Association between tetracycline consumption and tetracycline resistance in escherichia coli from healthy danish slaughter pigs

FOODBORNE PATHOGENS AND DISEASEVolume 6, Number 1, 2009ª Mary Ann Liebert, Inc.
DOI: 10.1089=fpd.2008.0152 Association Between Tetracycline Consumption and Tetracycline Resistance in Escherichia coli Antonio Roberto Vieira,1,2 Hans Houe,2 Henrik C. Wegener,1 Danilo M.A. Lo Fo Wong,3 and Hanne-Dorthe Emborg1 It has been recognized that exposure to antimicrobial agents can exert a selective pressure for the emergenceof antimicrobial resistance. The objective of this study was to investigate an association between theprobability of isolating a tetracycline-resistant Escherichia coli isolate from the intestinal tract of healthy pigsand patterns of tetracycline consumption in the herds of origin, together with other risk factors. Data onantimicrobial resistance, antimicrobial consumption, and pig herd demographics were obtained from dif-ferent Danish surveillance programs. Descriptive statistics were performed for the risk factors in relation tothe susceptibility status. Logistic regression analysis was performed to identify risk factors with significanteffect on the log odds of tetracycline resistance of E. coli isolates. The model showed that an increase in theinterval between last prescription and sampling date would decrease the probability of isolating a resistantE. coli isolate ( p-value ¼ 0.01). Also, a direct association between treatment incidence rate in a herd andprobability of resistance was detected ( p-value ¼ 0.03). Other risk factors found to have a significant effect inthe isolate susceptibility status were number of produced animals in the year and year of sampling. Otherantimicrobial consumption risk factors, such as number of prescriptions and amount prescribed, althoughnot included in the final model, presented indirect impact in the tetracycline resistance probability. From thisstudy, we can infer that tetracycline usage, the time span between last treatment and sampling date, togetherwith herd size and the proportion of animals being treated in a herd, increase the probability of obtaining aresistant isolate.
ing from higher healthcare costs up to increase inmortality rate due to treatment failure (Cohen, Antimicrobials can be broadly defined as 1992,1994;Wiseetal.,1998).Itisrecognizedthat substances that can destroy or inhibit the exposure to antimicrobial agents can exert a se- growth of other microorganisms. In the veteri- lective pressure for the emergence of resistance, nary field, antimicrobials are used for treatment, and a major risk factor for the development of prevention, and control of infectious diseases in resistance is thought to be misuse and overuse animal production. However, resistance to an- of antimicrobial drugs (Aarestrup et al., 2000a, timicrobials was observed soon after their usage 2000b; Witte, 2000; Aarestrup, 2005). It has was introduced, and this has been a matter of resulted in several risk management strate- public health concern, with consequences rang- gies being developed and partly implemented 1National Food Institute, Technical University of Denmark, Søborg, Denmark.
2Faculty of Life Science, University of Copenhagen, Frederiksberg, Denmark.
3World Health Organization, Geneva, Switzerland.
worldwide in order to attempt to limit the microbial resistance. In particular, it is important emergence and spread of resistant bacteria in the in order to be able to conduct risk assessments food production chain (World Health Organi- and to make recommendations on antimicro- bial administration strategies to minimize the The association between antimicrobial con- impact of antimicrobial consumption on the sumption and the occurrence of resistance to a development of resistance. The objective of this given drug has been difficult to establish for study was to quantify the correlation between a variety of reasons, mainly because resis- tetracycline consumption at herd level and tance is an evolutionary and biologically com- occurrence of resistance in commensal E. coli plex process and because detailed data on isolated from slaughter pigs, by detecting sig- antimicrobial consumption and resistance are nificant risk factors on antimicrobial usage.
necessary to investigate the problem in detail(Fluit et al., 2006). Surveillance systems for antimicrobial resistance among veterinary iso- lates have been developed in a number ofcountries (Marano et al., 2000; EFSA, 2008).
