Several case-control studies have investigated risk factors for human salmonellosis while others have used Salmonella subtyping to attribute human infections to different food and animal reservoirs. This study combined case-control and source attribution data into a single analysis to explore risk factors at the point of exposure for human salmonellosis originating from four putative food-producing animal reservoirs (pigs, cattle, broilers and layers/eggs) in the Netherlands. We confirmed that most human cases (∼90%) were attributable to layers/eggs and pigs. Layers/eggs and broilers were the most likely reservoirs of salmonellosis in adults, in urban areas, and in spring/summer, whereas pigs and cattle were the most likely reservoirs of salmonellosis in children, in rural areas, and in autumn/winter. Several reservoir-specific risk factors were identified. Not using a chopping board for raw meat only and consuming raw/undercooked meat were risk factors for infection with salmonellas originating from pigs, cattle and broilers. Consuming raw/undercooked eggs and by-products were risk factors for layer/egg-associated salmonellosis. Using antibiotics was a risk factor for pig- and cattle-associated salmonellosis and using proton-pump inhibitors for salmonellosis attributable to any reservoir. Pig- and cattle-associated infections were also linked to direct contact with animals and environmental exposure (e.g. playing in sandboxes). Eating fish, meat in pastry, and several non-meat foods (fruit, vegetables and pasteurized dairy products) were protective factors. Consuming pork and occupational exposure to animals and/or raw meats were protective against layer/egg-associated salmonellosis. We concluded that individuals acquiring salmonellosis from different reservoirs have different associated risk factors, suggesting that salmonellas may infect humans through various transmission pathways depending on their original reservoirs. The outcome of classical case-control studies can be enhanced by incorporating source attribution data and vice versa.
Citation: Mughini-Gras L, Enserink R, Friesema I, Heck M, van Duynhoven Y, van Pelt W (2014) Risk Factors for Human Salmonellosis Originating from Pigs, Cattle, Broiler Chickens and Egg Laying Hens: A Combined Case-Control and Source Attribution Analysis. PLoS ONE 9(2): e87933. https://doi.org/10.1371/journal.pone.0087933
Editor: Yung-Fu Chang, Cornell University, United States of America
Received: November 24, 2013; Accepted: January 1, 2014; Published: February 4, 2014
Copyright: © 2014 Mughini-Gras et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors have no support or funding to report.
Competing interests: The authors have declared that no competing interests exist.
Salmonellosis is a major cause of human bacterial gastroenteritis and the second most reported zoonosis in the European Union (EU), after campylobacteriosis . It has been estimated that more than 34,500 symptomatic infections with Salmonella spp. occur annually in the Netherlands (∼16.5 million population) , with Enteritidis and Typhimurium being the two most frequently isolated Salmonella enterica subsp. enterica serotypes, accounting for 29% (Enteritidis) and 47% (Typhimurium, including its monophasic variant 4,,12:i:-) of human cases in 2011 .
The overall number of culture-confirmed human Salmonella infections, particularly those caused by serotype Enteritidis, has decreased markedly over the last decades . Even so, salmonellosis continues to contribute heavily to the disease burden of foodborne pathogens, both in terms of Disability Adjusted Life Years (DALYs)  and cost of illness , . A particularly worrisome trend concerns the rapid emergence of S. 4,,12:i:-, which increased from 0.1% in 2004 to 20% in 2011 of all human Salmonella infections . Human infections with the classical strains of S. Typhimurium have remained relatively stable over recent years , albeit with major outbreaks at both local  and national ,  level. In addition, sudden explosions of serotypes of traditionally minor importance have been documented. An instance is the 2012 massive outbreak of S. Thompson linked to a rather unusual vehicle: smoked salmon . This shows that the epidemiology of salmonellas is complex and dynamic; thus, a multi-tiered approach to control is needed, taking into account the different reservoirs, pathways and risk factors involved.
Consumption of raw/undercooked eggs has consistently been identified as the primary risk factor for human S. Enteritidis infection , . This is in line with the relatively high prevalence of S. Enteritidis in layers , . In contrast, transmission pathways for S. Typhimurium are less clear. A variety of risk factors has been identified, including consumption of beef , , pork , dairy products made with raw milk , exposure to animals  or to raw meats , and playing in sandboxes . Yet, S. Typhimurium is mainly isolated in pigs and cattle , .
Salmonella source attribution is being performed in several countries to ascertain the main food-producing animal reservoirs towards which control efforts should be directed and to assess the impact of such interventions –. Classical case-control studies can only trace back the source of human infections up to the exposure (e.g. food consumption, contact with animals, etc.), which, however, may not point to the original reservoirs because of, for instance, cross-contamination. Combining source attribution and case-control data would allow us to reconstruct the underlying transmission pathway, from a given reservoir up to the point of exposure, providing more refined results than when performing separate analyses –.
The objectives of this study were: 1) to attribute human salmonellosis cases to four putative food-producing animal reservoirs (pigs, cattle, broilers and layers/eggs); and 2) to combine the results of the attribution analysis with the available case-control data  to explore risk factors at the point of exposure for human salmonellosis caused by Salmonella subtypes attributable to pigs, cattle, broilers and layers/eggs.
