Accepted Manuscript Title: Farm specific risk factors for Campylobacter colonisation in Danish and Norwegian broilers Author: B. Borck Høg H.M. Sommer L.S. Larsen A.I.V. Sørensen B. David M. Hofshagen H. Rosenquist PII: DOI: Reference:
S0167-5877(16)30107-6 http://dx.doi.org/doi:10.1016/j.prevetmed.2016.04.002 PREVET 4012
To appear in:
PREVET
Received date: Revised date: Accepted date:
23-4-2015 1-4-2016 4-4-2016
Please cite this article as: Hog, B.Borck, Sommer, H.M., Larsen, L.S., Sorensen, A.I.V., David, B., Hofshagen, M., Rosenquist, H., Farm specific risk factors for Campylobacter colonisation in Danish and Norwegian broilers.Preventive Veterinary Medicine http://dx.doi.org/10.1016/j.prevetmed.2016.04.002 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Farm specific risk factors for Campylobacter colonisation in Danish and Norwegian broilers
B. Borck Høga*), H.M. Sommera,b, L.S. Larsen1a, A.I.V. Sørensena, B. Davidc, M. Hofshagenc, H. Rosenquista
a
Risk Assessment and Nutrition, The National Food Institute, Technical University of
Denmark, Mørkhøj Bygade 19, DK-2860 Søborg, Denmark
b
Data Analysis and Statistics, DTU Compute, Technical University of Denmark,
building 324, DK-2800 Kgs. Lyngby, Denmark c
Norwegian Veterinary Institute, Pb 750 Sentrum, N-0106 Oslo, Norway
*) Corresponding author: Birgitte Borck Høg, DVM, Ph.D. The National Food Institute, Technical University of Denmark Mørkhøj Bygade 19 DK-2860 Søborg Denmark Tel: +45 35887066 Fax: +45 35887028 E-mail: [email protected]
1
Present address: Statens Serum Institut, 5 Artillerivej, 2300 Copenhagen S, Denmark
1
Highlights
Cross country survey of risk factors for Campylobacter colonization of broilers.
Common and country specific risk factors were identified.
Risk factors were generally related to breaches in biosecurity.
The broiler flock prevalence differed between countries (Denmark>Norway).
2
Abstract Campylobacteriosis has become the leading bacterial zoonosis in humans in the European Union and other developed countries. There are many sources of human Campylobacter infections, but broilers and broiler meat have been shown to be the most important. In order to implement effective interventions that reduce the probability of Campylobacter colonisation of broiler flocks, it is essential to fully understand the risk factors involved. We present a bi-national risk factor survey comprising Campylobacter data from more than 5,200 Danish and Norwegian indoor, conventional broiler flocks and the responses to a standardised questionnaire, with more than 40 explanatory variables from 277 Danish and Norwegian farms. We explored several models by using different combinations of the Danish and Norwegian data, including models with singlecountry datasets. All models were analysed using a generalized linear model using backwards elimination and forward selection. The results show that Norwegian broiler flocks had a lower risk of being colonised than Danish flocks. Farm specific variables that increased the risk of flocks becoming colonised with Campylobacter in both countries were: broiler houses older than five years; longer downtime (no. of days between flocks), probably a consequence of longer downtimes being associated with less focus on maintaining a high biosecurity level; broiler houses without a separate ante-room or barrier; and the use of the drinker nipples with cups or bells compared with nipples without cups. Additional country specific risk factors were also identified. For Norway, the risk of colonisation increased with increasing numbers of houses on a farm and when the water used for the broilers originated from surface water or bore holes instead
3
of mains. For Denmark, having boot dips or low stocking density increased the risk of a flock becoming Campylobacter positive. The different model approaches allowed us to explore the effect of having a large number of data available to identify the significant variables. To a large extent, the country specific models identified risk factors that were also found in the bi-national model. However, the bi-national model identified more risk factors than the country specific models. This indicated that combining the data sets from the two countries did not disrupt the results but was beneficial due to the greater strength achieved in the statistical analyses and the possibility of examining interactions terms with the variable Country.
Key words: Campylobacter, poultry, broiler flocks, risk factors, generalized linear model
4
1. Introduction For many years, campylobacteriosis has been the most common cause of foodborne gastrointestinal illness in the EU and the rest of the industrialized world (WHO, 2012; EFSA and ECDC, 2015). Broilers and broiler meat are the most important single source of infection in humans even though other sources have also been identified (Wingstrand et al., 2006; EFSA, 2010; Mughini-Gras et al., 2012; Muellner et al., 2013; Boysen et al., 2014). Therefore, reduction of Campylobacter in the broiler production is considered to be an important step towards reducing the number of human infections. Risk factors associated with the introduction of Campylobacter in broiler flocks have been studied in several countries since the 1990s. Lack of employment of hygiene measures and strict hygienic routines are very important risk factors for colonisation of broiler flocks with Campylobacter (Berndtson et al., 1996; van de Giessen et al., 1996; Hald et al., 2000; Adkin et al., 2006; Lyngstad et al., 2008). Other risk factors have also been identified, but many of these are somehow related to compromised biosecurity, or increased pressure on biosecurity due to increased load of Campylobacter in the farm environment. Factors that may increase the load of Campylobacter in the environment are e.g. presence of other animals on the farm or in close proximity of the farm, more than one broiler house on the farm (Refregier-Petton et al., 2001, Bouwknegt et al., 2004, Guerin et al., 2007a, McDowell et al., 2008; Chowdhury et al., 2012), and climate-related factors - especially temperature (Patrick et al., 2004, Guerin et al., 2008, McDowell et al., 2008; Rushton et al., 2009, Jore et al., 2010, Jonsson et al., 2012). Hald et al. (2007) and Bahrndorff et al. (2013) found that the climate related factors may partly be explained by flies transmitting Campylobacter into the broiler houses. Examples of factors that may
5
influence (compromise) biosecurity are; the number of persons tending the broilers (Refregier-Petton et al., 2001), thinning broiler flocks (Berndtson et al., 1996; Hald et al., 2000; Allen et al., 2008), presence of rodents at farms (McDowell et al., 2008; Sommer et al., 2013,), and increased ventilation in the summer (Newell and Fearnley, 2003). Other identified risk factors include age of the broilers at slaughter (Berndtson et al., 1996, Bouwknegt et al., 2004, Barrios et al., 2006, Chowdhury et al., 2012, Sommer et al., 2013), using drinkers with trays (Näther et al., 2009), flock size (Berndtson et al., 1996, Barrios et al., 2006, Guerin et al., 2007a), and hatchery (Bouwknegt el al., 2004). Many risk factor studies concerning Campylobacter in broiler flocks are based on data from a limited number of houses monitored for one or a few production cycles. Furthermore, the studies have typically included data from just one country. The studies have identified different risk factors and the findings may in some instances seem contradictory. However, some differences may simply be due to data limitations (limited number of variables/number of observations), which has reduced the power of the tests for significance. The conventional indoor broiler productions in Norway and Denmark share many common characteristics. For example, quality assurance programmes ensuring a high level of biosecurity on the farms have been implemented in both countries. Generally, all farms apply an all-in-all-out system with an “empty period” between flocks where the litter is removed and the houses cleaned. However, there are also a number of factors that differ between the countries, e.g. the farms in Norway generally have fewer houses per farm, (Høg et al., 2011). At the time of the study, approximately 100 million broilers were raised per year in Denmark. There were approximately 170 indoor conventional broiler farms
6
and the majority (>98%) of broilers were produced for two major broiler companies. In Norway, at the time of the study, approximately 60 million broilers were raised annually on approximately 600 farms and the majority of the farmers produced broilers for one of the three main companies. Together, these companies produced more than 98% of the Norwegian broilers at four different slaughterhouses. This study included Campylobacter surveillance results from more than 5,200 indoor, conventional broiler flocks, collected over a two-year period in two Scandinavian countries. The surveillance data were analysed together with the outcomes of a questionnaire survey answered by 277 Danish and Norwegian farms. By analysing data from both countries together we aimed a) to achieve a stronger basis for the statistical analyses by utilizing larger datasets and b) identify common and/or country specific risk factors for the occurrence of Campylobacter in Danish and Norwegian broiler flocks.
2. Material and methods 2.1. Farm specific variables For the survey, a questionnaire including 43 questions concerning production, farm management procedures, farm conditions etc. was developed. The questionnaire can be found in full in Høg et al. (2011). The questionnaire was translated into Danish and Norwegian and the survey took place between December 2010 and June 2011. This study aimed to identify risk factors for Campylobacter colonisation of conventional indoor broiler flocks and we aimed to obtain information from approximately 200 farms from each country. We used the official registers of animal
7
holdings in both Denmark and Norway to select farms for the study. In Denmark there were less than 200 conventional indoor broiler farms and thus questionnaires were sent out to all the farms. In Norway, farms were selected so that all 19 counties in were represented. However, due to unforeseen circumstances, farmers in one county of Norway did not receive questionnaires, (Høg et al., 2011). All questionnaires were distributed and returned by postal services and filled out by the broiler farmers themselves. Overall, questionnaires were returned by 119 and 183 farmers in Denmark and Norway, respectively, corresponding to response rates of 58% and 59% in the two countries (Høg et al., 2011). The no. of responding farms represented approximately 70% of all Danish indoor conventional broiler farms (ca. 170 farms) and 30% of all indoor conventional broiler farms in Norway (ca. 600 farms). The detailed descriptive outcomes of the questionnaire survey are presented in Høg et al., (2011) In Denmark and Norway the poultry farms and slaughterhouses are not integrated, meaning that farms are not owned by the broiler companies, but by the farmers themselves. However, the broilers are produced in accordance with delivery contracts made with the broiler companies. In both Denmark and Norway, the broiler industries have agreed upon and implemented guidelines and quality assurance systems according to which all indoor conventional broilers are produced; KIK (the Quality Assurance System in Danish Broiler Production) in Denmark, and KSL (Agricultural Quality System) in Norway. All of the farms included in this study, were indoor conventional broiler farms with a mean of seven crops produced annually. The mean downtime between crops was 8 days Denmark and 19 days in Norway.
8
2.2.Campylobacter status Both in Denmark and Norway, Campylobacter data were obtained through national surveillance programmes. In Denmark, Campylobacter data were obtained during the period from January 2010 to April 2012 for flocks on farms that had responded to the questionnaire. The flocks were tested for Campylobacter by analysing one pair of boot swabs collected in each broiler houses 7-10 days prior to slaughter using PCR as described by Lund et al., (2004). In Norway, Campylobacter data from farms that had responded to the questionnaire were obtained for the period May to October for the years 2010 and 2011. The reason for restricting sampling in Norway to the summer months was that the historical data from the winter months in the period 2001-2007 showed a very low Campylobacter prevalence. Campylobacter status was obtained using swab samples, 10 swabs from each flock collected at the farm as close to slaughter as possible (max. four days before slaughter). The laboratory method for analysing the swab samples was the same as was used in Denmark (Lund et al. 2004).