The Danish Veterinary Medicines Statistics However, only Denmark currently presents a Program (VetStat) was created with the purpose fully developed and established antimicrobial of monitoring antimicrobial usage in production animals. It includes all the data obtained from (Stege et al., 2003). This program was first im- pharmacies, veterinary prescriptions, and also from preparation of medicated feed-by-feed its objectives was the provision of data for mills, including information on active sub- pharmaco-epidemiological research. Later, sev- stances, amounts, target species, age group, and eral years of antimicrobial consumption data diagnosis, as well as herd identification number were collected, making possible investigations (Stege et al., 2003). These data are available on on the risk factors impacting the association animal species at herd level and can be used to between antimicrobial consumption and resis- calculate the prescription frequency and the amount of antimicrobial drugs prescribed to each herd in a period of time. Measuring anti- antimicrobial classes used in the Danish pig microbial usage in terms of weight of active production. In 2005, a total of 12,446 kg was substance naturally has limitations due to dif- prescribed for slaughter pigs in Denmark, ferences in antimicrobial groups and in their making tetracycline one of the priority drugs on potency and, therefore, dosages are needed as antimicrobial consumption behavior studies standard measures. Since there is no defined (DANMAP, 2005). At the same time, tetracy- daily dose in veterinary medicine, the veterinary cline resistance levels have been increasing over equivalent, defined animal daily dose (ADD) the past few years. Resistance to tetracyclines was used in the VetStat system to standardize among Escherichia coli has peaked in 2004, when antimicrobial consumption and allow compari- 44% of the pig isolates presented phenotypical son between different antimicrobial compounds resistance. Fecal E. coli are often used for epide- and age groups. To each antimicrobial agent, miological studies on antimicrobial resistance in ADD was defined as the average maintenance the food chain due to their tendency to easily dose for the main indication, and calculated for develop antimicrobial resistance, their ability to each age group within each animal species transfer resistance genes, and their potential to work as a resistance source (Turnidge and The number of slaughter pigs in each herd was calculated using data from the Danish Ser- Food safety concerns often include the anti- ological Salmonella Surveillance Program on microbial resistance issue, and epidemiological pigs. This database provided by the Danish research is needed in order to improve the slaughterhouses presented the accurate number understanding of the association between anti- of slaughter pigs delivered by each registered microbial consumption and occurrence of anti- TETRACYCLINE CONSUMPTION AND RESISTANCE AMONG E. COLI FROM PIGS ferent size categories, according to the number of by this study was the Danish Integrated Anti- animals slaughtered per year (Table 1). Cutpoints microbial Resistance Monitoring and Research used to determine the four categories were based on data from the herd size distribution of the represents a close collaboration between veteri- nary, food, and human health authorities in order to provide comparable data to investigate trends identification number that allows the fully inte- of antimicrobial resistance, monitor consumption gration of data on antimicrobial consumption of antimicrobials, and explore associations be- and resistance (Bager, 2000). For each herd, tween usage and occurrence of resistance at na- information on tetracycline prescriptions for tional and regional levels. Samples are collected slaughter pigs was extracted from the VetStat from animals, foodstuffs, and humans. For pigs, database and combined with the results of the cecal contents are collected by meat inspection susceptibility testing. For each prescription, the staff at the slaughterhouses and submitted for herd identifier, prescribed tetracycline, date of isolation and susceptibility testing. Only one prescription, the amount prescribed, adminis- sample per herd is collected in any sampling tration route, and the calculated ADD were occasion. E. coli was isolated by direct inocula- used. Since the VetStat database has the anti- tion of the material onto Drigalski plates (Statens microbials classified based on their anatomical Serum Institut, Copenhagen, Denmark) and in- therapeutic chemical veterinary classification cubation at 378C overnight. Presumptive E. coli system, active substances included in the group QJ01A (chlortetracycline, doxycycline, oxytet- racycline, and tetracycline) were summarized as France), and red colonies were identified as E.
