Microbial Risk Assessment: Salmonella in Peanut butter

Do you know the case of salmonella outbreak from peanut butter in America?

Background:

In 2008/2009, 9 people died and at least 714 people (half of them children) fell ill, all from food poisoning after eating products containing salmonella contaminated peanuts. It triggered the most extensive food recall in US involving 46 states, more than 360 companies, and more than 3,900 different products manufactured using Peanut Corporation of America (PCA) ingredients. In September 2015, Stewart Parnell was sentenced to 28 years in prison for his role in the nationwide outbreak.( for details check Wikipedia)

Microbial Risk Assessment

Let’s carry out microbial risk assessment of salmonella from peanut butter. The assumption of this case study is taken from the paper of Chen 2013.

The salmonella is not very common in peanut. Based on the historical data, the initial prevalence of salmonella in peanut during manufacturing is estimated to be 5.5 x 10^-6 (i.e. 55 units out of 10⁷ units could be contaminated with salmonella).
The initial concentration of salmonella in peanut can vary uniformly from -1.52 to 2.55 log CFU/g. So we will need to consider this variation during probabilistic risk assessment. During manufacturing a single batch contains 6.85x 10⁰⁶ g of peanut which is converted into peanut butter. After grinding and mixing, the peanut butter is directly filled in the unit mass of 250 g. The salmonella concentration does not change during grinding and mixing. However, the evidence shows that there can be reduction in the salmonella load in peanut butter during storage by 0.49 to 3.47 log CFU.

Let’s assume that people eat 30 g of peanut in a single dose (occasion). Peanut butter is very common in America. The data shows that Americans eat peanut butter once in a week in an average. The total current population of America is 330,689,970. Hence, the total eating occasions by total American people is 1.7 x 10¹⁰.

The dose-response model for salmonella is well established to follow beta-poisson model with the alpha = 0.1324 and beta =51.45 (Food and Agriculture Organization of the United Nations and World Health Organization. 2002, Risk assessments of Salmonella in eggs and broiler chickens). 

The health effects of salmonella in terms of DALYs (disability-adjusted life years) per case is estimated to be 0.019 (U.S. Food and Drug Administration. 2012. Salmonella species, p. 12–14).

The probabilistic Risk Assessment was carried out using R-software using mc2d package (two dimensional monte carlo simulation package). The code is not displayed here, as it won’t be easy to understand for you. The summary output of the risk assessment is as follows:

summary(Risk)

Risk_illness_dose :

          mean       sd Min 2.5%      25%      50%      75%    97.5%      Max   nsv Na's

NoUnc 1.97e-07 4.06e-07   0    0 3.29e-10 1.51e-08 1.48e-07 1.56e-06 2.35e-06 1e+05    0

total_illness_year :

      mean   sd Min 2.5% 25% 50%  75% 97.5%   Max   nsv Na's

NoUnc 3355 6895   0    0 5.6 256 2514 26527 39918 1e+05    0

DALYs_occastion :

          mean       sd Min 2.5%      25%      50%      75%    97.5%      Max   nsv Na's

NoUnc 3.75e-09 7.71e-09   0    0 6.26e-12 2.86e-10 2.81e-09 2.96e-08 4.46e-08 1e+05    0

DALYs_annum :

      mean  sd Min 2.5%   25%  50%  75% 97.5% Max   nsv Na's

NoUnc 63.8 131   0    0 0.106 4.87 47.8   504 758 1e+05    0

> print(Risk,digits = 2)

                node    mode    nsv nsu nva variate min    mean  median     max Nas type outm

1  Risk_illness_dose numeric 100001   1   1       1   0 2.0e-07 1.5e-08 2.3e-06   0    V each

2 total_illness_year numeric 100001   1   1       1   0 3.4e+03 2.6e+02 4.0e+04   0    V each

3    DALYs_occastion numeric 100001   1   1       1   0 3.8e-09 2.9e-10 4.5e-08   0    V each

4        DALYs_annum numeric 100001   1   1       1   0 6.4e+01 4.9e+00 7.6e+02   0    V each

The mean risk of illness is the average probability of illness from one serving. It was calculated through Monte Carlo simulations from the mean of the risk-per-serving distribution among contaminated servings and the final prevalence of contamination in the food.

The final prevalence of salmonella in peanut butter was calculated to be approximately 4x10^-6 (approximately 4 in 1 million) and the mean risk of illness per serving was approximately 2 x10^-7 (or 2 cases per 10 million servings). The total likely illness in whole American population in an average was 3355 persons per year. The DALYs per dose of peanut butter consumption was 3.75 x 10^-9 years. Hence, the DALYs per year was calculated to be 63.8 years.

The risk of salmonella from peanut butter seems small. But, it was significant enough to cause a historic outbreak. Since there was no thermal deactivation step in the processing of peanut, the probability of salmonella infection through consumption of peanut butter remained significant.

Whenever we talk about peanut butter, most of us would not even think that salmonella could be a problem. But history does not lie. 

Government, manufacturers, consumers and all the stakeholders should take the issue of food safety very seriously.