Hostname: page-component-7c8c6479df-r7xzm Total loading time: 0 Render date: 2024-03-29T05:45:54.478Z Has data issue: false hasContentIssue false

Associations between host characteristics and antimicrobial resistance of Salmonella Typhimurium

Published online by Cambridge University Press:  03 December 2013

I. RUDDAT*
Affiliation:
Department of Biometry, Epidemiology and Information Processing, WHO Collaborating Centre for Research and Training in Veterinary Public Health, University of Veterinary Medicine, Hannover, Germany
E. TIETZE
Affiliation:
Robert Koch-Institute, Wernigerode Branch, Division Enteropathogenic Bacteria and Legionella, National Reference Centre for Salmonellae and other Bacterial Enteric Pathogens, Wernigerode, Germany
D. ZIEHM
Affiliation:
Governmental Institute of Public Health of Lower Saxony, Hannover, Germany
L. KREIENBROCK
Affiliation:
Department of Biometry, Epidemiology and Information Processing, WHO Collaborating Centre for Research and Training in Veterinary Public Health, University of Veterinary Medicine, Hannover, Germany
*
*Author for correspondence: Mrs I. Ruddat, Department of Biometry, Epidemiology and Information Processing, University of Veterinary Medicine, Hannover, Bünteweg 2, D-30559 Hannover, Germany. (Email: inga.ruddat@tiho-hannover.de)
Rights & Permissions [Opens in a new window]

Summary

A collection of Salmonella Typhimurium isolates obtained from sporadic salmonellosis cases in humans from Lower Saxony, Germany between June 2008 and May 2010 was used to perform an exploratory risk-factor analysis on antimicrobial resistance (AMR) using comprehensive host information on sociodemographic attributes, medical history, food habits and animal contact. Multivariate resistance profiles of minimum inhibitory concentrations for 13 antimicrobial agents were analysed using a non-parametric approach with multifactorial models adjusted for phage types. Statistically significant associations were observed for consumption of antimicrobial agents, region type and three factors on egg-purchasing behaviour, indicating that besides antimicrobial use the proximity to other community members, health consciousness and other lifestyle-related attributes may play a role in the dissemination of resistances. Furthermore, a statistically significant increase in AMR from the first study year to the second year was observed.

Type
Original Papers
Copyright
Copyright © Cambridge University Press 2013 

INTRODUCTION

Antimicrobial use is seen to be the key reason for effective proliferation of resistant bacteria [Reference Acar and Röstel1]. Nonetheless, showing the association between antimicrobial use of hosts and antimicrobial resistance (AMR) of strains they carried at the population level can be complicated. There are studies which were able to show this association based on individual patient data [e.g. Reference Costelloe2, Reference Malhotra-Kumar3] and on surveillance data [Reference Goossens4]. However, other studies observed no association [Reference Seidman5, Reference Walson6]. Several studies even showed the presence of resistant strains in humans or animals not using antimicrobials [Reference Walson6, Reference Bartoloni7]. This demonstrates that the dissemination of AMR is very complex and many factors may play a role.

Zoonotic bacteria like Salmonella are of special concern in monitoring programmes because of EU Zoonoses Directive 2003/99/EC [8]. Due to the use of antimicrobial agents in animals and humans, they may be subject to variable selective pressures [Reference Schwarz, Kehrenberg and Walsh9]. In each host, various AMR genes may be acquired by horizontal transfer between strains of the same or different bacteria species [Reference Schwarz, Kehrenberg and Walsh9, Reference Aarestrup and Aarestrup10]. Underlining this, a recent study showed differences in resistance profiles of Salmonella Typhimurium DT104 occurring in animals and humans [Reference Mather11].

Concerning the spread of AMR between different individuals, it is assumed that resistance genes are transferred via the same routes as bacterial strains [Reference Aarestrup and Aarestrup10]. As meat is the main source for S. Typhimurium infection, indirect transmission of resistance genes between farm animals and humans via food is probable [Reference Sorum and L'Abee-Lund12]. This assumption is supported by the association between raw meat handling and AMR found in Escherichia coli [Reference Coleman13]. Transmission via direct contact of humans with animals is also possible, this having been shown for food animals for transmission of resistant S. Typhimurium [Reference Hendriksen14] and for companion animals for transmission of resistant Staphylococcus intermedius [Reference Guardabassi, Loeber and Jacobson15]. Regarding human populations only, person-to-person spread was shown in studies regarding persons from same households or from the same communities [Reference Melander16, Reference Miller17]. Sociodemographic attributes may reflect AMR patterns, e.g. proximity to other community members [Reference Walson6, Reference Bruinsma18], and contact with healthcare services [Reference Cantón and Bryan19] or daycare centres [Reference Reves20] were described as risk factors for AMR.