tetracyclines (Dahlin et al., 2001). Consumption coli after incubation at 378C overnight. Only one of tetracycline 12 months prior to the sampling E. coli colony from each sample was subcultured date was calculated for each herd as ADD. In and susceptibility tested using a commercially addition, the frequency of tetracycline prescrip- available minimum inhibitory concentration tion was calculated for the same period. Sam- (MIC) technique (Sensititre; Trek Diagnostic pling date was used to classify the isolate in one Systems, East Grinstead, UK). Data on antimi- of the 3 years included in the study, as well as crobial resistance for each isolate are collected in to calculate the interval between the date of a database (Bager, 2000). The DANMAP report the last prescription of tetracycline and the date is published annually containing the latest ob- of sampling (in days). In order to estimate the served trends regarding antimicrobial resistance number of slaughter pig-days per herd, for each in Denmark, as well as data on antimicrobial delivered slaughter pig a fattening period of 112 consumption provided by VetStat. Resistance data from indicator E. coli isolates from pig The proportion of tetracycline prescribed for herds in Denmark were analyzed for this study.
parenteral use in the 12 months prior to samplingwas assessed and included in the dataset as per-centage of ADD and frequency of ADD pre- scribed for parenteral administration. Last, the Indicator E. coli isolates from pig herds in treatment incidence rate in a herd (TI) was cal- Denmark collected through the DANMAP pro- culated as the total number of ADD prescribed gram between January 2003 and December 2005 within the 12-month period prior to the sampling were included. For each isolate, date of sampling, date divided by the estimated number of slaugh- herd identification, and the results of the sus- ter pig-days per herd during the same period.
ceptibility testing to tetracycline categorized assusceptible=resistant (susceptibility status) were available. The isolate susceptibility status was based on the standardized MIC breakpoint for tetracycline adopted by DANMAP (>8 mg=mL).
In addition, herds were classified into four dif- Table 1. Description of the Explanatory Variables for the Probability of Isolating a Tetracycline- Resistant Escherichia coli Isolate in Danish Slaughter Pigs Between 2003 and 2005 prescriptions in the 3 and12 months previous tosampling for slaughter pigs in the 3 and12 months previous to sampling parenteral use in the herd in the12 months previous to sampling MIC, minimum inhibitory concentration; ADD, defined animal daily dose.
prior to sampling date. A complete descriptionof all variables is given in Table 1.
Treatment incidence rate ðTIÞ ¼ Number of Following data collection and development of slaughter pigs could not be ascertained were variables, a dataset was generated containing excluded from the analysis. Also, for eight iso- resistance, consumption, and pig herd demo- lates the treatment incidence rate was unrealist- graphic data. For each isolate, the datasets con- ically high (TI above 20 ADD per 100 slaughter tained the results of the susceptibility test for that pig-days at risk) or one of the prescription dates antimicrobial drug (susceptibility status), sam- was not available and these observations were pling date, herd identification number, number deleted. The categorical variables consumption, of animals, herd size, prescribed amount (ADD), sampling year, and herd size were analyzed prescription frequency, interval between last with their frequency and prevalence calculated prescription and sampling, treatment incidence by their susceptibility status. Mean and stan- rate, and percentage of antimicrobials prescri- dard error of the mean were calculated for the bed to parenteral use (amount and frequency).
The variables consumption, prescription, and A logistic regression analysis was performed ADD were also calculated for a 3-month period with the susceptibility status as the model TETRACYCLINE CONSUMPTION AND RESISTANCE AMONG E. COLI FROM PIGS Table 2. Distributions of the Quantitative Explanatory Variables for the Probability of Isolating a Tetracycline-Resistant Escherichia coli Isolate in Danish Slaughter Pigs Between 2003 and 2005, by Susceptibility Status outcome. To avoid multicollinearity, the Spear- 2000). Also, the sensitivity and specificity of the man’s correlation coefficients between quan- logistic model was assessed through plotting a titative explanatory variables were initially receiver operating characteristic curve, and the assessed. In case of highly correlated variables area under the curve (AUC) was calculated.
