Trend Analysis


 Buck Black
 9 days ago
 Views:
Transcription
1 CODA CERVA Centrum voor Onderzoek in Diergeneeskunde en Agrochemie Centre de Recherches et d Etudes Vétérinaires et Agrochimiques Antimicrobial Resistance in commensal Escherichia coli from livestock in Belgium: Trend Analysis Unit Epidemiology and Risk Assessment (ERASURV) Context This report summarises the results of the trend analysis of the data related to antimicrobial resistance in Escherichia coli (E. coli) during six consecutive years ( ) regarding commensal intestinal flora of several livestock categories in Belgium:  Veal calves  Young beef cattle  Slaughter pigs  Broiler chickens Commensal E. coli is regarded as a general indicator for resistance amongst Gramnegative bacteria. It can be frequently isolated from all animal species and is therefore suitable for comparisons and surveillance programmes. During sampling, faecal material was taken at the slaughterhouse or directly at the farms depending on the animal category. E. coli were isolated and thereafter tested for their susceptibility to a panel of several antimicrobial substances. The objectives of this study were twofold:  To provide a trend analysis of the prevalence of resistant strains over the six consecutive years. The results were compared for the six years and then analysed to check whether the observed trends (increase or decrease) were statistically significant.  To evaluate the level of multiresistance and its trend over the same period: using the same data, we calculated for each animal category the proportion of multiresistant strains (i.e. resistance to more than two antimicrobials (= at least three) by the same strain) and checked whether there was a significant trend. 1
2 I. Material and Methods A. Sampling Samples of fresh faeces were collected each year by agents of the Federal Agency for the Safety of the Food Chain (FASFC) according to standardized technical sampling instructions (PRI codes) as part of a nationwide surveillance programme. Samples were taken from the following categories of foodproducing animals:  Veal calves: young cattle kept in specialized units for fattening and slaughtered at an average age of 8 months. In 2011, faecal samples were taken on the floor at the farm level (PRI516: 10 animals/farm of 7 months or younger), while in 2012, 2013, 2014, 2015, 2016 the samples were taken directly from the rectum of the animals at the slaughterhouse (PRI036: 1 animal sampled/farm)  Beef Cattle (meat production): young animals (7 months or younger) from farms raising beef cattle for meat production. Faecal samples were taken from the floor at the farm (PRI515:1 sample consisted of a pool of faeces collected from different spots on the floor representing 10 animals).  Broiler chickens: samples were taken at the slaughter house (PRI019: pools of pairs of caeca from 10 chickens /batch)  Fattening pigs: faecal samples of fattening pigs older than 3 months were taken from the rectum at the slaughterhouse (PRI035: 1 animal /origin farm). Following EFSA recommendations and in order to allow resistance trends to be detected with an acceptable confidence and precision ( the target sample size for each animal category was fixed to 170 isolates. In order to improve representativeness, the sampling was stratified by province proportionally to the number of registered herds or slaughterhouses. B. Isolation of the strains and antimicrobial susceptibility testing Isolates of E. coli strains were obtained from the faecal samples at the laboratories of the two animal health associations ARSIA and DGZ. The isolation methods are described in the annual reports of CODACERVA on antimicrobial resistance in E. Coli ( and were performed according to the standard operating procedures (SOP). The isolates were sent to the National Reference Laboratory (CODACERVA) for confirmation of identification and antimicrobial susceptibility testing. Susceptibility was tested by a microdilution technique (Trek Diagnostics) as it is described in the annual reports. The antimicrobials which were followed up in and 2016 and those common to the six years ( ) are all presented in Table A. For each strain and each antimicrobial substance, the Minimal Inhibitory Concentration (MIC) was recorded: MIC is defined as the lowest concentration by which no visible growth could be detected. MICs were semiautomatically recorded and stored in a database (Annex 1). 2