2.3.Data analysis 2.3.1. Data preparation Prior to any data analyses, the dataset was checked to ensure that only conventional farms were included, and data from farms with too few observations were excluded.
9
Data were explored and for practical reasons, some of the farm specific variables were modified to carry out the analyses. The raw data set consisted of a large number of explanatory variables and parameters, as well as a relatively large number of data records with one or more missing values, which made it impossible to run a model. Thus, the number of parameters had to be reduced. The result of the data preparation and the final input for statistical analysis is shown in Table A and B in the supplementary material. The final data set included data from 277 broiler farms (104 Danish and 173 Norwegian farms).
2.3.2. Statistical analysis Multivariable analysis was used to identify risk factors and the analyses were carried out using SAS (version 9.4, SAS Institute Inc.). The prevalence (pr) was the response variable and was defined as the proportion of Campylobacter positive flocks out of the total number of broiler flocks produced on a given broiler farm in the entire study period. The data from the two-year study period were merged since we did not expect a significant trend over just two years. Thus, the model used in the analysis was a non-hierarchical model and the unit of analysis was the mean prevalence per farm. Since the response variable was not normally distributed, a generalized linear model was applied and the response-variable was transformed using a logit link function. logit where
is the intercept,
⋯
∙
⋯
is the parameter for the variable i, i = 1,2,3,…,44 (numbers
referring to the question number given in Table A and B), the indexes j and k refers to
10
the categorical levels, and
∙
is the interaction term between variable i and
the variable Country. Not all variables could be included from the start of the analysis, and the interaction terms with Country could not be included in the model until the model had been reduced to consist only of variables with p-values < 0.10. The statistical procedure of backward elimination and forward selection described by Sommer et al. (2013) was applied. In brief, we wanted to apply a method that made use of as much information from the entire dataset as possible. This meant that we did not exclude variables with missing values (unless the inclusion meant that we could not run the model due to too many missing values). We allowed observations with missing values to enter into the model if 1) the variable for which the observation had a missing value was removed from the model in a backward elimination steps and 2) if the p-value was less than 0.10 in a forward selection step. As the model was reduced in number of variables, the data included in the model increased due to exclusion of variables with missing values. In the initial steps of analysis, the main model included 32 variables. The variables with the largest number of missing values were left out of the initial model, which was based on 153 out of 276 farms. The p-value for several variables changed drastically at some model reduction steps of the analysis. The majority of these changes was instigated by the ‘new’ data included in the analyses, and only for a few steps, was the change caused by collinearity with the variables removed from the model. For each model, p-values, effect-estimates, standard errors, odds ratios (OR), and 95% confidence intervals (CI) for the OR estimates, were estimated. The OR values for a risk factor
also included in an interaction term with
(a continuous variable) was
calculated as the following when comparing e.g. level 1 with 3
11
, ,
exp
∗
∗
where the value for the continuous variable is the mean value or a chosen fixed value. The main model included data from the time period, where both Danish and Norwegian data were available, i.e. the summer period from May to October in 2010 and 2011. However, due to the difference in the available Campylobacter data, i.e. all year in Denmark/ May-October in Norway, alternative model A) was explored: a model including all Danish data (January 2010 - April 2012 (all year)) plus all Norwegian data from the summer period (May-October in 2010 and 2011). The Danish and Norwegian data were also analysed separately: alternative model B) a model with only Norwegian data; and alternative model C) a model with Danish data only. The ‘goodness of fit’ was examined by use of the deviance residuals.
3. Results The significant variables in each of the explored generalized linear models (main model, alternative model A, B, and C) are shown in Table 1 and 2. The results from the main model were based on data from 3,083 flocks from 256 farms out of 276. In the main model the most important variable was Country. It was highly significant (p < 0.0001) and had the largest estimated effect (3.20) and Denmark had the highest risk estimate of the two countries. The estimated OR was also high, however since Country is also part of the interaction term (House * Country) the OR varies with the number of houses, in this case the OR was estimated at 1.5 house per farm. The prevalence of Campylobacter in broiler flocks tested from May to October in 2010 and 2011 was 21% in Denmark and 4% in Norway. Another very important variable was the
12
Downtime (p=0.001). In general, the longer the downtime, the higher the risk of the broiler flocks becoming colonised by Campylobacter. The variable Age of house was also important (p=0.002). The newest broiler houses had the lowest risk estimate. Broiler houses between age 6-15 years and age >15 years were not significantly different from each other (p=0.06). Furthermore, having anterooms and barriers in each broiler house (Anteroom/barrier) significantly reduced the risk (p=0.003), and farms using drinker systems (Drinkers) with nipples without cups had a significantly lower risk estimate (p=0.006). Nipples with cups and bells were not significantly different from each other (p=0.07). By using re-parameterisation we found that the number of houses only was significant for the Norwegian data. This was found by examining the p-values for the interaction term No. of houses x Country. The re-parameterisation was carried out by removing the variable No. of houses from the final model and then re-running the model. This way, all the variation that was explained by the main effect. No. of houses was now explained by the interaction term together with the ‘pure’ interaction effect. The results from the three alternative models are shown in Table 2. Model A, including all Danish and Norwegian data was based on data from 4,744 flocks from 253 farms out of 277. This model identified the same significant variables as the main model plus two more variables: Stocking density *Country and Boot dips. Interestingly, having boot dips increased the risk of a flock becoming Campylobacter positive. Furthermore, a high stocking density reduced the risk for the Danish data but not for the Norwegian data. In model B (Danish data only with 3,430 flocks, 93 farms out of 104 farms), Boot dip and Stocking density were also significant (p=0.025 and 0.026,
13
respectively). No. of houses, however, was not significant, which makes sense since this variable was only important for the Norwegian data (model A). When analysing Norwegian data alone (model C), only three variables became significant. Two variables were the same as identified in the other models (Age of house and Downtime) and one was new; Water source (p=0.020). Farms using mains as their water source had the lowest risk, second lowest were bore holes and those with the highest risk were farms using surface water. Model C was based on data from 1,342 flocks from 164 out of 173 Norwegian farms. The effect of Downtime was further explored in an attempt to understand how a longer Downtime could increase the risk of Campylobacter colonisation. We found that among the farms with a downtime of more than 30 days, 25% responded that they did not have a program for cleaning and disinfecting the houses between the flocks. In contrast, on farms with downtimes of less than 30 days, only 11% stated that they had no program for cleaning and disinfecting the houses between flocks, see Table 3. Furthermore, the number of farms with a high frequency of rodent control (1-2 times/month) dropped from 20% to 13% for farms with a downtime of more than 20 days. Some variables that were not found statistically significant, but had pvalues less than 0.10, are worth mentioning. In the main model the Stocking density was nearly significant (p=0.06) and so was Other animals on farm (p=0.07), where cattle and pigs had an increased estimate compared to the other animals. In model A, Other animals on farm was also nearly significant (p=0.06) as well as Closed ventilation sites (p=0.07) with a decreased risk when closed. Moreover, Fly screens (p=0.07) and Rodent control by company (p=0.08) were nearly significant indicating a reduced risk if
14
houses were screened or if rodent control was maintained by the farmer rather than by a company. The dataset consisting of the largest number of data identified the highest number of variables. For all models the deviance residuals had values less than 2, indicating that there is no overdispersion and that there is no (severe) lack of fit.
4. Discussion Risk factors Country. The variable ‘Country’ was included in the analysis to capture potential differences between the countries that were important in relation to describing the variation in the dataset and that could not be explained by the other variables in the analysis. This variable (Country) describes - in one variable - all the unknown factors that may significantly influence the prevalence and that are unequally distributed in the two countries and therefore do not even out. The overall prevalence of Campylobacter in broilers, as observed through the national surveillance programmes, has consistently been lower in Norway than in Denmark for many years (Anon, 2014a; Anon., 2014b; EFSA and ECDC 2015), and the current study also found that broilers raised in Denmark had a significantly higher risk of being Campylobacter positive, compared to broilers raised in Norway. The fact that Country is significant implies that there is more variation between the two countries than can be explained by the rest of the variables in the model. An important factor that may contribute to this difference in risk is the different climatic conditions in the two
15
countries (Jore et al., 2010, Jonsson et al., 2012). Other factors could be differences in action plans to prevent Campylobacter colonisation of broiler flocks (Hofshagen et al., 2005; Rosenquist et al., 2009), but also the difference in the structure of the production e.g. the age of the chickens at slaughter or factors such as distance to farms with chicken, cattle or livestock may result in a higher the risk of Campylobacter. Downtime. Another important risk factor in this study was the length of downtime between flocks. We found that a long downtime was associated with a higher risk of the flocks becoming colonised with Campylobacter. At first glance this was surprising, since a long downtime, in theory, should allow the broiler house to dry completely, creating the least favourable conditions for Campylobacter survival in the broiler house environment. However, it may be explained by other factors that were characteristic for the farms with long downtimes and could reflect that these farms had a less intensive production and less focus on maintaining a high level of biosecurity and hygiene. The variables Cleaning between flocks and Rodent control frequency were tested in the model both as main effects and as interaction terms with Downtime but were found to be insignificant. One theory could be that none of the two variables were strong enough alone, but together through a third variable (Downtime) significance was seen. There may be other factors characteristic for the farms with long downtimes that play a role in Campylobacter colonisation that we were unable to identify in this study Results of three other Scandinavian risk factor studies have also indicated that the length of downtime affected the risk. However, in contrast to our results these studies indicated that a period of less than 9-14 d between depopulation and restocking was associated with an increased risk of Campylobacter colonisation (Berndtson et al.,
16
1996; Hald et al., 2000; Lyngstad et al., 2008). Downtime appears to be an important risk factor, but should not be considered out of context. Thus, the downtime should be sufficiently long to allow the broiler houses to be efficiently emptied, cleaned and disinfected. However, if the downtime is too long the risk of introducing Campylobacter into the houses increases, especially if the biosecurity and hygiene level on the farm is not optimal. Age of house. The age of the broiler house also significantly influenced the risk of the broilers becoming colonised with Campylobacter. Farms with newer houses had a significantly lower risk of becoming colonised with Campylobacter than farms with houses older than five years. However, we could not show significant differences between the two categories with the oldest broiler houses. The effect of the broiler house age was also reported by Sommer et al. (2013). The importance of the age of the broiler house is likely connected to the fact that older houses may have more difficulty conforming to the modern biosecurity standards that have proven important. It may be speculated that older houses are more difficult to seal towards the surrounding environment and it is therefore easier for Campylobacter to be introduced to the inside environment via flies, rodents and birds that may all carry Campylobacter (Hald et al., 2007; Huneau-Salaun, 2007; McDowell et al., 2008). Anteroom/barriers. The significance of hygiene measures in the prevention of Campylobacter colonisation of broiler flocks has been well documented (Berndtson et al., 1996, van de Giessen et al., 1996, Hald et al., 2000; EFSA, 2010) and is supported by the outcome of the present study, where the presence of anterooms and barriers in each house decreased the risk of Campylobacter colonisation. In Denmark and Norway, the anterooms are usually small rooms, divided into a clean and a dirty
17
zone, through which all staff enter the broiler house. Typically a low barrier separates the dirty and clean zone. All clothes and footwear that has had any contact with the environment outside the broiler house stays in the dirty zone and coveralls and boots for use in the broiler house stay in the clean zone. Many more biosecurity and hygiene variables were included in the questionnaire, but could not be proven significant. However, at the time we conducted the survey, a high level of biosecurity, in concordance with the applied quality assurance schemes, was already employed on most farms and may be one reason that only few of the biosecurity variables could be shown to have an effect. Drinkers. The type of drinkers was shown to affect the occurrence of Campylobacter in both Danish and Norwegian broiler flocks; drinker systems with nipples without cups were better than nipples with cups and bells. Drinker systems with cups and bells both have reservoirs of standing water. Campylobacter may be introduced to these reservoirs via colonised birds or their droppings, whereby the standing water can become a dissemination vehicle, transmitting Campylobacter to other birds when they drink the water. Thus, this study indicates that it was better to use drinkers without a water reservoir for Campylobacter to survive in, and this is in concordance with the work of Näther and co-workers (2009). No. of houses. Increased number of broiler houses on the farm was a risk factor for Norway. The number of houses on a farm has previously been associated with an increased occurrence of Campylobacter on farms (Refregier-Petton et al., 2001, Bouwknegt et al., 2004, Guerin et al., 2007a, McDowell et al., 2008; Sommer et al., 2013).