(correlation coefficient jrj > 0.8), the biologically Finally, the probability of resistance determined more plausible variable was left in the model.
by each model was plotted against predictors Remaining variables were all entered in the in order to assess agreement between predictor statistical software SAS (Enterprise Guide 3.0;SAS Institute, Cary, NC) was used for the anal- ysis, and a binomial distribution and a logit linkwere assumed on the logistic regression analy- A total of 593 isolates were included in the sis. The repeated statement was used in order to study. Fifty-six isolates were excluded due to account for repeated measurements and clus- unrealistically high treatment incidence rate tering in the data. Manual backward elimination (TI), lacking information on total number of de- was used to select the independent variables to livered slaughter pigs, or one of the prescription be included in the regression analysis and all dates was missing. Isolates originated from 558 two-factor interactions, as well as higher order herds, where 26 herds were represented with 2 terms were tested during the selection process isolates, 3 herds with 3 isolates, and 1 herd with 4 for variables presenting p-value # 0.2. The cri- teria for keeping fixed effects in the final model A total of 199 (33.6%) of the E. coli isolates were were p-value # 0.05. Final model including all resistant to tetracycline. The distribution of the significant variables is presented at the Results MIC values revealed a clear bimodal distribu- tion, with few isolates having MIC values around the tetracycline breakpoint (>8 mg=mL).
(G-o-f) test was applied to the model to check From the MIC values, the distribution of the for lack of data fit. Linear relationship between susceptibility status by year was obtained. In the independent continuous factors and the logit 2004, the highest proportion of tetracycline- of the dependent was assessed by three different resistant E. coli was observed with 41.5% out methods: a univariable smoothed scatterplot of the 176 isolates included in the study being of the logit scale, design variables based on the resistant to tetracycline in that year.
quartiles of the distribution, and the method of Distributions of the categorical variables, fractional polynomials (Hosmer and Lemeshow, by the susceptibility status of the isolate, are Table 3. Distributions of the Categorical presented significant difference between the Explanatory Variables for the Probability means for the resistant and susceptible groups.
of Isolating a Tetracycline-Resistant Escherichia coli Isolate in Danish Slaughter Pigs Between strated a high level of correlation (jrj > 0.8) be-tween some of the independent variables, and these were not included in the model. Variables to be tested in the model were year, herd size,prescriptions in the prior 12 months, TI, ADD prescribed to the herd 12 months previous to sampling, and prescriptions via parenteral route. Because of a high correlation between prescriptions in the previous 3 and 12 months, it was decided that variables for 12 months would better represent the overall prescription pattern in the herds. Logistic regression using backward elimination resulted in the model þ b3(TI · Interval) þ Herd Sizecat þ Yearcat Due to the low number of repeated measure- ments within herds, the high level of indepen- presented in Table 3, while mean and standard error for quantitative variables are available in herd, and the lack of improvement in the model Table 2. Herds with tetracycline-resistant E. coli when accounting for repeated measurements, presented a higher mean amount of prescribed we have removed the repeated statement from ADD, as well as higher TI and prescription fre- the final model. Both continuous variables in- quency. The mean interval between last pre- cluded in the final model, treatment incidence scription and sampling date appears to be rate and interval between last prescription and smaller in herds where resistant isolates were sampling date, were checked for linearity in obtained. Statistical tests unadjusted for covari- the logits. Lowess-smoothed univariable logit ates demonstrated that only TI and interval plots were evaluated, and the plot for the vari- Table 4. Final Model for the Probability of Isolating a Tetracycline-Resistant Escherichia coli Isolate in Danish Slaughter Pigs Between 2003 and 2005 TETRACYCLINE CONSUMPTION AND RESISTANCE AMONG E. COLI FROM PIGS able interval presented a linear decrease of thesmoothed logit over time. However, the vari-able treatment incidence rate (TI) did not pres-ent the same behavior. It showed a peak in thesmoothed logit for value between 0 and 0.2,followed by some variation and then a steeperlinear increase. Design variables based on thequartiles of the distribution did not supportconclusively linearity in the logit for treatmentincidence rate and interval nor do they rule itout. Fractional polynomial analyses were testedfor each continuous variable, and the best non-linear transformations were not significantlydifferent from the linear model. Thus, the frac-tional polynomial analysis supported treating Predicted probability of isolating a tetracycline- both variables as linear in the logit.