3 Table A. Panel of antimicrobials tested during for E. coli (in blue: antimicrobials only tested in ; in green: only tested in ; in black: tested during the 6 consecutive years) Symbol AMP AZI CHL CIP COL FFN FOT GEN KAN MERO NAL SMX STR TAZ TGC TET TMP Antimicrobial Ampicillin Azithromycin Chloramphenicol Ciprofloxacin Colistin Florphenicol Cefotaxime Gentamicin Kanamycin Meropenem Nalidixic acid Sulphamethoxazole Streptomycin Ceftazidime Tigecycline Tetracycline Trimethoprim C. Data The datasets for were formatted in Excel files by the Department of Bacteriology of CODACERVA and validated by the FASFC. They included identification of the samples corresponding to each isolate recorded in the LIMS merged with the corresponding MIC value for each tested antibiotic. After several steps of crosschecking and cleaning of the data, six yearly distinct data sets were produced, imported and analysed in SAS 9.2 software and R freeware. Emphasis was put on verifying that the animal category of the sample was correct. If it could not be confirmed, the sample was excluded. The final annual datasets contained the following fields: i. isolate identification number, ii. animal category, iii. sampling date and iiii. MIC values for each of the tested antimicrobials. D. Statistical Methods All subsequent statistics were carried out using SAS 9.2 software and R freeware. 1. Prevalence Quantitative MIC values were converted into binary qualitative values (Resistant/Susceptible) based on the susceptibility breakpoints defined by the European Committee on Antimicrobial Susceptibility Testing (EUCAST). The ECOFFs (Epidemiological cutoffs values) were used in order to define strains as Resistant (R) or Susceptible (S) (Annex 1). 3
4 For each animal category and year, the proportion of resistant isolates (p) was calculated per tested antimicrobial (resistance prevalence), as well as the associated 95% confidence interval (CI). In order to avoid interval boundaries outside 01, which does not make sense for probabilities, CI were constructed for logit(p). 2. Trend Analysis The trends analysis aims at finding models to describe the variation of antimicrobial resistance over the years and to check if this variation is significant or not. Several statistical methods were used to model the probability of an isolate to be resistant: logistic regression models (in the univariate model each antimicrobial was considered separately), a linear Generalized Estimating Equations model (GEE) and a nonlinear mixed model (both multivariate models; taking into account the possible correlation between antimicrobial substances in a single model; assuming an unstructured correlation matrix in the GEE). The models with the smallest Akaike information criterion (AIC) were selected and presented in the results. The results are described in the form of Odds Ratio (OR), where an OR > 1 means that the probability to be resistant increases with time. Plots representing the log odds for each year were also produced for each antimicrobial and animal category. The odds represent here the probability to be resistant on the probability to be susceptible. In this study, the effects of the different antimicrobials were assessed on an individual level. Hence, the 5% significance levels were specified for each antimicrobial separately. If the interest is in making a statement on the entire pool of antimicrobials jointly, a family wise significance level should be specified. In order to adjust the pvalues and reduce the chances of obtaining falsepositive results (type I errors; i.e. detection of a trend when in reality there is no trend) when several dependent or independent statistical tests are being performed simultaneously on a single data set, both the Bonferroni s correction method and the linear stepup method of Benjamini and Hochberg (1995) were applied to the GEE (linear multivariate model) and the resulting corrected pvalues were produced and presented in annex for documentation. 3. Multiresistance Multiresistance was considered in this report as resistance by an isolate to at least three antimicrobials belonging to any three antimicrobial families as recommended by EFSA (2008, 2014). These antimicrobials were: ampicillin, cefotaxime and/or ceftazidime, chloramphenicol, ciprofloxacin and/or nalidixic acid, colistine, gentamycin, sulphonamides, tetracycline and trimethoprim. Thus, nine antimicrobials belonging to different classes were considered in this part of the analysis. Based on this, for each animal category, the estimate for the prevalence of multiresistant isolates was calculated together with the 95% CI, calculated using normal distribution. 4
