Source attribution of human campylobacteriosis at the point of exposure by combining comparative exposure assessment and subtype comparison based on comparative genomic fingerprinting

André Ravel; Matt Hurst; Nicoleta Petrica; Julie David; Steven K. Mutschall; Katarina Pintar; Eduardo N. Taboada; Frank Pollari

doi:10.1371/journal.pone.0183790

Abstract

Human campylobacteriosis is a common zoonosis with a significant burden in many countries. Its prevention is difficult because humans can be exposed to Campylobacter through various exposures: foodborne, waterborne or by contact with animals. This study aimed at attributing campylobacteriosis to sources at the point of exposure. It combined comparative exposure assessment and microbial subtype comparison with subtypes defined by comparative genomic fingerprinting (CGF). It used isolates from clinical cases and from eight potential exposure sources (chicken, cattle and pig manure, retail chicken, beef, pork and turkey meat, and surface water) collected within a single sentinel site of an integrated surveillance system for enteric pathogens in Canada. Overall, 1518 non-human isolates and 250 isolates from domestically-acquired human cases were subtyped and their subtype profiles analyzed for source attribution using two attribution models modified to include exposure. Exposure values were obtained from a concurrent comparative exposure assessment study undertaken in the same area. Based on CGF profiles, attribution was possible for 198 (79%) human cases. Both models provide comparable figures: chicken meat was the most important source (65–69% of attributable cases) whereas exposure to cattle (manure) ranked second (14–19% of attributable cases), the other sources being minor (including beef meat). In comparison with other attributions conducted at the point of production, the study highlights the fact that Campylobacter transmission from cattle to humans is rarely meat borne, calling for a closer look at local transmission from cattle to prevent campylobacteriosis, in addition to increasing safety along the chicken supply chain.

Citation: Ravel A, Hurst M, Petrica N, David J, Mutschall SK, Pintar K, et al. (2017) Source attribution of human campylobacteriosis at the point of exposure by combining comparative exposure assessment and subtype comparison based on comparative genomic fingerprinting. PLoS ONE 12(8): e0183790. https://doi.org/10.1371/journal.pone.0183790

Editor: Patrick Jon Biggs, Massey University, NEW ZEALAND

Received: September 21, 2016; Accepted: August 12, 2017; Published: August 24, 2017

Copyright: © 2017 Ravel 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.

Data Availability: All relevant data are within the paper and its Supporting Information files.

Funding: The authors received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Campylobacter is the leading bacterial cause of foodborne enteric disease in most developed countries, with an estimated total incidence of 213,749 domestically-acquired cases (90% credible interval 144,288–308,837) per year among the 32,500,000 Canadians in 2006 [1]. This bacterium is present in many animal reservoirs as well as in the water, to which the few outbreaks registered are often related. Case information illustrates that human infections of campylobacteriosis are mostly sporadic, which is what makes direct identification of the source of contamination difficult. The epidemiological studies conducted in the form of case-controls or intervention studies have helped identify poultry as the probable main source of Campylobacter [2]. However, there is some evidence that other sources may play a non-negligible role in Campylobacter aetiology. Therefore, source attribution, the quantification of the proportion of disease cases linked to potential sources, may prove helpful to determine the relative importance of the Campylobacter sources and hence direct public health efforts more efficiently [3, 4].

The microbial subtyping attribution approaches are currently one promising method to perform source attribution. They are data intensive and best suited to use data from a well-designed integrated surveillance system with an efficient (discriminatory, reproducible), standardised and systematically applied typing method [5]. The data on human cases are to be compared to the data on the sources, hence the necessity of a harmonised typing method to ensure comparability of those data. Moreover, representative isolates from all possible sources of human cases should be available to get reliable results. The data on the sources (reservoirs or vehicles) should thus be based on a representative sampling and cover as many sources as possible [5, 6]. There are two main sorts of microbial subtyping attribution models, the frequency matched attribution models, based on the comparison of human strain types and the distribution of those types in the sources, and the population genetic models based on modelling the organism’s evolutionary history. The frequency matched attribution models require a typing method that has sufficient discrimination while maintaining several types that are in both human cases and sources [3].

Human illness attribution can be undertaken at various points along the transmission route from the reservoirs to the humans, and is usually performed at the point of production (i.e. animal reservoir), point of distribution and point of exposure (i.e. consumption of contaminated food or unintended ingestion of the pathogen with water or because of unhygienic practice) [3, 4]. Most previous studies on Campylobacter or Salmonella source attribution based on subtype comparison were undertaken at the production point [5–9] and a few at the distribution point [10]. Some studies mixed subtype data of isolates collected in live animals or at slaughter with data of isolates detected at the distribution point (e.g. imported meat) [8].

Comparative exposure assessment is another approach to estimate human illness attribution. It has been more rarely undertaken with only two published studies on Campylobacter attribution to our knowledge [11, 12]. Combining the two approaches (comparative exposure assessment and subtyping comparison) was judged worthwhile because it allows one to inform risk management and prioritisation of control strategies for each of the different routes of the pathogen transmission instead of focusing on the reservoir [3].

The subtyping of Campylobacter, beyond the species level, for such studies is a challenge. The most widely used method so far is multi-locus sequence typing (MLST), which is highly discriminatory, reproducible and produces easy-to-interpret results [13]. This method has been used in several Campylobacter source attribution studies so far including in New Zealand [7, 14], and in various parts of Europe [8, 15–20]. As it is costly and time consuming, routine use of MLST at the scale required for integrated surveillance is impractical; a Dutch research team, for instance, has showed that the use of non-local data from far away countries may result in some geographical bias in the attribution estimates [21].

A new typing method, referred to as comparative genomic fingerprinting (CGF), was recently developed in Canada. This molecular method detects the presence/absence of 40 specific genes that target genetic variability in accessory genome content. This method has been shown to be highly concordant with MLST while having much greater discriminatory power [22]. Moreover, because it is a high-throughput, low-cost, and high-resolution method, CGF is amenable to deployment in a surveillance context [23]. Therefore, the CGF method could be a valuable alternative to MLST, especially in the frame of source attribution based on subtypes.

Finally, this study analyses data from FoodNet Canada, a comprehensive sentinel-based integrated surveillance system for enteric diseases in Canada. It was implemented to understand which sources are contributing to human diseases and to identify risk factor information that can be used to guide more effective interventions for food and water safety. This system collects both microbial and epidemiologic data (e.g. human/animal species, location, and for the human cases, travel information, outbreak implicated information, exposure information, etc.). Among other pathogens, it includes Campylobacter isolates from reported human cases, from foods at retail, from farm animals and from surface water within the same region (sentinel site). Within its first sentinel site, the Region of Waterloo, Ontario, a food flow analysis estimated that between 1% and 10% of the beef, pork, and poultry meat available at retail come from local farms [24]. In this context, food animal and their products at retail can be considered independent sources of Campylobacter for humans. This setting provided a unique standardized microbial and epidemiological dataset over a well-defined and narrow spatio-temporal frame, thus uniquely fulfilling data recommendations for source attribution by subtyping comparisons [3, 4], while allowing work at the point of exposure.

The study aims to derive source attribution estimates for human campylobacteriosis at the point of exposure using the subtyping comparison approach applied to Campylobacter subtyped thanks to CGF combined with the results of a comparative exposure assessment. A secondary objective was to explore the impact of the mathematical modelling on the source attribution estimates and parameters.

Material & methods

Sources of isolates

FoodNet Canada’s enhanced human surveillance component collects information on enteric disease cases in its sentinel sites from the local public health authorities. For this paper, Campylobacter data from the Region of Waterloo (ROW) sentinel site was analyzed. The ROW is located in southwestern Ontario, Canada and is composed of three urban municipalities and four rural townships with a total population of about 500,000. FoodNet Canada uses the existing laboratory-based surveillance system for reportable diseases as it is mandatory for clinical laboratories to report each case of reportable disease to the local public health authority. FoodNet Canada has enhanced this passive reporting system by implementing a standardized questionnaire on all cases of enteric disease that are reported to the local public health authority. This questionnaire includes detailed risk factor and exposure information (see in [25]. Laboratory results are consolidated with the questionnaire information by the public health authorities, who ultimately provide depersonalized epidemiological and microbiological data to FoodNet Canada. Ethics approval for the surveillance data collection was obtained through the Region of Waterloo Public Health Ethics Review Committee in 2005.

Active source surveillance is an integral part of FoodNet Canada; its three components test samples of uncooked retail meats, manure from local food animal producers, and untreated surface water from five points within the Grand River watershed [26]. Retail meats tested for Campylobacter include ground beef, chicken (chicken breast, ground chicken, and chicken nuggets), ground turkey, and pork chops. Food animal manures sampled include beef and dairy cattle, broiler chickens, and swine.

The human clinical isolates included 249 sporadic, domestically-acquired cases and one outbreak-related case randomly chosen among the 22 cases the outbreak encompassed. The selection of only one out of the 22 cases was chosen to avoid any bias towards the source that would be the cause of this outbreak.

Eight potential sources of Campylobacter were used: three reservoir (cattle manure, chicken manure, swine manure), four food vehicle (beef meat, chicken meat, pork meat, turkey meat) and one environmental vehicle (surface water).

Samples analyzed during the years 2006 to 2011 inclusively were used to achieve a sufficient number of isolates for rarer sources and some sources that were not sampled uniformly over time (Table 1).

Download:

Table 1. Distribution of Campylobacter isolates with CGF data by year and by origin available for source attribution.

https://doi.org/10.1371/journal.pone.0183790.t001

CGF analysis

Campylobacter isolates (n = 1,764) were retrieved from frozen glycerol stocks and subcultured onto blood agar plates under microaerophilic conditions at 42°C for 24–48 hours. Biomass was harvested for DNA extraction using the Qiagen EZ1 BioRobot, or EZ1 Advanced XL with the Qiagen Blood and Tissue Kit according to manufacturer’s instructions (Qiagen, Mississauga, Canada). CGF40 PCRs were performed as described previously [22]. PCR products were analyzed using the QIAxcel capillary electrophoresis system with the DNA Screening Kit. The 40 PCR targets were scored as binary data based on presence or absence of bands of expected size, which were then compiled to create a CGF pattern. CGF subtypes are routinely defined based on pattern similarity of isolates in the CGF database at three similarity thresholds (90%, 95% and 100%) and based on hierarchical position in the clustered database. Only the 100% similarity threshold was used to define CGF subtypes in this study (S1 File).

Data analysis

The number of different CGF subtypes were calculated for each origin (human cases or potential sources) and overall. Within each origin, the number of subtypes specific to this origin and the subtypes also found in other origins (common subtype) were tabulated. The subtypes found among human isolates and shared with a single source were also identified.