18
Boot dips. Having boot dips at the entrance to the broiler house was identified as a risk factor but only in the alternative models A and B, in which all the Danish data were included. In order for boot dips to work efficiently they need to be maintained properly, and the disinfectant has to be replenished at appropriate intervals, otherwise they may contribute to an increased risk rather than act as a protective factor (Berndtson et al., 1996, Guerin et al., 2007b, McDowell et al., 2008, Gibbens et al., 2001). Stocking density. We found a reduced risk associated with a high stocking density in Denmark, but not in Norway. Other studies that have looked at the risk associated with flock size, have indicated an increased risk with increasing size of flocks (Barrios et al., 2006, Guerin et al, 2007a). However, perhaps the farms in Denmark with the highest stocking density were also farms with strictest management routines and a highest level of biosecurity. Water source. Using surface water for drinking increased the risk compared to private boreholes, wells and mains. Using mains was the best. Surface water was only used in Norway, and water source was found significant only in model C (Norwegian data only) and was not highly significant. Only 6 out of 164 farms used surface water, so conclusions should be made very cautiously as to the effect of using surface water. An extra model was examined where surface water was merged with the category private bore holes and wells. Here water source was still significant though with a higher p-value (0.047) and mains were still the best. Water source and it’s interaction with Country was, however, not significant in the main model. Nonetheless, using surface water should not be neglected as a possible source for Campylobacter in domestic animals.
19
Other factors. The significance of ventilating has been demonstrated in other studies (Newell and Fearnley, 2003; Guerin et al., 2007a; Rushton et al., 2009). Our questionnaire included several questions concerning ventilation, none of which were found significant. However, some of the variables that related specifically to the broiler houses suffered somewhat from the design of the study, where questions were asked at the farm level rather than at the house level. In this study 35% of the participating farms had two or more broiler houses and for these farms, important details concerning the house specific variables were lost due to the design of the questionnaire. The significance of flies in the transmission of Campylobacter and the effect of using fly screens has been demonstrated by Hald and co-workers (Hald et al., 2007; Bahrndorff et al., 2013). Unfortunately, only three of the participating farms had installed fly screens on the broiler houses (all Danish farms) and therefore any results should be interpreted very carefully. Fly screens was not significant in the main model, but close to significant in alternative model A, perhaps due to the greater statistical power. Thinning was not included in this study, since the available Campylobacter data did not allow us to calculate prevalence estimates before and after thinning of the broiler flocks.
Models As described above, four models based on different datasets were explored. When evaluating and comparing the different models it is important to look at the power of the test statistics and the frequency of bias results. The first is evaluated by examining the number of data included in the models. Alternative model A, based on 4,744 flocks, seemed to have greater statistical power than the main model, based on
20
3,083 flocks. However, the estimate of Country in model A was biased due to the different sampling periods in the two countries in combination with a severe seasonality which was not accounted for, because the response variable was an aggregated prevalence estimate for a whole year. Apart from the biased estimate of Country, model A identified the same variables as the main model plus two extra. The additional variables likely appear as a consequence of the larger number of degrees of freedom in model A, compared with the main model. Model A identified the same six variables as the main model, and the estimates were within the same ranges, which indicates that the significant variables were only affected little or not at all by the different sampling periods for Denmark in the two datasets. This suggests that the significant variables are not driven by the season but more or less have the same influence on the prevalence all year around. The purpose of exploring both the main and the alternative models was to investigate the benefit of combining the Danish and the Norwegian data sets and thereby increase the statistical power of the analyses. We wanted to investigate which and how many risk factors could be identified given the larger joint dataset from both countries, compared with the smaller single-country datasets. In the joint models (main model + model A) the interactions terms with Country were included in the models, when possible, to allow for country specific risk factors. Furthermore, we wanted to investigate if expanding the Danish data, to also include the winter period, would drive the results of the model in another direction. In other words, we wanted to see whether the effect of the significant variables depended on the sampling period (summer/winter). The reason for not running a model with all data (Denmark: whole year + Norway: summer), where variables were split in two (one for the summer and one for winter),
21
was that it would result in twice as many parameters to estimate rendering the model not to run. One may argue that model B, compared to the main model, is the better choice for Denmark since this model is based on 347 more flocks. The number of farms in model B was, however, only 93 compared with 256 for the main model, and more than doubling the number of farms is attractive since it includes farms with other combinations of the variables and thus makes the design more complete. This is also the case when comparing model C with the main model. Model C, based on 164 Norwegian farms, resulted in three significant variables, whereas seven significant variables were identified in the main model containing data from 256 farms. If the risk factors identified in the main model were not relevant for Norway, this would have been expressed in the interaction term with Country in the main model and model A. All risk factors (except one) identified in model B and C were also identified in the main model and model A. All models in this study were non-hierarchical since the unit of analysis were the mean prevalence per farm. One could have expected the unit of the samples to be measurements per broiler house per farm which would have resulted in a hierarchical model structure. The houses would then be nested with the farms and houses from the same farm would most likely be correlated in some way which could be accounted for by applying an appropriate variance structure. However, such a model was not applied since we did not have information on house level for all data in the study. Even if we had information on house levels the results would probably not change much since for half of the dataset (Norway) most farms only had one broiler house.
22
Questionnaires. A number of lessons concerning questionnaires were learned. Designing a questionnaire well suited for a bi-national country survey requires extensive knowledge of the production in the participating countries. In the questionnaire designed for this survey, a number of responses were given for each question. This solution was chosen in an attempt to make it quick and easy for the respondent to answer the question, but also in order ensure homogenous inputs to the data. For the same reasons, very little free text was permitted. Nonetheless, a number of pitfalls were identified when data were returned, particularly in questions concerning ventilation systems, which are highly house specific, and therefore not well suited for this farm-level questionnaire. Also, a pilot survey to identify common pitfalls was not conducted in this survey, and could very likely have improved the questionnaire survey, had it been carried out. The questionnaires were distributed by postal service, filled out by the broiler farmers and returned by postal service, which was time consuming. For the future, and electronic questionnaire surveys would be preferred, if at all possible, since it less time consuming and would also greatly ease the process of entering data into a common dataset.
5. Conclusion Based on our analyses of Danish and Norwegian data on farm specific variables and Campylobacter prevalence in broiler flocks, we conclude that the risk of flocks becoming colonised with Campylobacter increased when: broilers were raised in Denmark compared with Norway; broiler houses were older than five years; the downtime was increased; broiler houses lacked a separate anteroom with barrier; and when the type of drinkers were nipples with cups or bells rather than with nipples
23
without cups. When analysed alone, the risk of Campylobacter colonisation in Norwegian flocks increased with increasing numbers of houses on a farm, and when surface water was used as water source. For the Danish data only, having boot dips or a low stocking density increased the risk of a flock becoming Campylobacter positive. Designing questionnaires well suited for bi-country surveys is a discipline in itself. Even though much effort was put into designing an optimal questionnaire for our purpose, it proved time-consuming and challenging. Conducting a pilot survey to identify common pitfalls, and using electronic questionnaires is highly recommended. The different model approaches allowed us to explore the effect of having a large number of data available in order to identify the significant variables. It also allowed us to explore if the effect of the variables in the final models were significantly influenced by the season and last but not least, allowed us to ensure that the results were not driven by one country alone.
Acknowledgements This research was part of the CamCon project; Campylobacter control novel approaches in primary poultry production, funded by the European Community’s Seventh Framework Programme (FP7/2007-2013) under Grant Agreement no. 244547. The assistance of all participating Danish and Norwegian broiler farmers, typing assistants from Division of Nutrition, Technical University of Denmark and Mie Nielsen Blom from the Danish Agricultural Council is gratefully acknowledged, without their contributions, this study could not have been carried out.
