Quantitative Microbial Risk Assessment for Consumption of Paneer contaminated with Escherichia coli O157:H7, Enteropathogenic Escherichia coli and Non-Typhoidal Salmonella spp in Mumbai, India

Hazard Identification

Thirty samples of paneer cheese were collected from retail establishments in Mumbai, India. Eight of these sources were packaged, and therefore considered to be produced in large-scale facilities (the “organized” sector). The remaining twenty-two unpackaged samples were considered to be produced in small-scale facilities (the “unorganized” sector). Additionally, 50% of the packaged and unpackaged samples were analyzed after collection, and subsequently stored for 30 days at -20°C prior and re-analyzed (stored samples), while the remaining 50% of the packaged and unpackaged samples were analyzed immediately after collection (fresh samples) but did not undergo further analysis or storage. All of the samples were analyzed for concentration levels of total (generic) Escherichia coli and E. coli O157:H7, and presence or absence of Salmonella spp. The prevalence of microbial contamination in the paneer samples is described in Table 1.

Table 1: Prevalence of microbial contamination in Paneer samples

The source of these organisms is fecal contamination, suspected to be of human origin. However, all of these pathogens are considered to also be zoonotic, according to Todar. The majority of the contamination is assumed to have occurred during the packaging process (Shrivastava, In preparation), due to the high temperatures that are involved in other processes including the production of paneer from milk. During the first heat-treating step, the milk is heated up to 90°C for 10-15 min, and during the coagulation process, it is again heated up to 60-90°C, according to Khan and Pal. Salmonella spp infection can result in a range of outcomes, particularly salmonellosis, which is characterized by fever, diarrhea, vomiting and abdominal cramps. S. typhi and S. paratyphi (A, B, C) can result in severe enteric fevers, according to Lacey. Bronze and Greenfield and Ryan and Ray state that the infectious dose for non-typhoidal Salmonella has been observed to be as low as 103 organisms.

“Food poisoning” from Enteropathogenic Escherichia coli (EPEC) often results in similar health outcomes as that of Salmonellosis, which can be acquired from ingesting 106 organisms, according to Todar. Infection with Escherichia coli O157:H7 can be acquired from a dose of 10-100 organisms (Todar, 2008) and is typically characterized by the same set of aforementioned symptoms and hemorrhagic colitis. However, O157:H7 infection may result in a more severe condition known as hemolytic uremic syndrome (HUS). This is seen in approximately 5-10% of all O157:H7 cases (Centers for Disease Control and Prevention). The attack rates and incubation times for the bacterial isolates can vary depending on population and organism (Table 2).

Table 2: Attack rates and incubation times for E. coli, E. coli O157:H7 and Salmonella spp.

The specific mortality of Salmonella spp has been estimated to be as high as 24% in developing countries (Chimalizeni et al, 2010). However, the overall burden of these specific diseases in these countries appears largely unknown due to the high incidence of acute diarrheal illnesses reported, with no specific cause. The mortality of diarrheal diseases in children under 5 years is estimated to be 10% of the total deaths in India, according to WHO. In 2012, the World Health Organization reported that almost 587,000 deaths could be attributed to diarrhea among the entire population (WHO).

Exposure Assessment

The goal of the exposure assessment was to characterize both the probability of exposure to pathogens and the amount of pathogens ingested among paneer consumers in India, in accordance with methods described by Hass et al.. The spoilage of paneer is mainly due to bacterial action. Rao et al. concluded in their study that fresh paneer prepared under strict conditions did not contain infectious organisms, suggesting that all pathogens were completely destroyed during heating of milk, although they may reappear in the paneer through different routes if proper sanitation and hygiene practices are not followed during chilling or packaging. Other possible sources of contamination include workers who handle paneer, and the conditions under which it is stored and transported. Amongst the three main routes of exposure (oral, inhalation, and dermal) the oral route via ingestion is the most relevant with regard to paneer consumption. In the current assessment, the exposure to pathogens among paneer consumers was considered to be a function of the microbial concentration of paneer and the ingestion rate of paneer. Based on unpublished data from Shrivastava et al., the mean consumption rate for paneer has been estimated as 10 g/person-day. Due to data constraints, it was not possible to consider all the possible influencers of microbial contamination, particularly those that occur during the production and preparation processes. Accordingly, it was decided to limit the exposure assessment to cover parameters that influence microbial contamination that occur from retail to consumption. Thus, the microbial concentration of paneer was considered to be a function of storage time, storage temperature, and whether the paneer was sourced from the organized or unorganized sector (i.e. whether the paneer is packaged or unpackaged, respectively). The exposure framework addressed by our exposure models is described in Figure 2 below.