Source attribution estimations

Three methods, falling under the general subtyping comparison approach, were applied to the Campylobacter isolates characterized by their CGF subtype to calculate source attribution estimates: the proportional similarity index, the Dutch model, and the Hald model [8, 14, 27]. The models and their equations are explained in the following paragraphs. For sake of clarity and consistency between the methods’ equations, we define several parameters in a manner that is in alignment with previous papers as much as possible. They are:

h_i = number of human cases with subtype i
b_ij = the number of Campylobacter positive isolates for subtype i and source j (no restriction is made on whether there are human cases with subtype i or not)
r_ij = b_ij / ∑_i b_ij, the proportion of Campylobacter-positive isolates of subtype i in source j
s_j = the number of samples testing negative for Campylobacter in source j
p_ij = b_ij / [(∑_i b_ij)+s_j], the proportion of Campylobacter-positive samples of subtype i in source j, referred to as prevalence

Proportional similarity index.

The proportional similarity index (PSI) was used to measure the degree of overlap of the frequency distributions of CGF subtypes between the human isolates and a group of non-human isolates [27]. It was computed as follows: (1) with h_i and r_ij as previously defined. PSI_j ranges from zero to one, where one indicates that the two groups are identical and zero means they have no types in common. 95% confidence intervals were computed using bias-corrected and accelerated non-parametric bootstrap, as implemented in Stata (StataCorp. 2013. Stata: Release 13. Statistical Software. College Station, TX) using the bootstrap command with the bca option. The bootstrap command was applied to a custom estimation program that calculates the PSI (available on request).

Dutch model.

Source attribution estimates based on the Dutch model were computed as follows [28]: (2) where k_ij is the relative occurrence of subtype i in source j, a parameter explained later, and h_i as explained earlier.

λ_ij is the estimated number of cases of subtype i from source j. The sum across subtypes gave the total number of cases from source j, λ_j: (3)

Final source attribution proportions, z_j were then calculated (Eq 4). Confidence intervals were calculated using the same custom estimation program discussed for the PSI. Hypothesis tests for differences between attribution estimates also used this program and are based on the same bootstrap method.

(4)

Hald model.

The model was adapted from the Campylobacter attribution model by Boysen et al. [8], which is an adaptation of a Salmonella attribution model [29]. It uses a Bayesian framework to estimate the number of human cases attributed to the various sources under study. It is described according to the following equations: (5) where λ_ij is the expected number of cases of subtype i from source j, k_ij is the occurrence of subtype i in source j and is defined below, q_i is the subtype-dependent parameter (interpreted as the specific ability of subtype i to cause disease), and a_j the source-dependent parameter (interpreted as the specific ability of source j to transfer the pathogen to humans). The q_i and a_j are unknown. These parameters were set with a hierarchical prior [8] and an exponential prior distribution, respectively. The log of q_i was set to follow a normal distribution N(0, τ) with the prior distribution for τ being the Gamma(0.01, 0.01) distribution as suggested by Mullner [7]. The prior for a_j was assigned an Exponential (0.02, 0.02), also suggested by Mullner [7]. Following previous work [30], the value for each q_i “anchor” was set to fixed value for each subtype that was source-specific, meaning those subtypes i that were present from human isolates and found in only one source. In those cases the q_i value was: (6)

The models were estimated using OpenBugs (http://www.openbugs.net/w/FrontPage), with 5000 burn-in iterations followed by 40,000 iterations.

Defining the occurrence k_ij.

For the primary objective of source attribution at the point of exposure, we set the k_ij parameter to be r_ij*E_j, where r_ij was defined above and E_j is the exposure value for source j. The exposure value E_j for the eight sources tested were derived from a concurrent Campylobacter comparative exposure assessment study [12]. This work estimated the average population exposure of Campylobacter to humans (measured in terms of number of organisms ingested/person/day) from 13 different sources including seven of the eight sources of interest here. It was conducted for the province of Ontario that includes the area from which the surveillance data used in this study originate. The E_j values for chicken, beef and pork retail meat was directly taken from the comparative exposure assessment study results. The exposure for the turkey meat was adapted from the chicken meat by using the relevant parameter values whenever available (i.e. turkey net consumption, retail ground turkey contamination rate for Campylobacter). The exposure value for contact with animals was available from the exposure assessment study for all food animals together and for exposure from either living on a farm or visiting a farm. Thanks to a previous survey of healthy humans on their behavior and exposure in the same area [31], we were able to breakdown the estimated value of exposure through food animals by living on a farm for the three animal species of interest (chicken, cattle and swine) proportionally to the species that people living on farm have contact with and similarly for the exposure through visiting a farm. Finally the values for exposure through living on a farm and through visiting a farm were summed up by animal species. The exposure values used in this study as E_j are presented in Table 2.

Download:

Table 2. Exposure value used in the exposure-based attribution models [12].

https://doi.org/10.1371/journal.pone.0183790.t002

Because of uncertainty in the exposure values, a sensitivity analysis was undertaken for the exposure-based Dutch model to assess to what extent the source attribution values estimated by the model change when the exposure value for a given source is increased or decreased, the values for all other parameters being kept constant. Probability intervals for exposure estimates varied between 1 and 2 log from the mean estimate to include all possibilities, even extreme scenarios. Based on this, we used sensitivity inputs that are -2, -1.5, -1, -0.5, 0, 0.5, 1, 1.5 and 2 (log₁₀ units) times the mean value to explore the association between exposure values E_j and source attribution estimates λ_j.

For the secondary objective of exploring the impact of including exposure values, we set the k_ij parameters to be r_ij (proportion-based model) and p_ij (prevalence-based model), ran the Dutch and the Hald model accordingly and visually compared the outcomes. We also explore the impact of the structure of the Hald model on the source-dependent parameters (a_j) and subtype-dependent parameters (q_i). The a_j parameters are supposed to be source specific, implying a unique value for a given source. The original Hald model designed for attribution at the point of production included a food consumption parameter M_j [29]. Later studies using this model dropped the consumption parameter on the basis that the changes in the model structure (with or without the food consumption) are absorbed by the a_j parameters [7, 32, 33]. The evidences for such absorption is limited [7]. Furthermore, recent studies proposed or used a modified Hald model where its structure includes another parameter: the proportion of food consumed raw or undercooked [33]. In order to better document the absorption of the changes in the model structure by the a_j parameters, we had the posterior distribution of k_ij*a_j computed when running the three Hald models and looked at the correlation between their mean values. Our hypothesis was that the value of k_ij*a_j would be identical or at least very close if the absorption effect mentioned held true. We also computed the relative change in their mean values between the three models using the prevalence-based model as the reference. Finally, the Hald model assumes the existence of subtype-dependent parameters (q_i). According to this, the q_i values for a given subtype should be the same independently of the model structure. We explored this assumption by looking at the correlation between the proportion, prevalence and exposure-based model in their posterior mean values.

Results

Description of samples, isolates and subtypes

Table 3 describes the isolates available from the different origins as well as their prevalence in the potential sources as derived from the FoodNet Canada active surveillance. Overall, 453 different CGF subtypes were found among the 1768 isolates analyzed. The number of isolates per subtype ranged from 1 to 71. Just over eighty percent of all subtypes (81.2%, 368/453) were unique to an origin (Table 3). Unique subtypes were found among all sampling origins except for chicken manure and beef meat.

Download:

Table 3. Distribution of samples, isolates and CGF subtypes by sampling origin.

https://doi.org/10.1371/journal.pone.0183790.t003

The 250 human isolates were distributed among 104 CGF subtypes, 48 of these subtypes (53 isolates) being unique to the human cases, and the remaining 56 subtypes (197 isolates) being found in at least one of the studied sources, most often among chicken meat (49 subtypes encompassing 187 human isolates) and cattle manure (33 subtypes encompassing 157 human isolates) (Table 2). Sixteen subtypes encompassing 23 human isolates were present in only a single source: one subtype was common to chicken manure, four to cattle manure, 17 to chicken meat, and one to surface water.

Ten subtypes were found in chicken manure, none being unique, and were most often found in human cases (9 subtypes) and chicken meat (8 subtypes), and one subtype each in turkey meat and surface water (Table 3). The 624 cattle manure isolates were distributed among 135 CGF subtypes with 85 being unique to this source and 50 subtypes (497 isolates) being found in all the other sources including human cases (33 subtypes) and chicken meat (39 subtypes). The 330 swine manure isolates were grouped into 128 subtypes, 109 being unique, and 19 (98 isolates) were found in human cases (5 subtypes) and in other studied sources, mostly chicken meat (17 subtypes) and cattle manure (12 subtypes), but not in chicken manure, beef meat and turkey meat.

The 446 chicken meat isolates fell into 166 subtypes, 94 unique and the other 72 (322 isolates) covering all of the other origins studied, mostly among human cases (49 subtypes) and cattle manure (39 subtypes). The three beef meat isolates were of three different subtypes that were also common to human cases, cattle manure and chicken meat: The six pork meat isolates were each of a different subtype, two were unique to pork meat whereas the other four isolates were present in human cases, cattle manure and chicken meat. Turkey meat isolates were of 17 different subtypes, five unique and 12 (21 isolates) being found in the other origins, most often in chicken meat (11 subtypes) and human cases (10 subtypes), but not in swine manure and beef meat. About two-thirds of the subtypes found in surface water were unique (23 subtypes) whereas the other third were found in all the other origins except beef meat, and more frequently in chicken meat (nine subtypes), cattle manure (eight subtypes), and human cases (seven subtypes).

The CGF subtypes among the chicken meat (PSI = 0.421) and the cattle manure (PSI = 0.356) were moderately similar to the human subtypes, whereas the similarity with human isolates was low for surface water (PSI = 0.110), beef meat (PSI = 0.076), pork meat (PSI = 0.060), and swine manure (PSI = 0.019) (Table 4).

Download:

Table 4. Proportional Similarity Index (PSI) between human campylobacteriosis isolates and isolates from potential sources based on their Comparative Genomic Fingerprinting subtypes.

https://doi.org/10.1371/journal.pone.0183790.t004

Attribution at the point of exposure

The exposure-based Hald model runs converged without any issues and the posterior mean number of cases by subtype closely matched the actual number of cases by less than one case for the majority of the subtypes. Chicken meat was by far the main source of human cases of campylobacteriosis (69%) followed by cattle manure (14%). Cases were almost uniformly spread across the remaining sources, except for swine manure (0.02%), with attribution estimates ranging between 2 and 5% (Fig 1).

Download:

Fig 1. Source attribution results (estimated proportion of human campylobacteriosis cases attributed to each source for 197 attributable cases) with 95% credible interval according to the proportion-, prevalence- and exposure-based Hald models.

https://doi.org/10.1371/journal.pone.0183790.g001

According to the Dutch exposure-based model, chicken meat was by far the largest contributor to human cases at 65%, with cattle manure coming in second at 19%, chicken manure third at 8.0%, followed by the small contributors of beef meat (3.3%), turkey meat (2.9%), surface water (0.9%), pork meat (0.3%), and swine manure (0.0%) (Fig 2).