Materials and Methods
Data of the so-called “CaSa” study, a case-control study on risk factors for human salmonellosis conducted in the Netherlands between April 2002 and April 2003 , formed the basis of the present study. In the CaSa study, a total of 1194 culture-confirmed cases of human salmonellosis (533 and 437 of which caused by S. Enteritidis and S. Typhimurium, respectively) were identified by the Dutch Regional Public Health Laboratories (RPHLs) through passive surveillance. Controls (n = 3165) were selected by frequency matching by age group (0–4, 5–17, 18–29, 30–44, 45–59, ≥60 years), sex, quarter of the year, and degree of urbanization (urban: >2500 addresses/km2; urbanized: 500–2500 addresses/km2 ; rural: <500 addresses/km2). For both cases and controls, information regarding food consumption, eating habits, kitchen hygiene, contact with animals, occupation, recreational activities, medication use, history of chronic diseases, and contact with people with gastroenteritis, was collected using a standardized questionnaire. Questions covered the seven days prior to symptom onset (cases) and questionnaire completion (controls). Missing values were handled using multiple imputation as described previously .
Case isolates were sent to the Dutch National Institute for Public Health and the Environment (RIVM), which serves as the EU and national reference laboratory for Salmonella, for serotyping and further phage typing of S. Enteritidis and S. Typhimurium isolates as described elsewhere . Of the identified cases, 414 (168 and 197 of which caused by S. Enteritidis and S. Typhimurium, respectively) were enrolled in the study. Exclusion criteria were: 1) having travelled abroad with at least one overnight stay; 2) living outside the Netherlands; or 3) not returning the abovementioned questionnaire.
Salmonella isolates from four food-producing animal reservoirs, i.e. pigs (n = 1595 isolates), cattle (n = 734), broilers (n = 2930) and layers/eggs (n = 822) (available as Supporting Information, Table S1) were collected between 2001 and 2004 as part of diagnostic activities and a diversity of surveillance programmes on farms, slaughterhouses and at retail. These isolates were sent to the RIVM by the Regional Veterinary Services for serotyping and further phage typing of S. Enteritidis and S. Typhimurium isolates, as was done for the human isolates.
Source attribution analysis
For the purposes of source attribution, all serotyped/phage typed Salmonella isolates from human cases that had occurred in the Netherlands between January 2002 and December 2003 (n = 3735 isolates), were used. Of these cases, 271 (7.3%) and 350 (9.4%) were discarded from the analysis because they were travel- or outbreak-related, respectively. Another 170 cases (4.6%) were discarded because their serotypes/phage types were only found in humans and not in any of the considered reservoirs. The modified Dutch model for source attribution  was then used to attribute the remaining 2944 sporadic and domestic cases to pigs, cattle, broilers and layers/eggs. These cases were infected with 164 different Salmonella subtypes that were found in at least one of the considered reservoirs. These subtypes included 94 serotypes along with 51 S. Typhimurium subtypes and 19 S. Enteritidis subtypes (Table S1).
The modified Dutch model has been presented in detail previously . Briefly, the expected number of human infections caused by subtype i from reservoir j, denoted as λij, is given by
where pij is the prevalence of subtype i from reservoir j, ei is the frequency of human salmonellosis cases of subtype i, and mj is the per capita annual food consumption (kg/person per year) for reservoir j. A sum over subtypes gives the total number of cases attributable to each reservoir. Details of model parameters are reported in Table 1.
It has been pointed out that the amount of food consumed does not in itself account for the actual degree of exposure to a given source, as some foods are more likely to be eaten raw/undercooked than others . An additional parameter (cj), expressing the probability for foods from reservoir j to be eaten raw/undercooked by the population, was therefore included in the model. Parameter cj was modelled as Beta(αj+1, β j+1), where αj corresponded to the number of controls in the CaSa study reporting to have consumed raw/undercooked foods of pig (n = 278), cattle (n = 780), broiler (n = 244) and layer/egg (n = 1244) origin, and βj is the total number of controls enrolled (n = 3165) minus αj. The final form of the modified Dutch model used here was
The model was implemented in @RISK (Palisade Corp., USA) by setting 100,000 iterations with the Latin hypercube sampling technique and a seed of 1. For every subtype, the model estimated a relative probability (Pr) to originate from each of the four reservoirs, as given by λij/∑jλij. Differences in Pr values for pigs, cattle, broilers and layers/eggs were tested between age groups, urbanization degrees, gender and seasons (spring/summer, i.e. April–September 2002 vs. autumn/winter, i.e. October 2002–March 2003) of the 414 salmonellosis cases enrolled in the CaSa study (Table S2) using Kruskal-Wallis or Mann-Whitney U tests, as appropriate (α-level: 0.05).
Risk factor analysis
One-hundred and thirty-one putative risk factors were tested for association with the occurrence of salmonellosis of probable pig, cattle, broiler and layer/egg origin using Pr-weighted logistic regression models. A separate model for each reservoir was constructed. An equal probability weight of one was assigned to the controls, while the salmonellosis cases were weighted according to the reservoir Pr values for the subtype in question. Weights are often used in regression analysis to account for unequal probabilities of selection of sample units, as well as non-coverage, non-response and other unbalances of the sampled population that might lead to departures of the sample from the population . Thus, by weighting the cases in our regression models according to their Pr values, the (sample) distribution of cases in the case-control study was adjusted to make it reflect the (population) distribution of cases that were more likely to originate from each of the reservoirs, as revealed by the source attribution analysis. As a result, factors associated with the occurrence of human salmonellosis in the Pr-weighted models could be interpreted as reservoir-specific risk factors.