24
References Adkin, A., Harnett, E., Jordan, L., Newell, D., Davison, H., 2006. Use of a systematic review to assist the development of Campylobacter control strategies in broilers. J. Appl. Microbiol. 100, 306-315. Allen, V.M., Weaver, H., Ridley, A.M., Harris, J.A., Sharma, M., Emery, J., Sparks, N., Lewis, M., Edge, S., 2008. Source of spread of thermophilic Campylobacter spp. during partial depopulation of broiler chicken flocks. J. Food Prot. 71, 264-270. Anonymous, 2014a. Annual Report on Zoonoses in Denmark 2013, National Food Institute, Technical University of Denmark. Anonymous, 2014b. Norway - Trends and sources of zoonoses and zoonotic agents in humans, foodstuffs, animals and feeding stuffs in 2013. Bahrndorff, S., Rangstrup-Christensen, L., Nordentoft, S., Hald, B.,2013. Foodborne disease prevention and broiler chickens with reduced Campylobacter infection. Emerg. Infect. Dis. 19, 425-430. Barrios, P.R., Reiersen J., Lowman R., Bisaillon J.R., Michel P., Fridriksdottir V., Gunnarsson, E., Stern, N., Berke, O., McEwen, S., Martin, W., 2006. Risk factors for Campylobacter spp. colonization in broiler flocks in Iceland. Prev. Vet. Med. 74, 264278. Berndtson, E., Emanuelson, U., Engvall, A., Danielsson-Tham, M.-L.,1996. A 1-year epidemiological study of Campylobacter in 18 Swedish chicken farms. Prev. Vet. Med. 26, 167-185.
25
Bouwknegt, M., van de Giessen, A.W., Dam-Deisz, W.D., Havelaar, A.H., Nagelkerke, N.J., Henken, A.M., 2004. Risk factors for the presence of Campylobacter spp. in Dutch broiler flocks. Prev. Vet. Med. 62, 35-49. Boysen, L., Rosenquist, H., Larsson, J.T., Nielsen, E.M., Sørensen, G., Nordentoft, S., Hald, T., 2014. Source attribution of human campylobacteriosis in Denmark. Epidemiol. Infect. Volume 142, 1599-1608. Chowdhury, S., Sandberg, M., Themudo, G.E., Ersbøll, A.K., 2012. Risk factors for Campylobacter infection in Danish broiler chickens. Poult. Sci. 91, 2701-2709. EFSA and ECDC (European Food Safety Authority and European Centre for Disease Prevention and Control), 2015. The European Union Summary Report on Trends and Sources of Zoonoses, Zoonotic Agents and Food-borne Outbreaks in 2013. EFSA Journal 13, 3991, 162 pp. EFSA Panel on Biological Hazards (BIOHAZ), 2010. Scientific Opinion on Quantification of the risk posed by broiler meat to human campylobacteriosis in the EU. EFSA Journal 8, 1437, 89 pp. Ellis-Iversen J., Jorgensen F., Bull S., Powell L., Cook A.J., Humphrey T.J., 2009 Risk factors for campylobacter colonisation during rearing of broiler flocks in Great Britain. Prev. Vet. Med. 89(3-4), 178-184. Gibbens J.C., Pascoe S.J., Evans S.J., Davies R.H., Sayers A.R., 2001. A trial of biosecurity as a means to control campylobacter infection of broiler chickens. Prev. Vet. Med. 48, 85-99.
26
Guerin, M.T., Martin, W., Reiersen, J., Berke, O., McEwen, S.A., Bisaillon, J.-R., Lowman, R., 2007a. A farm-level study of risk factors associated with the colonization of broiler flocks with Campylobacter spp. in Iceland, 2001-2004. Acta. Vet. Scand. 49,18. Guerin, M.T., Martin, W., Reiersen, J., Berke, O., McEwen, S.A., Bisaillon, J.-R., Lowman, R., 2007b. House-level risk factors associated with the colonization of broiler flocks with campylobacter spp. in Iceland, 2001 - 2004. BMC Vet. Res. 2007 Nov 12;3:30. Guerin, M.T., Martin, S.W., Reiersen, J., Berke, O., McEwen, S.A., McEwen, S.A., Bisaillon, J.-R., Lowman, R., Frioriksdottir, V., Berke, O., 2008.Temperature-related risk factors associated with the colonization of broiler-chicken flocks with Campylobacter spp. in Iceland, 2001-2004. Prev. Vet. Med. 86, 14-29. Hald, B., Sommer, H.M., Skovgard, H., 2007. Use of fly screens to reduce Campylobacter spp. introduction in broiler houses. Emerg. Infect. Dis. 13, 1951-1953. Hald, B., Wedderkopp, A., Madsen, M., 2000. Thermophilic Campylobacter spp. in Danish broiler production: A cross-sectional survey and a retrospective analysis of risk factors for occurrence in broiler flocks. Avian Pathol. 29, 123-31. Hofshagen, M., Kruse, H., 2005. Reduction in flock prevalence of Campylobacter spp. in broilers in Norway after implementation of an action plan. J. Food Prot. 68, 22202223.
27
Huneau-Salaun A, Denis M, Balaine L, Salvat G., 2007. Risk factors for campylobacter spp. colonization in french free-range broiler-chicken flocks at the end of the indoor rearing period. Prev. Vet. Med. 80(1),34-48. Høg, B.B., Rosenquist, H., Sørensen, A.I.V., Larsen, L.S., Osek, J., Wieczorek, K., Kusyk, P., Cerdà-Cuéllar, M., Dolz, R., Urdaneta, S., David, B., Hofshagen, M., Wagenaar, J.A., Bolder, N., Jørgensen, F., Williams, N., Merga, Y., Humphrey, T., 2011. Questionnaire survey among broiler producers in six European countries. http://www.camcon-eu.net/wp-content/uploads/2015/05/D-1.1.2-Report-on-broilers-inEurope.pdf Jonsson, M.E., Chriél, M., Norström, M., Hofshagen, M., 2012. Effect of climate and farm environment on Campylobacter spp. colonisation in Norwegian broiler flocks. Prev. Vet. Med. 107, 95-104. Jore, S., Viljugrein, H., Brun, E., Heier, B.T., Borck, B., Ethelberg, S., Hakkinen, M., Kuusi, M., Reiersen, J., Hansson, I., Engvall, E.O., Løfdahl, M., Wagenaar, J.A., van Pelt, W., Hofshagen, M., 2010. Trends in Campylobacter incidence in broilers and humans in six European countries, 1997-2007. Prev. Vet. Med. 93, 33-41.
Kapperud G, Skjerve E, Vik L, Hauge K, Lysaker A, Aalmen I, et al., 1993. Epidemiological investigation of risk factors for campylobacter colonization in norwegian broiler flocks. Epidemiol Infect. 111(2), 245-55.
Lund, M., Nordentoft, S., Pedersen, K., Madsen, M., 2004. Detection of Campylobacter spp. in Chicken Fecal Samples by Real-Time PCR. J. Clin. Microbiol. 42, 5125–5132.
28
Lyngstad, T.M., Jonsson, M.E., Hofshagen, M., Heier, B.T., 2008. Risk factors associated with the presence of Campylobacter species in Norwegian broiler flocks. Poult. Sci. 87, 1987-1994. McDowell, S.W., Menzies, F.D., McBride, S.H., Oza, A.N., McKenna, J.P., Gordon, A.W., Neill, S.D., 2008. Campylobacter spp. in conventional broiler flocks in Northern Ireland: Epidemiology and risk factors. Prev. Vet. Med. 84, 261-76. Muellner, P., Pleydell, E., Pirie, R., Baker, M.G., Campbell, D., Carter, P.E., French, N.P., 2013. Molecular-based surveillance of campylobacteriosis in New Zealand – from source attribution to genomic epidemiology. Euro Surveill. 18, 20365.
Mughini-Gras, L., Smid, J.H., Wagenaar, J.A., de Boer, A.G., Havelaar, A.H., Friesema, I.H.M., French, N.P., Busani, L., van Pelt, W., (2012). Risk Factors for Campylobacteriosis of Chicken, Ruminant, and Environmental Origin: A Combined Case-Control and Source Attribution Analysis. PLoS ONE 7(8): e42599.
Näther, G., Alter, T., Martin, A., Ellerbroek, L., 2009. Analysis of risk factors for Campylobacter species infection in broiler flocks. Poult. Sci. 88, 1299-1305.
Newell, D.G., Fearnley, C., 2003. Source of Campylobacter colonization in broiler chickens. Appl. Environ. Microbiol. 69, 4343-4351. Patrick, M.E., Christiansen, L.E., Waino, M., Ethelberg, S., Madsen, H., Wegener, H.C., 2004. Effects of climate on incidence of Campylobacter spp. in humans and prevalence in broiler flocks in Denmark. Appl. Environ. Microbiol. 70, 7474-7480.
29
Refregier-Petton, J., Rose, N., Denis, M., Salvat, G., 2001. Risk factors for Campylobacter spp. contamination in French broiler-chicken flocks at the end of the rearing period. Prev. Vet. Med. 50, 89-100. Rosenquist, H., Boysen, L., Galliano, C., Nordentoft, S., Ethelberg, S., Borck, B., 2009. Danish strategies to control Campylobacter in broilers and broiler meat: facts and effects. Epidemiol. Infect. 137, 1742-1750. Rushton SP, Humphrey TJ, Shirley MD, Bull S, Jorgensen F., 2009. Campylobacter in housed broiler chickens: A longitudinal study of risk factors. Epidemiol. Infect.;137(8), 1099-1110. Sommer, H.M., Heuer, O.E., Sørensen, A.I., Madsen, M., 2013. Analysis of factors important for the occurrence of Campylobacter in Danish broiler flocks. Prev. Vet. Med. 111, 100-111. van de Giessen, A.W., Bloemberg, B.P., Ritmeester, W.S., Tilburg, J.J., 1996. Epidemiological study on risk factors and risk reducing measures for Campylobacter infections in Dutch broiler flocks. Epidemiol. Infect. 117, 245-250. Wingstrand, A., Neimann, J., Engberg, J., Nielsen, E. M., Gerner-Smidt, P., Wegener, H. C., Molbak, K., 2006. Fresh chicken as main risk factor for campylobacteriosis, Denmark. Emerg. Infect. Dis. 12, 280-285. WHO, 2012. The Global View of Campylobacteriosis. Report of an expert consultation, Utrecht, Netherlands, 9-11 July 2012. http://www.who.int/iris/bitstream/10665/80751/1/9789241564601_eng.pdf
30
Table 1. Significant explanatory variables for the main model of Campylobacter prevalence in Danish and Norwegian broiler farms. The pvalue is given for each significant variable together with the effect-estimate, standard error, odds ratio and OR confidence intervals.