Download:

Fig 2. Source attribution results (estimated proportion of human campylobacteriosis cases attributed to each source for 197 attributable cases) with 95% confidence interval according to the proportion-, prevalence- and exposure-based Dutch models.

https://doi.org/10.1371/journal.pone.0183790.g002

The sensitivity analysis undertaken on the exposure-based Dutch model showed that the proportion of human cases attributed to each source varied greatly from changes in the exposure estimates (Ei). Varying the chicken meat exposure value only from 0.01 times its mean value to 100 times changed the percentage of cases attributed to chicken meat from 13% to 94%, with cattle manure losing the most (see Fig 3 for chicken meat). Varying cattle manure exposure values only by these amounts produced the second greatest change in attributed values: 2.4% to 69% with the vast majority of this increase coming from losses in chicken meat. The third greatest change was in chicken manure, changing from 0.8% to 27%. Chicken meat lost the most from this change, with cattle manure changing very little. Turkey meat was next, varying from 0.0% to 18% with most of this coming from chicken meat. The remaining sources varied less than 10%.

Download:

Fig 3. Results of the sensitivity analysis of the exposure-based Dutch model for chicken meat.

https://doi.org/10.1371/journal.pone.0183790.g003

Comparing results from the proportion-, prevalence-, and exposure-based models

All proportion- and prevalence-based Hald model runs converged without any issues and the posterior mean number of cases by subtype closely matched the actual number of cases by less than one case for the majority of the subtypes. The outcomes of the three proportion-, prevalence-, and exposure-based models were different, especially for the latter one (Fig 1). The proportion-based model showed that cattle manure was the primary source of campylobacteriosis (30%), with chicken meat (22%) and chicken manure (13%) coming next, then the meats (beef meat (8.9%), turkey meat (7.9%), and pork meat (7.1%)) and surface water (7.7%) while 2.9% cases were attributed to swine manure. Results for the prevalence-based model indicated that cattle manure was the primary source of campylobacteriosis (47%), with chicken meat (21%), chicken manure (11%) coming next, then turkey meat (6.8%) while 2–5% were attributed to swine manure, pork meat, beef meat and surface water each.

Using the Dutch model, important changes in attribution were also observed between the exposure-, proportion- and prevalence-based models (Fig 2). Results for the proportion-based model indicated that chicken meat, chicken manure and cattle manure were the primary sources of campylobacteriosis, and to a lesser extent turkey, beef and pork meat. Chicken meat (27%) and cattle manure (27%) contributed equally with chicken manure a close third at 18%. The prevalence-based model results indicated cattle manure was the largest contributor at 47%, chicken meat second at 30% and chicken manure third at 10%.

Impact of the Hald model structure on the a and q parameters

Forty q_i parameters (56 subtypes– 12 for which the value was fixed by the data) were estimated by each proportion, prevalence-, and exposure-based Hald model. Their posterior mean values had the same range, between 0.5 and 4, for the three models. The mean values estimated through the proportion-based model were highly correlated with the prevalence-based model (r = 0.96) and exposure-based model (r = 0.72) estimates (S1 Fig). The mean of the posterior values for a_j varied between Hald models with correlation of 0.90 between proportion- and prevalence-based models, of 0.31 between proportion- and exposure-based models, and of 0.12 between prevalence- and exposure-based models (S2 Fig). The sets of posterior mean values of k_ij*a_j showed two distinct and strong correlation patterns between the proportion- and the prevalence-based models (S3 Fig). One correlation was apparent for some k_ij*a_j values between the exposure- and the prevalence-based model, whereas a lack of correlation was shown for the other values (S4 Fig). Finally, the k_ij*a_j mean values changed significantly between the models (S5 Fig). The relative changes were similar within each source between the exposure- and the prevalence-based models with the exception of the chicken meat where a large variation in the k_ij*a_j values was observed. The relative changes in the k_ij*a_j values were high between and within sources with the exception of the chicken meat.

Discussion

This study includes several original and innovative features. It aimed at providing source attribution estimates for human campylobacteriosis at the point of exposure, whereas most studies have been undertaking at the point of production or distribution. It also combined two source attribution approaches: the comparative exposure assessment and the microbial subtyping comparison. This combination of approaches has been suggested [3] but nothing similar has been published to our knowledge. For the microbial subtyping comparison, we use the Comparative Genotype Fingerprinting method to define the subtypes based on the presence or absence of 40 genes [22]. Finally, another strength and novelty was the quality and comprehensiveness of the isolates and data used: the isolates came from a comprehensive sentinel site surveillance system encompassing enhanced surveillance of human clinical cases and active surveillance in food animal, retail meat and surface water within the same geographic area (http://www.phac-aspc.gc.ca/foodnetcanada/index-eng.php). This allowed for a focus on domestically-acquired cases of campylobacteriosis, along with the ability to capture subtypes at the reservoir level (through isolates from chicken, beef and dairy cattle, and swine manure samples), at the vehicle level (through isolates from chicken, turkey, beef and pork retail meat) and from surface water. All the surveillance isolates used for this study came from the same geographic area, thus reducing the likely bias introduced when using disperse geographic origins of isolates [21]. Similarly for time, the effort in source sampling and testing was intense enough that all isolates used came from the same, relatively small period of time. As a result, the isolates and data available for attribution greatly fulfilled the requirement for microbial subtyping comparison [3, 4].

Campylobacteriosis attribution at the point of exposure

According to this study, chicken meat is by far the most important source of human campylobacteriosis (accounting for two-third of the attributable cases) whereas exposure to cattle is the second source (accounting for 14–19% of attributable cases), the other sources (live chicken, live pigs, chicken, turkey, and pork meat, surface water) being minor sources of campylobacteriosis. The particular design and settings of this study limit its direct comparability with other studies on microbial subtype comparison for source attribution undertaken on a larger (usually national) scale and at the point of production or distribution. Nevertheless, their general findings agree with ours about the very first importance of the chicken commodity with bovine source ranking second [8, 14, 17, 21, 34–36]. Our main findings are consistent with the findings of previous Canadian studies on campylobacteriosis attribution, most having used approaches other than subtyping comparison or comparative exposure assessment [37–42].

Importance of proximity to cattle

With its focus on point of exposure, the present study is valuable for detangling the importance of various pathways in bringing Campylobacter form its natural reservoirs to humans. While other studies with a point of production focus found that cattle is the second important reservoir for campylobacteriosis, our findings specify that the transmission from cattle is merely not meat borne but it seems more related to proximity with live animals. In another Canadian study, 25.8% of C. jejuni infection cases were attributed to cattle (fecal samples), second after chicken meat (64.5%) based on Bayesian clustering method that used multilocus subtypes [42]. The authors hypothesized that unpasteurized milk consumption may explain this attribution to cattle as raw milk and unpasteurized dairy products have been identified as the cause of several campylobacteriosis outbreaks (for example [43] or [44] and as a risk factor for sporadic infection [2]. Another hypothesis is the local contamination of water sources and wells in rural area by cattle manure whereby people may be contaminated locally [45–48], in contrast to a wider spread of Campylobacter through the watershed that our water samples would have detected. These results should encourage public health authorities, food safety and water quality stakeholders to target the cattle farms for the prevention of Campylobacter transmission to humans through contact with cattle, consumption of raw milk or well water contaminated or not adequately treated, in addition to continue their efforts to reduce Campylobacter contamination along the poultry supply chain.

Importance of water as a source of Campylobacter

The study deals with surface water as an exposure for humans to Campylobacter considering that this water is used as recreational water at some places within the study area and is the intake of drinking water system for a large part of the population. Our attribution estimates for water (<5%) is lower than the results of a study that attributed 7.4% of cases to water in another region of Canada based on Bayesian clustering method that used multilocus subtypes [42], and than the 9% of campylobacteriosis being waterborne according to a Canadian expert elicitation study [40]. The watershed from which our samples were taken was found regularly contaminated by Campylobacter according to a microbial water contamination study, with some contamination coming from waterfowl [49]. Most CGF profiles found in water were unique to water and few were common to humans or to beef, chicken or pigs manure (Table 3) implying that this water is little impacted by animal production activities or human wastes in comparison with what has been found in other studies [50]. It also implies that if contamination by waterfowl or other wild animals occurs, it does not necessarily lead to many human cases as observed elsewhere [51]. Considering our results attributing relatively more campylobacteriosis cases to cattle than to surface water and the discussion above, we concluded that the attribution to water should include isolates from a more representative sampling of water, notably covering small and individual sources of drinking water.

The value of CGF for source attribution purpose

This study was the second to use the new method of Comparative Genotype Fingerprinting to define subtypes of Campylobacter as the basis of the comparison between human cases and the investigated sources. Using CGF, Deckert et al. showed that urban campylobacteriosis cases were more likely to be infected by Campylobacter strains found in retail chicken compared to rural cases [52]. Previous studies on source attribution based on subtypes comparison have used species, antimicrobial resistance, MLST, PFGE and other typing alone or in combination to define subtypes and derive their source attribution estimates [17, 34, 35, 42, 53–56]. Our study showed that CGF is an alternative method. CGF defined enough subtypes among the clinical isolates and among the source isolates, some being unique but most being shared with at least one other origin, so that microbial source comparison can be performed. We used the 100% similarity threshold across the 40 genes to define CGF subtypes. Two more relaxed thresholds (95% and 90%) were tested on the three Dutch models and the attribution results agreed quite closely across the thresholds indicating some robustness of the attribution results to the CGF subtypes definition. Because the CGF method tests for the presence or absence of 40 genes, it does not provide as detailed genotype information as Multilocus Sequence Typing, hence it is not suitable for attributing cases to source based on detailed phylogenetic differences as performed by the asymmetric island attribution method.

Relevance of the exposure-based model structure

Both the Hald and Dutch models modified with the inclusion of exposure yielded similar results, increasing our confidence in the findings and in the model structure. Obviously, the inclusion of exposure in any model makes a large difference in the attribution estimates, hence it should be highly recommended to include exposure value (E_j) whenever attributing source at the point of exposure is the purpose. Estimating exposure to Campylobacter or any enteric zoonotic pathogens however is a difficult endeavour facing several uncertainties resulting in wide confidence interval for the exposure estimates [11, 12]. Nevertheless, even an imprecise measurement of exposures is worthwhile to enter in any exposure-based model because the large differences in exposure (in terms of a few to several log units) between the sources will still be reflected in the attribution estimates, as shown by our sensitivity analysis. This may differ when modelling attribution at the production or distribution point. The inclusion of food consumption as a parameter was part of the original Hald model [29]; the addition of another parameter, the proportion of food consumed raw or undercooked for each source, was proposed and has been used [33, 57]. Whether to include or not these two parameters in the Dutch or Hald models is not that clear with some studies providing evidence that they are not necessary for the Hald model whereas other studies used one (food consumption) or the two. Noteworthy, the consideration for these two parameters is aligned with the concept of exposure, at least for any foodborne exposure. Actually, the comparative exposure assessment study from which we used some results included food (portion) consumption and the impact of any treatment that may reduce the pathogen (e.g. cooking), among its variables to define exposure. This exposure assessment went further down in including other variables (e.g. pathogen concentration, cross-contamination) to define and quantify exposure and dealt with non-foodborne exposure in a similar, structured and systematic way. We consider the exposure assessment model used as a generalization of the ways for taking into account the relative weight of sources that have been proposed (including food consumption and the proportion of food consumed raw or undercooked) in previous models, going further in detail by including more parameters and being expendable to none foodborne exposure. For example, it allows dealing with exposure to pets or wild birds, two under looked exposures to Campylobacter, something that we could not perform because of lack of isolates from these two sources.