Figure 2: Exposure Framework Flow Chart

Exposure Model Algorithm

\begin{align*} ln(\frac{N}{N_0}) = a exp[-exp(b-ct)] - a exp[-exp(b)] \end{align*}(1)

For the exposure model, the Gompertz Growth Equation (Equation 1) was used to determine the growth of the organisms of interest at specific temperatures over time. The input parameters for this equation for each organism were assumed to be similar to those found in queso fresco cheese, according to Kasrazadeh and Genigeorgi and Kasrazadeh and Genigeorgi. These parameters included the specific growth rate for the organisms, the lag time and generation time. The initial concentrations (N0) were determined by the distribution for the appropriate organisms. The ingestion rate for this risk assessment was modeled as a normal distribution with a mean of 10g/person/day based on the available data.

Dose Response

In order to calculate a risk probability based upon the dose of pathogen received from consuming paneer, mathematical models characterizing the relationship between dose and infection or disease incidence were selected for each pathogen. The QMRA Wiki website (available at https://qmrawiki.org/index.php?title=Quantitative_Microbial_Risk_Assessment_(QMRA)_Wiki) was utilized to evaluate the available models for each pathogen. Models and associated model parameters for each pathogen were selected based upon common route of exposure, common species (human or related animal species preferred), number of doses tested, and similarity between the pathogen tested in the dose-response studies and the pathogens evaluated in this assessment. A descriptive summary is provided in Table 3.

Table 3: Descriptive Summary of Dose Response Models

a From QMRAWiki study # 38, 39, 40, 42, 99, 144

b From QMRAWiki study # 213

c From QMRAWiki study # 232, 233

The model chosen for the Escherichia coli O157:H7 data is an exponential model developed from a pig feeding trial in which 3 dose levels of the pathogen were evaluated, according to Cornick and Helgerson. There were no relevant dose-response studies of E. coli O157:H7 conducted in humans; thus, this model was chosen as pigs are more similar to humans with respect to microbial dose-response compared with the other available animal models. Although the study used to develop this model was not relevant to cheese or dairy products, because the animals were exposed orally, the study is consistent with the primary route of exposure for paneer consumption. Fecal shedding of the pathogen was used as a proxy for infection, which may underestimate the actual incidence of infection as some pigs may be intermittent shedders.

The model chosen for the Escherichia coli generic data is a beta-Poisson model developed from several human studies, according to DuPont et al., Haas et al., Graham et al., and June et al., in which the dose-response data were pooled for a total of 15 doses evaluated. These studies measured the transmission of E. coli among people drinking milk, which is ideal for simulating the consumption of soft Paneer cheese. The outcome that was measured in the pooled data was occurrence of mild to severe diarrhea in humans, which results in an underestimation of the infection rate, as disease outcomes do not account for individuals who are infected but asymptomatic. The E. coli strains tested included Enteroinvasive E. coli (EIEC) and Enterotoxigenic E. coli (ETEC), which excludes both non-pathogenic E. coli as well as the most pathogenic E. coli strains, including the Shiga Toxin-producing E. coli (STEC) strains.

The model chosen for the Salmonella data is a beta-Poisson model developed from a dose-response study in humans, according to McCullough and Elsele. This study is a human feeding trial of Salmonella enterica subspecies enterica serotype Anatum in eggnog, which is a similar dairy product to soft Paneer cheese. Although the Salmonella concentration data in Paneer was not serotyped and is likely made up of multiple different Salmonella serotypes, Salmonella Anatum is a commonly reported bovine Salmonella serotype that can be associated with food products derived from cattle, according to CDC. This study measured 16 different doses of Salmonella and evaluated positive fecal cultures as a proxy for infection. As stated previously, this method may underestimate the true infection rate as some participants may be infected but not actively shedding the pathogen.

Risk Model

Probability functions for infection risk were calculated by a Monte Carlo simulation using Crystal Ball, a Microsoft Excel-based software package. To create a sample distribution of the E. coli concentration data for unpackaged Paneer (n = 22) for use in the Monte Carlo simulations, the mean and the standard deviation for the sample distribution were estimated by maximizing the sum of the likelihood values associated with each data point using the solver function in Excel. The sample distribution for the E. coli concentration data for packaged Paneer (n = 8) was derived by fitting the data to a range of stochastic probability models using the software package R, and selecting the distributions with the lowest AIC (Akaike’s Information Criterion) and BIC (Bayesian Information Criterion). Lastly, the sample distribution for the Salmonella dichotomous (presence/absence) data for unpackaged Paneer was derived by calculating the mean and standard deviation of the Most Probable Number (MPN) using methods described by the FDA (2010). The sample distribution for Salmonella concentrations in unpackaged Paneer was assumed to be lognormal, with a mean and standard deviation defined as the mean and standard deviation of the MPN, respectively. Since no Salmonella was detected in any of the packaged Paneer samples, we were unable to create a sample distribution for Salmonella concentrations in packaged Paneer.