resistant Escherichia coli isolate and the interval between Log odds of resistance was inversely associ- last prescription and sampling, for large herds (1001–3000slaughter pigs=year). The treatment incidence rate is fixed ated with interval (in days), meaning that an at 1 ADD=100 slaughter pig-days at risk and year in 2004.
increase in days between prescription and sam-pling resulted in a decreased probability of col- test did not indicate any lack of fit in the model lecting a resistant isolate. A positive association between probability of resistance and treatmentincidence rate (TI) in the previous 12 months was also significant, meaning that in herds wheremore animals were treated with tetracycline, it Due to the Danish efforts on the collection of was more likely to obtain a tetracycline-resistant data on antimicrobial consumption and resis- isolate. The interaction between these two risk tance, research assessing their association can factors was significant and was included in the now be performed, highlighting the importance model. Herd size ( p-value ¼ 0.03) was also sig- of an established antimicrobial usage and nificant, indicating that it was more likely toobtain tetracycline-resistant E. coli isolates in thesmallest herd size category, with up to 200 ani-mals produced per year. As the herd size cate-gory increased, the probability of isolating aresistant E. coli isolate decreased. Finally, year ofsampling ( p-value < 0.01) was also significant inthe final model (Table 4).
The predicted probability of isolating a resis- tant E. coli isolate and the interval between lastprescription and sampling is shown on Fig. 1.
Due to the interaction between interval andtreatment incidence rate, the latter was fixed in 1ADD=100 slaughter pig-days at risk. Similar in-terpretation can be drawn from Fig. 2 wherepredicted probability of isolating a resistantE. coli isolate is plotted against treatment inci-dence rate. Interval was fixed in 30 days forthis graphic. For both figures, estimates for Predicted probability of isolating a tetracycline- herds classified as large herds, in the year of resistant Escherichia coli isolate and treatment incidencerate (TI), for large herds (1001–3000 slaughter pigs=year).
2004, were applied. Finally, AUC was calculated Interval between last prescription date and sampling date as 62%, and the Hosmer and Lemeshow G-o-f is fixed at 30 days and year in 2004.
resistance monitoring program. In this study, we borg et al. (2007) demonstrated that the percent- investigated the association between tetracy- age of kilograms of pigs treated with tetracycline cline consumption risk factors and the proba- was associated with tetracycline-resistant Salmo- bility of isolating a tetracycline-resistant E. coli isolate from healthy pigs in the herds. The study The tetracycline resistance prevalence was showed that an increase in the proportion of 29% among E. coli isolates from herds where animals treated with tetracycline (treatment in- tetracyclines were not prescribed to slaughter cidence rate) and a decrease in the interval be- pigs within the 12 months before the sampling tween the last tetracycline prescription date and date. In another study, a significant difference in the sampling date were both associated with an the prevalence of fecal E. coli resistant to at least increased probability of isolating tetracycline- one antimicrobial between pig farms with non- resistant E. coli from pigs. Also, herd size and therapeutic usage of antimicrobials (68%) and year of sampling presented significant associa- antimicrobial-free farms (22%) was identified tion with tetracycline resistance status.
(Akwar et al., 2008). However, the same study Several studies have tried to investigate the presented very similar prevalences of tetracy- link between antimicrobial consumption and re- cline resistance for herds with and without sistance, by comparing the occurrence of anti- antimicrobial usage. Isolates obtained during a microbial resistance among bacteria isolated from 20-month period from adult pigs from a herd different sources with different usage patterns of where antimicrobials were discontinued for antimicrobial agents. However, these are often 126 months presented a tetracycline resistance limited by lack of comprehensive data on both prevalence of 34% (Langlois et al., 1988). A sim- consumption and resistance to be analyzed.