Another important point regarding resistance dissemination is that different resistance genes may be genetically linked, e.g. located together on a plasmid, and thus could be transferred together to other bacterial strains [Reference Aarestrup and Aarestrup10]. This increases the occurrence of resistance genes even if the particular antimicrobial was not used, which hampers the analysis of an association between resistance against a single antimicrobial and the use of that particular antimicrobial. Phenotypic susceptibility data are typically analysed without information on the presence of resistance genes or mobile genetic elements. Therefore, to avoid biased results and misinterpretation, it is beneficial to regard the whole resistance profile as outcome and include correlations between resistance properties by performing a multivariate analysis.

In this study, we conducted an explorative risk factor analysis for AMR of S. Typhimurium isolates collected within a case-control study of sporadic salmonellosis cases in humans from Lower Saxony, Germany [Reference Ziehm21, Reference Ziehm22]. Individual patient data were available for sociodemographic factors, medical history, animal contact and food habits. Therefore, a detailed description of host behaviour covering the time before infection was given and associations between these factors and AMR were analysed. Investigations on linkages between resistance properties have been published for a part of the study data [Reference Ruddat23]. These confirm a high level of association between resistance properties within the test population and support the microbiological observation that resistance genes are often genetically linked and/or occur together. Therefore, in this study, entire resistance profiles were investigated as multivariate outcome in the risk-factor models using a non-parametric approach, performing between-group comparisons by permutation tests. Calculations were based on minimum inhibitory concentration (MIC) values.

MATERIALS AND METHODS

Study design

Case data were available from a case-control study on sporadic Salmonella enterica infections in humans in Lower Saxony, Germany, conducted between June 2008 and May 2010 by the State Health Department of Lower Saxony (Niedersächsisches Landesgesundheitsamt; NLGA) [Reference Ziehm21, Reference Ziehm22]. Notified sporadic cases of salmonellosis that were not part of apparent local outbreaks were recruited and standardized questionnaires were completed via telephone interviews conducted by trained interviewers from NLGA and local public health departments of the districts. Salmonella isolates were made available through cooperating sentinel laboratories and were sent to the German National Reference Centre for Salmonella at the Robert Koch-Institute, Wernigerode, where the serotype, phage type and antimicrobial susceptibility profile were determined. As isolates with different serotypes show different resistance distributions, in this study we focused on S. Typhimurium; this was, with 52% (n = 383), the most frequent serotype in the sampling period.

Host data

The questionnaires comprised questions concerning risk factors for salmonellosis, which were divided into four groups: general patient information, medical history, animal-related factors, and food-related factors. Analysing categorical variables which had a minimum of eight observations in each category, a total of 63 factors were investigated which are listed in Table 1.

Table 1. List of host factors analysed to investigate the association with antimicrobial resistance

Factors with significant association with antimicrobial resistance are underlined (see Table 4).

a Female, male.

b 0 to ⩽2 years, 2 to ⩽14 years, 14 years.

c Conurbation area, urbanized area, rural area (according to a region type classification in [37]).

d June 2008 to May 2009, June 2009 to May 2010.

e Yes, no.

f Concerning 3 days before onset of disease.

g Concerning 4 weeks before onset of disease.

Phenotypic data

Serotyping of S. Typhimurium was conducted according to the White–Kauffmann–LeMinor scheme by slide agglutination with O- and H-antigen-specific sera [Reference Grimont and Weill24]. Phage typing of S. Typhimurium isolates was performed in compliance with previous work [Reference Rabsch25, Reference Anderson26].

Antimicrobial susceptibility tests were conducted by broth microdilution according to document 58940-8 of the German Institute for Standardisation (Deutsches Institut für Normung; DIN) [27] obtaining MIC values for 13 antimicrobial agents (Table 2). E. coli ATCC® 25 922 served as quality control strain. For calculations with MIC values, we used the following conventions: if the lowest test concentration of an antimicrobial agent already inhibited growth, the corresponding MIC value was recorded as being equal to this concentration. If the highest test concentration of an antimicrobial agent did not inhibit growth, the corresponding MIC value was recorded as the next concentration in a twofold dilution series.

