
of contamination expected for an average serving of 28 g
(made up of 2 slices, 1 mm thick, cut orthogonally to the
ham’s surface) was calculated.
2.2.3. Simulation of the extraction of servings
The first step was to generate the theoretical distribution
of surface contamination, by extraction of the two param-
eters (mean and standard deviation) from the distribution
of uncertainty, and according to the estimated correlation
coefficient between the two parameters:
Mean ¼ Normalðll rl Þ
St: Dev ¼ Normalðlr rr Þ correlated to the Mean
according to r
The second step was the extraction of the log (MPN/g) of
surface contamination from the above extracted
distribution:
logðMPN=gÞ ¼ NormalðMean; St: Dev:Þ
The third step was the conversion of log (MPN/g) of sur-
face contamination into MPN/serving and rounding the re-
sult to the closest integer.
A total of 100,000 iterations were run in @Risk environ-
ment to obtain convergence of simulation results.
2.2.4. Estimation of the frequency distribution of the
contamination of servings
The simulated distribution of contamination of servings
was assumed to be a log-normal distribution like that of
surface contamination. The best fitting pair of l and r,
was chosen by the same Kolmogorov–Smirnov test for
goodness-of-fit (Siegel & Castellan, 1988) as described
above.
Uncertainty of the parameters of the theoretical distri-
bution was estimated by non-parametric bootstrap. One-
thousand iterations of the bootstrap sampling were
performed using @Risk and the pairs of values for mean
and standard deviation recorded.
2.3. Dose–response assessment
2.3.1. Dose response assessment immediately after
production
A first dose–response assessment was performed under
the assumption of the absence of bacterial growth.
It was based on an exponential curve for the following
reasons:
• The use of the Exponential model meets the recommen-
dations of the draft ‘‘WHO/FAO Guidelines on Hazard
Characterization for Pathogens in Food and Water for
the selection of dose–response models for infectious
microorganisms’’ (FAO/WHO, 2003).
• The Exponential model is a non-threshold model, which
implies that there is no ‘‘minimum infectious dose’’.
• A key attribute of the model is its log-linearity (log dose
vs. log probability of illness) at low doses; this implies
that at low doses, a single serving with a specified level
of contamination has the same public health impact as
10 servings with 10-fold fewer organisms.
Exponential curve:
P ¼ 1  eRN
where P is the probability of illness; N is the number of L.
monocytogenes consumed; R is the parameter that defines
the dose–response relation for the population being
considered.
Values for R were obtained from the ‘‘Joint FAO/WHO
Expert Consultation on Risk Assessment of Microbiologi-
cal Hazards in Foods – Risk characterization of Salmonella
spp. in eggs and broiler chickens and L. monocytogenes in
ready-to-eat foods’’ (WHO/FAO, 2001). Various risk pop-
ulations were considered: normal adult population, organ
transplant, AIDS, dialysis, bladder cancer, gynecological
cancer patients, elderly over 60 years of age, perinatal
exposure.
For each risk sub-population the relevant dose–response
curve was calculated.
Then, based on the distribution of exposure levels, the
expected incidence of listeriosis per serving consumed by
each risk sub-population was calculated. The expected inci-
dence of listeriosis per serving was calculated as the average
of the probability of infection, weighted on the entire sim-
ulated distribution of exposure and on the uncertainty (i.e.
the probability distribution) of the parameters of the log-
normal distribution of exposure, and generated by
bootstrap.
2.3.2. Simulation of Listeria growth during storage
Growth of L. monocytogenes was simulated using the
deterministic model developed by the USDA and
implemented in the USDA Pathogen Modeling Program
(PMPWin version 6.1.0), under the following set of
assumptions:
• Initial contamination equal to the maximum observed
contamination level in the sampled hams (5 organisms/
serving).
• Water activity aw = 0.94 (the maximum observed in
Parma ham).
• Duration of storage between 1 and 30 days.
• Storage temperature between 4 and 10 C.
• Storage at higher, less suitable temperatures of between
10 and 25 C.
Finally, the expected incidence of listeriosis per serving
consumed by each risk sub-population was recalculated,
based on the new exposure levels after bacterial growth
during storage.
 A. Giovannini et al. / Food Control 18 (2007) 789–799 793