5 In addition, logistic regression models were used to check whether there was a significant trend over the years regarding the prevalence of multiresistant strains, for each animal category. In addition, a diversity index was calculated for multiresistance: Diversity index: Weighted entropy This indice is calculated to describe the degree of diversity of multiresistance for a specific year and a specific animal category. The weighted entropy index takes into account order and will take higher values when multiresistance is more frequent for large number of antimicrobials. Therefore, a higher weighted entropy index reflects a shift to multiresistance to a greater number of antibiotics. This latter index was calculated using R software based on the formula of Guiasu (1971). II. Results A. Prevalence The following table summarizes the data obtained in 2011, 2012, 2013, 2014, 2015 and 2016 regarding prevalence of resistant isolates for each animal category and each tested antimicrobial substance:  N = number of tested samples  Prevalence of resistant isolates and confidence intervals 5
6 6
7 Resistance (%) CODACERVA B. Trend Analysis N.B: Detailed outputs of the multiple comparisons corrections are presented in Annex 2. N.B.2: In this report the adjective high was used in case of a prevalence of resistant strains higher than 50%. However, the significance of a given level of resistance will depend on the particular antimicrobial and its importance in human and veterinary medicine. a) Veal Calves: (N= 34 (2011); 181 (2012); 202 (2013); 188 (2014); 196 (2015);174 (216) As shown in figure 1, high levels of resistance (>50%) were observed for the six consecutive years for TET, SMX, AMP. Resistance was > 40% for the six consecutive years for TMP. 100 Resistance strains prevalence Veal calves  E. coli AMP CHL CIP COL FOT GEN NAL SMX TAZ TET TMP figure 1 Based on the results of the linear multivariate model (GEE), statistically significant decrease of resistance over time was observed for all tested substances except for FOT (figure 2). 7
8 Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper CODACERVA figure 2 The detailed odds ratios obtained from the nonlinear multivariate model are shown in table 1 and the log odds of the logistic regression are plotted in figure 3. We notice that a constant significant decrease that began in 2012 but slowed down and stopped for all substances in table 1 (OR1= year 2012 vs 2011: OR2= year 2013 vs 2012;OR3= year 2014 vs 2013; OR4= year 2015 vs 2014; OR5= year 2015 vs 2016) Substance OR1 OR2 OR3 OR4 OR5 AMP 0,66 0,49 0,84 0,72 0,57 0,87 0,80 0,72 0,89 0,90 0,79 1,00 1,00 0,77 1,23 CHL 0,65 0,48 0,81 0,70 0,56 0,84 0,79 0,71 0,87 0,89 0,77 1,01 1,00 0,75 1,25 CIP 0,64 0,47 0,80 0,69 0,55 0,82 0,76 0,68 0,84 0,84 0,72 0,96 0,93 0,69 1,18 COL 0,57 0,30 0,83 0,60 0,38 0,81 0,65 0,50 0,80 0,71 0,43 0,98 0,77 0,26 1,28 FOT 0,58 0,33 0,82 0,64 0,39 0,89 0,78 0,62 0,94 0,95 0,65 1,26 1,16 0,49 1,83 GEN 0,69 0,38 1,00 0,74 0,50 0,98 0,81 0,67 0,95 0,89 0,66 1,11 0,97 0,52 1,42 NAL 0,66 0,48 0,84 0,70 0,56 0,84 0,77 0,69 0,85 0,84 0,72 0,96 0,92 0,67 1,16 SMX 0,60 0,08 0,75 0,66 0,07 0,80 0,76 0,04 0,85 0,89 0,06 1,00 1,03 0,13 1,28 TAZ 0,59 0,14 0,87 0,64 0,12 0,88 0,72 0,08 0,88 0,82 0,14 1,10 0,93 0,29 1,49 TET 0,72 0,16 1,04 0,76 0,09 0,93 0,83 0,05 0,93 0,91 0,06 1,03 1,00 0,13 1,25 TMP 0,63 0,09 0,80 0,67 0,07 0,80 0,72 0,04 0,79 0,77 0,05 0,87 0,83 0,10 1,03 8
9 9