Methods of Risk Management Evaluation

Possible risk management strategies aimed at reducing public health risks associated with STEC/Salmonella and paneer were evaluated in terms of number of illnesses per year in India. Applications of these strategies were modeled by varying input values or distributions in the developed QMRA model. Resulting risks were compared with the baseline model estimates to evaluate the effectiveness of proposed risk management strategies.

Reducing shelf life

Salmonella and E. coli O157:H7 have the ability to survive for an extended period of time under refrigerated temperatures, and can exhibit growth under extreme temperature, according to Wahi et al. and Shrivastava et al.. Thus, paneer is recommended to be consumed within a given shelf life. The maximum storage time before consumption of paneer was truncated to 4 days (96 hours) in the risk management scenario to calculate the relative risk reduction compared to the baseline model.

Cost-benefit analysis

Analysis was performed to compare the cost-benefit of implementation of alternative MCs compared to the current MC. By implementation of alternative MCs, a certain number of paneer batches will be rejected due to the application of more strict sampling plans. The percentage of paneer rejection resulting from more strict sampling plans was estimated using the ICMSF sampling tool. The additional costs of implementing the alternative MCs were calculated by multiplying the percentage of rejection, annual consumption of paneer, and the cost of paneer. The cost-benefit of implementation of alternative MCs was evaluated by comparing the cost increases due to rejection and cost decreases due to illness reduction.

Risk Characterization

Exposure Model

The results of the exposure model are presented in Figures 3A and 3B below. The respective predicted E. coli and Salmonella growth in soft cheese and Spanish cheese was heavily influenced by storage temperature and storage time. At 12°C, both pathogens were predicted to persist but exhibit no additional growth over a 250 hour time period, while at 14°C, a 3-log growth was predicted for both pathogens after approximately 125 hours of storage. At 24°C and 28°C, a 3-log growth was predicted for both pathogens at time 0.

Figures 3. Predicted Microbial Growth in Soft (A) and Spanish (B) Cheeses

Risk Characterization

Results from the model indicate that risk of infection with the pathogens of interest is highly dependent on source of paneer (packaged versus loose), time between purchase and consumption, and temperature of storage. In loose paneer kept at 4 Celsius the median risks of infection for EPEC, E. coli O157, and Salmonella at 24 hours are 55, 2,638, and 102 in 100,000, respectively. For packaged paneer kept at 4°C the median risks of infection for EPEC, and E. coli O157 at 24 hours are 1 and 459 in 100,000 respectively. At 4°C the median risk of infection for each pathogen remains constant over time because the growth model parameters do not predict any increase in pathogen concentration over time. When both loose and packaged paneer are stored at 14°C, median risks of infection remain relatively constant at the values previously stated for 4°C (due to initial concentrations of pathogens in the paneer) until approximately 96 hours when concentrations begin increasing due to growth. At the maximum expected storage of 6 days at 14°C, median risks of infection for EPEC, E. coli O157, and Salmonella in loose paneer are 3,686, 8,478, and 50,747 in 100,000, respectively. For package paneer at 6 days of storage at 14°C median risks of infection for EPEC, and E. coli O157 are 53 and 27,825 in 100,000 respectively. When stored at 24°C, risks of infection from both packaged and loose paneer increase rapidly due to microbial growth. After 24 hours of storage of loose paneer at 24°C median risks of infection for EPEC, E. coli O157, and Salmonella are 33,909, 100,000, and 51,362 in 100,000 respectively. After 24 hours of storage of packaged paneer at 24°C median risks of infection for EPEC, and E. coli O157 are 2,078, and 100,000 in 100,000 respectively. The relationship between paneer source, storage temperature, and time is summarized for each pathogen in the figure 4.

Since the pathogen occurrence data set indicated that pathogens were present in greater numbers in loose paneer, the risk model was also used to characterize the relative risk of consuming loose paneer versus packaged paneer. Relative risks could only be compared for EPEC and E. coli O157 since Salmonella occurrence data indicated that Salmonella was not present in packaged paneer. Relative risks were compared for consumption 24 hours after purchase at varying temperatures. In every instance except one, consumption of loose paneer had a higher risk of infection (3.82 to 55 times greater) than consumption of packaged paneer. The only exception was for E. coli O157 when stored at 24 Celsius where microbial growth in both loose and packaged paneer made the risk of infection following consumption of paneer from both sources equal at 24 hours. Relative risks and confidence intervals for EPEC and E. coli O157 at 24 hours at varying temperatures of storage are listed in table 4.