Main model - DK and NO (summer)
Type 3 p-value
Intercept
-
Age of house
0-5 year
Anteroom/ barrier
Country
Standard Error
Estimate
Odds Ratio
95% Confidence intervals [lower; upper]
-
-
0.63
[0.41 ; 0.98]
-
-1.83
0.80
0.002
-0.46
0.22
6-15 years
-
0.25
0.13
1.28
> 15 years
-
0
0
-
Yes
0.003
-0.37
0.13
0.69
No
-
0
0
-
DK
S0167-5877(16)30107-6 http://dx.doi.org/doi:10.1016/j.prevetmed.2016.04.002 PREVET 4012
To appear in:
PREVET
Received date: Revised date: Accepted date:
23-4-2015 1-4-2016 4-4-2016
Please cite this article as: Hog, B.Borck, Sommer, H.M., Larsen, L.S., Sorensen, A.I.V., David, B., Hofshagen, M., Rosenquist, H., Farm specific risk factors for Campylobacter colonisation in Danish and Norwegian broilers.Preventive Veterinary Medicine http://dx.doi.org/10.1016/j.prevetmed.2016.04.002 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Farm specific risk factors for Campylobacter colonisation in Danish and Norwegian broilers
B. Borck Høga*), H.M. Sommera,b, L.S. Larsen1a, A.I.V. Sørensena, B. Davidc, M. Hofshagenc, H. Rosenquista
a
Risk Assessment and Nutrition, The National Food Institute, Technical University of
Denmark, Mørkhøj Bygade 19, DK-2860 Søborg, Denmark
b
Data Analysis and Statistics, DTU Compute, Technical University of Denmark,
building 324, DK-2800 Kgs. Lyngby, Denmark c
Norwegian Veterinary Institute, Pb 750 Sentrum, N-0106 Oslo, Norway
*) Corresponding author: Birgitte Borck Høg, DVM, Ph.D. The National Food Institute, Technical University of Denmark Mørkhøj Bygade 19 DK-2860 Søborg Denmark Tel: +45 35887066 Fax: +45 35887028 E-mail: [email protected]
1
Present address: Statens Serum Institut, 5 Artillerivej, 2300 Copenhagen S, Denmark
1
Highlights
Cross country survey of risk factors for Campylobacter colonization of broilers.
Common and country specific risk factors were identified.
Risk factors were generally related to breaches in biosecurity.
The broiler flock prevalence differed between countries (Denmark>Norway).
2
Abstract Campylobacteriosis has become the leading bacterial zoonosis in humans in the European Union and other developed countries. There are many sources of human Campylobacter infections, but broilers and broiler meat have been shown to be the most important. In order to implement effective interventions that reduce the probability of Campylobacter colonisation of broiler flocks, it is essential to fully understand the risk factors involved. We present a bi-national risk factor survey comprising Campylobacter data from more than 5,200 Danish and Norwegian indoor, conventional broiler flocks and the responses to a standardised questionnaire, with more than 40 explanatory variables from 277 Danish and Norwegian farms. We explored several models by using different combinations of the Danish and Norwegian data, including models with singlecountry datasets. All models were analysed using a generalized linear model using backwards elimination and forward selection. The results show that Norwegian broiler flocks had a lower risk of being colonised than Danish flocks. Farm specific variables that increased the risk of flocks becoming colonised with Campylobacter in both countries were: broiler houses older than five years; longer downtime (no. of days between flocks), probably a consequence of longer downtimes being associated with less focus on maintaining a high biosecurity level; broiler houses without a separate ante-room or barrier; and the use of the drinker nipples with cups or bells compared with nipples without cups. Additional country specific risk factors were also identified. For Norway, the risk of colonisation increased with increasing numbers of houses on a farm and when the water used for the broilers originated from surface water or bore holes instead
3
of mains. For Denmark, having boot dips or low stocking density increased the risk of a flock becoming Campylobacter positive. The different model approaches allowed us to explore the effect of having a large number of data available to identify the significant variables. To a large extent, the country specific models identified risk factors that were also found in the bi-national model. However, the bi-national model identified more risk factors than the country specific models. This indicated that combining the data sets from the two countries did not disrupt the results but was beneficial due to the greater strength achieved in the statistical analyses and the possibility of examining interactions terms with the variable Country.
Key words: Campylobacter, poultry, broiler flocks, risk factors, generalized linear model
4
1. Introduction For many years, campylobacteriosis has been the most common cause of foodborne gastrointestinal illness in the EU and the rest of the industrialized world (WHO, 2012; EFSA and ECDC, 2015). Broilers and broiler meat are the most important single source of infection in humans even though other sources have also been identified (Wingstrand et al., 2006; EFSA, 2010; Mughini-Gras et al., 2012; Muellner et al., 2013; Boysen et al., 2014). Therefore, reduction of Campylobacter in the broiler production is considered to be an important step towards reducing the number of human infections. Risk factors associated with the introduction of Campylobacter in broiler flocks have been studied in several countries since the 1990s. Lack of employment of hygiene measures and strict hygienic routines are very important risk factors for colonisation of broiler flocks with Campylobacter (Berndtson et al., 1996; van de Giessen et al., 1996; Hald et al., 2000; Adkin et al., 2006; Lyngstad et al., 2008). Other risk factors have also been identified, but many of these are somehow related to compromised biosecurity, or increased pressure on biosecurity due to increased load of Campylobacter in the farm environment. Factors that may increase the load of Campylobacter in the environment are e.g. presence of other animals on the farm or in close proximity of the farm, more than one broiler house on the farm (Refregier-Petton et al., 2001, Bouwknegt et al., 2004, Guerin et al., 2007a, McDowell et al., 2008; Chowdhury et al., 2012), and climate-related factors - especially temperature (Patrick et al., 2004, Guerin et al., 2008, McDowell et al., 2008; Rushton et al., 2009, Jore et al., 2010, Jonsson et al., 2012). Hald et al. (2007) and Bahrndorff et al. (2013) found that the climate related factors may partly be explained by flies transmitting Campylobacter into the broiler houses. Examples of factors that may
5
influence (compromise) biosecurity are; the number of persons tending the broilers (Refregier-Petton et al., 2001), thinning broiler flocks (Berndtson et al., 1996; Hald et al., 2000; Allen et al., 2008), presence of rodents at farms (McDowell et al., 2008; Sommer et al., 2013,), and increased ventilation in the summer (Newell and Fearnley, 2003). Other identified risk factors include age of the broilers at slaughter (Berndtson et al., 1996, Bouwknegt et al., 2004, Barrios et al., 2006, Chowdhury et al., 2012, Sommer et al., 2013), using drinkers with trays (Näther et al., 2009), flock size (Berndtson et al., 1996, Barrios et al., 2006, Guerin et al., 2007a), and hatchery (Bouwknegt el al., 2004). Many risk factor studies concerning Campylobacter in broiler flocks are based on data from a limited number of houses monitored for one or a few production cycles. Furthermore, the studies have typically included data from just one country. The studies have identified different risk factors and the findings may in some instances seem contradictory. However, some differences may simply be due to data limitations (limited number of variables/number of observations), which has reduced the power of the tests for significance. The conventional indoor broiler productions in Norway and Denmark share many common characteristics. For example, quality assurance programmes ensuring a high level of biosecurity on the farms have been implemented in both countries. Generally, all farms apply an all-in-all-out system with an “empty period” between flocks where the litter is removed and the houses cleaned. However, there are also a number of factors that differ between the countries, e.g. the farms in Norway generally have fewer houses per farm, (Høg et al., 2011). At the time of the study, approximately 100 million broilers were raised per year in Denmark. There were approximately 170 indoor conventional broiler farms
6
and the majority (>98%) of broilers were produced for two major broiler companies. In Norway, at the time of the study, approximately 60 million broilers were raised annually on approximately 600 farms and the majority of the farmers produced broilers for one of the three main companies. Together, these companies produced more than 98% of the Norwegian broilers at four different slaughterhouses. This study included Campylobacter surveillance results from more than 5,200 indoor, conventional broiler flocks, collected over a two-year period in two Scandinavian countries. The surveillance data were analysed together with the outcomes of a questionnaire survey answered by 277 Danish and Norwegian farms. By analysing data from both countries together we aimed a) to achieve a stronger basis for the statistical analyses by utilizing larger datasets and b) identify common and/or country specific risk factors for the occurrence of Campylobacter in Danish and Norwegian broiler flocks.
2. Material and methods 2.1. Farm specific variables For the survey, a questionnaire including 43 questions concerning production, farm management procedures, farm conditions etc. was developed. The questionnaire can be found in full in Høg et al. (2011). The questionnaire was translated into Danish and Norwegian and the survey took place between December 2010 and June 2011. This study aimed to identify risk factors for Campylobacter colonisation of conventional indoor broiler flocks and we aimed to obtain information from approximately 200 farms from each country. We used the official registers of animal
7
holdings in both Denmark and Norway to select farms for the study. In Denmark there were less than 200 conventional indoor broiler farms and thus questionnaires were sent out to all the farms. In Norway, farms were selected so that all 19 counties in were represented. However, due to unforeseen circumstances, farmers in one county of Norway did not receive questionnaires, (Høg et al., 2011). All questionnaires were distributed and returned by postal services and filled out by the broiler farmers themselves. Overall, questionnaires were returned by 119 and 183 farmers in Denmark and Norway, respectively, corresponding to response rates of 58% and 59% in the two countries (Høg et al., 2011). The no. of responding farms represented approximately 70% of all Danish indoor conventional broiler farms (ca. 170 farms) and 30% of all indoor conventional broiler farms in Norway (ca. 600 farms). The detailed descriptive outcomes of the questionnaire survey are presented in Høg et al., (2011) In Denmark and Norway the poultry farms and slaughterhouses are not integrated, meaning that farms are not owned by the broiler companies, but by the farmers themselves. However, the broilers are produced in accordance with delivery contracts made with the broiler companies. In both Denmark and Norway, the broiler industries have agreed upon and implemented guidelines and quality assurance systems according to which all indoor conventional broilers are produced; KIK (the Quality Assurance System in Danish Broiler Production) in Denmark, and KSL (Agricultural Quality System) in Norway. All of the farms included in this study, were indoor conventional broiler farms with a mean of seven crops produced annually. The mean downtime between crops was 8 days Denmark and 19 days in Norway.
8
2.2.Campylobacter status Both in Denmark and Norway, Campylobacter data were obtained through national surveillance programmes. In Denmark, Campylobacter data were obtained during the period from January 2010 to April 2012 for flocks on farms that had responded to the questionnaire. The flocks were tested for Campylobacter by analysing one pair of boot swabs collected in each broiler houses 7-10 days prior to slaughter using PCR as described by Lund et al., (2004). In Norway, Campylobacter data from farms that had responded to the questionnaire were obtained for the period May to October for the years 2010 and 2011. The reason for restricting sampling in Norway to the summer months was that the historical data from the winter months in the period 2001-2007 showed a very low Campylobacter prevalence. Campylobacter status was obtained using swab samples, 10 swabs from each flock collected at the farm as close to slaughter as possible (max. four days before slaughter). The laboratory method for analysing the swab samples was the same as was used in Denmark (Lund et al. 2004).