10 figure 3 (year0: 0=2011; 1=2012; 2=2013; 3=2014; 4=2015 5=2016) b) Beef cattle: N= 154 (2011) ; 175 (2012) ; 204 (2013); 164 (2014); 180 (2015); 176 (2016)) Globally, significantly lower prevalences of resistance were observed in E. coli from beef cattle compared to veal calves: resistance prevalences are smaller than 45 %. However, the highest resistance prevalences were observed against the same substances than for veal calves: AMP, SMX, TET and TMP (figure 4). 10
11 Resistance (%) CODACERVA 50 Resistance strains prevalence Beef cattle  E. coli AMP CHL CIP COL FOT GEN NAL SMX TAZ TET TMP figure 4 Based on the results of the linear multivariate model (GEE), statistically significant decrease of resistance over time was observed for all tested substances except for COL and GEN (figure 5). 11
12 figure 5 As we can see from the results of the NL mixed model (table 2), a decreasing trend has started in 2012 for AMP. However, in 2016 all upper confidence s are close to 1. Results of the logistic regression are plotted in Figure 6. 12
13 13
14 Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper CODACERVA figure 6 (where year0: 0=2011; 1=2012; 2=2013; 3=2014; 4=2015; 5=2016) table 2 (OR1= year 2012 vs 2011: OR2= year 2013 vs 2012, OR3= year 2014 vs 2013 and OR4= year 2015 vs 2014; OR5= 2016 vs 2015 ) Substance OR1 OR2 OR3 OR4 OR5 AMP 0,87 0,65 1,08 0,85 0,72 0,97 0,82 0,75 0,90 0,80 0,67 0,93 0,78 0,57 0,99 CHL 1,14 0,76 1,51 1,00 0,83 1,32 0,89 0,80 0,99 0,80 0,65 0,95 0,71 0,49 0,94 CIP 0,90 0,58 1,21 0,83 0,67 1,00 0,77 0,67 0,88 0,72 0,54 0,90 0,66 0,39 0,94 COL 1,170,24 2,59 0,85 0,39 1,31 0,63 0,29 0,97 0,47 0,00 0,94 0,35 0 0,90 FOT 0,94 0,40 1,47 0,85 0,59 1,11 0,77 0,59 0,94 0,70 0,41 0,98 0,63 0,21 1,05 GEN 1,79 0,58 2,99 1,35 0,88 1,81 1,07 0,86 1,28 0,85 0,60 1,11 0,68 0,33 1,03 NAL 0,83 0,54 1,11 0,79 0,63 0,94 0,75 0,64 0,85 0,71 0,52 0,89 0,67 0,38 0,96 SMX 0,92 0,71 1,12 0,90 0,78 1,02 0,89 0,82 0,96 0,88 0,76 1,00 0,87 0,67 1,07 TAZ 0,97 0,36 1,57 0,83 0,56 1,09 0,71 0,52 0,89 0,60 0,31 0,90 0,52 0,12 0,91 TET 0,89 0,55 1,23 0,92 0,79 1,05 0,89 0,81 0,97 0,86 0,73 0,99 0,83 0,61 1,04 TMP 0,97 0,70 1,25 0,89 0,75 1,02 0,81 0,73 0,89 0,74 0,61 0,87 0,68 0,47 0,88 14
15 Resistance (%) CODACERVA c) Broiler Chickens ( N= 420 (2011) ; 320 (2012) ; 234 (2013); 158 (2014); 152 (2015); 167 (2016)) A high prevalence of resistance (>50%) was observed for broiler chickens for the six consecutive years for several substances: AMP (the highest level all substances and species included), SMX, CIP, and TMP (figure 7). Prevalence for NAL was >60% for five consecutive years and was >40% in Resistance strains prevalence Chickens  E.coli AMP CHL CIP COL FOT GEN NAL SMX TAZ TET TMP figure 7 Based on the results of the linear multivariate model (GEE), statistically significant decrease of resistance over time was observed for all tested substances except for CIP, AMP (substance with high levels of resistance) and GEN (figure 8). 15
16 Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper CODACERVA figure 8 For CIP the NL mixed and the logistic regression models indicate a significant increasing trend followed by a significant decreasing trend starting in The odds ratio based on the NL mixed model are presented in table 3 and those based on the logistic regression are plotted in figure 9. Table 3 (OR1= year 2012 vs 2011; OR2= year 2013 vs 2012; OR3= year 2014 vs 2013; OR4= year 2015 vs 2014; OR5= year 2016 vs 2015) Substance OR1 OR2 OR3 OR4 OR5 AMP 0,78 0,63 0,92 0,85 0,74 0,95 0,94 0,87 1,02 1,04 0,89 1,20 1,16 0,87 1,44 CHL 1,22 0,98 1,47 1,03 0,93 1,14 0,88 0,82 0,95 0,76 0,65 0,86 0,65 0,50 0,80 CIP 1,51 1,19 1,83 1,15 1,04 1,26 0,91 0,85 0,97 0,72 0,64 0,81 0,57 0,46 0,69 COL 5,56 < ,07 0,34 < ,71 0,01 < ,04 0,00 0,00 0,00 0,00 0,00 0,00 FOT 0,80 0,62 0,97 0,77 0,68 0,86 0,74 0,66 0,82 0,72 0,57 0,86 0,69 0,47 0,91 GEN 1,35 0,78 1,93 1,15 0,91 1,40 1,00 0,85 1,15 0,87 0,63 1,11 0,76 0,42 1,09 NAL 1,36 1,08 1,64 1,06 0,96 1,17 0,86 0,80 0,91 0,69 0,61 0,77 0,56 0,45 0,67 SMX 0,80 0,67 0,94 0,84 0,76 0,93 0,89 0,83 0,96 0,95 0,83 1,07 1,00 0,80 1,21 TAZ 0,84 0,64 1,03 0,79 0,69 0,89 0,75 0,66 0,83 0,70 0,55 0,85 0,66 0,44 0,89 TET 0,79 0,60 0,97 0,83 0,75 0,91 0,85 0,79 0,90 0,86 0,76 0,96 0,87 0,70 1,04 TMP 0,88 0,74 1,02 0,89 0,81 0,97 0,90 0,84 0,96 0,91 0,81 1,02 0,92 0,74 1,10 16
17 17
18 figure 9 (where year0: 0=2011; 1=2012; 2=2013; 3=2014; 4=2015; 5=2016) 18