Hald model structure and parameters

We explored the validity of some features of the Hald model. First, the model includes a subtype-specific parameter (q_i) that should be constant independently of the model structure [8, 29]. Our results provided empirical evidence for the validity of such subtype-specific parameter. Second, it was reasoned that the food consumption parameter that was part of the original Hald model formula was not a necessity and could be dropped from the model without impact on attribution estimates [7]. It was explained by the fact that the value of the source-specific parameter would change and absorb the lack of the consumption value in the model. The impact of including or not the food consumption parameter was formally tested in the context of human salmonellosis attribution [33]. This test showed that the attribution estimates were similar with or without food consumption when using the Hald model, but different when using the Dutch model. Our results contrast with this findings since our attribution estimates varied between the proportion-, the prevalence-, and the exposure-based models for both the Dutch and the Hald approaches. Furthermore, our assessment of the k_ij*a_j quantity, that should be similar over the three models if the absorption phenomenon exists, showed that this absorption is plausible between the proportion and the prevalence-based models, but is less likely true for the exposure-based-model. It should be noted that the exposure values we used for the sources varied considerably between the sources, by several units on a log scale, whereas the food consumption values used by Mughini-Gras and Van Pelt were of similar order of magnitude across their four sources (pig, cattle, broilers, layers/eggs): 42.2, 19.2, 17.3, and 13.9 kg/person/year, respectively [33]. The large range among our exposure values might allow for an extreme assessment of the absorption phenomenon compared to the previous study. We conclude that the absorption phenomenon claimed for the Hald model does not hold true for all model structures, and cautions should be exercised when deciding on the model structure.

Study limitations

The study limitations include the inclusion of all isolates independently of the Campylobacter species. This provides a global figure of source attribution of human campylobacteriosis cases (mainly due to C. jejuni infection), masking probable differences between species as demonstrated elsewhere [17, 35, 56]. The number of isolates usable per source investigated was variable. For two sources (beef and pork meat) this number was below 25, the minimum number of isolates for a given source for a source attribution estimation based on the asymetric island model [21], meaning a reduced capacity (statistical power) in attributing cases to these two sources in our study. On the other hand these low numbers reflected the rare presence of Campylobacter in these sources despite intensive sampling. In addition, the study was carried out in a small area of half million people, which precludes the extrapolation to other regions or even nationally because of its specific settings. Other factors that can influence the relative importance of the sources of Campylobacter have not been considered and should be in future works, including the patient’s place of residency (rural vs. urban) [34], the patient’s age, and the season [36]. No isolates from pets or wildlife were available from the study area over the time period. Wildlife, especially wild birds, and pets, especially dogs, have been shown to be source of campylobacteriosis [17, 34, 58–61] and should be considered in future attribution endevours. Like previous work, the study does not consider the influence of person-to-person transmission, either directly (human to human) or indirectly (human to food or water to human). Finally, the study considered exposure but it is recognized that exposure does not systematically lead to infection and disease. Such limitations are general and are applicable to all other campylobacteriosis source attribution efforts.

Conclusion

This study successfully combined comparative exposure assessment and microbial subtype comparison with subtypes defined by comparative genomic fingerprinting to quantify the relative contribution of eight sources of human campylobacteriosis at the point of exposure in a Canadian area that provided high quality data thanks to an integrated surveillance system. It showed that chicken meat is the primary source of human campylobacteriosis, with 65 or 69% (depending on models) of attributable cases assigned to it, followed by live cattle (14 or 19%). Live chicken and pigs, pork, beef, and turkey meat and water are minor sources for human campylobacteriosis. Prevention of human campylobacteriosis should benefit of further studies on the exact contamination pathways between live cattle and humans. Strengthening efforts to improve food safety along the chicken supply chain is critical to reduce the burden of campylobacteriosis in Canada. Attention to the other pathways, particularly contact with animals and water, should be maintained and even reinforced for an overall reduction of the campylobacteriosis burden.

Supporting information

S1 Fig. Correlations between the mean of the posterior q_i subtype specific parameter values (n = 40 subtypes) estimated by the prevalence-based Hald model and those estimated by the proportion- and the exposure-based Hald models.

https://doi.org/10.1371/journal.pone.0183790.s001

(TIF)

S2 Fig. Correlations between the mean of the posterior a_j source specific parameter values (n = 8 sources) estimated by the prevalence-based Hald model and those estimated by the proportion- and the exposure-based Hald models.

https://doi.org/10.1371/journal.pone.0183790.s002

(TIF)

S3 Fig. Correlation of the mean of the posterior values of the k_ija_j values (n = 40 subtypes 8 sources) estimated by the proportion- and the prevalence-based Hald models.

https://doi.org/10.1371/journal.pone.0183790.s003

(TIF)

S4 Fig. Correlation of the mean of the posterior values of the k_ija_j values (n = 40 subtypes 8 sources) estimated by the exposure- and the prevalence-based Hald models.

https://doi.org/10.1371/journal.pone.0183790.s004

(TIF)

S5 Fig. Relative changes by source in the mean of the posterior values of the k_ija_j (n = 40 subtypes 8 sources) estimated by the proportion-, the prevalence-, and the exposure-based Hald models.

https://doi.org/10.1371/journal.pone.0183790.s005

(TIF)

S1 File. Campylobacter CGF data analyzed.

It includes the isolate identification number, the Campylobacter species, the origin of the isolate, the year of the sample collection, the CGF pattern, and the CGF subtype number based on the 100% similarity.

https://doi.org/10.1371/journal.pone.0183790.s006

(XLSX)

Acknowledgments

The authors thank the staff of the Region of Waterloo Public Health for providing the surveillance case data and the staff at the National Microbiology Laboratory at Lethbridge for the laboratory analysis.