Regression models were built in stepwise fashion. Variables showing a p-value ≤0.10 for the association with the outcome in the single-variable logistic regression analysis were selected for inclusion in a multivariable logistic regression model for each reservoir. Next, variables were dropped one by one if they showed a p-value ≥0.05 and their exclusion from the models did not influence the association of the other covariates. The population attributable risk (PAR) of each significant risk factor from the final model was then calculated based on adjusted odds ratios (ORs) and the prevalence of exposure in cases. Similarly, 95% confidence intervals (95%CI) of PARs were derived from the CIs of the adjusted ORs , . All final regression models showed an overall statistical significance (likelihood ratio χ2 test, p<0.05) and goodness-of-fit (Hosmer-Lemeshow test, p>0.05). The level of education (categorized like in , ) was always included in the regression models to control for confounding in addition to the frequency-matched variables. Similar to previous analyses , –, all possible two-way interactions between each of the significant risk factors of the reduced multivariable model and the age group, gender, season, and urbanization degree were tested for significance in a separate logistic regression model. This model included an interaction term at a time along with the corresponding original variables, while holding all other frequency-matched variables and the level of education constant. However, as no significant interaction was found, the reduced multivariable model was not expanded to include any of the tested interaction terms. In addition to the Pr-weighted models for each reservoir, an overall unweighted model was developed to identify risk factors for human salmonellosis as a whole. Finally, the multivariate shared relationships between the significant risk factors and the reservoir Prs were explored altogether using canonical correlation analysis (CCA). Statistical analyses were performed using STATA 13 (StataCorp, USA).
Attribution of human salmonellosis
The modified Dutch model estimated that the majority of the 2944 sporadic, domestic cases of salmonellosis that occurred in the Netherlands during 2002–2003, originated from layers/eggs (50.1%, 95%CI: 47.8–51.7%), followed by pigs (39.9%, 95%CI: 38.3–41.8%), cattle (6.2%, 95%CI: 5.5–7.7%) and broilers (3.8%, 95%CI: 2.9–4.6%). Cases infected with the two most prevalent serotypes, Enteritidis (n = 1361) and Typhimurium (n = 1009), were predominantly attributed to layers/eggs (89.5%, 95%CI: 85.1–92.9%) and pigs (81.8%, 95%CI: 78.5–84.3%), respectively. The remaining 574 cases infected with other serotypes were mainly attributed to pigs (49.24.8%, 95%CI: 46.45–51.99%) and layers/eggs (33.7%, 95%CI: 30.7–36.7%).
Median percent Pr values by age group, gender, season and urbanization degree are reported in Table 2. Significant age effects were found in Pr values for pigs (p = 0.003), cattle (p = 0.003), and layers/eggs (p = 0.001). The probability for pigs and cattle to act as reservoirs of salmonellosis was highest among children and adolescents, whereas that for layers/eggs was highest among adults. The probability to acquire infection from cattle was significantly higher for males than for females (p = 0.006). Significant seasonal effects were found in Prs for pigs, cattle and layers/eggs. In autumn/winter vs. spring/summer, people had a significantly higher probability to acquire Salmonella infection from pigs (p<0.001) and cattle (p<0.001), whereas the inverse was found for layers/eggs (p<0.001). Significant effects of urbanization were found for all reservoirs. The probability for pigs and cattle to be reservoirs of salmonellosis was highest in rural areas (pigs, p<0.001; cattle, p = 0.001), whereas that for broilers and layers/eggs was highest in urban areas (broilers, p = 0.034; layers/eggs, p<0.001) (Table 2).
Reservoir-specific risk factors
Significant food-related risk factors for human salmonellosis as a whole were: not cleaning chopping board when using it for raw meat and other foods (PAR 11.6%), consuming raw/undercooked meat (7.1%), and consuming raw/undercooked eggs (3.6%) (Tables 3 and 4). Looking at the Pr-weighted models, not cleaning the chopping board remained a significant risk factor for salmonellosis attributable to pigs (PAR 14.4%) and cattle (17.2%), while changing kitchen rags less than once a week became a significant risk factor for broiler-associated salmonellosis (12.4%). Consuming raw/undercooked meat remained a risk factor for salmonellosis attributable to all three meat-producing animal reservoirs, i.e. pig (8.6%), cattle (9.4%) and broilers (8.7%), whilst consuming raw/undercooked eggs or products containing raw/undercooked eggs were risk factors for layer/egg-associated salmonellosis only (3.6% and 5.8%, respectively).
Among the significant non-food-related risk factors for salmonellosis as a whole, using proton-pump inhibitors (PPIs) showed the largest PAR (7.9%), followed by contacting people with gastroenteritis outside the household (6.6%), using antibiotics (3.7%), and using H2-receptor antagonists (1.7%) (Tables 3 and 4). In the Pr-weighted models, using PPIs remained a significant risk factor for salmonellosis attributable to any of the four reservoirs, with PARs ranging from 7.8% (layers/eggs) to 8.7% (broilers). Using H2-receptor antagonists remained significant for salmonellosis attributable to layers/eggs (1.9%) and broilers (2.1%), and using antibiotics remained significant for salmonellosis of pig (4.7%) and cattle (5.0%) origin. Contacting people with gastroenteritis outside the household remained a significant risk factor for salmonellosis attributable to meat-producing animals, with PARs ranging from 7.8% (broilers) to 8.8% (pigs), whereas contacting people with gastroenteritis within the household was associated with salmonellosis of layer/egg origin (6.1%) (Tables 3 and 4). Four additional risk factors became significant in the Pr-weighted models: owning a puppy for salmonellosis of pig origin (2.0%); owning >1 dog and at least one puppy for salmonellosis of layer/egg (6.8%) and broiler (7.4%) origin; playing in a sandbox for salmonellosis of pig (10.7%) and cattle (13.2%) origin; and occupational exposure to animals/raw meats, which was a significant risk factor for cattle-associated salmonellosis (0.4%), but a protective factor for layer/egg-associated salmonellosis.