2.3.Data analysis 2.3.1. Data preparation Prior to any data analyses, the dataset was checked to ensure that only conventional farms were included, and data from farms with too few observations were excluded.
9
Data were explored and for practical reasons, some of the farm specific variables were modified to carry out the analyses. The raw data set consisted of a large number of explanatory variables and parameters, as well as a relatively large number of data records with one or more missing values, which made it impossible to run a model. Thus, the number of parameters had to be reduced. The result of the data preparation and the final input for statistical analysis is shown in Table A and B in the supplementary material. The final data set included data from 277 broiler farms (104 Danish and 173 Norwegian farms).
2.3.2. Statistical analysis Multivariable analysis was used to identify risk factors and the analyses were carried out using SAS (version 9.4, SAS Institute Inc.). The prevalence (pr) was the response variable and was defined as the proportion of Campylobacter positive flocks out of the total number of broiler flocks produced on a given broiler farm in the entire study period. The data from the two-year study period were merged since we did not expect a significant trend over just two years. Thus, the model used in the analysis was a non-hierarchical model and the unit of analysis was the mean prevalence per farm. Since the response variable was not normally distributed, a generalized linear model was applied and the response-variable was transformed using a logit link function. logit where
is the intercept,
⋯
∙
⋯
is the parameter for the variable i, i = 1,2,3,…,44 (numbers
referring to the question number given in Table A and B), the indexes j and k refers to
10
the categorical levels, and
∙
is the interaction term between variable i and
the variable Country. Not all variables could be included from the start of the analysis, and the interaction terms with Country could not be included in the model until the model had been reduced to consist only of variables with p-values < 0.10. The statistical procedure of backward elimination and forward selection described by Sommer et al. (2013) was applied. In brief, we wanted to apply a method that made use of as much information from the entire dataset as possible. This meant that we did not exclude variables with missing values (unless the inclusion meant that we could not run the model due to too many missing values). We allowed observations with missing values to enter into the model if 1) the variable for which the observation had a missing value was removed from the model in a backward elimination steps and 2) if the p-value was less than 0.10 in a forward selection step. As the model was reduced in number of variables, the data included in the model increased due to exclusion of variables with missing values. In the initial steps of analysis, the main model included 32 variables. The variables with the largest number of missing values were left out of the initial model, which was based on 153 out of 276 farms. The p-value for several variables changed drastically at some model reduction steps of the analysis. The majority of these changes was instigated by the ‘new’ data included in the analyses, and only for a few steps, was the change caused by collinearity with the variables removed from the model. For each model, p-values, effect-estimates, standard errors, odds ratios (OR), and 95% confidence intervals (CI) for the OR estimates, were estimated. The OR values for a risk factor
also included in an interaction term with
(a continuous variable) was
calculated as the following when comparing e.g. level 1 with 3
11
, ,
exp
∗
∗
where the value for the continuous variable is the mean value or a chosen fixed value. The main model included data from the time period, where both Danish and Norwegian data were available, i.e. the summer period from May to October in 2010 and 2011. However, due to the difference in the available Campylobacter data, i.e. all year in Denmark/ May-October in Norway, alternative model A) was explored: a model including all Danish data (January 2010 - April 2012 (all year)) plus all Norwegian data from the summer period (May-October in 2010 and 2011). The Danish and Norwegian data were also analysed separately: alternative model B) a model with only Norwegian data; and alternative model C) a model with Danish data only. The ‘goodness of fit’ was examined by use of the deviance residuals.
3. Results The significant variables in each of the explored generalized linear models (main model, alternative model A, B, and C) are shown in Table 1 and 2. The results from the main model were based on data from 3,083 flocks from 256 farms out of 276. In the main model the most important variable was Country. It was highly significant (p < 0.0001) and had the largest estimated effect (3.20) and Denmark had the highest risk estimate of the two countries. The estimated OR was also high, however since Country is also part of the interaction term (House * Country) the OR varies with the number of houses, in this case the OR was estimated at 1.5 house per farm. The prevalence of Campylobacter in broiler flocks tested from May to October in 2010 and 2011 was 21% in Denmark and 4% in Norway. Another very important variable was the
12
Downtime (p=0.001). In general, the longer the downtime, the higher the risk of the broiler flocks becoming colonised by Campylobacter. The variable Age of house was also important (p=0.002). The newest broiler houses had the lowest risk estimate. Broiler houses between age 6-15 years and age >15 years were not significantly different from each other (p=0.06). Furthermore, having anterooms and barriers in each broiler house (Anteroom/barrier) significantly reduced the risk (p=0.003), and farms using drinker systems (Drinkers) with nipples without cups had a significantly lower risk estimate (p=0.006). Nipples with cups and bells were not significantly different from each other (p=0.07). By using re-parameterisation we found that the number of houses only was significant for the Norwegian data. This was found by examining the p-values for the interaction term No. of houses x Country. The re-parameterisation was carried out by removing the variable No. of houses from the final model and then re-running the model. This way, all the variation that was explained by the main effect. No. of houses was now explained by the interaction term together with the ‘pure’ interaction effect. The results from the three alternative models are shown in Table 2. Model A, including all Danish and Norwegian data was based on data from 4,744 flocks from 253 farms out of 277. This model identified the same significant variables as the main model plus two more variables: Stocking density *Country and Boot dips. Interestingly, having boot dips increased the risk of a flock becoming Campylobacter positive. Furthermore, a high stocking density reduced the risk for the Danish data but not for the Norwegian data. In model B (Danish data only with 3,430 flocks, 93 farms out of 104 farms), Boot dip and Stocking density were also significant (p=0.025 and 0.026,
13
respectively). No. of houses, however, was not significant, which makes sense since this variable was only important for the Norwegian data (model A). When analysing Norwegian data alone (model C), only three variables became significant. Two variables were the same as identified in the other models (Age of house and Downtime) and one was new; Water source (p=0.020). Farms using mains as their water source had the lowest risk, second lowest were bore holes and those with the highest risk were farms using surface water. Model C was based on data from 1,342 flocks from 164 out of 173 Norwegian farms. The effect of Downtime was further explored in an attempt to understand how a longer Downtime could increase the risk of Campylobacter colonisation. We found that among the farms with a downtime of more than 30 days, 25% responded that they did not have a program for cleaning and disinfecting the houses between the flocks. In contrast, on farms with downtimes of less than 30 days, only 11% stated that they had no program for cleaning and disinfecting the houses between flocks, see Table 3. Furthermore, the number of farms with a high frequency of rodent control (1-2 times/month) dropped from 20% to 13% for farms with a downtime of more than 20 days. Some variables that were not found statistically significant, but had pvalues less than 0.10, are worth mentioning. In the main model the Stocking density was nearly significant (p=0.06) and so was Other animals on farm (p=0.07), where cattle and pigs had an increased estimate compared to the other animals. In model A, Other animals on farm was also nearly significant (p=0.06) as well as Closed ventilation sites (p=0.07) with a decreased risk when closed. Moreover, Fly screens (p=0.07) and Rodent control by company (p=0.08) were nearly significant indicating a reduced risk if
14
houses were screened or if rodent control was maintained by the farmer rather than by a company. The dataset consisting of the largest number of data identified the highest number of variables. For all models the deviance residuals had values less than 2, indicating that there is no overdispersion and that there is no (severe) lack of fit.
4. Discussion Risk factors Country. The variable ‘Country’ was included in the analysis to capture potential differences between the countries that were important in relation to describing the variation in the dataset and that could not be explained by the other variables in the analysis. This variable (Country) describes - in one variable - all the unknown factors that may significantly influence the prevalence and that are unequally distributed in the two countries and therefore do not even out. The overall prevalence of Campylobacter in broilers, as observed through the national surveillance programmes, has consistently been lower in Norway than in Denmark for many years (Anon, 2014a; Anon., 2014b; EFSA and ECDC 2015), and the current study also found that broilers raised in Denmark had a significantly higher risk of being Campylobacter positive, compared to broilers raised in Norway. The fact that Country is significant implies that there is more variation between the two countries than can be explained by the rest of the variables in the model. An important factor that may contribute to this difference in risk is the different climatic conditions in the two
15
countries (Jore et al., 2010, Jonsson et al., 2012). Other factors could be differences in action plans to prevent Campylobacter colonisation of broiler flocks (Hofshagen et al., 2005; Rosenquist et al., 2009), but also the difference in the structure of the production e.g. the age of the chickens at slaughter or factors such as distance to farms with chicken, cattle or livestock may result in a higher the risk of Campylobacter. Downtime. Another important risk factor in this study was the length of downtime between flocks. We found that a long downtime was associated with a higher risk of the flocks becoming colonised with Campylobacter. At first glance this was surprising, since a long downtime, in theory, should allow the broiler house to dry completely, creating the least favourable conditions for Campylobacter survival in the broiler house environment. However, it may be explained by other factors that were characteristic for the farms with long downtimes and could reflect that these farms had a less intensive production and less focus on maintaining a high level of biosecurity and hygiene. The variables Cleaning between flocks and Rodent control frequency were tested in the model both as main effects and as interaction terms with Downtime but were found to be insignificant. One theory could be that none of the two variables were strong enough alone, but together through a third variable (Downtime) significance was seen. There may be other factors characteristic for the farms with long downtimes that play a role in Campylobacter colonisation that we were unable to identify in this study Results of three other Scandinavian risk factor studies have also indicated that the length of downtime affected the risk. However, in contrast to our results these studies indicated that a period of less than 9-14 d between depopulation and restocking was associated with an increased risk of Campylobacter colonisation (Berndtson et al.,
16
1996; Hald et al., 2000; Lyngstad et al., 2008). Downtime appears to be an important risk factor, but should not be considered out of context. Thus, the downtime should be sufficiently long to allow the broiler houses to be efficiently emptied, cleaned and disinfected. However, if the downtime is too long the risk of introducing Campylobacter into the houses increases, especially if the biosecurity and hygiene level on the farm is not optimal. Age of house. The age of the broiler house also significantly influenced the risk of the broilers becoming colonised with Campylobacter. Farms with newer houses had a significantly lower risk of becoming colonised with Campylobacter than farms with houses older than five years. However, we could not show significant differences between the two categories with the oldest broiler houses. The effect of the broiler house age was also reported by Sommer et al. (2013). The importance of the age of the broiler house is likely connected to the fact that older houses may have more difficulty conforming to the modern biosecurity standards that have proven important. It may be speculated that older houses are more difficult to seal towards the surrounding environment and it is therefore easier for Campylobacter to be introduced to the inside environment via flies, rodents and birds that may all carry Campylobacter (Hald et al., 2007; Huneau-Salaun, 2007; McDowell et al., 2008). Anteroom/barriers. The significance of hygiene measures in the prevention of Campylobacter colonisation of broiler flocks has been well documented (Berndtson et al., 1996, van de Giessen et al., 1996, Hald et al., 2000; EFSA, 2010) and is supported by the outcome of the present study, where the presence of anterooms and barriers in each house decreased the risk of Campylobacter colonisation. In Denmark and Norway, the anterooms are usually small rooms, divided into a clean and a dirty
17
zone, through which all staff enter the broiler house. Typically a low barrier separates the dirty and clean zone. All clothes and footwear that has had any contact with the environment outside the broiler house stays in the dirty zone and coveralls and boots for use in the broiler house stay in the clean zone. Many more biosecurity and hygiene variables were included in the questionnaire, but could not be proven significant. However, at the time we conducted the survey, a high level of biosecurity, in concordance with the applied quality assurance schemes, was already employed on most farms and may be one reason that only few of the biosecurity variables could be shown to have an effect. Drinkers. The type of drinkers was shown to affect the occurrence of Campylobacter in both Danish and Norwegian broiler flocks; drinker systems with nipples without cups were better than nipples with cups and bells. Drinker systems with cups and bells both have reservoirs of standing water. Campylobacter may be introduced to these reservoirs via colonised birds or their droppings, whereby the standing water can become a dissemination vehicle, transmitting Campylobacter to other birds when they drink the water. Thus, this study indicates that it was better to use drinkers without a water reservoir for Campylobacter to survive in, and this is in concordance with the work of Näther and co-workers (2009). No. of houses. Increased number of broiler houses on the farm was a risk factor for Norway. The number of houses on a farm has previously been associated with an increased occurrence of Campylobacter on farms (Refregier-Petton et al., 2001, Bouwknegt et al., 2004, Guerin et al., 2007a, McDowell et al., 2008; Sommer et al., 2013).