Table 2. Tested antimicrobial agents, dilution ranges, clinical breakpoints (μg/ml) and proportions of resistant strains for Salmonella Typhimurium isolates (n = 383)

S, Susceptible; R, resistant.

Furthermore, the isolates were categorized as resistant (R) and non-resistant (NR, summarizing susceptible and intermediate isolates) using clinical breakpoints recommended in DIN 58940-4 [27]. If DIN breakpoints were not available, breakpoints given in document M100-S20 [28] of the Clinical Laboratory and Standards Institute (CLSI) were used. This was done for nalidixic acid and kanamycin. As no internationally accepted breakpoints exist for streptomycin we used the breakpoints accepted by the Robert Koch-Institute in their long-term Salmonella monitoring programmes. The breakpoints used are listed in Table 2. Isolates were classified as multi-resistant when they exhibited resistance to at least three antimicrobial agents of different classes/subclasses [Reference Schwarz29].

Statistical models

Risk-factor analyses with multivariate MIC profiles as outcome were conducted using distance-based permutation tests suggested by Anderson [Reference Anderson30] and McArdle & Anderson [Reference McArdle and Anderson31]. This approach is non-parametric and appropriate for unbalanced and multifactorial designs. For this case, the n × p data matrix Y, containing the 2-log-transformed MIC values on n isolates for p antimicrobial agents, was transferred into the n × n distance matrix D, whose elements d ij reflect the distance between isolate i and isolate j with respect to the p 2-log-transformed MIC values (for distance calculation see below). Summing up the distances between all possible pairs of isolates gives an estimator for total variability. The total variability can be partitioned into the variability between groups and the variability within groups by estimating the variance components based on distances only. The test statistic for conducting between-group comparisons of MIC profiles is then the ratio of both variance components, which follows the same philosophy as in analysis of variance (ANOVA) for univariate data. In the case of a multifactorial model the total variation is partitioned into several variance components, again directly from the distance matrix, and appropriate test statistics are formalized. In the case of only one outcome variable (p = 1) and calculating distances as Euclidean distance, the test statistic for testing one risk factor is equal to the common F statistic in a one-way ANOVA. However, for this approach any metric or non-metric distance measure can be chosen, and it can be applied to variables with any measurement levels [Reference Anderson30, Reference McArdle and Anderson31]. Appropriate P values can be obtained using permutation tests.

Distances between MIC profiles were calculated as follows: as tested concentration steps for generating MIC values were limited, observed MIC distributions occurred as truncated and a large number of ties were possible. Therefore, the level of measurement for MIC values was understood as ordinal and the Manhattan distance was used. The distance d ij between isolate i and isolate j, with i, j = 1, …, n and ij, was calculated by

$$d_{ij} = \sum\limits_{k = 1}^p {\displaystyle{{\left| {y_{ik} - y_{\,jk}} \right|} \over {c_k}}} ,$

where y ik and y jk denote the 2-log MIC for antimicrobial agent k, k = 1, …, p of the ith and jth isolate, and c k is the number of tested concentration steps for antimicrobial agent k. Scaling the distances by the particular test ranges, leads to a maximum distance of 1 for each antimicrobial agent and to a maximum distance of p for d ij .

For each host characteristic (see Table 1) a model was generated to investigate the association with the multivariate MIC profiles. As resistance distributions depend on phage type, this was considered as a confounder variable in each model with six categories of isolates: DT104, DT120 monophasic, DT120 biphasic, DT193 monophasic, DT193 biphasic, and others (see Table 3). Therefore, each host characteristic was examined within a two-factor model involving phage type, the host characteristic and an interaction term as fixed factors. Analyses were conducted with R version 2.15.2 [32] with function adonis, package vegan version 2.0–4 [Reference Oksanen33]. For all tests a subset of 9999 permutations was used. In this study P values ⩽0·05 were interpreted as statistically significant. In cases where two-way interactions were not meaningful (less than three observations per group) or did not appear to be statistically significant, models with main effects only were investigated. If there was a statistically significant result for testing a factor or an interaction with three or more groups, we conducted pairwise comparisons between groups. As this study is of an exploratory nature, no α-adjustment to account for multiple hypotheses testing was applied. For interpreting significant differences between groups, mean number of resistances (MNR) were calculated to quantify the degree of resistance per group. To visualize group differences in MIC distributions MIC10, MIC50 and MIC90 are displayed, which are the 10%, 50% and 90% distribution percentiles [Reference Schwarz29].