References

1. Thomas MK, Murray R, Flockhart L, Pintar K, Pollari F, Fazil A, et al. Estimates of the burden of foodborne illness in Canada for 30 specified pathogens and unspecified agents, circa 2006. Foodborne Pathogens and Disease. 2013;10(7):639–48. pmid:23659355; PubMed Central PMCID: PMCPMC3696931.
- View Article
- PubMed/NCBI
- Google Scholar
2. Domingues AR, Pires SM, Halasa T, Hald T. Source attribution of human campylobacteriosis using a meta-analysis of case-control studies of sporadic infections. Epidemiology and Infection. 2012;140(6):970–81. pmid:22214729
- View Article
- PubMed/NCBI
- Google Scholar
3. European Food Safety Authority. Overview of methods for source attribution for human illness from food-borne microbiological hazards—Scientific Opinion of the Panel on Biological Hazards. EFSA Journal. 2008;6(7). Epub 764.
- View Article
- Google Scholar
4. Pires SM, Evers EG, van Pelt W, Ayers T, Scallan E, Angulo FJ, et al. Attributing the Human Disease Burden of Foodborne Infections to Specific Sources. Foodborne Pathogens and Disease. 2009;6(4):417–24. pmid:19415971
- View Article
- PubMed/NCBI
- Google Scholar
5. Barco L, Barrucci F, Olsen JE, Ricci A. Salmonella source attribution based on microbial subtyping. International Journal of Food Microbiology. 2013;163(2–3):193–203. pmid:23562696
- View Article
- PubMed/NCBI
- Google Scholar
6. David JM, Sanders P, Bemrah N, Granier SA, Denis M, Weill FX, et al. Attribution of the French human Salmonellosis cases to the main food-sources according to the type of surveillance data. Preventive Veterinary Medicine. 2013;110(1):12–27. pmid:23453456
- View Article
- PubMed/NCBI
- Google Scholar
7. Mullner P, Jones G, Noble A, Spencer SEF, Hathaway S, French NP. Source Attribution of Food-Borne Zoonoses in New Zealand: A Modified Hald Model. Risk Analysis. 2009;29(7):970–84. pmid:19486473
- View Article
- PubMed/NCBI
- Google Scholar
8. Boysen L, Rosenquist H, Larsson JT, Nielsen EM, Sorensen G, Nordentoft S, et al. Source attribution of human campylobacteriosis in Denmark. Epidemiology and Infection. 2014;142(8):1599–608. pmid:24168860
- View Article
- PubMed/NCBI
- Google Scholar
9. Mughini-Gras L, Smid JH, Wagenaar JA, de Boer AG, Havelaar AH, Friesema IHM, et al. Risk Factors for Campylobacteriosis of Chicken, Ruminant, and Environmental Origin: A Combined Case-Control and Source Attribution Analysis. PLoS ONE. 2012;7(8). pmid:22880049
- View Article
- PubMed/NCBI
- Google Scholar
10. Guo C, Hoekstra RM, Schroeder CM, Pires SM, Ong KL, Hartnett E, et al. 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 Pathogens and Disease. 2011;8(4):509–16. pmid:21235394
- View Article
- PubMed/NCBI
- Google Scholar
11. Evers E, Van Der Fels-Klerx H, Nauta M, Schijven J, Havelaar A.H. Campylobacter source attribution by exposure assessment. International Journal of Risk Assessment and Management. 2008;8(1/2):174–90.
- View Article
- Google Scholar
12. Pintar K, Thomas K CT, Otten A, Nesbitt A, Marshall B, Pollari P, Hurst M, Ravel A. A Comparative Exposure Assessment of Campylobacter in Ontario, Canada. Risk Analysis. 2016. pmid:27641939
- View Article
- PubMed/NCBI
- Google Scholar
13. Dingle KE, Colles FM, Wareing DR, Ure R, Fox AJ, Bolton FE, et al. Multilocus sequence typing system for Campylobacter jejuni. J. Clin. Microbiol. 2001;39(1):14–23. Epub 2001/01/04. pmid:11136741; PubMed Central PMCID: PMC87672.
- View Article
- PubMed/NCBI
- Google Scholar
14. Mullner P, Spencer SE, Wilson DJ, Jones G, Noble AD, Midwinter AC, et al. Assigning the source of human campylobacteriosis in New Zealand: a comparative genetic and epidemiological approach. Infect Genet Evol. 2009;9(6):1311–9. Epub 2009/09/26. pmid:19778636.
- View Article
- PubMed/NCBI
- Google Scholar
15. Sheppard SK, Dallas JF, Strachan NJC, MacRae M, McCarthy ND, Wilson DJ, et al. Campylobacter Genotyping to Determine the Source of Human Infection. Clinical Infectious Diseases. 2009;48(8):1072–8. pmid:19275496
- View Article
- PubMed/NCBI
- Google Scholar
16. Strachan NJC, Gormley FJ, Rotariu O, Ogden ID, Miller G, Dunn GM, et al. Attribution of Campylobacter Infections in Northeast Scotland to Specific Sources by Use of Multilocus Sequence Typing. J Infect Dis. 2009;199(8):1205–8. pmid:19265482
- View Article
- PubMed/NCBI
- Google Scholar
17. Kittl S, Heckel G, Korczak BM, Kuhnert P. Source attribution of human Campylobacter isolates by MLST and fla-typing and association of genotypes with quinolone resistance. PLoS ONE. 2013;8(11). http://dx.doi.org/10.1371/journal.pone.0081796. pmid:20143053537.
- View Article
- PubMed/NCBI
- Google Scholar
18. Mughini-Gras L, Smid JH, Wagenaar JA, de Boer AG, Havelaar AH, Friesema IHM, et al. Risk factors for campylobacteriosis of chicken, ruminant, and environmental origin: a combined case-control and source attribution analysis. PLoS ONE. 2012;7(8):e42599–e. pmid:22880049.
- View Article
- PubMed/NCBI
- Google Scholar
19. Wilson DJ, Gabriel E, Leatherbarrow AJH, Cheesbrough J, Gee S, Bolton E, et al. Tracing the Source of Campylobacteriosis. Plos Genet. 2008;4(9). pmid:18818764
- View Article
- PubMed/NCBI
- Google Scholar
20. Mossong J, Mughini-Gras L, Penny C, Devaux A, Olinger C, Losch S, et al. Human Campylobacteriosis in Luxembourg, 2010–2013: A Case-Control Study Combined with Multilocus Sequence Typing for Source Attribution and Risk Factor Analysis. Sci Rep-Uk. 2016;6. pmid:26860258
- View Article
- PubMed/NCBI
- Google Scholar
21. Smid JH, Mughini-Gras L, de Boer AG, French NP, Havelaar AH, Wagenaar JA, et al. Practicalities of using non-local or non-recent multilocus sequence typing data for source attribution in space and time of human campylobacteriosis. PLoS ONE. 2013;8(2):e55029. Epub 2013/02/14. pmid:23405107; PubMed Central PMCID: PMC3566096.
- View Article
- PubMed/NCBI
- Google Scholar
22. Taboada EN, Ross SL, Mutschall SK, MacKinnon JM, Roberts MJ, Buchanan CJ, et al. Development and Validation of a Comparative Genomic Fingerprinting Method for High-Resolution Genotyping of Campylobacter jejuni. Journal of Clinical Microbiology. 2012;50(3):788–97. pmid:22170908
- View Article
- PubMed/NCBI
- Google Scholar
23. Taboada EN, Clark CG, Sproston EL, Carrillo CD. Current methods for molecular typing of Campylobacter species. J Microbiol Meth. 2013;95(1):24–31. pmid:23871858
- View Article
- PubMed/NCBI
- Google Scholar
24. Anonymous. Region of Waterloo Food Flow Analysis Study Guelph, ON: Harry Cummings and Associates inc., 2005 November 2005. Avaliable from http://chd.region.waterloo.on.ca/en/researchResourcesPublications/resources/FoodFlow_Analysis.pdf
25. Ravel A, Pintar K, Nesbitt A, Pollari F. Non food-related risk factors of campylobacteriosis in Canada: a matched case-control study. BMC Public Health. 2016;16(1):1016. pmid:27677338; PubMed Central PMCID: PMCPMC5039884.
- View Article
- PubMed/NCBI
- Google Scholar
26. FoodNet Canada. 2010 Annual Report. Public Health of Canada. Guelph, ON. 2014. p. 102.
27. Rosef O, Kapperud G, Lauwers S, Gondrosen B. Serotyping of Campylobacter-Jejuni, Campylobacter-Coli, and Campylobacter-Laridis from Domestic and Wild Animals. Applied and Environmental Microbiology. 1985;49(6):1507–10. pmid:4015088
- View Article
- PubMed/NCBI
- Google Scholar
28. Van Pelt W, van de Giessen A, van Leeuwen W, Wannet W, Henken A, Evers E, et al. Oorsprong, omvang en kosten van humane salmonellose. Deel 1. Oorsprong van human salmonellose met betrekking tot varken, rund, kip, ei en overige bronnen. Infectieziekten Bulletin. 1999;10(2):240–3.
- View Article
- Google Scholar
29. Hald T, Vose D, Wegener HC, Koupeev T. A Bayesian approach to quantify the contribution of animal-food sources to human salmonellosis. Risk Analysis. 2004;24(1):255–69. pmid:15028016
- View Article
- PubMed/NCBI
- Google Scholar
30. David JM, Guillemot D, Bemrah N, Thebault A, Brisabois A, Chemaly M, et al. The Bayesian Microbial Subtyping Attribution Model: Robustness to Prior Information and a Proposition. Risk Analysis. 2013;33(3):397–408. pmid:22882110
- View Article
- PubMed/NCBI
- Google Scholar
31. David JM, Ravel A, Nesbitt A, Pintar K, Pollari F. Assessing multiple foodborne, waterborne and environmental exposures of healthy people to potential enteric pathogen sources: effect of age, gender, season, and recall period. Epidemiology and infection. 2014;142(1):28–39. pmid:23731678.
- View Article
- PubMed/NCBI
- Google Scholar
32. Wahlstrom H, Andersson Y, Plym-Forshell L, Pires SM. Source attribution of human Salmonella cases in Sweden. Epidemiology and Infection. 2011;139(8):1246–53. pmid:20943003
- View Article
- PubMed/NCBI
- Google Scholar
33. Mughini-Gras L, van Pelt W. Salmonella source attribution based on microbial subtyping: Does including data on food consumption matter? International Journal of Food Microbiology. 2014;191:109–15. pmid:25261828
- View Article
- PubMed/NCBI
- Google Scholar
34. Strachan NJ, Gormley FJ, Rotariu O, Ogden ID, Miller G, Dunn GM, et al. Attribution of Campylobacter infections in northeast Scotland to specific sources by use of multilocus sequence typing. J Infect Dis. 2009;199(8):1205–8. Epub 2009/03/07. pmid:19265482; PubMed Central PMCID: PMC3985119.
- View Article
- PubMed/NCBI
- Google Scholar
35. Mughini-Gras L, Smid JH, Wagenaar JA, de Boer AG, Havelaar AH, Friesema IH, et al. Risk factors for campylobacteriosis of chicken, ruminant, and environmental origin: a combined case-control and source attribution analysis. PLoS ONE. 2012;7(8):e42599. Epub 2012/08/11. pmid:22880049; PubMed Central PMCID: PMC3411806.
- View Article
- PubMed/NCBI
- Google Scholar
36. Bessell PR, Rotariu O, Innocent GT, Smith-Palmer A, Strachan NJ, Forbes KJ, et al. Using sequence data to identify alternative routes and risk of infection: a case-study of campylobacter in Scotland. BMC Infect Dis. 2012;12:80. pmid:22462563; PubMed Central PMCID: PMCPMC3340322.
- View Article
- PubMed/NCBI
- Google Scholar
37. Ravel A, Greig J, Tinga C, Todd E, Campbell G, Cassidy M, et al. Exploring historical Canadian foodborne outbreak data sets for human illness attribution. J Food Prot. 2009;72(9):1963–76. pmid:19777901.
- View Article
- PubMed/NCBI
- Google Scholar
38. Ravel A, Davidson VJ, Ruzante JM, Fazil A. Foodborne proportion of gastrointestinal illness: estimates from a Canadian expert elicitation survey. Foodborne Pathog Dis. 2010;7(12):1463–72. pmid:20704505.
- View Article
- PubMed/NCBI
- Google Scholar
39. Davidson VJ, Ravel A, Nguyen TN, Fazil A, Ruzante JM. Food-Specific Attribution of Selected Gastrointestinal Illnesses: Estimates from a Canadian Expert Elicitation Survey. Foodborne Pathogens and Disease. 2011;8(9):983–95. pmid:21561379
- View Article
- PubMed/NCBI
- Google Scholar
40. Butler AJ, Thomas MK, Pintar KDM. Expert Elicitation as a Means to Attribute 28 Enteric Pathogens to Foodborne, Waterborne, Animal Contact, and Person-to-Person Transmission Routes in Canada. Foodborne Pathogens and Disease. 2015;12(4):335–44. pmid:25835810
- View Article
- PubMed/NCBI
- Google Scholar
41. Butler AJ, Pintar KDM, Thomas MK. Estimating the Relative Role of Various Subcategories of Food, Water, and Animal Contact Transmission of 28 Enteric Diseases in Canada. Foodborne Pathogens and Disease. 2016;13(2):57–64. pmid:26863428
- View Article
- PubMed/NCBI
- Google Scholar
42. Levesque S, Fournier E, Carrier N, Frost E, Arbeit RD, Michaud S. Campylobacteriosis in Urban versus Rural Areas: A Case-Case Study Integrated with Molecular Typing to Validate Risk Factors and to Attribute Sources of Infection. PLoS ONE. 2013;8(12). pmid:24386265
- View Article
- PubMed/NCBI
- Google Scholar
43. Hauri AM, Just M, McFarland S, Schweigmann A, Schlez K, Krahn J. Campylobacteriosis outbreaks in the state of Hesse, Germany, 2005–2011: raw milk yet again. Deut Med Wochenschr. 2013;138(8):357–61. pmid:23404322
- View Article
- PubMed/NCBI
- Google Scholar
44. Longenberger AH, Palumbo AJ, Chu AK, Moll ME, Weltman A, Ostroff SM. Campylobacter jejuni Infections Associated With Unpasteurized Milk-Multiple States, 2012. Clinical Infectious Diseases. 2013;57(2):263–6. pmid:23575199
- View Article
- PubMed/NCBI
- Google Scholar
45. Kemp R, Leatherbarrow AJH, Williams LJ, Hart CA, Clough HE, Turner J, et al. Prevalence and genetic diversity of Campylobacter spp. in environmental water samples from a 100-square-kilometer predominantly dairy farming area. Applied and Environmental Microbiology. 2005;71(4):1876–82. pmid:15812015
- View Article
- PubMed/NCBI
- Google Scholar
46. Murphy HM, Thomas MK, Schmidt PJ, Medeiros DT, McFadyen S, Pintar KDM. Estimating the burden of acute gastrointestinal illness due to Giardia, Cryptosporidium, Campylobacter, E. coli O157 and norovirus associated with private wells and small water systems in Canada. Epidemiology and Infection. 2016;144(7):1355–70. pmid:26564479
- View Article
- PubMed/NCBI
- Google Scholar
47. Cha W, Henderson T, Collins J, Manning SD. Factors associated with increasing campylobacteriosis incidence in Michigan, 2004–2013. Epidemiology and Infection. 2016;144(15):3316–25. pmid:27488877
- View Article
- PubMed/NCBI
- Google Scholar
48. Galanis E, Mak S, Otterstatter M, Taylor M, Zubel M, Takaro TK, et al. The association between campylobacteriosis, agriculture and drinking water: a case-case study in a region of British Columbia, Canada, 2005–2009. Epidemiology and Infection. 2014;142(10):2075–84. pmid:24892423
- View Article
- PubMed/NCBI
- Google Scholar
49. Van Dyke MI, Morton VK, McLellan NL, Huck PM. The occurrence of Campylobacter in river water and waterfowl within a watershed in southern Ontario, Canada. Journal of Applied Microbiology. 2010;109(3):1053–66. pmid:20408936
- View Article
- PubMed/NCBI
- Google Scholar
50. Mughini-Gras L, Penny C, Ragimbeau C, Schets FM, Blaak H, Duirn B, et al. Quantifying potential sources of surface water contamination with Campylobacter jejuni and Campylobacter coli. Water Res. 2016;101:36–45. pmid:27244295
- View Article
- PubMed/NCBI
- Google Scholar
51. Carter PE, McTavish SM, Brooks HJL, Campbell D, Collins-Emerson JM, Midwinter AC, et al. Novel Clonal Complexes with an Unknown Animal Reservoir Dominate Campylobacter jejuni Isolates from River Water in New Zealand. Applied and Environmental Microbiology. 2009;75(19):6038–46. pmid:19648367
- View Article
- PubMed/NCBI
- Google Scholar
52. Deckert AE, Taboada E, Mutschall S, Poljak Z, Reid-Smith RJ, Tamblyn S, et al. Molecular Epidemiology of Campylobacter jejuni Human and Chicken Isolates from Two Health Units. Foodborne Pathogens and Disease. 2014;11(2):150–5. pmid:24219173
- View Article
- PubMed/NCBI
- Google Scholar
53. Mullner P, Shadbolt T, Collins-Emerson JM, Midwinter AC, Spencer SEF, Marshall J, et al. Molecular and spatial epidemiology of human campylobacteriosis: source association and genotype-related risk factors. Epidemiol Infect. 2010;138(10):1372–83. pmid:20141645
- View Article
- PubMed/NCBI
- Google Scholar
54. Mullner P, Collins-Emerson JM, Midwinter AC, Carter P, Spencer SE, van der Logt P, et al. Molecular epidemiology of Campylobacter jejuni in a geographically isolated country with a uniquely structured poultry industry. Appl Environ Microbiol. 2010;76(7):2145–54. Epub 2010/02/16. pmid:20154115; PubMed Central PMCID: PMC2849254.
- View Article
- PubMed/NCBI
- Google Scholar
55. Gilpin BJ, Walsh G, On SL, Smith D, Marshall JC, French NP. Application of molecular epidemiology to understanding campylobacteriosis in the Canterbury region of New Zealand. Epidemiology and Infection. 2013;141(6):1253–66. pmid:22906314
- View Article
- PubMed/NCBI
- Google Scholar
56. Roux F, Sproston E, Rotariu O, Macrae M, Sheppard SK, Bessell P, et al. Elucidating the aetiology of human Campylobacter coli infections. PLoS ONE. 2013;8(5):e64504. Epub 2013/06/05. pmid:23734204; PubMed Central PMCID: PMC3667194.
- View Article
- PubMed/NCBI
- Google Scholar
57. Mughini-Gras L, Smid J, Enserink R, Franz E, Schouls L, Heck M, et al. Tracing the sources of human salmonellosis: A multi-model comparison of phenotyping and genotyping methods. Infection Genetics and Evolution. 2014;28:251–60. pmid:25315490
- View Article
- PubMed/NCBI
- Google Scholar
58. Mughini-Gras L, Smid JH, Wagenaar JA, Koene MG, Havelaar AH, Friesema IH, et al. 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. 2013;141(12):2526–35. Epub 2013/03/01. pmid:23445833.
- View Article
- PubMed/NCBI
- Google Scholar
59. Cody AJ, McCarthy ND, Bray JE, Wimalarathna HML, Colles FM, van Rensburg MJJ, et al. Wild bird-associated Campylobacter jejuni isolates are a consistent source of human disease, in Oxfordshire, United Kingdom. Env Microbiol Rep. 2015;7(5):782–8. pmid:26109474
- View Article
- PubMed/NCBI
- Google Scholar
60. Di Giannatale E, Garofolo G, Alessiani A, Di Donato G, Candeloro L, Vencia W, et al. Tracing Back Clinical Campylobacter jejuni in the Northwest of Italy and Assessing Their Potential Source. Front Microbiol. 2016;7. ARTN 887 pmid:27379033
- View Article
- PubMed/NCBI
- Google Scholar
61. Griekspoor P, Engvall EO, Akerlind B, Olsen B, Waldenstrom J. Genetic diversity and host associations in Campylobacter jejuni from human cases and broilers in 2000 and 2008. Vet Microbiol. 2015;178(1–2):94–8. pmid:25960333
- View Article
- PubMed/NCBI
- Google Scholar