ilar study detected tetracycline resistance deter- Many of the studies regarding antimicrobial minants in 50% of the sampled animals in a herd consumption among animals have been restric- where no antimicrobials were prescribed for ted to a reduced sample population where con- 146 months, while 71% of the isolates presented tetracycline resistance determinants in a herd personal interview with herd owners or drug with regular usage of chlortetracycline (Lee et al., suppliers. Other studies have their limitation on 1993). These findings indicate the persistence the homogeneity of the evaluated herds included of resistant bacteria within the pig population, in the study or in the incapacity to link isolates even under less selective pressure conditions, obtained by an antimicrobial resistance surveil- and that the tetracycline resistance prevalence lance program and the site where the isolate may be co-maintained by other risk factors, in originated from. Emborg et al. (2004) demon- addition to antimicrobial consumption.
strated an association between the consumption The bimodal MIC distribution helps to clarify of antimicrobial growth promoters and the oc- that few isolates had their MIC around the tet- currence of resistance among Enterococcus faecium racycline breakpoint (>8 mg=mL), and, because isolated from broilers in Denmark. In this study, of that, misclassification in relation to isolate the probability of resistance to Avilamycin, status is not expected. The effect of the signifi- Avoparcin, Erythromycin, and Virginiamycin cant variables must be interpreted carefully due decreased as the time span between sampling to the presence of an interaction term. Both in- and the last date of consumption increased. For crease in the interval and decrease in the treat- Dunlop et al. (1998), farms that added antimicro- ment incidence rate resulted in the decrease of bials to starter rations had higher tetracycline the probability of resistant E. coli isolates. This resistance than farms that did not, and group relation can also be inferred from Figs. 1 and 2, treatments with tetracycline administered during where, due to the interaction, values were fixed the grower phase were associated with increa- for the interacting variable in order to ease the sed prevalence of tetracycline-resistant E. coli.
interpretation. Further analysis of the final However, the same study reports high levels of model estimates reveals a larger impact of the tetracycline-resistant E. coli for farms where tet- interval between last prescription and sampling racycline was not administrated (Dunlop et al., date in the probability of tetracycline resistance.
1998). Analyzing Danish surveillance data, Em- For instance according to the model, an increase TETRACYCLINE CONSUMPTION AND RESISTANCE AMONG E. COLI FROM PIGS of 200 days in the interval would have a higher year. Ideally, if samples were yearly collected impact in the reduction of the probability of re- from the same set of herds, an analysis ac- sistance than a treatment incidence rate decrease counting for repeated measurements over time from 2 to 1 ADD per 100 slaughter pig-days at would provide useful data on the variation be- risk. Also, the herd size effect is a risk factor of tween sampled herds and allow better general- key impact. According to the model, smaller ization of the results. However, these data were herds present significant higher probability of not available due to the sampling strategy tetracycline resistance. This can be correlated to adopted by the Danish antimicrobial resistance the larger treatment incidence rates associated to monitoring program. Since this study is not these smaller herds where, even when few pigs meant to define an antimicrobial resistance sus- are medicated, they represent a larger propor- ceptibility status to each herd, the analysis of the tion of the total number of slaughter pigs pro- consumption and pig herd demographic data was carried out with the purpose of associating a In order to explore the scale of the continuous ‘‘past’’ to each isolate and to determine factors independent variables, interval, and treatment associated with the probability of isolating a re- incidence rate, three different methods were sistant E. coli isolate. For this reason, intraherd used. While the quartile-based design variable variation was not assessed, and all the conclu- analysis suggested categorization of both vari- sions were done based on the isolate resistance ables, fractional polynomial analysis supported probability, rather than on herd resistance treating them as linear in the logit, since none of probability. Also, we have to consider the po- the nonlinear transformations tested were able to tential limitations of measuring the explanatory provide a significantly better model. The uni- variables, such as herd size, frequency, and variable smoothed logit plot showed a linear amount of tetracycline consumed, at the herd decrease for the variable interval, differently from level and the fact that we were not able to assess what was observed using the quartile-based de- antimicrobial consumption at the animal level sign variable analysis. Because the quartile anal- for the pigs providing the susceptibility tested ysis is not an especially powerful diagnostic tool, samples. It leaves space for uncertainties on the it was decided to support the findings of the other outcome of the susceptibility tests, and a resis- two applied methods and assume linearity for tant isolate may arise either due to chance or this variable. Similar discussion can be drawn for due to the evaluated risk factors. These risk the variable treatment incidence rate. However, it factors can impact the sampled animals directly shows an initial peak in the logit between 0 and or indirectly, by increasing or decreasing the 0.2 ADDs=100 slaughter pig-days at risk. This chances of antimicrobial consumption or envi- was followed by a nonconsistent trend and then a ronmental exposure to tetracycline. Still, al- nearly linear increase in the range of 1.6–8. It was though a sampled pig may not be one of the of particular interest here to determine whether medicated animals from a herd, an increase in the initial peak in the logit was significant or the treatment incidence or a shorter period with- simply the result of a few isolates concentrated in out tetracycline treatment in this herd will in- that interval, or the effect of the interaction be- crease the probability that an isolate from a pig tween interval and treatment incidence rate.