19 Resistance (%) CODACERVA d) Pigs: (N= 157 (2011) ; 217 (2012) ; 206 (2013); 184 (2014); 186 (2015); 173 (2016)) The prevalence of resistance for SMX and TET was above 40% during the six consecutive years (figure 10). 70 Resistance strain prevalence Pigs  E. coli AMP CHL CIP COL FOT GEN NAL SMX TAZ TET TMP figure 10 Based on the results of the linear multivariate model (GEE) (figure 11), a significant decrease of resistance over time was observed for SMX, TMP,TET, CIP,NAL. For COL and GEN, prevalences of these two substances are very low (<4%). 19
20 Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper Estimate Lower Upper CODACERVA figure 11 Based on the NL mixed model (table 4), only NAL has continued to significantly decreasing since Results of the logistic regression are plotted in Figure 12. Table 4 (OR1= year 2012 vs 2011; OR2= year 2013 vs 2012; OR3= year 2014 vs 2013; OR4= year 2015 vs 2014; OR5= year 2016 vs 2015) Substance OR1 OR2 OR3 OR4 OR5 AMP 0,82 0,67 0,98 0,88 0,77 0,98 0,94 0,87 1,01 1,01 0,89 1,13 1,08 0,86 1,30 CHL 0,99 0,75 1,23 0,96 0,83 1,10 0,94 0,86 1,01 0,91 0,78 1,03 0,88 0,67 1,09 CIP 0,54 0,42 0,65 0,59 0,47 0,71 0,73 0,62 0,83 0,89 0,65 1,14 1,10 0,60 1,60 COL 1,070,01 2,25 0,97 0,40 1,53 0,88 0,54 1,21 0,79 0,26 1,34 0,720,10 1,54 FOT 0,51 0,40 0,61 0,56 0,34 0,77 0,83 0,64 1,02 1,24 0,69 1,79 1,86 0,44 3,28 GEN 0,82 0,24 1,40 0,82 0,48 1,17 0,82 0,59 1,06 0,83 0,39 1,26 0,83 0,11 1,54 NAL 0,59 0,38 0,80 0,58 0,45 0,71 0,57 0,43 0,70 0,55 0,30 0,80 0,54 0,16 0,91 SMX 0,86 0,69 1,03 0,88 0,77 0,99 0,90 0,84 0,96 0,92 0,81 1,03 0,94 0,75 1,14 TAZ 0,53 0,38 0,69 0,60 0,39 0,82 0,81 0,63 0,99 1,09 0,62 1,57 1,47 0,39 2,55 TET 0,92 0,66 1,17 0,83 0,73 0,93 0,87 0,81 0,94 0,92 0,81 1,03 0,97 0,77 1,16 TMP 0,93 0,74 1,13 0,93 0,81 1,04 0,92 0,85 0,98 0,91 0,80 1,02 0,90 0,72 1,09 20
21 21
22 Figure 12 (where year0: 0=2011; 1=2012; 2=2013; 3=2014; 4=2015; 5=2016) 22
23 Prevalence CODACERVA E. Multiresistance Prevalence of multiresistance The proportion of multiresistant strains (= strains resistant to at least three antimicrobials) was very high for broiler chickens (>62%) and high for veal calves (>50%) during the six consecutive years (Table 5 and Figure 13). Figure 14 displays the distribution of multiresistance patterns per animal category (i.e, number of isolates resistant to 0, 1.9 of the antimicrobial tested). Between 12%26%, 48% 71%, 5%12%, 25%41%, of, respectively, calves, cattle, chicken and pig isolates, were fully susceptible (=no resistance) to all tested antimicrobials. Table 5: proportion of multiresistant strains (+95% CI) Veal calves Beef cattle Chickens Pigs ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ( ) ( ) ( ) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% veal calves beef cattle chickens pigs figure 13 23
24 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Veal calves % 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Beef cattle
25 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Chickens % 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Pigs figure 14 25
26 Table 6 and 7 present the OR (the ratio of the odds for a one unit increase in the time) for the probability to be multiresistance obtained from the logistic models. For veal calves, the decrease becomes significant from 2012 onwards and one year later for pigs and broiler chickens (Table 7). In all cases, there was still a decreasing trend in 2014 which stopped in 2015 for veal calves and pigs. In 2016, the probability to be multi resistant is only significantly decreasing in beef cattle in which category a decreasing trend has been observed since 2014 but confident interval is close to 1. Table 6: OR and CI95% regarding to probability to be multiresistant (logistic regression, ) Species OR 95%CI veal calves beef cattle chickens pigs Table 7: OR and CI95% regarding to probability to be multiresistant (logistic regression, year by year) Years compared veal calves beef cattle chicken pig 2012 vs ( ) 0.94 ( ) 0.86 ( ) 0.85 ( ) 2013 vs ( ) 0.89 ( ) 0.87 ( ) 0.87 ( ) 2014 vs ( ) 0.85 ( ) 0.87 ( ) 0.89 ( ) 2015 vs ( ) 0.81 ( ) 0.88 ( ) 0.91 ( ) 2016 vs ( ) 0.76 ( ) 0.89 ( ) 0.94 ( ) Indice of diversity: Weighted Entropy : As shown in table 8, the value of the indices decreased over time for all species. The index is globally lower for pigs compared with other species, meaning that for this species, multiresistance to a high number of antibiotics is less frequent than for the others. Table 8 year Veal calves Beef cattle Chickens Pigs