Table 3. Resistance characteristics for Salmonella Typhimurium subtypes analysing antimicrobial resistance against 13 antimicrobial agents

MNR, Mean number of resistances; MR, multi-resistant isolates; NR, fully non-resistant pattern; RP, resistance profile.

For evaluating MIC profile-based results, each model was recalculated replacing the MIC profile with the binary-coded resistance profile based on the information resistant or non-resistant. For this, the distance-based permutation tests were performed using the simple matching coefficient to obtain distances between the binary-coded profiles.

RESULTS

During the whole study period 867 notified sporadic cases with identified serovar Typhimurium were recorded with a median age of 13 years and 53·9% male participants [Reference Ziehm22]. Of these, 540 cases were recorded in the first study year (median age 13 years, 56·9% males) and 327 cases were recorded in the second year (median age 11 years, 48·9% males). Phage typing and susceptibility tests were conducted for 383 of these isolates (first year, 250 isolates; second year, 133 isolates). In this study population the median age was 12 years (first year, 12 years; second year, 12 years) and the proportion of males was 53·0% (first year, 56·8%; second year, 45·9%).

Frequencies of observed phage types are given in Table 3. Testing the association between these phage-type categories and the multivariate MIC profiles in a one-factor model resulted in a global P value of <0·001. Each pairwise comparison showed statistically significant differences except for the comparison between DT120 monophasic and DT120 biphasic isolates (P = 0·155) as well as the comparison between DT120 biphasic isolates and the group of other phage-type isolates (P = 0·636).

Group MNRs and P values are listed in Table 4 for host factors, which showed statistically significant associations with multivariate MIC profiles in the two-way models. Corresponding MIC distributions are shown in Figure 1.

Fig. 1. Comparison of MIC distributions between groups of significant host characteristics. (In cases of significant interactions only significant phage-type subgroups are presented. The symbols represent locations of MIC50. The vertical lines represent the range between MIC10 and MIC90 in each group. Horizontal lines represent the clinical breakpoints used. The concentration ranges tested are those contained in the white area. Values above this range are due to isolates which were not inhibited by the largest tested concentration, values corresponding to the lowest concentration are due to isolates which were already inhibited by the smallest tested concentration.)

Table 4. Descriptive information on host-specific factors with significant results (P ⩽0·05) in two-factorial models with adjustment for phage types analysing the association with antimicrobial resistance for Salmonella Typhimurium isolates (n = 383)

a MNR, Mean number of resistances.

b In the case of significant phage-type interaction only the P value is given and MNRs are given at the phage-type level.

c Information is given at phage-type level, if the phage-type interaction was significant.

d Significant main factor, interaction was not significant.

e Significant phage-type interaction.

f CA, Conurbation area; UA, urbanized area; RA, rural area.

Concerning variables on general patient information, differences between study years and region types were found to be statistically significant. Comparing the MIC distributions and the MNRs, the second study year tended to have higher MIC values and higher numbers of resistances than the first year independent of phage type. Region type was statistically significant for phage-type DT120 biphasic isolates, comparing conurbation with urbanized areas and conurbation with rural areas and for phage-type DT193 monophasic isolates, comparing conurbation with urbanized areas. In these cases isolates of patients living in conurbation areas tended to have higher MIC values than the other two groups.

For medical history, antimicrobial consumption had a statistically significant association with AMR for phage-type DT120 biphasic isolates, where the group with consumption showed an increased MNR and higher MIC values. For other subgroups no significant associations were observed.

While no animal-related factor was statistically significant, three food-related factors were detected, all referring to egg-purchasing behaviour. The association between AMR and the factors direct producer delivery and purchasing free-range eggs depended on phage type. For both, DT120 biphasic isolates of patients with that particular behaviour had lower MIC values and lower MNRs than those of patients without the behaviour. The other phage-type subgroups were not significant. Purchasing barn eggs was associated with MIC values independent of phage type. Higher values and higher numbers of resistances were found in isolates of patients that purchased barn eggs.

DISCUSSION

The objective of this study was to identify and assess the importance of factors associated with AMR in S. Typhimurium isolates from sporadic salmonellosis cases in humans from Lower Saxony, Germany. Lower Saxony is a federal state where agriculture and livestock farming are of major importance. It is located in north-west Germany and the total population is 7·9 million with a population density of 166·5 residents/km2 [34].