3. Results Table 2
Ham samples positive and subdivided by consortium and product type
3.1. Collection and analysis of samples Product type Parma San Not #Positive
Daniele identified samples
for L.m.
A total of 637 samples were collected (Table 1). Only the
490 samples collected from de-boned ham were considered Bone-in ham (P) 0 0 0 0
De-boned ham (D) 15 5 0 20
in the following analysis because the methods for sample Sliced ham, 0 0 0 0
collections were different in the case of sliced hams; also, vacuum
the ability of this sampling procedure for detecting a sur- packed [V]
face contamination was lower. Total 15 5 0 20
Twenty positive samples were detected, 15 in Parma
hams and 5 in San Daniele hams (Table 2). The total fre-
quency of contamination was 4.08% (95% confidence inter- points were included within the 95% confidence limits of
val: 2.67–6.22%). the distribution, as determined by bootstrap.
Each MPN/g of surface contamination was equivalent
3.2. Exposure assessment to 3.10 MPN/serving (Table 3).
The results of the simulation of the extraction of serv-
The frequency distribution of observed contamination ings are shown in Fig. 4. The frequency distribution of
levels (MPN/g of surface contamination) was fitted by a the simulated levels of contamination of servings was fitted
log-normal distribution with the mean 4.8 (95% confi- by a lognormal distribution with the mean 4.6 (95% con-
dence limits: 6 to 4) and the standard deviation 1.9 fidence limits: 4.4 to 4.8), the standard deviation 2.7
(95% confidence limits: 1.5–2.4) (Fig. 3); the mean and (95% confidence limits: 2.6–2.8) and the Pearson correla-
the standard deviation were significantly correlated (Pear- tion coefficient between the mean and the standard devia-
son correlation coefficient r = 0.70). All observed data tion r = 0.996 (Fig. 5). All simulated data points were
included within the 95% confidence interval. The result of
the exposure assessment (Fig. 5) is the probability distribu-
tion of the doses ingested by consumers of the portion of
Table 1 the ham not covered by the skin; this is considered the
Ham samples collected and subdivided by consortium and product type
worst-case scenario.
Product type Parma San Not Total
Daniele identified
3.3. Dose–response assessment
Bone-in ham (P) 0 0 1 1
De-boned ham (D) 306 181 3 490
Sliced ham, vacuum packed [V] 16 130 0 146
The results of the dose–response assessment, compared
to the results of the exposure assessment (probability of
Total 322 311 4 637
exposure to the various doses of L. monocytogenes) are

Fig. 3. Observed cumulative distribution of the log (MPN/g) of Listeria surface contamination, the theoretical lognormal distribution, and its 99%
confidence limits.
794 A. Giovannini et al. / Food Control 18 (2007) 789–799

Table 3 shown in Figs. 6–8. The dose able to give a 1% probability

Frequency distribution of the levels of Listeria contamination per slice and of clinical listeriosis in the normal adult population is
per serving
between 1011 and 1013 CFU of L. monocytogenes and, given
Results of sampling MPN/slice MPN/serving Number of the distribution of contamination on hams, its probability
(MPN/g less than) samples
can be estimated between 3.5 · 108 and 5 · 1010. The
0.04 0.06 0.12 470 most likely levels of contamination have a negligible ill-
0.3 0.47 0.93 9
0.6 0.93 1.86 6
ness-inducing probability (Fig. 6), i.e. the probability of
0.9 1.40 2.79 2 clinical listeriosis after ingesting a serving with the maxi-
1.2 1.86 3.73 1 mum observed level of contamination is between
1.5 2.33 4.66 2 1 · 1014 and 1 · 1012. Based on the distribution of expo-

Fig. 4. Frequency distribution of the simulated levels of Listeria contamination in extracted servings.

Fig. 5. Cumulative probability distribution of the doses of Listeria ingested by consumers of the portion of the ham not covered by the skin (worst-case
scenario): simulated values, theoretical distribution, and 95% confidence limits.
 A. Giovannini et al. / Food Control 18 (2007) 789–799 795

sure levels, the expected incidence of listeriosis per serving
consumed by each risk sub-population, was calculated
(Fig. 9).
The dry cured ham has to be stored at refrigeration
temperatures unless it is vacuum packed or packed in a
modified atmosphere. Therefore, the dose–response curves
after growth during storage at temperatures ranging from
4 to 10 C and for periods ranging from 1 to 30 days are
shown in Figs. 10 and 11. A further set of dose–response
curves has been calculated for the normal population after
storage under higher temperature conditions (tempera-
tures ranging from 10 to 25 C, Fig. 12). Since the esti-
mated probability distribution of contamination is a
lognormal distribution, with a theoretical range from
1 to +1, it could overestimate the very high contami-
nation levels. For this reason, the growth of L. monocytog-
enes during storage was simulated with reference to the
maximum observed contamination level. In Fig. 10, results
for the normal adult population are shown: during the
first 6 days of storage at 10 C the expected increase of
the risk of clinical listeriosis is 1 logarithm (from the base-
line 1 · 1013 to 1 · 1012); subsequently, the expected
increase is around 1 logarithm every 2 days; in the
instance of storage at 4 C, around 16 days are required
to increase the risk of the first logarithm with the subse-
quent increase at 1 logarithm every week. In Fig. 11 the