18
Boot dips. Having boot dips at the entrance to the broiler house was identified as a risk factor but only in the alternative models A and B, in which all the Danish data were included. In order for boot dips to work efficiently they need to be maintained properly, and the disinfectant has to be replenished at appropriate intervals, otherwise they may contribute to an increased risk rather than act as a protective factor (Berndtson et al., 1996, Guerin et al., 2007b, McDowell et al., 2008, Gibbens et al., 2001). Stocking density. We found a reduced risk associated with a high stocking density in Denmark, but not in Norway. Other studies that have looked at the risk associated with flock size, have indicated an increased risk with increasing size of flocks (Barrios et al., 2006, Guerin et al, 2007a). However, perhaps the farms in Denmark with the highest stocking density were also farms with strictest management routines and a highest level of biosecurity. Water source. Using surface water for drinking increased the risk compared to private boreholes, wells and mains. Using mains was the best. Surface water was only used in Norway, and water source was found significant only in model C (Norwegian data only) and was not highly significant. Only 6 out of 164 farms used surface water, so conclusions should be made very cautiously as to the effect of using surface water. An extra model was examined where surface water was merged with the category private bore holes and wells. Here water source was still significant though with a higher p-value (0.047) and mains were still the best. Water source and it’s interaction with Country was, however, not significant in the main model. Nonetheless, using surface water should not be neglected as a possible source for Campylobacter in domestic animals.
19
Other factors. The significance of ventilating has been demonstrated in other studies (Newell and Fearnley, 2003; Guerin et al., 2007a; Rushton et al., 2009). Our questionnaire included several questions concerning ventilation, none of which were found significant. However, some of the variables that related specifically to the broiler houses suffered somewhat from the design of the study, where questions were asked at the farm level rather than at the house level. In this study 35% of the participating farms had two or more broiler houses and for these farms, important details concerning the house specific variables were lost due to the design of the questionnaire. The significance of flies in the transmission of Campylobacter and the effect of using fly screens has been demonstrated by Hald and co-workers (Hald et al., 2007; Bahrndorff et al., 2013). Unfortunately, only three of the participating farms had installed fly screens on the broiler houses (all Danish farms) and therefore any results should be interpreted very carefully. Fly screens was not significant in the main model, but close to significant in alternative model A, perhaps due to the greater statistical power. Thinning was not included in this study, since the available Campylobacter data did not allow us to calculate prevalence estimates before and after thinning of the broiler flocks.
Models As described above, four models based on different datasets were explored. When evaluating and comparing the different models it is important to look at the power of the test statistics and the frequency of bias results. The first is evaluated by examining the number of data included in the models. Alternative model A, based on 4,744 flocks, seemed to have greater statistical power than the main model, based on
20
3,083 flocks. However, the estimate of Country in model A was biased due to the different sampling periods in the two countries in combination with a severe seasonality which was not accounted for, because the response variable was an aggregated prevalence estimate for a whole year. Apart from the biased estimate of Country, model A identified the same variables as the main model plus two extra. The additional variables likely appear as a consequence of the larger number of degrees of freedom in model A, compared with the main model. Model A identified the same six variables as the main model, and the estimates were within the same ranges, which indicates that the significant variables were only affected little or not at all by the different sampling periods for Denmark in the two datasets. This suggests that the significant variables are not driven by the season but more or less have the same influence on the prevalence all year around. The purpose of exploring both the main and the alternative models was to investigate the benefit of combining the Danish and the Norwegian data sets and thereby increase the statistical power of the analyses. We wanted to investigate which and how many risk factors could be identified given the larger joint dataset from both countries, compared with the smaller single-country datasets. In the joint models (main model + model A) the interactions terms with Country were included in the models, when possible, to allow for country specific risk factors. Furthermore, we wanted to investigate if expanding the Danish data, to also include the winter period, would drive the results of the model in another direction. In other words, we wanted to see whether the effect of the significant variables depended on the sampling period (summer/winter). The reason for not running a model with all data (Denmark: whole year + Norway: summer), where variables were split in two (one for the summer and one for winter),
21
was that it would result in twice as many parameters to estimate rendering the model not to run. One may argue that model B, compared to the main model, is the better choice for Denmark since this model is based on 347 more flocks. The number of farms in model B was, however, only 93 compared with 256 for the main model, and more than doubling the number of farms is attractive since it includes farms with other combinations of the variables and thus makes the design more complete. This is also the case when comparing model C with the main model. Model C, based on 164 Norwegian farms, resulted in three significant variables, whereas seven significant variables were identified in the main model containing data from 256 farms. If the risk factors identified in the main model were not relevant for Norway, this would have been expressed in the interaction term with Country in the main model and model A. All risk factors (except one) identified in model B and C were also identified in the main model and model A. All models in this study were non-hierarchical since the unit of analysis were the mean prevalence per farm. One could have expected the unit of the samples to be measurements per broiler house per farm which would have resulted in a hierarchical model structure. The houses would then be nested with the farms and houses from the same farm would most likely be correlated in some way which could be accounted for by applying an appropriate variance structure. However, such a model was not applied since we did not have information on house level for all data in the study. Even if we had information on house levels the results would probably not change much since for half of the dataset (Norway) most farms only had one broiler house.
22
Questionnaires. A number of lessons concerning questionnaires were learned. Designing a questionnaire well suited for a bi-national country survey requires extensive knowledge of the production in the participating countries. In the questionnaire designed for this survey, a number of responses were given for each question. This solution was chosen in an attempt to make it quick and easy for the respondent to answer the question, but also in order ensure homogenous inputs to the data. For the same reasons, very little free text was permitted. Nonetheless, a number of pitfalls were identified when data were returned, particularly in questions concerning ventilation systems, which are highly house specific, and therefore not well suited for this farm-level questionnaire. Also, a pilot survey to identify common pitfalls was not conducted in this survey, and could very likely have improved the questionnaire survey, had it been carried out. The questionnaires were distributed by postal service, filled out by the broiler farmers and returned by postal service, which was time consuming. For the future, and electronic questionnaire surveys would be preferred, if at all possible, since it less time consuming and would also greatly ease the process of entering data into a common dataset.
5. Conclusion Based on our analyses of Danish and Norwegian data on farm specific variables and Campylobacter prevalence in broiler flocks, we conclude that the risk of flocks becoming colonised with Campylobacter increased when: broilers were raised in Denmark compared with Norway; broiler houses were older than five years; the downtime was increased; broiler houses lacked a separate anteroom with barrier; and when the type of drinkers were nipples with cups or bells rather than with nipples
23
without cups. When analysed alone, the risk of Campylobacter colonisation in Norwegian flocks increased with increasing numbers of houses on a farm, and when surface water was used as water source. For the Danish data only, having boot dips or a low stocking density increased the risk of a flock becoming Campylobacter positive. Designing questionnaires well suited for bi-country surveys is a discipline in itself. Even though much effort was put into designing an optimal questionnaire for our purpose, it proved time-consuming and challenging. Conducting a pilot survey to identify common pitfalls, and using electronic questionnaires is highly recommended. The different model approaches allowed us to explore the effect of having a large number of data available in order to identify the significant variables. It also allowed us to explore if the effect of the variables in the final models were significantly influenced by the season and last but not least, allowed us to ensure that the results were not driven by one country alone.
Acknowledgements This research was part of the CamCon project; Campylobacter control novel approaches in primary poultry production, funded by the European Community’s Seventh Framework Programme (FP7/2007-2013) under Grant Agreement no. 244547. The assistance of all participating Danish and Norwegian broiler farmers, typing assistants from Division of Nutrition, Technical University of Denmark and Mie Nielsen Blom from the Danish Agricultural Council is gratefully acknowledged, without their contributions, this study could not have been carried out.