Testing the association between phage type and AMR resulted in a highly statistically significant P value. Nearly all DT104 and DT193 monophasic and biphasic isolates were multi-resistant (Table 3). On the other hand, for DT120 biphasic isolates and the group of other phage types the most common resistance profile was the fully susceptible pattern. These phage-type-specific resistance distributions imply that it is necessary to analyse resistance data at the phage-type level, not only on serotype level as is commonly done. In cases where phage types are distributed differently in the risk factors of interest, the estimated effect will be biased. In this study, we included phage type as a confounder in each model and tested for interactions instead of conducting a stratified analysis. In cases of no significant interaction we were able to interpret the results at the serotype level, thereby gaining a higher sample size and more confidence in the results.

For interpreting statistically significant host characteristics we assumed that persons with risk factors have an increased probability of having a higher frequency of resistant bacteria in their faecal flora and/or that the S. Typhimurium isolates transmitted to those persons have a higher probability of having increased MIC values.

Regarding antimicrobial use, it was known whether a person in the study population had taken any antimicrobials during the previous 4 weeks before onset of the disease, but no further specifications concerning agent or duration of use were available. Antimicrobial use was associated with increased AMR in DT120 biphasic isolates.

In the 2-year study period higher MIC values were observed in the second year. This trend was observed continually between 1999 and 2008 in German Salmonella monitoring [35]. Although absolute numbers of prescriptions in Germany remained almost constant from 1991 to 2011, the proportion of reserve antimicrobials, e.g. fluoroquinolones and cephalosporins, increased continuously [Reference Schroeder36]. In agreement with this finding, in our study the MIC90 increased for nalidixic acid in the second year.

Region type was the factor with the lowest P value when testing for the association with AMR. Increased MIC values and number of resistances were observed for conurbation areas for DT120 biphasic and DT193 monophasic isolates. The region categories were defined based on structural characteristics for settlements combining aspects of population density and accessibility of centres [37]. A descriptive comparison between region type and district-based ambulant antimicrobial prescription data for children (aged 0–17 years) for 2010 [Reference Glaeske38] showed no clear association. For adults no German data at district level have been published. Assuming that prescriptions for adults are correlated with those for children, we do not suppose that antimicrobial use may explain the observed region association. Effects on AMR were suggested for contact with other community members in isolated populations in Nepal [Reference Walson6] and for population density comparing populations of three cities in Canada, Greece and The Netherlands [Reference Bruinsma18]. These factors are associated with these region types by definition and may play a role in this study, too. In general, assuming that different region types represent different lifestyles of persons, many factors may be associated with it and this may be the real risk factor behind region type for AMR. The only conclusion we can draw is that the overall exposure to AMR was estimated to be larger for persons living in conurbation areas.

No animal-related factor was identified as a risk factor for AMR. The smallest P value was observed for having contact with dogs, with P = 0·090 for testing the interaction with phage type. It should be noted that in the study population only a few individuals had contact with cattle, pigs, horses or reptiles. Therefore, phage-type adjustment was not possible and results for these species are not very reliable.

The only statistically significant results for food-related factors were observed for the three factors concerning egg-purchasing behaviour, where purchasing barn eggs was associated with higher MIC values and increased number of resistances while free-range eggs and direct egg delivery were associated with lower MIC values and decreased number of resistances. As eggs are not a typical infection source for S. Typhimurium, the spread of resistance while handling or consuming eggs was not assumed to be important here. A study on Canadian egg consumers found that health-conscious consumers were willing to pay more for eggs when they had special attributes, e.g. free-range or Omega-3 eggs [Reference Goddard39]. As the proportion of sold cage eggs is very low and decreasing further, barn eggs can be seen as the cheapest egg class available in German food retailing [40]. Purchasing free-range eggs or purchasing eggs directly from the producer would then, assuming the same association here, reflect a greater health consciousness. As these considerations are speculative, more specific data are needed for further investigations.