Significant protective factors for salmonellosis as a whole were: consumption of pasteurized dairy products other than milk and cheese (e.g. yoghurt), chocolate, raw vegetables, fruit, meat in pastry, nuts, fish, and cooked vegetables (Tables 3 and 4). These protective factors remained significant in most of the Pr-weighted models. Of note, eating fish and eating meat in pastry remained significant for pig- and cattle-associated salmonellosis. Two additional protective factors, i.e. drinking pasteurized milk and eating salad, became significant in the models for cattle and broilers, respectively. Moreover, consuming chicken became protective for layer/egg-associated salmonellosis, while consuming pork became protective for both layer/egg- and broiler-associated salmonellosis.
Canonical correlation analysis
The CCA defined three canonical dimensions which explained 82%, 15% and 3% of the total variance in the data, respectively. These canonical dimensions represent orthogonal, linear combinations of the original variables within each of the two sets (i.e. reservoir Pr values and risk factors) that best explain the variability both within and between these two sets of variables. In the scatterplot of the first two canonical dimensions (Figure 1), risk factors significantly associated with salmonellosis of pig, cattle, broiler and layer/egg origin are represented as arrows directed to where their correlations with the reservoir specificity are maximal. Like in a case-case comparison, risk factors in common to most reservoirs (e.g. PPIs) tend to disappear in the centre of the plot. In general, the CCA summarized and confirmed all the associations found in the logistic regression analysis. The point cloud of S. Enteritidis subtype Prs was closest to layers/eggs and its density was relatively higher than that of S. Typhimurium, which was closest to pigs.
Blue squares: S. Typhimurium subtypes. Red dots: S. Enteritidis subtypes. Green triangles: other Salmonella subtypes. Yellow stars: reservoir centroids. Total variance explained by the first and second canonical dimensions is 82% and 15%, respectively.
This is the first case-control study in which risk factors for human salmonellosis are investigated in relation to the attributable reservoirs using subtyping. Previous studies have mainly examined risk factors for salmonellosis according to different Salmonella subtypes –, , ,  and/or to specific populations at risk , . Cummings et al.  have examined risk factors for human salmonellosis caused by bovine-associated subtypes in a similar way, although their attribution relied solely on the crude one-to-one matching (by geographical location, serotype and PFGE pattern) of human isolates with the contemporary bovine isolates, and a case-case approach was used. In the present study, quantitative risk modelling was used to attribute stochastically human cases to reservoirs, and the Pr-weighted case-control study design allowed for the identification of general risk factors for salmonellosis in addition to those factors that were reservoir-specific. These reservoir-specific risk factors provided an indication of the underlying transmission pathways by which salmonellas may reach humans from the different reservoirs.
Reservoirs of human salmonellosis
Layers were estimated to be the most important reservoir of human salmonellosis in the Netherlands in 2002–2003, accounting for ∼50% of domestic and sporadic infections. This is in line with previous estimates based on the original Dutch model  and with other studies conducted in European countries using data from around that period , . However, because of the significant drop in S. Enteritidis infections, the percentage of cases attributable to layers/eggs has decreased over the last decade in most European countries , , including the Netherlands, where layers/eggs in 2011 accounted for almost the same percentage of cases attributable to pigs (∼34%) . Attribution analysis also confirmed that subtypes of S. Enteritidis are strongly associated with layers/eggs and those of S. Typhimurium with pigs. Yet, not all S. Typhimurium strains come from pigs and a considerable variation in host association within this serotype was found (Figure 1), a possible reflection of its ubiquitous nature.
The probability for human infections to originate from each of the reservoirs, as expressed by their Pr values, exhibited significant variation over age groups, gender, seasons and degrees of urbanization. For instance, adults were more likely to be infected by layer/egg-associated salmonellas, whereas pig- and cattle-associated salmonellosis was more likely to occur in the young. Moreover, cattle-associated salmonellosis was more likely to be acquired by males than by females. Although these differences are based on univariate analyses and should thus be interpreted with caution, they may reflect that men are generally more likely to be occupationally exposed to livestock and that different chances of exposure to foodborne and environmental sources exist at varying ages . We also found that poultry-associated salmonellosis was more likely to occur in the warmest months of the year, whereas pig- and cattle-associated salmonellosis were more likely to occur in the coldest ones. The reasons of these seasonal differences are not entirely known. They may reflect a combination of factors, including the parallel Salmonella shedding trends in animal reservoirs , , a relatively larger effect of temperature on products with contaminated eggs combined with at-risk behaviours such as insufficient refrigeration or mishandling of foods , and the enhanced fitness to extreme environmental conditions of certain salmonellas such as S. Typhimurium, which is capable of rapidly recommencing normal growth after prolonged cold stress , .
Previous studies have found that chicken- and cattle-associated campylobacteriosis have differing urban-rural gradients, with chicken being the most likely reservoir for (foodborne) campylobacteriosis in people living in urban areas and cattle being the most likely reservoir for those living in rural areas, especially when considering children aged 0–4 years , , . This was also reported for Shiga toxin-producing Escherichia coli (STEC) O157 infection , whereby suggesting a prominent role of environment-mediated transmission pathways for enteropathogens putatively shed by cattle in the countryside. We report here a similar effect of urbanization on salmonellosis, as pig- and cattle-associated salmonellas were more likely to infect people living in rural areas whereas poultry-associated salmonellosis was more likely to occur in urban areas.