[ref1] 1. Thomas MK, Murray R, Flockhart L, Pintar K, Pollari F, Fazil A, et al. Estimates of the burden of foodborne illness in Canada for 30 specified pathogens and unspecified agents, circa 2006. Foodborne Pathogens and Disease. 2013;10(7):639–48. pmid:23659355; PubMed Central PMCID: PMCPMC3696931.
View Article
PubMed/NCBI
Google Scholar

[2] View Article

[3] PubMed/NCBI

[4] Google Scholar

[ref2] 2. Domingues AR, Pires SM, Halasa T, Hald T. Source attribution of human campylobacteriosis using a meta-analysis of case-control studies of sporadic infections. Epidemiology and Infection. 2012;140(6):970–81. pmid:22214729
View Article
PubMed/NCBI
Google Scholar

[6] View Article

[7] PubMed/NCBI

[8] Google Scholar

[ref3] 3. European Food Safety Authority. Overview of methods for source attribution for human illness from food-borne microbiological hazards—Scientific Opinion of the Panel on Biological Hazards. EFSA Journal. 2008;6(7). Epub 764.
View Article
Google Scholar

[10] View Article

[11] Google Scholar

[ref4] 4. Pires SM, Evers EG, van Pelt W, Ayers T, Scallan E, Angulo FJ, et al. Attributing the Human Disease Burden of Foodborne Infections to Specific Sources. Foodborne Pathogens and Disease. 2009;6(4):417–24. pmid:19415971
View Article
PubMed/NCBI
Google Scholar

[13] View Article

[14] PubMed/NCBI

[15] Google Scholar

[ref5] 5. Barco L, Barrucci F, Olsen JE, Ricci A. Salmonella source attribution based on microbial subtyping. International Journal of Food Microbiology. 2013;163(2–3):193–203. pmid:23562696
View Article
PubMed/NCBI
Google Scholar

[17] View Article

[18] PubMed/NCBI

[19] Google Scholar

[ref6] 6. David JM, Sanders P, Bemrah N, Granier SA, Denis M, Weill FX, et al. Attribution of the French human Salmonellosis cases to the main food-sources according to the type of surveillance data. Preventive Veterinary Medicine. 2013;110(1):12–27. pmid:23453456
View Article
PubMed/NCBI
Google Scholar

[21] View Article

[22] PubMed/NCBI

[23] Google Scholar

[ref7] 7. Mullner P, Jones G, Noble A, Spencer SEF, Hathaway S, French NP. Source Attribution of Food-Borne Zoonoses in New Zealand: A Modified Hald Model. Risk Analysis. 2009;29(7):970–84. pmid:19486473
View Article
PubMed/NCBI
Google Scholar

[25] View Article

[26] PubMed/NCBI

[27] Google Scholar

[ref8] 8. Boysen L, Rosenquist H, Larsson JT, Nielsen EM, Sorensen G, Nordentoft S, et al. Source attribution of human campylobacteriosis in Denmark. Epidemiology and Infection. 2014;142(8):1599–608. pmid:24168860
View Article
PubMed/NCBI
Google Scholar

[29] View Article

[30] PubMed/NCBI

[31] Google Scholar

[ref9] 9. Mughini-Gras L, Smid JH, Wagenaar JA, de Boer AG, Havelaar AH, Friesema IHM, et al. Risk Factors for Campylobacteriosis of Chicken, Ruminant, and Environmental Origin: A Combined Case-Control and Source Attribution Analysis. PLoS ONE. 2012;7(8). pmid:22880049
View Article
PubMed/NCBI
Google Scholar

[33] View Article

[34] PubMed/NCBI

[35] Google Scholar

[ref10] 10. Guo C, Hoekstra RM, Schroeder CM, Pires SM, Ong KL, Hartnett E, et al. 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 Pathogens and Disease. 2011;8(4):509–16. pmid:21235394
View Article
PubMed/NCBI
Google Scholar

[37] View Article

[38] PubMed/NCBI

[39] Google Scholar

[ref11] 11. Evers E, Van Der Fels-Klerx H, Nauta M, Schijven J, Havelaar A.H. Campylobacter source attribution by exposure assessment. International Journal of Risk Assessment and Management. 2008;8(1/2):174–90.
View Article
Google Scholar

[41] View Article

[42] Google Scholar

[ref12] 12. Pintar K, Thomas K CT, Otten A, Nesbitt A, Marshall B, Pollari P, Hurst M, Ravel A. A Comparative Exposure Assessment of Campylobacter in Ontario, Canada. Risk Analysis. 2016. pmid:27641939
View Article
PubMed/NCBI
Google Scholar

[44] View Article

[45] PubMed/NCBI

[46] Google Scholar

[ref13] 13. Dingle KE, Colles FM, Wareing DR, Ure R, Fox AJ, Bolton FE, et al. Multilocus sequence typing system for Campylobacter jejuni. J. Clin. Microbiol. 2001;39(1):14–23. Epub 2001/01/04. pmid:11136741; PubMed Central PMCID: PMC87672.
View Article
PubMed/NCBI
Google Scholar

[48] View Article

[49] PubMed/NCBI

[50] Google Scholar

[ref14] 14. Mullner P, Spencer SE, Wilson DJ, Jones G, Noble AD, Midwinter AC, et al. Assigning the source of human campylobacteriosis in New Zealand: a comparative genetic and epidemiological approach. Infect Genet Evol. 2009;9(6):1311–9. Epub 2009/09/26. pmid:19778636.
View Article
PubMed/NCBI
Google Scholar

[52] View Article

[53] PubMed/NCBI

[54] Google Scholar

[ref15] 15. Sheppard SK, Dallas JF, Strachan NJC, MacRae M, McCarthy ND, Wilson DJ, et al. Campylobacter Genotyping to Determine the Source of Human Infection. Clinical Infectious Diseases. 2009;48(8):1072–8. pmid:19275496
View Article
PubMed/NCBI
Google Scholar