27 III. Discussion and conclusions To summarize the findings, data from six consecutive years (2011 to 2016) indicate an overall decreasing trend of E. coli antimicrobial resistance in Belgian production animals, especially since 2013 but in 2016 an increase of resistance prevalence is noticed, principally in pigs and chickens. Nevertheless, as described in this report, the increase of resistance prevalence is not significant. TET prevalence in calves increases the most in 2016 (+9% resistant E. coli strains). GEN is the only substance which prevalence decreases in 2016 in all animal categories. There is globally a high level of resistance to AMP, SMX, TET and TMP in all animal species, but to a lesser extent in beef cattle. The common patterns of resistance to AMP, SMX, TMP and TET and combinations thereof often feature as a component of multiresistance patterns, and are probably related to the presence of class 1 or class 2 integrons, which generally carry genes conferring resistance to these antimicrobials (Marchant et al., 2013; EFSA and ECDC, 2015). Although other risk factors have been described, antimicrobial use is recognized as the main selector for antimicrobial resistance and a correlation with resistance was pointed out in Belgium (Cargnel et al., 2017). In Belgium, antimicrobial sales data for use in animals are being collected on an annual basis since 2009 (BelVetSAC, 2016). In 2016, a decrease of 20,0% in the sales of antimicrobials has been observed since The AIC criteria of the nonlinear multivariate model and from the logistic regression are similar (a little bit lower for the last one) and they can be considered for the final results. The levels of antimicrobial resistance are very high in veal calves for AMP, SMX and TET (more than 50% of isolates are resistant during the six consecutive years). Although the linear multivariate analysis highlight a significant decrease in resistance, except for FOT (nonsignificant but prevalence are extremely low), it cannot be affirmed by the nonlinear analysis that the significant decreases observed for from 2012 to 2015, depending on the substance, continued afterward. In beef cattle, resistance prevalence is globally lower than in other species and a significant decrease of resistance for more than 80% substances since 2014 by both univariate and the NL mixed models is observed. Based on the logistic regression, resistance decreases in all substances since 2012, even not always at a significant level, it can be assessed that in 2016, the decrease is only significant for CHL and TMP. Chicken present substance with a high level a resistance (e.i. AMP, SMX, CIP, TMP are >50% resistance during the 6 years). An increasing trend which had been previously detected for in this species for CIP but no other increase in any substance was noticed after An increase of resistance prevalence was detected for 7/11 substances in 2016 compared to 2015, but was not statistically significant. Despite these decreasing trends of resistance in the previous year, in 2016, still more than 90% of E. coli chicken strains are resistant to at least one of the antimicrobials in the panel. The high resistance to quinolones 27
28 in chicken is especially worrisome because of a higher resistance percentage for ciprofloxacin compared to nalidixic acid, suggesting the presence of plasmid mediated quinolone resistance (Strahilevitz et al., 2009). The collection of antimicrobial use data in the entire poultry chain might expose why selection pressure on the intestinal flora in broiler chickens remains high. In pigs, except for SMX, for the second year antimicrobial resistance is lower than 50% for all tested substance but the prevalence increased in all substance except for GEN. Resistance in chickens and pigs should be monitored because the nonlinear models reveals nonsignificant OR >1. The proportions of multiresistant strains remained high during the six consecutive years, especially in chickens and veal calves (>65% and >55%, respectively). Moreover, in chickens, an increase was seen again in 2015, resulting in a borderline significant decreasing trend for the 5 consecutive years. E. coli strains from broiler chickens were also resistant to the highest number of antimicrobials (highest entropy), whereas E. coli from pigs had the lowest entropy. In a previous report a significant decreasing trend of the multiresistant strains of E. coli was found from 2012 onwards by linear and nonlinear analysis. A decreasing trend is observed with the linear analysis but was borderline in 2015 compared to 2014 and is nonsignificant in 2016 compared to 2015 for calves, chicken and pig. Again, multiresistance and entropy reveal the broiler chicken production as the animal production sector where additional measures should be taken in terms of the use of antimicrobials. Multiresistance rapidly decreased since 2012 in veal calves, but increased in 2016 and is still resistant to a high number of different antimicrobial classes (entropy. E. coli from beef cattle remain the less multiresistant for 6 consecutive years. 28
29 IV. References BelVetSAC. Belgian Veterinary Surveillance of Antibacterial Consumption, National consumption report Available online: CARGNEL M., SARRAZIN S., CALLENS B., WATTIAU P., WELBY S., Assessing the possible association between veterinary antimicrobial consumption and resistance in indicator E. coli isolated from farm animals in Belgium. Poster presented at the SVEPM Conference & Annual General Meeting of the Society 2017, Inverness, Scotland. Available at: EFSA and ECDC. European Food Safety Authority and European Centre for Disease Prevention and Control, EU Summary Report on antimicrobial resistance in zoonotic and indicator bacteria from humans, animals and food in EFSA Journal 2015;13(2):4036, 178 pp. Available at: Marchant, M., Vinue, L., Torres, C., Moreno, M.A., Change of integrons over time in Escherichia coli isolates recovered from healthy pigs and chickens. Vet Microbiol 163, Strahilevitz, J., Jacoby, G.A., Hooper, D.C., Robicsek, A., Plasmidmediated quinolone resistance: a multifaceted threat. Clin Microbiol Rev 22,
30 ANNEX 1: List of antimicrobials tested over and Epidemiological cutoff values (ECOFF) Resistant strain if MIC value of the isolate > Cutoff Symbol Antimicrobial Cutoff value (mg/ml) AMP Ampicillin 8 CHL Chloramphenicol 16 CIP Ciprofloxacin 0,064 COL Colistin 2 FOT Cefotaxime 0,25 GEN Gentamicin 2 NAL Nalidixic acid 16 SMX Sulphonamide 64 TAZ Ceftazidime 0,5 TET Tetracycline 8 TMP Trimethoprim 2 Outputs of the univariate logistic regression model ( ) The LOGISTIC Procedure SPECIES= calve year0 at substance=amp year0 at substance=chl year0 at substance=cip year0 at substance=col year0 at substance=fot year0 at substance=gen year0 at substance=nal year0 at substance=smx year0 at substance=taz
31 year0 at substance=tet year0 at substance=tmp The LOGISTIC Procedure SPECIES= cattle year0 at substance=amp year0 at substance=chl year0 at substance=cip year0 at substance=col year0 at substance=fot year0 at substance=gen year0 at substance=nal year0 at substance=smx year0 at substance=taz year0 at substance=tet year0 at substance=tmp The LOGISTIC Procedure SPECIES= chicken year0 at substance=amp year0 at substance=chl year0 at substance=cip year0 at substance=col
32 year0 at substance=fot year0 at substance=gen year0 at substance=nal year0 at substance=smx year0 at substance=taz year0 at substance=tet year0 at substance=tmp The LOGISTIC Procedure SPECIES= pig year0 at substance=amp year0 at substance=chl year0 at substance=cip year0 at substance=col year0 at substance=fot year0 at substance=gen year0 at substance=nal year0 at substance=smx year0 at substance=taz year0 at substance=tet year0 at substance=tmp
33 Outputs of the univariate logistic regression model, year by year The LOGISTIC Procedure SPECIES= calve AMP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= CHL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= CIP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0=
34 COL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= FOT year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= GEN year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0=