Since a model where isolates belonging to that We cannot exclude other risk factors evalu- interval were excluded presented similar esti- ated by this study as relevant and of potential mates for its variables and no model was signifi- impact for the occurrence of resistance. The cantly better than the linear model, it was decided model aims to find the optimal balance between to treat the variable treatment incidence rate as the estimates associated to the different risk factors, and the exclusion of one or more risk Although the model had a hierarchical struc- factors from the final model can be attributed to ture, it contained a limited number of repeated model prioritization or significance hierarchy.
As an example, the number of prescriptions in more than one animal sampled over the 4 years each herd in the months before sampling was and never more than one animal sampled per not significant, meaning that it did not influence the probability that a randomly selected E. coli tetracycline was prescribed to a herd and the day isolate was resistant in this model. However, of sampling seem to play a major role in this such a risk factor could easily become significant in a study where interval between last pre-scription date and sampling date is not included.
For instance, even though ADD amount andnumber of prescriptions were risk factors not No competing financial interests exist.
included in the final model, the variables inter-val and treatment incidence rate accounted for data enclosed in those variables. From interval, Aarestrup FM. Veterinary drug usage and antimicrobial for instance, it is possible to infer the presence of resistance in bacteria of animal origin. Basic Clin. Phar- consumption over a time period. Amount con- sumed (corrected by the number of animals) was Aarestrup FM, Bager F, and Andersen JS. Association be- covered by the treatment incidence rate. The tween the use of avilamycin for growth promotion and year of sampling was significant in the model the occurrence of resistance among Enterococcus faeciumfrom broilers: epidemiological study and changes over and might account for a set of uncertainties that time. Microb. Drug Resist. 2000a;6:71–75.
were not assessed by this study. Several hy- Aarestrup FM, Kruse H, Tast E, et al. Associations between potheses can be drawn to explain the year esti- the use of antimicrobial agents for growth promotion mate variation, ranging from complex seasonal and the occurrence of resistance among Enterococcus variation of the E. coli subtypes, where some faecium from broilers and pigs in Denmark, Finland, and would be more likely to develop resistance Norway. Microb. Drug Resist. 2000b;6:63–70.
Akwar HT, Poppe C, Wilson J, et al. Prevalence and mechanisms, until simple sample selection bias patterns of antimicrobial resistance of fecal Escherichia over the years. Other consumption factors such coli among pigs on 47 farrow-to-finish farms with dif- as trade or contact between herds are also po- ferent in-feed medication policies in Ontario and British tential explanations. However, the problem of Columbia. Can. J. Vet. Res. 2008;72:195–201.
antimicrobial resistance is complex and involves Bager F. DANMAP: monitoring antimicrobial resistance in Denmark. Int. J. Antimicrob. Agents 2000;14:271–274.
not only antimicrobial consumption, but also Cohen ML. Epidemiology of drug resistance: implications interaction of multiple factors related to the for a post-antimicrobial era. Science 1992;257:1050–1055.
particular organism and the environment. Even Cohen ML. Antimicrobial resistance: prognosis for public health. Trends Microbiol. 1994;2:422–425.
examples of antimicrobial resistance and con- Dahlin A, Eriksson A, Kjartansdottir T, et al. The ATCvet classification system for veterinary medicinal products.