The multivariate non-parametric model approach used here was first introduced for applications in ecology [Reference Anderson30, Reference McArdle and Anderson31]. As calculations are based on distances only (and any distance measure can be chosen), this method can be applied where in principle a multivariate analysis of variance (MANOVA) design is available but the outcome variables do not fulfil the assumption of being normally distributed. In this study, we assessed the MIC values for 13 antimicrobial agents as outcome. These are ordinal and the Manhattan distance was used. Alternatively, resistance profiles using the classified information ‘resistant’ and ‘non-resistant’ may be analysed with this approach using, e.g. the simple matching coefficient to measure distances. Analysing the associations between host characteristics and AMR based on these binary-coded resistance profiles with adjustment for phage type resulted in identifying the same statistically significant factors, interactions and pairwise comparisons with similar P values as in the analysis of MIC profiles except for study year, for which testing the main factor resulted in P = 0·265. This shows that variation in MIC profiles corresponded basically with variation in the binary-coded resistance profiles in this study. For study year, the group MNRs show slight differences (Table 4), but the MIC distributions in Figure 1 show that group differences in MIC10, MIC50 or MIC90 values were observed below the resistance breakpoint, and cannot be reflected in the binary-coded profiles. Therefore, the advantage of analysing MIC values is that variation below and above the breakpoint is accounted for. For Salmonella and other species, changes concerning low resistance antimicrobials like ciprofloxacin or the cephalosporins are of special interest. For these, slight increases are relevant, even when values remain below the breakpoint.

Like MANOVA (and even univariate ANOVA) this distance-based approach is based on the assumption of homoscedasticity, which is that group variances have to be equal [Reference McArdle and Anderson31]. If this assumption does not hold, a factor may be declared significant due to differences in location or differences in dispersion. In our opinion, for susceptibility data analysis any kind of differences associated with a special host characteristic may be interesting for collecting information on the dissemination of resistance genes. Therefore, the test results are interesting in both a homoscedastic and a heteroscedastic situation, but for the interpretation a description of location and variation within groups has to be considered. This was done here by displaying the MIC50 and the range between MIC10 and MIC90 per group and antimicrobial agent.

Limitations of this study arose due to sample size. All results observed are based on two-way risk-factor models, screening for significant associations between single host characteristics (Table 1) and multivariate resistance profiles with adjustment for phage type of isolates. An adjustment for further covariates and analysing multifactorial models would have provided further insight into associations with AMR. In this study, it was not possible to do this as subgroups would have been too small for meaningful comparisons with adequate power. Furthermore, as this study was of an exploratory nature, no α-adjustment was made for multiple pairwise comparisons between groups. Therefore, this study should be considered for generating hypotheses and all results reported here need to be confirmed in future studies.

In summary, by analysing multivariate resistance profiles of MIC values we determined statistically significant phage-type adjusted associations with antimicrobial intake, study year, region type and three factors on egg-purchasing behaviour for S. Typhimurium isolates collected in Lower Saxony, Germany. We conclude that besides antimicrobial use other host characteristics such as proximity to other community members, health consciousness and other lifestyle-related factors may play an important role in the dissemination of resistance genes. As this study was of an exploratory nature, all observed results should be interpreted as initial evidence, and further research is needed for validation.

ACKNOWLEDGEMENTS

This study is part of the FBI Zoo Consortium, financially supported by the German Federal Ministry of Education and Research (BMBF) through the German Aerospace Centre (DLR), grant nos. 01KI07128 and 01KI1012F.

DECLARATION OF INTEREST

None.