24
References Adkin, A., Harnett, E., Jordan, L., Newell, D., Davison, H., 2006. Use of a systematic review to assist the development of Campylobacter control strategies in broilers. J. Appl. Microbiol. 100, 306-315. Allen, V.M., Weaver, H., Ridley, A.M., Harris, J.A., Sharma, M., Emery, J., Sparks, N., Lewis, M., Edge, S., 2008. Source of spread of thermophilic Campylobacter spp. during partial depopulation of broiler chicken flocks. J. Food Prot. 71, 264-270. Anonymous, 2014a. Annual Report on Zoonoses in Denmark 2013, National Food Institute, Technical University of Denmark. Anonymous, 2014b. Norway - Trends and sources of zoonoses and zoonotic agents in humans, foodstuffs, animals and feeding stuffs in 2013. Bahrndorff, S., Rangstrup-Christensen, L., Nordentoft, S., Hald, B.,2013. Foodborne disease prevention and broiler chickens with reduced Campylobacter infection. Emerg. Infect. Dis. 19, 425-430. Barrios, P.R., Reiersen J., Lowman R., Bisaillon J.R., Michel P., Fridriksdottir V., Gunnarsson, E., Stern, N., Berke, O., McEwen, S., Martin, W., 2006. Risk factors for Campylobacter spp. colonization in broiler flocks in Iceland. Prev. Vet. Med. 74, 264278. Berndtson, E., Emanuelson, U., Engvall, A., Danielsson-Tham, M.-L.,1996. A 1-year epidemiological study of Campylobacter in 18 Swedish chicken farms. Prev. Vet. Med. 26, 167-185.
25
Bouwknegt, M., van de Giessen, A.W., Dam-Deisz, W.D., Havelaar, A.H., Nagelkerke, N.J., Henken, A.M., 2004. Risk factors for the presence of Campylobacter spp. in Dutch broiler flocks. Prev. Vet. Med. 62, 35-49. Boysen, L., Rosenquist, H., Larsson, J.T., Nielsen, E.M., Sørensen, G., Nordentoft, S., Hald, T., 2014. Source attribution of human campylobacteriosis in Denmark. Epidemiol. Infect. Volume 142, 1599-1608. Chowdhury, S., Sandberg, M., Themudo, G.E., Ersbøll, A.K., 2012. Risk factors for Campylobacter infection in Danish broiler chickens. Poult. Sci. 91, 2701-2709. EFSA and ECDC (European Food Safety Authority and European Centre for Disease Prevention and Control), 2015. The European Union Summary Report on Trends and Sources of Zoonoses, Zoonotic Agents and Food-borne Outbreaks in 2013. EFSA Journal 13, 3991, 162 pp. EFSA Panel on Biological Hazards (BIOHAZ), 2010. Scientific Opinion on Quantification of the risk posed by broiler meat to human campylobacteriosis in the EU. EFSA Journal 8, 1437, 89 pp. Ellis-Iversen J., Jorgensen F., Bull S., Powell L., Cook A.J., Humphrey T.J., 2009 Risk factors for campylobacter colonisation during rearing of broiler flocks in Great Britain. Prev. Vet. Med. 89(3-4), 178-184. Gibbens J.C., Pascoe S.J., Evans S.J., Davies R.H., Sayers A.R., 2001. A trial of biosecurity as a means to control campylobacter infection of broiler chickens. Prev. Vet. Med. 48, 85-99.
26
Guerin, M.T., Martin, W., Reiersen, J., Berke, O., McEwen, S.A., Bisaillon, J.-R., Lowman, R., 2007a. A farm-level study of risk factors associated with the colonization of broiler flocks with Campylobacter spp. in Iceland, 2001-2004. Acta. Vet. Scand. 49,18. Guerin, M.T., Martin, W., Reiersen, J., Berke, O., McEwen, S.A., Bisaillon, J.-R., Lowman, R., 2007b. House-level risk factors associated with the colonization of broiler flocks with campylobacter spp. in Iceland, 2001 - 2004. BMC Vet. Res. 2007 Nov 12;3:30. Guerin, M.T., Martin, S.W., Reiersen, J., Berke, O., McEwen, S.A., McEwen, S.A., Bisaillon, J.-R., Lowman, R., Frioriksdottir, V., Berke, O., 2008.Temperature-related risk factors associated with the colonization of broiler-chicken flocks with Campylobacter spp. in Iceland, 2001-2004. Prev. Vet. Med. 86, 14-29. Hald, B., Sommer, H.M., Skovgard, H., 2007. Use of fly screens to reduce Campylobacter spp. introduction in broiler houses. Emerg. Infect. Dis. 13, 1951-1953. Hald, B., Wedderkopp, A., Madsen, M., 2000. Thermophilic Campylobacter spp. in Danish broiler production: A cross-sectional survey and a retrospective analysis of risk factors for occurrence in broiler flocks. Avian Pathol. 29, 123-31. Hofshagen, M., Kruse, H., 2005. Reduction in flock prevalence of Campylobacter spp. in broilers in Norway after implementation of an action plan. J. Food Prot. 68, 22202223.
27
Huneau-Salaun A, Denis M, Balaine L, Salvat G., 2007. Risk factors for campylobacter spp. colonization in french free-range broiler-chicken flocks at the end of the indoor rearing period. Prev. Vet. Med. 80(1),34-48. Høg, B.B., Rosenquist, H., Sørensen, A.I.V., Larsen, L.S., Osek, J., Wieczorek, K., Kusyk, P., Cerdà-Cuéllar, M., Dolz, R., Urdaneta, S., David, B., Hofshagen, M., Wagenaar, J.A., Bolder, N., Jørgensen, F., Williams, N., Merga, Y., Humphrey, T., 2011. Questionnaire survey among broiler producers in six European countries. http://www.camcon-eu.net/wp-content/uploads/2015/05/D-1.1.2-Report-on-broilers-inEurope.pdf Jonsson, M.E., Chriél, M., Norström, M., Hofshagen, M., 2012. Effect of climate and farm environment on Campylobacter spp. colonisation in Norwegian broiler flocks. Prev. Vet. Med. 107, 95-104. Jore, S., Viljugrein, H., Brun, E., Heier, B.T., Borck, B., Ethelberg, S., Hakkinen, M., Kuusi, M., Reiersen, J., Hansson, I., Engvall, E.O., Løfdahl, M., Wagenaar, J.A., van Pelt, W., Hofshagen, M., 2010. Trends in Campylobacter incidence in broilers and humans in six European countries, 1997-2007. Prev. Vet. Med. 93, 33-41.
Kapperud G, Skjerve E, Vik L, Hauge K, Lysaker A, Aalmen I, et al., 1993. Epidemiological investigation of risk factors for campylobacter colonization in norwegian broiler flocks. Epidemiol Infect. 111(2), 245-55.
Lund, M., Nordentoft, S., Pedersen, K., Madsen, M., 2004. Detection of Campylobacter spp. in Chicken Fecal Samples by Real-Time PCR. J. Clin. Microbiol. 42, 5125–5132.
28
Lyngstad, T.M., Jonsson, M.E., Hofshagen, M., Heier, B.T., 2008. Risk factors associated with the presence of Campylobacter species in Norwegian broiler flocks. Poult. Sci. 87, 1987-1994. McDowell, S.W., Menzies, F.D., McBride, S.H., Oza, A.N., McKenna, J.P., Gordon, A.W., Neill, S.D., 2008. Campylobacter spp. in conventional broiler flocks in Northern Ireland: Epidemiology and risk factors. Prev. Vet. Med. 84, 261-76. Muellner, P., Pleydell, E., Pirie, R., Baker, M.G., Campbell, D., Carter, P.E., French, N.P., 2013. Molecular-based surveillance of campylobacteriosis in New Zealand – from source attribution to genomic epidemiology. Euro Surveill. 18, 20365.
Mughini-Gras, L., Smid, J.H., Wagenaar, J.A., de Boer, A.G., Havelaar, A.H., Friesema, I.H.M., French, N.P., Busani, L., van Pelt, W., (2012). Risk Factors for Campylobacteriosis of Chicken, Ruminant, and Environmental Origin: A Combined Case-Control and Source Attribution Analysis. PLoS ONE 7(8): e42599.
Näther, G., Alter, T., Martin, A., Ellerbroek, L., 2009. Analysis of risk factors for Campylobacter species infection in broiler flocks. Poult. Sci. 88, 1299-1305.
Newell, D.G., Fearnley, C., 2003. Source of Campylobacter colonization in broiler chickens. Appl. Environ. Microbiol. 69, 4343-4351. Patrick, M.E., Christiansen, L.E., Waino, M., Ethelberg, S., Madsen, H., Wegener, H.C., 2004. Effects of climate on incidence of Campylobacter spp. in humans and prevalence in broiler flocks in Denmark. Appl. Environ. Microbiol. 70, 7474-7480.
29
Refregier-Petton, J., Rose, N., Denis, M., Salvat, G., 2001. Risk factors for Campylobacter spp. contamination in French broiler-chicken flocks at the end of the rearing period. Prev. Vet. Med. 50, 89-100. Rosenquist, H., Boysen, L., Galliano, C., Nordentoft, S., Ethelberg, S., Borck, B., 2009. Danish strategies to control Campylobacter in broilers and broiler meat: facts and effects. Epidemiol. Infect. 137, 1742-1750. Rushton SP, Humphrey TJ, Shirley MD, Bull S, Jorgensen F., 2009. Campylobacter in housed broiler chickens: A longitudinal study of risk factors. Epidemiol. Infect.;137(8), 1099-1110. Sommer, H.M., Heuer, O.E., Sørensen, A.I., Madsen, M., 2013. Analysis of factors important for the occurrence of Campylobacter in Danish broiler flocks. Prev. Vet. Med. 111, 100-111. van de Giessen, A.W., Bloemberg, B.P., Ritmeester, W.S., Tilburg, J.J., 1996. Epidemiological study on risk factors and risk reducing measures for Campylobacter infections in Dutch broiler flocks. Epidemiol. Infect. 117, 245-250. Wingstrand, A., Neimann, J., Engberg, J., Nielsen, E. M., Gerner-Smidt, P., Wegener, H. C., Molbak, K., 2006. Fresh chicken as main risk factor for campylobacteriosis, Denmark. Emerg. Infect. Dis. 12, 280-285. WHO, 2012. The Global View of Campylobacteriosis. Report of an expert consultation, Utrecht, Netherlands, 9-11 July 2012. http://www.who.int/iris/bitstream/10665/80751/1/9789241564601_eng.pdf
30
Table 1. Significant explanatory variables for the main model of Campylobacter prevalence in Danish and Norwegian broiler farms. The pvalue is given for each significant variable together with the effect-estimate, standard error, odds ratio and OR confidence intervals.
Main model - DK and NO (summer)
Type 3 p-value
Intercept
-
Age of house
0-5 year
Anteroom/ barrier
Country
Standard Error
Estimate
Odds Ratio
95% Confidence intervals [lower; upper]
-
-
0.63
[0.41 ; 0.98]
-
-1.83
0.80
0.002
-0.46
0.22
6-15 years
-
0.25
0.13
1.28
> 15 years
-
0
0
-
Yes
0.003
-0.37
0.13
0.69
No
-
0
0
-
DK