35 NAL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= SMX year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= TAZ year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0=
36 TET year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= TMP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= The LOGISTIC Procedure species=cattle AMP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0=
37 CHL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= CIP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= COL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0=
38 FOT year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= GEN year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= NAL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= SMX 38
39 year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= TAZ year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= TET year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= TMP 39
40 TMP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= The LOGISTIC Procedure species=chicken AMP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= CHL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0=
41 CHL year0 at year0= CIP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= COL year0 at year0= year0 at year0= year0 at year0= year0 at year0=3 <0.001 < year0 at year0=4 <0.001 < year0 at year0=5 <0.001 < FOT year0 at year0= year0 at year0= year0 at year0= year0 at year0=
42 FOT year0 at year0= year0 at year0= GEN year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= NAL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= SMX year0 at year0= year0 at year0= year0 at year0=
43 SMX year0 at year0= year0 at year0= year0 at year0= TAZ year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0=
44 TET year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= TMP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= The LOGISTIC Procedure species=pig AMP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0=
45 CHL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= CIP year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= COL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= FOT 45
46 year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= GEN year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= NAL year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= SMX 46
47 SMX year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= TAZ year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= TET year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= TMP 47
48 year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= year0 at year0= ANNEX 2: GEE linear model with multiple comparisons corrections (pvalues) CALVES Test probz Bonferroni Linear Stepup AMP CHL < CIP < < < COL FOT GEN NAL < < < SMX < TAZ TET TMP < < <
49 CATTLE Test probz Bonferroni Linear Stepup AMP < < CHL CIP < COL FOT GEN NAL < < SMX TAZ TET TMP < < CHICKEN Test probz Bonferroni Linear Stepup AMP CHL CIP COL < < < FOT < < < GEN NAL SMX
50 Test probz Bonferroni Linear Stepup TAZ < < < TET < < < TMP PIG Test probz Bonferroni Linear Stepup AMP CHL CIP < COL FOT GEN NAL < < < SMX TAZ TET TMP
51 Results of the univariate (logistic regression) and multivariate (GEE) analysis are summarized hereafter in a table using simple symbols in order to get an overall picture of the situation over the six consecutive years and to easily make comparisons between animal categories. All indicated trends (, ) were statistically significant (p = 0.05) both in univariate (logistic regression) and multivariate (GEE) analysis, even after using correction methods for multiple testing (Bonferroni and Linear stepup method), unless otherwise mentioned. Veal Calves Beef Cattle Chickens Pigs AMP CHL 1 CIP ++ COL 3 FOT 2 1 GEN 1 NAL SMX TAZ 1 2 TET TMP 1 ++ = High prevalence (> 50%) for the 6 consecutive years = decreasing trend of resistance detected* 1=Trend not significant after p value adjustment with Bonferroni method 2= Trend not significant in multivariate analysis (GEE) 3=Trend not significant in univariate analysis (logistic regression) but significant in multivariate analysis (GEE). *statistically significant trend (5% significance level) detected at least once during the 6 years 51