References

REFERENCES

1. Acar, J, Röstel, B. Antimicrobial resistance: an overview. Revue Scientifique et Technique de l'Office International des Epizooties 2001; 20: 797810.Google Scholar
2. Costelloe, C, et al. Effect of antibiotic prescribing in primary care on antimicrobial resistance in individual patients: systematic review and meta-analysis. British Medical Journal 2010; 340: c2096.Google Scholar
3. Malhotra-Kumar, S, et al. Effect of azithromycin and clarithromycin therapy on pharyngeal carriage of macrolide-resistant streptococci in healthy volunteers: a randomised, double-blind, placebo-controlled study. Lancet 2007; 369: 482490.CrossRefGoogle Scholar
4. Goossens, H, et al. Outpatient antibiotic use in Europe and association with resistance: a cross-national database study. Lancet 2005; 365: 579587.Google Scholar
5. Seidman, JC, et al. Risk factors for antibiotic-resistant E. coli in children in a rural area. Epidemiology and Infection 2009; 137: 879888.Google Scholar
6. Walson, JL, et al. Carriage of antibiotic-resistant fecal bacteria in nepal reflects proximity to Kathmandu. Journal of Infectious Diseases 2001; 184: 11631169.Google Scholar
7. Bartoloni, A, et al. Multidrug-resistant commensal Escherichia coli in children, Peru and Bolivia. Emerging Infectious Diseases 2006; 12: 907913.Google Scholar
8. The European Parliament and the Council of the European Union. Directive 2003/99/EC on the monitoring of zoonoses and zoonotic agents, amending Council Decision 90/424/EEC and repealing Council Directive 92/117/EEC. Official Journal of the European Union (http://eur-lex.europa.eu/LexUriServ/site/en/oj/2003/l_325/l_32520031212en00310040.pdf). 2003.Google Scholar
9. Schwarz, S, Kehrenberg, C, Walsh, TR. Use of antimicrobial agents in veterinary medicine and food animal production. International Journal of Antimicrobial Agents 2001; 17: 431437.Google Scholar
10. Aarestrup, FM. The origin, evolution, and local and global dissemination of antimicrobial resistance. In: Aarestrup, FM, ed. Antimicrobial Resistance in Bacteria of Animal Origin. Washington D.C.: ASM Press, 2006, pp. 339359.Google Scholar
11. Mather, AE, et al. An ecological approach to assessing the epidemiology of antimicrobial resistance in animal and human populations. Proceedings of the Royal Society of London, Series B: Biological Sciences 2012; 279: 16301639.Google Scholar
12. Sorum, H, L'Abee-Lund, TM. Antibiotic resistance in food-related bacteria – a result of interfering with the global web of bacterial genetics. International Journal of Food Microbiology 2002; 78: 4356.Google Scholar
13. Coleman, BL, et al. The role of drinking water in the transmission of antimicrobial-resistant E. coli. Epidemiology and Infection 2012; 140: 633642.Google Scholar
14. Hendriksen, SW, et al. Animal-to-human transmission of Salmonella Typhimurium DT104A variant. Emerging Infectious Diseases 2004; 10: 22252227.Google Scholar
15. Guardabassi, L, Loeber, ME, Jacobson, A. Transmission of multiple antimicrobial-resistant Staphylococcus intermedius between dogs affected by deep pyoderma and their owners. Veterinary Microbiology 2004; 98: 2327.Google Scholar
16. Melander, E, et al. Frequency of penicillin-resistant pneumococci in children is correlated to community utilization of antibiotics. Pediatric Infectious Disease Journal 2000; 19: 11721177.Google Scholar
17. Miller, YW, et al. Sequential antibiotic therapy for acne promotes the carriage of resistant staphylococci on the skin of contacts. Journal of Antimicrobial Chemotherapy 1996; 38: 829837.Google Scholar
18. Bruinsma, N, et al. Influence of population density on antibiotic resistance. Journal of Antimicrobial Chemotherapy 2003; 51: 385390.Google Scholar
19. Cantón, R, Bryan, J. Global antimicrobial resistance: from surveillance to stewardship. Part 1: surveillance and risk factors for resistance (Meeting report of 22nd European Congress of Clinical Microbiology and Infectious Diseases). Expert Review of Anti-infective Therapy 2012; 10: 12691271.Google Scholar
20. Reves, RR, et al. Risk factors of fecal colonization with trimethoprim-resistant and multiresistant Escherichia coli among children in day-care centers in Houston, Texas. Antimicrobial Agents and Chemotherapy 1990; 34: 14291434.Google Scholar
21. Ziehm, D, et al. Risk factors associated with sporadic salmonellosis in adults: a case-control study. Epidemiology and Infection 2012; 141: 284292.Google Scholar
22. Ziehm, D, et al. Subtype specific risk factor analyses for sporadic human salmonellosis: a case-case comparison in Lower Saxony, Germany. International Journal of Hygiene and Environmental Health 2012; 216: 428434.Google Scholar
23. Ruddat, I, et al. A quantitative approach to analyse linkages between antimicrobial resistance properties in Salmonella Typhimurium isolates. Epidemiology and Infection 2012; 140: 157167.Google Scholar