Reservoir-specific risk factors
The primary outcome of this study was to explore reservoir-specific, rather than subtype-specific, risk factors for salmonellosis. This was also pursued in the CCA, which not only confirmed the main findings of the regression analysis, but it also illustrated in a visually appealing way that risk factors are multidimensionally related to Salmonella subtype reservoir specificity. All risk factors were associated in an epidemiologically plausible way according to the reservoir in question. For instance, consumption of raw/undercooked meat (of any origin) was a risk factor for infections attributable to meat-producing animals (but not to layers/eggs), while consumption of raw/undercooked eggs and by-products was a risk factor for infections attributable to layers/eggs (but not to meat-producing animals). Moreover, consuming pork and chicken, together with occupational exposure to animals/raw meats, were protective against layer/egg-associated infections. Plausibly, a person may be somewhat more “protected” against infection from contaminated eggs when exposed to reservoirs other than layers, such as meat-producing animals. Furthermore, substandard kitchen hygiene and exposure to environmental sources such as playing in a sandbox were risk factors for salmonellosis attributable to meat-producing animals, particularly cattle and pigs. This agrees with our other finding that cattle- and pig-associated salmonellosis are more likely to occur in children and indicates that these salmonellas, the majority of which belongs to S. Typhimurium, are not only confined to foods, but may be widespread in the environment or be more capable of infecting people upon direct hand-mouth contact . Indeed, occupational exposure to animals/raw meats was a risk factor for cattle-associated salmonellosis, supporting previous evidence  that contact with cattle increases the risk of bovine-associated salmonellosis and that more infections of cattle origin might result from direct contact than previously recognized.
Owning a puppy was a risk factor for pig-associated salmonellosis, but owning several dogs and at least one puppy was a risk factor for salmonellosis of poultry origin. The association between human salmonellosis and contact with dogs is already known , as dogs can carry (often asymptomatically) salmonellas, especially when kennelled or fed with raw foods . Yet, the different associations found here are difficult to interpret; they may be either due to the different diets/feeding regimes of dogs or to at-risk behaviours undertaken by dog owners. For instance, feeding a household dog with (contaminated) dry food has been reported to be a risk factor for human S. Schwarzengrund infection , and attending veterinary facilities (as either an employee or a client) was linked to a multidrug-resistant S. Typhimurium outbreak in the United States, with splash exposure to dog faeces being the suspected means of transmission . Further analyses considering the contribution of both food and non-food sources to human infections are needed in order to clarify the epidemiological role of pets in the directionality of transmission of salmonellas and other enteropathogens .
Using PPIs was a risk factor for salmonellosis of any origin, while using H2-receptor antagonists increased the risk of acquiring poultry-associated salmonellas only. Moreover, using antibiotics was a risk factor for salmonellosis of pig and cattle origins. While neutralization of gastric acidity by antisecretory drugs may enhance the survival of salmonellas during their passage through the stomach, antibiotics may alter the intestinal microflora in a way that favours infection , , . However, it is also possible that receiving these medications is just a marker of people with gastrointestinal problems and increased susceptibility to infection in general. Besides, these people may be under more frequent medical attention and diagnostic thoroughness so that Salmonella infections are more likely to be detected. Nevertheless, differences in salmonellosis origin associated with different medications are difficult to discern. They might be due to the different therapeutic indications, side-effects and power/duration of effects of PPIs versus H2-receptor antagonists, as well as to the relatively stronger selection pressure by antibiotics towards salmonellas in meat-producing animals than in layers, for which antibiotic use in the EU is highly restrictive (Commission Regulation EC 1177/2006). This hypothesis is also supported by the much higher incidence of antibiotic resistance and emergence of multidrug-resistant strains in S. Typhimurium and S. 4,,12:i:- than in S. Enteritidis , .
Contact with people with gastroenteritis inside the household was a risk factor for layer/egg-associated infections, whereas contact outside the household was associated with infections attributable to meat-producing animals. Considering that contact with other gastroenteritis cases within the household may merely be a proxy for a common source of infection among the family members rather than real person-to-person transmission, this association further supports the notion of the more foodborne nature of S. Enteritidis and the ability of S. Typhimurium to spread through non-food pathways.
Consumption of fish, meat in pastry and several non-meat foods, including fruits, vegetables and dairy products (mainly yoghurt), were protective against infections attributed to different reservoirs. Possibly, people who had consumed fish or meat in pastry were less likely to eat other (potentially more contaminated) meat products. Although a diet rich in fruits, vegetables and dairy probiotics may have genuinely beneficial effects on general health by inhibiting bacterial growth, enhancing general immunity to infection, and altering the intestinal flora in a way that prevents infection, most controls returning the questionnaire may just be particularly motivated people with a generally healthier lifestyle.
The main limitation of the present study is related to the age of the data. From 2002–2003 to date, many changes have occurred in Salmonella epidemiology worldwide; thus, what was identified as risk factors here might not reflect entirely the actual situation anymore. Although our data are relatively outdated in order to identify potential targets for today’s Salmonella control activities, they represent a large, comprehensive set of information to develop our combined case-control and source attribution analysis in a robust way. The analytical approach we propose to study reservoir-specific risk factors for human salmonellosis will be of particular interest when more recent data will become available. This will allow the analysis to be updated as to corroborate and compare results. Moreover, the analytical approach presented here can, in principle, be extended to virtually all other zoonotic pathogens for which combining epidemiological and source attribution data can make a difference. This seems to be the case of Campylobacter, for which similar analyses have been performed using multilocus sequence typing –. However, the attempt to use Pr weights was unsuccessful with Campylobacter (personal communication of the authors). This may be due to the fact that associations of Campylobacter sequence types (STs) with specific reservoirs are much more dynamic and transient in time and space than those of Salmonella serotypes/phage types , which tend to adapt to specific reservoirs more consistently, showing a generally greater discriminatory power between sources than Campylobacter STs. For this reason it was only possible to run the analysis for Campylobacter using those cases infected with rarely occurring and highly reservoir-specific STs, whereas for Salmonella it was possible to use the whole (Pr-weighted) dataset.