Fig. 6. Listeria exponential dose–response curves for the normal adult population, compared to the results of exposure assessment.

Fig. 7. Listeria exponential dose–response curves for some at-risk sub-populations, compared to the results of exposure assessment.
796 A. Giovannini et al. / Food Control 18 (2007) 789–799

Fig. 8. Listeria exponential dose–response curves for some at-risk sub-populations, compared to the results of exposure assessment.

Fig. 9. Expected incidence of listeriosis per serving in various at-risk sub-populations.

results for the most susceptible population (organ trans-
plant patients) are shown: the increase of the risk of clin-
ical listeriosis follows the same trend observed for the
normal adult population, beginning from an initial level
around 109. In Fig. 12, results after the storage of ham
at higher temperatures are shown for the normal adult
population: during storage at 25 C the expected increase
of the risk of clinical listeriosis is about 3 logarithms per
day (from a baseline of 1 · 1013) during the first three
days of storage; subsequently the increase is slower until
the stationary phase of the bacterial growth is reached
on the 5th day. The time required for bacterial growth
to stabilize increases decreasing storage temperature; at
10 C stabilization is reached on the 30th day of storage.
Once the maximum bacterial concentration (log(CFU)/g
around 9.5) is reached, the probability of infection per
serving for the normal adult population is around 105
(Fig. 12).
4. Discussion
Sampling was conducted at holdings exporting hams to
the USA and was stratified proportionally to the amount
of product (weight) exported in 2001. Nevertheless, the
results represent the entire production of the two consortia
since their exports to the USA is only a small fraction of
their entire production, most of which is for the domestic
and European markets.
 A. Giovannini et al. / Food Control 18 (2007) 789–799 797

Fig. 10. Dose–response curve after Listeria growth during refrigerated storage for the normal adult population; initial contamination = 5 cells per serving;
aw = 0.94.

Fig. 11. Dose–response curve after Listeria growth during refrigerated storage for organ transplant patients; initial contamination = 5 cells per serving;
aw = 0.94.