[56] View Article

[57] PubMed/NCBI

[58] Google Scholar

[ref16] 16. Strachan NJC, Gormley FJ, Rotariu O, Ogden ID, Miller G, Dunn GM, et al. Attribution of Campylobacter Infections in Northeast Scotland to Specific Sources by Use of Multilocus Sequence Typing. J Infect Dis. 2009;199(8):1205–8. pmid:19265482
View Article
PubMed/NCBI
Google Scholar

[60] View Article

[61] PubMed/NCBI

[62] Google Scholar

[ref17] 17. Kittl S, Heckel G, Korczak BM, Kuhnert P. Source attribution of human Campylobacter isolates by MLST and fla-typing and association of genotypes with quinolone resistance. PLoS ONE. 2013;8(11). http://dx.doi.org/10.1371/journal.pone.0081796. pmid:20143053537.
View Article
PubMed/NCBI
Google Scholar

[64] View Article

[65] PubMed/NCBI

[66] Google Scholar

[ref18] 18. Mughini-Gras L, Smid JH, Wagenaar JA, de Boer AG, Havelaar AH, Friesema IHM, et al. Risk factors for campylobacteriosis of chicken, ruminant, and environmental origin: a combined case-control and source attribution analysis. PLoS ONE. 2012;7(8):e42599–e. pmid:22880049.
View Article
PubMed/NCBI
Google Scholar

[68] View Article

[69] PubMed/NCBI

[70] Google Scholar

[ref19] 19. Wilson DJ, Gabriel E, Leatherbarrow AJH, Cheesbrough J, Gee S, Bolton E, et al. Tracing the Source of Campylobacteriosis. Plos Genet. 2008;4(9). pmid:18818764
View Article
PubMed/NCBI
Google Scholar

[72] View Article

[73] PubMed/NCBI

[74] Google Scholar

[ref20] 20. Mossong J, Mughini-Gras L, Penny C, Devaux A, Olinger C, Losch S, et al. Human Campylobacteriosis in Luxembourg, 2010–2013: A Case-Control Study Combined with Multilocus Sequence Typing for Source Attribution and Risk Factor Analysis. Sci Rep-Uk. 2016;6. pmid:26860258
View Article
PubMed/NCBI
Google Scholar

[76] View Article

[77] PubMed/NCBI

[78] Google Scholar

[ref21] 21. Smid JH, Mughini-Gras L, de Boer AG, French NP, Havelaar AH, Wagenaar JA, et al. Practicalities of using non-local or non-recent multilocus sequence typing data for source attribution in space and time of human campylobacteriosis. PLoS ONE. 2013;8(2):e55029. Epub 2013/02/14. pmid:23405107; PubMed Central PMCID: PMC3566096.
View Article
PubMed/NCBI
Google Scholar

[80] View Article

[81] PubMed/NCBI

[82] Google Scholar

[ref22] 22. Taboada EN, Ross SL, Mutschall SK, MacKinnon JM, Roberts MJ, Buchanan CJ, et al. Development and Validation of a Comparative Genomic Fingerprinting Method for High-Resolution Genotyping of Campylobacter jejuni. Journal of Clinical Microbiology. 2012;50(3):788–97. pmid:22170908
View Article
PubMed/NCBI
Google Scholar

[84] View Article

[85] PubMed/NCBI

[86] Google Scholar

[ref23] 23. Taboada EN, Clark CG, Sproston EL, Carrillo CD. Current methods for molecular typing of Campylobacter species. J Microbiol Meth. 2013;95(1):24–31. pmid:23871858
View Article
PubMed/NCBI
Google Scholar

[88] View Article

[89] PubMed/NCBI

[90] Google Scholar

[ref24] 24. Anonymous. Region of Waterloo Food Flow Analysis Study Guelph, ON: Harry Cummings and Associates inc., 2005 November 2005. Avaliable from http://chd.region.waterloo.on.ca/en/researchResourcesPublications/resources/FoodFlow_Analysis.pdf

[ref25] 25. Ravel A, Pintar K, Nesbitt A, Pollari F. Non food-related risk factors of campylobacteriosis in Canada: a matched case-control study. BMC Public Health. 2016;16(1):1016. pmid:27677338; PubMed Central PMCID: PMCPMC5039884.
View Article
PubMed/NCBI
Google Scholar

[93] View Article

[94] PubMed/NCBI

[95] Google Scholar

[ref26] 26. FoodNet Canada. 2010 Annual Report. Public Health of Canada. Guelph, ON. 2014. p. 102.

[ref27] 27. Rosef O, Kapperud G, Lauwers S, Gondrosen B. Serotyping of Campylobacter-Jejuni, Campylobacter-Coli, and Campylobacter-Laridis from Domestic and Wild Animals. Applied and Environmental Microbiology. 1985;49(6):1507–10. pmid:4015088
View Article
PubMed/NCBI
Google Scholar

[98] View Article

[99] PubMed/NCBI

[100] Google Scholar

[ref28] 28. Van Pelt W, van de Giessen A, van Leeuwen W, Wannet W, Henken A, Evers E, et al. Oorsprong, omvang en kosten van humane salmonellose. Deel 1. Oorsprong van human salmonellose met betrekking tot varken, rund, kip, ei en overige bronnen. Infectieziekten Bulletin. 1999;10(2):240–3.
View Article
Google Scholar

[102] View Article

[103] Google Scholar

[ref29] 29. Hald T, Vose D, Wegener HC, Koupeev T. A Bayesian approach to quantify the contribution of animal-food sources to human salmonellosis. Risk Analysis. 2004;24(1):255–69. pmid:15028016
View Article
PubMed/NCBI
Google Scholar

[105] View Article

[106] PubMed/NCBI

[107] Google Scholar

[ref30] 30. David JM, Guillemot D, Bemrah N, Thebault A, Brisabois A, Chemaly M, et al. The Bayesian Microbial Subtyping Attribution Model: Robustness to Prior Information and a Proposition. Risk Analysis. 2013;33(3):397–408. pmid:22882110
View Article
PubMed/NCBI
Google Scholar

[109] View Article

[110] PubMed/NCBI

[111] Google Scholar

[ref31] 31. David JM, Ravel A, Nesbitt A, Pintar K, Pollari F. Assessing multiple foodborne, waterborne and environmental exposures of healthy people to potential enteric pathogen sources: effect of age, gender, season, and recall period. Epidemiology and infection. 2014;142(1):28–39. pmid:23731678.
View Article
PubMed/NCBI
Google Scholar

[113] View Article

[114] PubMed/NCBI

[115] Google Scholar

[ref32] 32. Wahlstrom H, Andersson Y, Plym-Forshell L, Pires SM. Source attribution of human Salmonella cases in Sweden. Epidemiology and Infection. 2011;139(8):1246–53. pmid:20943003
View Article
PubMed/NCBI
Google Scholar

[117] View Article

[118] PubMed/NCBI

[119] Google Scholar

[ref33] 33. Mughini-Gras L, van Pelt W. Salmonella source attribution based on microbial subtyping: Does including data on food consumption matter? International Journal of Food Microbiology. 2014;191:109–15. pmid:25261828
View Article
PubMed/NCBI
Google Scholar

[121] View Article

[122] PubMed/NCBI

[123] Google Scholar

[ref34] 34. Strachan NJ, Gormley FJ, Rotariu O, Ogden ID, Miller G, Dunn GM, et al. Attribution of Campylobacter infections in northeast Scotland to specific sources by use of multilocus sequence typing. J Infect Dis. 2009;199(8):1205–8. Epub 2009/03/07. pmid:19265482; PubMed Central PMCID: PMC3985119.
View Article
PubMed/NCBI
Google Scholar

[125] View Article

[126] PubMed/NCBI

[127] Google Scholar

[ref35] 35. Mughini-Gras L, Smid JH, Wagenaar JA, de Boer AG, Havelaar AH, Friesema IH, et al. Risk factors for campylobacteriosis of chicken, ruminant, and environmental origin: a combined case-control and source attribution analysis. PLoS ONE. 2012;7(8):e42599. Epub 2012/08/11. pmid:22880049; PubMed Central PMCID: PMC3411806.
View Article
PubMed/NCBI
Google Scholar

[129] View Article

[130] PubMed/NCBI

[131] Google Scholar

[ref36] 36. Bessell PR, Rotariu O, Innocent GT, Smith-Palmer A, Strachan NJ, Forbes KJ, et al. Using sequence data to identify alternative routes and risk of infection: a case-study of campylobacter in Scotland. BMC Infect Dis. 2012;12:80. pmid:22462563; PubMed Central PMCID: PMCPMC3340322.
View Article
PubMed/NCBI
Google Scholar

[133] View Article

[134] PubMed/NCBI

[135] Google Scholar

[ref37] 37. Ravel A, Greig J, Tinga C, Todd E, Campbell G, Cassidy M, et al. Exploring historical Canadian foodborne outbreak data sets for human illness attribution. J Food Prot. 2009;72(9):1963–76. pmid:19777901.
View Article
PubMed/NCBI
Google Scholar

[137] View Article

[138] PubMed/NCBI

[139] Google Scholar

[ref38] 38. Ravel A, Davidson VJ, Ruzante JM, Fazil A. Foodborne proportion of gastrointestinal illness: estimates from a Canadian expert elicitation survey. Foodborne Pathog Dis. 2010;7(12):1463–72. pmid:20704505.
View Article
PubMed/NCBI
Google Scholar

[141] View Article

[142] PubMed/NCBI

[143] Google Scholar

[ref39] 39. Davidson VJ, Ravel A, Nguyen TN, Fazil A, Ruzante JM. Food-Specific Attribution of Selected Gastrointestinal Illnesses: Estimates from a Canadian Expert Elicitation Survey. Foodborne Pathogens and Disease. 2011;8(9):983–95. pmid:21561379
View Article
PubMed/NCBI
Google Scholar

[145] View Article

[146] PubMed/NCBI

[147] Google Scholar

[ref40] 40. Butler AJ, Thomas MK, Pintar KDM. Expert Elicitation as a Means to Attribute 28 Enteric Pathogens to Foodborne, Waterborne, Animal Contact, and Person-to-Person Transmission Routes in Canada. Foodborne Pathogens and Disease. 2015;12(4):335–44. pmid:25835810
View Article
PubMed/NCBI
Google Scholar

[149] View Article

[150] PubMed/NCBI

[151] Google Scholar

[ref41] 41. Butler AJ, Pintar KDM, Thomas MK. Estimating the Relative Role of Various Subcategories of Food, Water, and Animal Contact Transmission of 28 Enteric Diseases in Canada. Foodborne Pathogens and Disease. 2016;13(2):57–64. pmid:26863428
View Article
PubMed/NCBI
Google Scholar

[153] View Article

[154] PubMed/NCBI

[155] Google Scholar

[ref42] 42. Levesque S, Fournier E, Carrier N, Frost E, Arbeit RD, Michaud S. Campylobacteriosis in Urban versus Rural Areas: A Case-Case Study Integrated with Molecular Typing to Validate Risk Factors and to Attribute Sources of Infection. PLoS ONE. 2013;8(12). pmid:24386265
View Article
PubMed/NCBI
Google Scholar