As a last point, it has been postulated that repeated exposures to enteropathogens may lead to sufficient immunity to provide protection against (severe) clinical illness , . In case-control studies, this "protective immunity" would lead to misclassification, as some controls could have been infected asymptomatically. As cases were identified by passive surveillance, they were likely to represent the most severe, symptomatic infections that occurred in the population. Thus, the risk factors identified here especially represent risk factors for severe salmonellosis.
Several case-control studies have investigated risk factors for human salmonellosis caused by specific subtypes while other studies have used Salmonella subtyping to attribute human infections to different food and animal reservoirs. This study combined both these types of studies into a single analysis to explore risk factors at the point of exposure for human salmonellosis attributable to four putative food-producing animal reservoirs.
We confirmed that salmonellosis in the Netherlands in 2002–2003 was mainly attributable to layers/eggs, followed by pigs. However, we also found that poultry is the primary reservoir of salmonellosis for adults, for people living in urban areas, and during spring/summer time, whereas pigs and cattle are the primary reservoir for children, for people living in rural areas, and during autumn/winter.
Besides the identification of some universal risk factors for gastrointestinal infections in general (e.g. using PPIs), this study shows that risk factors for salmonellosis vary according to the attributable reservoir, suggesting that salmonellas may infect humans through various transmission pathways depending on their original reservoirs. This study also shows that the outcome of classical case-control studies can be enhanced by incorporating source attribution data, providing a valuable approach for supporting and generating hypotheses. Results also indicate that the general concept of Salmonella source attribution modelling makes sense epidemiologically and that reservoir-specific risk factors for salmonellosis are mirroring, to some extent, the underlying differences in the epidemiology of the two most important serotypes, S. Typhimurium and S. Enteritidis, albeit in a more refined and informed way.
Frequencies of Salmonella subtypes in humans (2002-2003) and animal reservoirs (2001-2004), The Netherlands.
We are grateful to the participating Regional Public Health Laboratories for their contribution to the data collection, especially to F. Vlaspolder, J.H. Sloos, J. Spaargaren, J. Peereboom, M.A. Schouten, R.W. Brimicombe, F.W. Sebens, Ph.H. Rothbarth, L.J.M. Sabbe, H. Mulder, Veenendaal, E. Ijzerman, J.H.T. Wagenvoort, J.H. van Zeijl, B.M. de Jongh, M. Tersmette, P. Voorn, A.M. Horrevorts, J. Buitenwerf, B.G.A. Hendrickx, M. Peeters and A.R. Jansz. We also are grateful to H.M.E. Maas and A.J. Verbruggen for serotyping and phage typing of the Salmonella isolates.
Conceived and designed the experiments: LMG WvP. Performed the experiments: MH YvD. Analyzed the data: LMG RE IF WvP. Contributed reagents/materials/analysis tools: LMG RE IF MH YvD WvP. Wrote the paper: LMG RE IF MH YvD WvP.
- 1. European Food Safety Agency (EFSA), European Centre for Disease Control (ECDC) (2012) The European Union summary report on trends and sources of zoonoses, zoonotic agents and food-borne outbreaks in 2010. EFSA J 10: 2597. Available: http://www.efsa.europa.eu/en/efsajournal/doc/2597.pdf. Accessed 2013 Nov 21.
- 2. Havelaar AH, Haagsma JA, Mangen MJJ, Kemmeren JM, Verhoef LPB, et al. (2012) Disease burden of foodborne pathogens in the Netherlands, 2009. Int J Food Microbiol 156: 231–238.
- 3. Rijksinstituut voor Volksgezondheid en Milieu (RIVM) (2012) Staat van zoönosen 2011. Bilthoventhe Netherlands: RIVM. 68 p.
- 4. Mangen M-JJ, Bouwknegt M, Friesema IHM, Kortbeek LM, van Pelt W, et al.. (2013) Disease burden and cost-of-illness of food-related pathogens in the Netherlands, 2011. Bilthoventhe Netherlands: Rijksinstituut voor Volksgezondheid en Milieu (RIVM). 95 p.
- 5. van den Brandhof WE, De Wit GA, de Wit MAS, van Duynhoven YTHP (2004) Costs of gastroenteritis in The Netherlands. Epidemiol Infect 132: 211–221.
- 6. Friesema IHM, Schimmer B, Ros JA, Ober HJ, Heck MEOC, et al. (2012) A regional Salmonella enterica serovar Typhimurium outbreak associated with raw beef products, The Netherlands, 2010. Foodborne Pathog Dis 9: 102–107.
- 7. van Duynhoven YTHP, Isken LD, Borgen K, Besselse M, Soethoudt K, et al. (2009) A prolonged outbreak of Salmonella Typhimurium infection related to an uncommon vehicle: hard cheese made from raw milk. Epidemiol Infect 137: 1548–1557.
- 8. Whelan J, Noel H, Friesema I, Hofhuis A, de Jager CM, et al.. (2010) National outbreak of Salmonella Typhimurium (Dutch) phage-type 132 in the Netherlands, October to December 2009. Euro Surveill 15: pii 19705.
- 9. Friesema IH, de Jong AE, Fitz James IA, Heck ME, van den Kerkhof JH, et al. (2012) Outbreak of Salmonella Thompson in the Netherlands since July 2012. Euro Surveill 17: 20303.