Sampling was done at the end of the production process
just before dispatch. Sampling at this stage was considered
to be representative of the possible exposure risk to the
consumer, because contamination by L. monocytogenes is
restricted to that part of the surface of the muscle not cov-
ered by skin (aw = 0.92) and where it cannot reach deeper
tissues (aw = 0.94) until de-boning, when cross-contamina-
tion is possible. The de-boning and skinning of ham usually
takes place in the retail market or in the production plant
just before shipment.
798 A. Giovannini et al. / Food Control 18 (2007) 789–799
Fig. 12. Dose–response curve after Listeria growth during non-refrigerated storage for the normal adult population; initial contamination = 5 cells per
serving; aw = 0.94.
The exposure assessment that was performed is a worst-
case scenario because the only part of the ham originally
contaminated is the surface of the muscle not covered by
the skin because contamination of the other parts occurs
passively during skinning and de-boning. Other parts,
therefore, may have a contamination level that is less than
or, at maximum, equal to the contamination level of the
sampled portion of the ham. Bacterial growth may occur
only in cross-contaminated deep parts close to the bone
because the whole of the external surface has a water activ-
ity of less than 0.92. The simulation of bacterial growth
during storage (at retail or at the consumer’s home) was
performed considering a water activity of 0.94 for the
whole ham; this simulation is also a worst-case scenario.
The exponential dose–response curve was chosen as rec-
ommended by the Joint FAO/WHO Expert Consultation
on Risk Assessment of Microbiological Hazards in Foods
(WHO/FAO, 2001). This model, however, is based on
some stringent assumptions that only in some instances
can be met under real-world conditions, namely (i) the
probability of infection is assumed to have a constant value
for any given host and any given strain of the pathogen,
and (ii) the distribution of the organisms in the inoculum
is assumed to be random, and characterized by a Poisson
distribution (FAO/WHO, 2003). The first assumption does
not take account of the biological variability of both hosts
and pathogen strains. This shortcoming could be overcome
by the use of a Beta-Poisson model, that assumes a proba-
bility of initiating an infection to be different for every
organism in every host, and where the probability of infec-
tion is assumed to follow a beta distribution. The second
assumption excludes spatial clustering of cells in the inocu-
lum. Clustering can be represented by a negative binomial
distribution or any other contagious distribution. Cluster-
ing, however, has little effect on the shape of the dose–
response relationship. Not enough data on natural out-
breaks have been found in the available literature to
attempt to fit the data to a Beta-Poisson model as was
made in a previous risk assessment (Giovannini et al.,
2004).
It was not possible to use a stochastic approach for the
dose–response assessment because of the very little expected
number of cases according to the exponential model: even
with 100,000 iterations the model was just able to generate
a single case of listeriosis. The power of the hardware
employed did not allow us to run the millions of iterations
required to have an output probability distribution. There-
fore, it was necessary to use a deterministic model.
The probability of infection estimated after bacte-
rial growth at refrigeration temperature was lower
(3.9 · 108) than the probability of infection estimated
without bacterial growth (6.1 · 107). This is an artifact
due to the need to choose a value for the inoculum that ini-
tiates bacterial growth; the initial value chosen was 5 CFU/
serving (the maximum observed contamination level). On
the contrary, the theoretical range of the estimated proba-
bility distribution of contamination varies from 1 to
+1, and the simulations performed gave a maximum sim-
ulated exposure of 2,491,743 CFU/serving. The probability
of infection calculated without bacterial growth was the
average of the probability of infection, weighted on the
entire simulated distribution of exposure and on the uncer-
tainty (i.e. the probability distribution) of the parameters
of the lognormal distribution of exposure. On the contrary,
the probability of infection calculated after bacterial
growth refers to a single value of exposure. Therefore,
the risk calculated after bacterial growth, to be properly
interpreted, has to be considered in terms of its increase
 A. Giovannini et al. / Food Control 18 (2007) 789–799
with temperature and time of storage, rather than in abso-
lute terms of probability of infection.
The risk of listeriosis associated with the consumption of
dry cured ham has the following characteristics:
• A low probability of infection per serving (4.7 · 1010
cases per serving in the case of normal adult population,
6.1 · 107 in the case of organ transplanted patients –
the most susceptible risk sub-population, Fig. 9).
• A low annual number of expected cases: the total pro-
duction in 2002 of Parma and San Daniele hams was
around 41 million kg, equivalent to 1.4 billion servings.
This would lead, with a 95% probability, to a total
annual number of infections of less than 2 for the nor-
mal adult population. The expected number of cases in
the organ transplant patients would be 1 in every 1.6
million servings.
• A slight effect of the bacterial growth, at least when