J. Vet. Pharmacol. Ther. 2001;24:141–142.
might not give a full understanding of the dy- DANMAP. Use of antimicrobial agents and occurrence of namics of selection for antimicrobial resistance, antimicrobial resistance in bacteria from food animals, because there is a large number of environmen- foods and humans in Denmark. Available at www tal and host factors involved that are difficult to address. Still, the model was able to detect Dunlop RH, McEwen SA, Meek AH, et al. Associations among antimicrobial drug treatments and antimicrobial risk factors that would impact the probability of resistance of fecal Escherichia coli of swine on 34 farrow- to-finish farms in Ontario, Canada. Prev. Vet. Med. 1998;34:283–305.
EFSA. Harmonised monitoring of antimicrobial resistance in Salmonella and Campylobacter isolates from food ani- The study provides a descriptive overview of mals in the European Union. Clin. Microbiol. Infect.
2008;14:522–533.
Emborg HD, Andersen JS, Seyfarth AM, et al. Relations data in relation to susceptibility status, and we between the consumption of antimicrobial growth pro- can conclude that there is an association between moters and the occurrence of resistance among Entero- tetracycline treatment incidence rate in a herd coccus faecium isolated from broilers. Epidemiol. Infect.
and the probability of isolating a tetracycline- Emborg HD, Vigre H, Jensen VF, et al. Tetracycline con- resistant E. coli from the intestinal content of sumption and occurrence of tetracycline resistance in healthy pigs. Also, number of animals, year of Salmonella typhimurium phage types from Danish pigs.
sampling, and the interval between the last time Microb. Drug Resist. 2007;13:289–294.
TETRACYCLINE CONSUMPTION AND RESISTANCE AMONG E. COLI FROM PIGS Fluit AC, van der Bruggen JT, Aarestrup FM, et al. Prio- Turnidge J, and Christiansen K. Antibiotic use and rities for antibiotic resistance surveillance in Europe.
resistance—proving the obvious. Lancet 2005;365:548– Clin. Microbiol. Infect. 2006;12:410–417.
Hosmer DW, and Lemeshow S. Assessing the fit of the Wise R, Hart T, Cars O, et al. Antimicrobial resistance. Is model. In. Applied Logistic Regression. New York: Wiley, a major threat to public health. BMJ 1998;317:609–610.
Witte W. Selective pressure by antibiotic use in live- Jensen VF, Jacobsen E, and Bager F. Veterinary stock. Int. J. Antimicrob. Agents 2000;16 Suppl 1:S19– antimicrobial-usage statistics based on standardized measures of dosage. Prev. Vet. Med. 2004;64:201–215.
World Health Organization. WHO Global Principles for Langlois BE, Dawson KA, Leak I, et al. Effect of age and the containment of antimicrobial resistance in ani- housing location on antibiotic resistance of fecal coli- mals intended for food. Available at www.who.int= forms from pigs in a non-antibiotic-exposed herd. Appl.
salmsurv=links=en=GSSGlobalPrinciples2000.pdf.
Environ. Microbiol. 1988;54:1341–1344.
Lee C, Langlois BE, and Dawson KA. Detection of tetra- cycline resistance determinants in pig isolates from three Antonio Roberto Vieira, D.V.M., M.P.H.
herds with different histories of antimicrobial agent ex- Department of Microbiology and Risk Assessment posure. Appl. Environ. Microbiol. 1993;59:1467–1472.
Marano NN, Rossiter S, Stamey K, et al. The National Antimicrobial Resistance Monitoring System (NARMS) for enteric bacteria, 1996–1999: surveillance for action.
J. Am. Vet. Med. Assoc. 2000;217:1829–1830.
Stege H, Bager F, Jacobsen E, et al. VETSTAT—the Danish system for surveillance of the veterinary use of drugs forproduction animals. Prev. Vet. Med. 2003;57:105–115.

Source: http://www.dpi.inpe.br/Miguel/RefsEUREQA/Viera_et_al_AssocEschColiResis_FoodbornePathDisease_2009.pdf