24. Grimont, PAD, Weill, F. Antigenic formulae of the Salmonella serovars. Paris, France: WHO Collaborating Centre for Reference and Research on Salmonella, Institut Pasteur (http://www.pasteur.fr/ip/portal/action/WebdriveActionEvent/oid/01s-000036-089). Accessed 14 July 2013. 2007.Google Scholar
25. Rabsch, W. Salmonella Typhimurium phage typing for pathogens. Methods in Molecular Biology 2007; 394: 177211.Google Scholar
26. Anderson, ES, et al. Bacteriophage-typing designations of Salmonella Typhimurium. Journal of Hygiene 1977; 78: 297300.Google Scholar
27. German Institute for Standardization (DIN). Methods for susceptibility testing of bacterial pathogens against chemotherapeutics [in German]. In: Medical Microbiology and Immunology; Diagnostic Procedures. Berlin: Beuth, 2004, pp. 245430.Google Scholar
28. Clinical and Laboratory Standards Institute. Performance Standards for Antimicrobial Susceptibility Testing, Twentieth Informational Supplement M100-S20: Wayne, PA, 2010.Google Scholar
29. Schwarz, S, et al. Assessing the antimicrobial susceptibility of bacteria obtained from animals. Veterinary Microbiology 2010; 141: 14.Google Scholar
30. Anderson, MJ. A new method for non-parametric multivariate analysis of variance. Austral Ecology 2001; 26: 3246.Google Scholar
31. McArdle, BH, Anderson, MJ. Fitting multivariate models to community data: a comment on distance-based redundancy analysis Ecology 2001; 82: 290297.Google Scholar
32. R Core Team. R: A language and environment for statistical computing (http://www.R-project.org). Vienna, Austria: R Foundation for Statistical Computing, 2012.Google Scholar
33. Oksanen, J, et al. Package ‘vegan’ – Community Ecology Package. 2012.Google Scholar
34. Lower Saxony State Office of Statistics and Communications Technology (LSKN). Population of municipalities on 30 June 2010 (www.lskn.niedersachsen.de/portal/live.php?navigation_id=25688&article_id=87679&-psmand=40). Accessed 20 April 2013.Google Scholar
35. GERMAP 2010. Report on the consumption of antibiotics and the spread of antibiotic resistance in human and veterinary medicine in Germany [in German] (http://www.p-e-g.org/econtext/germap). Accessed 12 January 2012.Google Scholar
36. Schroeder, H. Hands away from reserve antibiotics [in German]. Gesundheit und Gesellschaft 2011; 7–8.Google Scholar
37. Federal Institute for Research on Building Urban Affairs and Spatial Development. Ongoing spatial monitoring – region types based on settlement structures [in German] (http://www.bbsr.bund.de/cln_016/nn_103086/BBSR/DE/Raumbeobachtung/Werkzeuge/Raumabgrenzungen/SiedlungsstrukturelleGebietstypen/Regionstypen/regionstypen.html). Accessed 14 August 2010.Google Scholar
38. Glaeske, G, et al. Antibiotic prescriptions for children [in German] (http://www.bertelsmann-stiftung.de/cps/rde/xbcr/SID-36F662C7-73499B9C/bst/xcms_bst_dms_35589_35590_2.pdf). Accessed 20 April 2013. 2012.Google Scholar
39. Goddard, EW, et al. Consumer attitudes, willingness to pay and revealed preferences for different egg production attributes: analysis of Canadian egg consumers (http://purl.umn.edu/52087), 2007. Accessed 15 April 2013.Google Scholar
40. Federal Ministry of Food Agriculture and Consumer Protection. Data on German egg market [in German] (http://www.bmelv.de/SharedDocs/Downloads/Landwirtschaft/Tier/TierzuchtTierhaltung/KennzahlenEiermarkt-Maerz2013.pdf). Accessed 11 June 2013.Google Scholar
Figure 0

Table 1. List of host factors analysed to investigate the association with antimicrobial resistance

Figure 1

Table 2. Tested antimicrobial agents, dilution ranges, clinical breakpoints (μg/ml) and proportions of resistant strains for Salmonella Typhimurium isolates (n = 383)

Figure 2

Table 3. Resistance characteristics for Salmonella Typhimurium subtypes analysing antimicrobial resistance against 13 antimicrobial agents

Figure 3

Fig. 1. Comparison of MIC distributions between groups of significant host characteristics. (In cases of significant interactions only significant phage-type subgroups are presented. The symbols represent locations of MIC50. The vertical lines represent the range between MIC10 and MIC90 in each group. Horizontal lines represent the clinical breakpoints used. The concentration ranges tested are those contained in the white area. Values above this range are due to isolates which were not inhibited by the largest tested concentration, values corresponding to the lowest concentration are due to isolates which were already inhibited by the smallest tested concentration.)

Figure 4

Table 4. Descriptive information on host-specific factors with significant results (P ⩽0·05) in two-factorial models with adjustment for phage types analysing the association with antimicrobial resistance for Salmonella Typhimurium isolates (n = 383)