[157] View Article

[158] PubMed/NCBI

[159] Google Scholar

[ref43] 43. Hauri AM, Just M, McFarland S, Schweigmann A, Schlez K, Krahn J. Campylobacteriosis outbreaks in the state of Hesse, Germany, 2005–2011: raw milk yet again. Deut Med Wochenschr. 2013;138(8):357–61. pmid:23404322
View Article
PubMed/NCBI
Google Scholar

[161] View Article

[162] PubMed/NCBI

[163] Google Scholar

[ref44] 44. Longenberger AH, Palumbo AJ, Chu AK, Moll ME, Weltman A, Ostroff SM. Campylobacter jejuni Infections Associated With Unpasteurized Milk-Multiple States, 2012. Clinical Infectious Diseases. 2013;57(2):263–6. pmid:23575199
View Article
PubMed/NCBI
Google Scholar

[165] View Article

[166] PubMed/NCBI

[167] Google Scholar

[ref45] 45. Kemp R, Leatherbarrow AJH, Williams LJ, Hart CA, Clough HE, Turner J, et al. Prevalence and genetic diversity of Campylobacter spp. in environmental water samples from a 100-square-kilometer predominantly dairy farming area. Applied and Environmental Microbiology. 2005;71(4):1876–82. pmid:15812015
View Article
PubMed/NCBI
Google Scholar

[169] View Article

[170] PubMed/NCBI

[171] Google Scholar

[ref46] 46. Murphy HM, Thomas MK, Schmidt PJ, Medeiros DT, McFadyen S, Pintar KDM. Estimating the burden of acute gastrointestinal illness due to Giardia, Cryptosporidium, Campylobacter, E. coli O157 and norovirus associated with private wells and small water systems in Canada. Epidemiology and Infection. 2016;144(7):1355–70. pmid:26564479
View Article
PubMed/NCBI
Google Scholar

[173] View Article

[174] PubMed/NCBI

[175] Google Scholar

[ref47] 47. Cha W, Henderson T, Collins J, Manning SD. Factors associated with increasing campylobacteriosis incidence in Michigan, 2004–2013. Epidemiology and Infection. 2016;144(15):3316–25. pmid:27488877
View Article
PubMed/NCBI
Google Scholar

[177] View Article

[178] PubMed/NCBI

[179] Google Scholar

[ref48] 48. Galanis E, Mak S, Otterstatter M, Taylor M, Zubel M, Takaro TK, et al. The association between campylobacteriosis, agriculture and drinking water: a case-case study in a region of British Columbia, Canada, 2005–2009. Epidemiology and Infection. 2014;142(10):2075–84. pmid:24892423
View Article
PubMed/NCBI
Google Scholar

[181] View Article

[182] PubMed/NCBI

[183] Google Scholar

[ref49] 49. Van Dyke MI, Morton VK, McLellan NL, Huck PM. The occurrence of Campylobacter in river water and waterfowl within a watershed in southern Ontario, Canada. Journal of Applied Microbiology. 2010;109(3):1053–66. pmid:20408936
View Article
PubMed/NCBI
Google Scholar

[185] View Article

[186] PubMed/NCBI

[187] Google Scholar

[ref50] 50. Mughini-Gras L, Penny C, Ragimbeau C, Schets FM, Blaak H, Duirn B, et al. Quantifying potential sources of surface water contamination with Campylobacter jejuni and Campylobacter coli. Water Res. 2016;101:36–45. pmid:27244295
View Article
PubMed/NCBI
Google Scholar

[189] View Article

[190] PubMed/NCBI

[191] Google Scholar

[ref51] 51. Carter PE, McTavish SM, Brooks HJL, Campbell D, Collins-Emerson JM, Midwinter AC, et al. Novel Clonal Complexes with an Unknown Animal Reservoir Dominate Campylobacter jejuni Isolates from River Water in New Zealand. Applied and Environmental Microbiology. 2009;75(19):6038–46. pmid:19648367
View Article
PubMed/NCBI
Google Scholar

[193] View Article

[194] PubMed/NCBI

[195] Google Scholar

[ref52] 52. Deckert AE, Taboada E, Mutschall S, Poljak Z, Reid-Smith RJ, Tamblyn S, et al. Molecular Epidemiology of Campylobacter jejuni Human and Chicken Isolates from Two Health Units. Foodborne Pathogens and Disease. 2014;11(2):150–5. pmid:24219173
View Article
PubMed/NCBI
Google Scholar

[197] View Article

[198] PubMed/NCBI

[199] Google Scholar

[ref53] 53. Mullner P, Shadbolt T, Collins-Emerson JM, Midwinter AC, Spencer SEF, Marshall J, et al. Molecular and spatial epidemiology of human campylobacteriosis: source association and genotype-related risk factors. Epidemiol Infect. 2010;138(10):1372–83. pmid:20141645
View Article
PubMed/NCBI
Google Scholar

[201] View Article

[202] PubMed/NCBI

[203] Google Scholar

[ref54] 54. Mullner P, Collins-Emerson JM, Midwinter AC, Carter P, Spencer SE, van der Logt P, et al. Molecular epidemiology of Campylobacter jejuni in a geographically isolated country with a uniquely structured poultry industry. Appl Environ Microbiol. 2010;76(7):2145–54. Epub 2010/02/16. pmid:20154115; PubMed Central PMCID: PMC2849254.
View Article
PubMed/NCBI
Google Scholar

[205] View Article

[206] PubMed/NCBI

[207] Google Scholar

[ref55] 55. Gilpin BJ, Walsh G, On SL, Smith D, Marshall JC, French NP. Application of molecular epidemiology to understanding campylobacteriosis in the Canterbury region of New Zealand. Epidemiology and Infection. 2013;141(6):1253–66. pmid:22906314
View Article
PubMed/NCBI
Google Scholar

[209] View Article

[210] PubMed/NCBI

[211] Google Scholar

[ref56] 56. Roux F, Sproston E, Rotariu O, Macrae M, Sheppard SK, Bessell P, et al. Elucidating the aetiology of human Campylobacter coli infections. PLoS ONE. 2013;8(5):e64504. Epub 2013/06/05. pmid:23734204; PubMed Central PMCID: PMC3667194.
View Article
PubMed/NCBI
Google Scholar

[213] View Article

[214] PubMed/NCBI

[215] Google Scholar

[ref57] 57. Mughini-Gras L, Smid J, Enserink R, Franz E, Schouls L, Heck M, et al. Tracing the sources of human salmonellosis: A multi-model comparison of phenotyping and genotyping methods. Infection Genetics and Evolution. 2014;28:251–60. pmid:25315490
View Article
PubMed/NCBI
Google Scholar

[217] View Article

[218] PubMed/NCBI

[219] Google Scholar

[ref58] 58. Mughini-Gras L, Smid JH, Wagenaar JA, Koene MG, Havelaar AH, Friesema IH, et al. 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. 2013;141(12):2526–35. Epub 2013/03/01. pmid:23445833.
View Article
PubMed/NCBI
Google Scholar

[221] View Article

[222] PubMed/NCBI

[223] Google Scholar

[ref59] 59. Cody AJ, McCarthy ND, Bray JE, Wimalarathna HML, Colles FM, van Rensburg MJJ, et al. Wild bird-associated Campylobacter jejuni isolates are a consistent source of human disease, in Oxfordshire, United Kingdom. Env Microbiol Rep. 2015;7(5):782–8. pmid:26109474
View Article
PubMed/NCBI
Google Scholar

[225] View Article

[226] PubMed/NCBI

[227] Google Scholar

[ref60] 60. Di Giannatale E, Garofolo G, Alessiani A, Di Donato G, Candeloro L, Vencia W, et al. Tracing Back Clinical Campylobacter jejuni in the Northwest of Italy and Assessing Their Potential Source. Front Microbiol. 2016;7. ARTN 887 pmid:27379033
View Article
PubMed/NCBI
Google Scholar

[229] View Article

[230] PubMed/NCBI

[231] Google Scholar

[ref61] 61. Griekspoor P, Engvall EO, Akerlind B, Olsen B, Waldenstrom J. Genetic diversity and host associations in Campylobacter jejuni from human cases and broilers in 2000 and 2008. Vet Microbiol. 2015;178(1–2):94–8. pmid:25960333
View Article
PubMed/NCBI
Google Scholar

[233] View Article

[234] PubMed/NCBI

[235] Google Scholar

Figures

Abstract

Introduction

Material & methods

Sources of isolates

CGF analysis

Data analysis

Source attribution estimations

Proportional similarity index.

Dutch model.

Hald model.

Defining the occurrence kij.

Results

Description of samples, isolates and subtypes

Attribution at the point of exposure

Comparing results from the proportion-, prevalence-, and exposure-based models

Impact of the Hald model structure on the a and q parameters

Discussion

Campylobacteriosis attribution at the point of exposure

Importance of proximity to cattle

Importance of water as a source of Campylobacter

The value of CGF for source attribution purpose

Relevance of the exposure-based model structure

Hald model structure and parameters

Study limitations

Conclusion

Supporting information

S1 Fig. Correlations between the mean of the posterior qi subtype specific parameter values (n = 40 subtypes) estimated by the prevalence-based Hald model and those estimated by the proportion- and the exposure-based Hald models.

S2 Fig. Correlations between the mean of the posterior aj source specific parameter values (n = 8 sources) estimated by the prevalence-based Hald model and those estimated by the proportion- and the exposure-based Hald models.

S3 Fig. Correlation of the mean of the posterior values of the kij*aj values (n = 40 subtypes * 8 sources) estimated by the proportion- and the prevalence-based Hald models.

S4 Fig. Correlation of the mean of the posterior values of the kij*aj values (n = 40 subtypes * 8 sources) estimated by the exposure- and the prevalence-based Hald models.

S5 Fig. Relative changes by source in the mean of the posterior values of the kij*aj (n = 40 subtypes * 8 sources) estimated by the proportion-, the prevalence-, and the exposure-based Hald models.

S1 File. Campylobacter CGF data analyzed.

Acknowledgments

References

Defining the occurrence k_ij.

S1 Fig. Correlations between the mean of the posterior q_i subtype specific parameter values (n = 40 subtypes) estimated by the prevalence-based Hald model and those estimated by the proportion- and the exposure-based Hald models.

S2 Fig. Correlations between the mean of the posterior a_j source specific parameter values (n = 8 sources) estimated by the prevalence-based Hald model and those estimated by the proportion- and the exposure-based Hald models.

S3 Fig. Correlation of the mean of the posterior values of the k_ija_j values (n = 40 subtypes 8 sources) estimated by the proportion- and the prevalence-based Hald models.

S4 Fig. Correlation of the mean of the posterior values of the k_ija_j values (n = 40 subtypes 8 sources) estimated by the exposure- and the prevalence-based Hald models.

S5 Fig. Relative changes by source in the mean of the posterior values of the k_ija_j (n = 40 subtypes 8 sources) estimated by the proportion-, the prevalence-, and the exposure-based Hald models.