In order to determine if the difference in risk between packaged and unpackaged data are statistically significantly different, a two tailed t-test was performed on the median risk values for each temperature. The null hypothesis for these tests was that the mean of the packaged data and the mean of the unpackaged data are the same. The results from the t-tests (Table 5) showed that the difference in risks varied with the temperature and in some cases the organisms.

Table 4: Relative risk of infection for packaged and unpackaged paneer

Figure 4: Relationship between infection risk, storage time, and temperature for E. coli and Salmonella in paneer.

Table 5: Two-way T-test to assess the difference in risk between unorganized sector paneer and organized sector paneer

Uncertainty and Sensitivity Analysis

For scenarios characterized by an absence of pathogen growth (storage at 4°C and 14°C), uncertainty in the estimate could be attributed to variability in the dose-response parameters of the model and the initial pathogen concentrations in both sources of paneer. When pathogens began growing (storage at 24°C and later times when stored at 14°C), the growth model parameters multiplied the variability associated with the initial pathogen concentrations, which produce wider confidence intervals. The median predicted risk of infection and confidence intervals for EPEC in both packaged and loose paneer are shown in figure 5. The uncertainty of the risk of infection associated with EPEC is representative of the uncertainty for E. coli O157, and Salmonella.

Sensitivity of the model to different parameters was assessed at 24 hours for each storage temperature considered (4°C, 14°C, and 24°C). For E. coli O157, risks of infection in all scenarios were most sensitive to initial concentration, ingestion, and the exponential dose-response parameter k. For Salmonella and EPEC, which are characterized by beta-poisson dose-response models, at temperatures where microbial growth was limited or non-existent, predicted risks were most sensitive to the beta-poisson dose-response parameters, alpha, and median infectious dose, followed by ingestion. At high temperatures where microbial growth occurred, salmonella and EPEC predicted risks were most sensitive to initial concentrations followed by median infectious dose, alpha, and ingestion. This indicates that at higher temperatures microbial growth becomes an important factor in risks associated with consumption of both packaged and loose paneer.

Figure 5: Confidence intervals for infection risk associated with packaged and loose paneer

Risk Management & Communication

Results and discussion from scenario analysis of risk management

As shown in Table 6, by implementation of MCs, the mean and median predicted annual number of illnesses related to non-O157 E. coli and packaged paneer were considerably reduced. Comparing to the baseline model prediction, risk (in term of annual number of illness cases) was reduced by more than 99%. No significance differences (data not shown) were observed between current MC and two alternative MCs. On the other hand, by reducing the shelf life to up to four days, risk was reduced by 11.9% comparing to the baseline model (shelf life of up to six days). Similar trends were observed for risk associated with E. coli O157:H7 or Salmonella and loose paneer. The results indicated that implementation of current and alternative MCs are effective in reducing public health risks associated with paneer contaminated with pathogenic E. coli or Salmonella.

Table 6. Comparison of predicted annual illnesses related to non-O157 E. coli and packaged paneer

The cost-benefit analysis predicted that implementation of more strict sampling plans will increase the annual cost by approximately 18.8 million USD and 56.4 million USD for MC1 and MC2 respectively after taking into account of decreased cost of illness and increased cost of implementation (Table 3). As a conclusion, the implementation of more strict MCs is not cost effective compared with the current MC.

Risk Communication Strategy

Our risk communication strategy is described in Table 7. The strategy is intended to engage all stakeholders, including Indian regulatory authorities, paneer producers, and paneer consumers in the shared goal of minimizing the risk of microbial infection and associated disease.

Key findings from this QMRA including estimated public health risk of paneer consumption in Mumbai, relative risks of different combinations of paneer and pathogens (packaged paneer contaminated with pathogenic E. coli, loose paneer contaminated with pathogenic E. coli, packaged paneer contaminated with Salmonella, and loose paneer contaminated with Salmonella), and effectiveness and cost benefit of implementation of current and alternative MCs will be communicated to the regulatory authority (Food Safety Standards Authority of India, FSSAI) in the form of QMRA reports. Educational information and recommendations related to public health risk of paneer will be conveyed to consumers through public announcement, social media, and brochure. Websites and Mobile apps will be developed to provide a platform for consumers to interact with regulatory authorities such as report of illness and FAQs. Key information and recommendations to consumers include explanation of potential risks associated with paneer, recommendation of consume before “best-before” date, adequate refrigeration, and heat treatment (e.g., oil-saute, fry, blanching or adding the paneer to adequately hot curries). Information and recommendations will be conveyed to producers and manufacturers through public announcement, social media, TV, and newspapers. For organized producers, key information is to ensure HACCP is strict followed during raw material receiving, processing, packaging, transportation, and storage of paneer to avoid potential loss due to rejection of product. For small-unorganized producers, basic training or workshop in HACCP will be provided to improve food safety management.

Table 7: Risk Communication Strategy for Paneer Consumption