- 10. Doorduyn Y, van den Brandhof WE, van Duynhoven YTHP, Wannet WJB, van Pelt W (2006) Risk factors for Salmonella Enteritidis and Typhimurium (DT104 and non-DT104) infections in The Netherlands: predominant roles for raw eggs in Enteritidis and sandboxes in Typhimurium infections. Epidemiol Infect 134: 617–626.
- 11. Marcus R, Varma JK, Medus C, Boothe EJ, Anderson BJ, et al. (2007) Re-assessment of risk factors for sporadic Salmonella serotype Enteritidis infections: a case-control study in five FoodNet Sites, 2002-2003. Epidemiol Infect 135: 84–92.
- 12. Delarocque-Astagneau E, Bouillant C, Vaillant V, Bouvet P, Grimont PA, et al. (2000) Risk factors for the occurrence of sporadic Salmonella enterica serotype Typhimurium infections in children in France: a national case-control study. Clin Infect Dis 31: 488–492.
- 13. Kuhn KG, Sørensen G, Torpdahl M, Kjeldsen MK, Jensen T, et al.. (2012) A long-lasting outbreak of Salmonella Typhimurium U323 associated with several pork products, Denmark, 2010. Epidemiol Infect. DOI: https://doi.org/http://dx.doi.org/10.1017/S0950268812000702.
- 14. Doré K, Buxton J, Henry B, Pollari F, Middleton D, et al. (2004) Risk factors for Salmonella Typhimurium DT104 and non-DT104 infection: a Canadian multi-provincial case-control study. Epidemiol Infect 132: 485–493.
- 15. Mughini-Gras L, Barrucci F, Smid JH, Graziani C, Luzzi I, et al.. (2013) Attribution of human Salmonella infections to animal and food sources in Italy (2002-2010): adaptations of the Dutch and modified Hald source attribution models. Epidemiol Infect. DOI: https://doi.org/http://dx.doi.org/10.1017/S0950268813001829.
- 16. Hald T, Vose D, Wegener HC, Koupeev T (2004) A Bayesian approach to quantify the contribution of animal-food sources to human salmonellosis. Risk Anal 24: 255–269.
- 17. Wahlström H, Andersson Y, Plym-Forshell L, Pires SM (2011) Source attribution of human Salmonella cases in Sweden. Epidemiol Infect 139: 1246–1253.
- 18. Guo C, Hoekstra RM, Schroeder CM, Pires SM, Ong KL, et al. (2011) Application of Bayesian techniques to model the burden of human salmonellosis attributable to U.S. food commodities at the point of processing: adaptation of a Danish model. Foodborne Pathog Dis 8: 509–516.
- 19. Mullner P, Jones G, Noble A, Spencer SEF, Hathaway S, et al. (2009) Source attribution of food-borne zoonoses in New Zealand: a modified Hald model. Risk Anal 29: 970–984.
- 20. David JM, Guillemot D, Bemrah N, Thébault A, Brisabois A, et al. (2013) The Bayesian microbial subtyping attribution model: robustness to prior information and a proposition. Risk Anal 33: 397–408.
- 21. Mughini Gras L, Smid JH, Wagenaar JA, de Boer AG, Havelaar AH, et al. (2012) Risk factors for campylobacteriosis of chicken, ruminant, and environmental origin: a combined case-control and source attribution analysis. PloS One 7: e42599.
- 22. Mughini Gras L, Smid JH, Wagenaar JA, Koene MGJ, Havelaar AH, et al. (2013) Increased risk for Campylobacter jejuni and C. coli infection of pet origin in dog owners and evidence for genetic association between strains causing infection in humans and their pets. Epidemiol Infect 141: 2526–2535.
- 23. Mughini-Gras L, Smid JH, Wagenaar JA, DE Boer A, Havelaar AH, et al.. (2013) Campylobacteriosis in returning travellers and potential secondary transmission of exotic strains. Epidemiol Infect. DOI: https://doi.org/http://dx.doi.org/10.1017/S0950268813002069.
- 24. van Duijkeren E, Wannet WJB, Houwers DJ, van Pelt W (2002) Serotype and phage type distribution of salmonella strains isolated from humans, cattle, pigs, and chickens in the Netherlands from 1984 to 2001. J Clin Microbiol 40: 3980–3985.
- 25. Winship C, Radbill L (1994) Sampling weights and regression analysis. Sociol Methods Res 23: 230–257.
- 26. Bassal R, Reisfeld A, Nissan I, Agmon V, Taran D, et al.. (2013) Risk factors for sporadic infection with Salmonella Infantis: a matched case-control study. Epidemiol Infect. DOI: https://doi.org/http://dx.doi.org/10.1017/S0950268813001799.
- 27. Hennessy TW, Cheng LH, Kassenborg H, Ahuja SD, Mohle-Boetani J, et al. (2004) Egg consumption is the principal risk factor for sporadic Salmonella serotype Heidelberg infections: a case-control study in FoodNet sites. Clin Infect Dis 38: S237–243.
- 28. Thomson CN, Phan VTM, Le TPT, Pham TNT, Hoang LP, et al. (2013) Epidemiological features and risk factors of Salmonella gastroenteritis in children resident in Ho Chi Minh City, Vietnam. Epidemiol Infect 141: 1604–1613.
- 29. Cummings KJ, Warnick LD, Davis MA, Eckmann K, Grohn YT, et al. (2012) Farm animal contact as risk factor for transmission of bovine-associated Salmonella subtypes. Emerg Infect Dis 18: 1929–1936.