properly stored and for the normal adult population,
due to the low levels of initial contamination and to
the low value of water activity even in the deeper parts
of the ham:
– In the case of the maximum observed contamination,
the expected probability of infection doubles after 8
days at 4 C and increases 51-fold after 8 days at
10 C; this means a probability of infection per serv-
ing of up to 3.9 · 108 in the case of organ transplant
patients when the initial contamination was equal to
the maximum observed level.
– It is worth remembering that due to dehydration
sliced ham is no longer palatable after 8 days in a
refrigerator, unless it has been vacuum packed.
Acknowledgements
The authors are indebted to Rudy Meiswinckel for his
revision of the final text and his suggestions to make the
text understandable to general readers without a specific
statistical background.
The research was financed by the Italian Ministry of
health, as part of the research project ‘‘Definizione di un
modello sperimentale di analisi del rischio, che preveda lo
sviluppo di uno strumento operative, applicabile in ambito.
decisionale, per il Servizio Sanitario Nazionale’’.
References
Altekruse, S. F., Timbo, B. B., Mowbray, J. C., Bean, N. H., & Potter, M.
E. (1998). Cheese-associated outbreaks of human illness in the United
States, 1973 to 1992 – sanitary manufacturing practices protect
consumers [Review]. Journal of Food Protection, 61, 405–1407.
Bemrah, N., Sanaa, M., Cassin, M. H., Griffiths, M. W., & Cerf, O.
(1998). Quantitative risk assessment of human listeriosis from
consumption of soft cheese made from raw milk. Preventive Veterinary
Medicine, 37, 129–145.
799
Brett, M. S. Y., Short, P., & McLaughlin, J. (1998). A small outbreak of
listeriosis associated with smoked mussels. International Journal of
Food Microbiology, 43, 223–229.
CFSAN/FSIS. (2003). Quantitative assessment of relative risk to public
health from foodborne Listeria monocytogenes among selected cate-
gories of ready-to-eat foods. Available from <http://www.cfsan.fda.
gov/~dms/lmr2-toc.html>. Consulted on May 25, 2006.
Dalton, C. B., Austin, C. C., Sobel, J., Hayes, P. S., Bibb, W. F., Graves,
L. M., et al. (1997). An outbreak of gastroenteritis and fever due to
Listeria monocytogenes in milk. The New England Journal of Medicine,
336, 100–105.
Ericsson, H., Eklow, A., Danielsson-Tham, M. L., Loncarevic, S.,
Mentzing, L. O., Persson, I., et al. (1997). An outbreak of listeriosis
suspected to have been caused by rainbow trout. Journal of Clinical
Microbiology, 3(11), 2904–2907.
FAO. (1999). FAO expert consultation on the trade impact of Listeria in
fish products. Amherst, MA, USA, 17–20 May 1999. FAO Fisheries
Report No. 604. Available from <http://www.fao.org/DOCREP/003/
X3018E/X3018E05.HTM#4.5%20Risk%20characterization>. Con-
sulted on May 25, 2006.
FAO/WHO. (2003). Hazard characterization for pathogens in food and
water. Guidelines. Microbiological Risk Assessment Series, No. 3.
Farber, J. M., Ross, W. H., & Harwig, J. (1996). Health risk assessment of
Listeria monocytogenes in Canada. International Journal of Food
Microbiology, 30, 145–156.
FSIS. (2002). Isolation and identification of Listeria monocytogenes
from Red Meat, Poultry, Egg and Environmental samples. Revision
3 of the 29th of April 2002. Now available as Revision 4 from <http://
www.fsis.usda.gov/Ophs/Microlab/Mlg_8_04.pdf>. Effective from
September 13, 2005. Consulted on May 25, 2006.
Gill, C. O., McGinnis, J. C., Rahn, K., & Houde, A. (1996). The hygienic
condition of manufacturing of beef destined for the manufacture of
hamburger patties. Food Microbiology, 13, 391–396.
Gilot, P., Hermans, C., Yde, M., Gigi, J., Janssen, M., Genicot, A., et al.
(1997). Sporadic case of listeriosis associated with the consumption of
a Listeria monocytogenes-contaminated camembert cheese. Journal of
Infection, 35, 195–197.
Giovannini, A., Prencipe, V., Conte, A., Marino, L., Petrini, A., Pomilio,
F., et al. (2004). Quantitative risk assessment of Salmonella spp.
infection for the consumer of pork products in an Italian region. Food
Control, 15, 139–144.
Gombas, D. E., Chen, Y., Clavero, R. S., & Scott, V. N. (2003). Survey of
Listeria monocytogenes in ready-to-eat foods. Journal of Food Protec-
tion, 66, 570–577.
Kilsby, D. C., & Pugh, M. E. (1981). The relevance of the distribution of
micro-organisms within batches of food to the control of microbio-
logical hazards from foods. Journal of Applied Bacteriology, 51,
345–354.
Miettinen, M. K., Siitonen, A., Heiskanen, P., Haajanen, H., Bjarkroth,
K. J., & Korkeala, H. J. (1999). Molecular epidemiology of an
outbreak of febrile gastroenteritis caused by Listeria monocytogenes in
cold smoked rainbow trout. Journal of Clinical Microbiology, 3(7),
2358–2360.
Siegel, S., & Castellan, N. J. Jr., (1988). Nonparametric statistics for the
behavioral sciences (2nd ed.). Statistics series. NY, USA: McGraw-Hill
International Editions.
UNI ISO 2859:1993 part 1 and part 2. Sampling procedures for inspection
by attributes – Sampling plans indexed by acceptable quality level
(AQL) for lot-by-lot inspection.
Vose, D. (2000). Risk analysis. A quantitative guide (2nd ed.). Chichester,
England, UK: Wiley, pp. 181–191.
WHO/FAO. (2001). Joint FAO/WHO expert consultation on risk
assessment of microbiological hazards in foods – risk characterization
of Salmonella spp. in eggs and broiler chickens and Listeria monocyt-
ogenes in ready-to-eat foods. FAO Headquarters, Rome, Italy, 30
April–4 May 2001.