- 30. David JM, Sanders P, Bemrah N, Granier SA, Denis M, et al. (2013) Attribution of the French human salmonellosis cases to the main food-sources according to the type of surveillance data. Prev Vet Med 110: 12–27.
- 31. Pires SM, de Knegt L, Hald T (2011) Estimation of the relative contribution of different food and animal sources to human Salmonella infections in the European Union. ParmaItaly: European Food Safety Agency (EFSA). 80 p.
- 32. Graziani C, Mughini-Gras L, Owczarek S, Dionisi A, Luzzi I, et al.. (2013) Distribution of Salmonella enterica isolates from human cases in Italy, 1980 to 2011. Euro Surveill 18: pii 20519.
- 33. Wales A, Breslin M, Carter B, Sayers R, Davies R (2007) A longitudinal study of environmental Salmonella contamination in caged and free-range layer flocks. Avian Pathol 36: 187–197.
- 34. Nielsen LR, Schukken YH, Gröhn YT, Ersbøll AK (2004) Salmonella Dublin infection in dairy cattle: risk factors for becoming a carrier. Prev Vet Med 65: 47–62.
- 35. Shah J, Desai PT, Chen D, Stevens JR, Weimer BC (2013) Pre-adaptation to cold-stress in Salmonella Typhimurium increases survival during subsequent acid-stress exposure. Appl Environ Microbiol 79: 7281–7289.
- 36. Humphrey S, Clark LF, Humphrey TJ, Jepson MA (2011) Enhanced recovery of Salmonella Typhimurium DT104 from exposure to stress at low temperature. Microbiology 157: 1103–1114.
- 37. Strachan NJC, Gormley FJ, Rotariu O, Ogden ID, Miller G, et al. (2009) Attribution of Campylobacter infections in northeast Scotland to specific sources by use of multilocus sequence typing. J Infect Dis 199: 1205–1208.
- 38. Mullner P, Shadbolt T, Collins-Emerson JM, Midwinter AC, Spencer SEF, et al. (2010) Molecular and spatial epidemiology of human campylobacteriosis: source association and genotype-related risk factors. Epidemiol Infect 138: 1372–1383.
- 39. Friesema IHM, van de Kassteele J, de Jager CM, Heuvelink AE, van Pelt W (2010) Geographical association between livestock density and human Shiga toxin-producing Escherichia coli O157 infections. Epidemiol Infect. DOI: https://doi.org/http://dx.doi.org/10.1017/S0950268810002050.
- 40. Varga C, Middleton D, Walton R, Savage R, Tighe M-K, et al. (2012) Evaluating risk factors for endemic human Salmonella Enteritidis infections with different phage types in Ontario, Canada using multinomial logistic regression and a case-case study approach. BMC Public Health 12: 866.
- 41. Leonard EK, Pearl DL, Finley RL, Janecko N, Peregrine AS, et al. (2011) Evaluation of pet-related management factors and the risk of Salmonella spp. carriage in pet dogs from volunteer households in Ontario (2005-2006). Zoonoses Public Health 58: 140–149.
- 42. Behravesh CB, Ferraro A, Deasy M, Dato V, Moll M, et al. (2010) Human Salmonella infections linked to contaminated dry dog and cat food, 2006-2008. Pediatrics 126: 477–483.
- 43. Centers for Disease Control and Prevention (CDC) (2001) Outbreaks of multidrug-resistant Salmonella Typhimurium associated with veterinary facilities --- Idaho, Minnesota, and Washington, 1999. MMWR Morb Mortal Wkly Rep 50: 701–704.
- 44. Doorduyn Y, van den Brandhof WE, van Duynhoven YTHP, Wannet WJB, van Pelt W (2006) Risk factors for Salmonella Enteritidis and Typhimurium (DT104 and non-DT104) infections in The Netherlands: predominant roles for raw eggs in Enteritidis and sandboxes in Typhimurium infections. Epidemiol Infect 134: 617–626.
- 45. Yang SJ, Park KY, Seo KS, Besser TE, Yoo HS, et al. (2001) Multidrug-resistant Salmonella Typhimurium and Salmonella Enteritidis identified by multiplex PCR from animals. J Vet Sci 2: 181–188.
- 46. NETHMAP MARAN (2013) Consumption of antimicrobial agents and antimicrobial resistance among medically important bacteria in the Netherlands - Monitoring of antimicrobial resistance and antibiotic usage in animals in the Netherlands. 126 p. Available: http://www.rivm.nl/dsresource?objectid=rivmp:181194&type=org&disposition=inline&ns_nc=1. Accessed 2013 Nov 22.
- 47. Smid JH, Mughini Gras L, de Boer AG, French NP, Havelaar AH, et al. (2013) Practicalities of using non-local or non-recent multilocus sequence typing data for source attribution in space and time of human campylobacteriosis. PloS One 8: e55029.
- 48. Swift L, Hunter PR (2004) What do negative associations between potential risk factors and illness in analytical epidemiological studies of infectious disease really mean? Eur J Epidemiol 19: 219–223.
- 49. Bouwknegt M, Dam-Deisz W, Wannet W, van Pelt W, Visser G, et al.. (2004) Surveillance of zoonotic bacteria in farm animals in The Netherlands - Results from January 1998 until December 2002. BilthovenThe Netherlands: Rijksinstituut voor Volksgezondheid en Milieu (RIVM). 58 p.
- 50. van Horne P (2007) Production and consumption of poultry meat and eggs in the European Union. European Commission, project number SSPE-CT-2004-513737 “Healthy Poultry”. 21 p. Available: http://www.healthy-poultry.org/Results%20of%20the%20project/chapter3.pdf. Accessed 2013 Nov 22.