Preventive Veterinary Medicine

Preventive Veterinary Medicine 130 (2016) 41 50 Contents lists available at ScienceDirect Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed Quantitative and qualitative antimicrobial usage patterns in farrow-to-finish pig herds in Belgium, France, Germany and Sweden M. Sjölund a,b,, M. Postma c, L. Collineau d,e, S. Lösken f, A. Backhans b, C. Belloc e,g, U. Emanuelson b, E.Gro e Beilage f, K. Stärk d, J. Dewulf c, on behalf of the MINAPIG consortium a Department of Animal Health and Antimicrobial Strategies, National Veterinary Institute, SVA, SE-751 89 Uppsala, Sweden b Department of Clinical Sciences, Swedish University of Agricultural Sciences, P.O. Box 7054, SE-750 07 Uppsala, Sweden c Veterinary Epidemiology Unit, Department of Reproduction, Obstetrics and Herd Health, Faculty of Veterinary Medicine, Ghent University, Salisburylaan 133, 9820 Merelbeke, Belgium d SAFOSO AG, Waldeggstrasse 1, CH 3097 Bern Liebefeld, Switzerland e UMR1300 BioEpAR, LUNAM Université, Oniris, INRA, F-44307 Nantes, France f Field Station for Epidemiology, University of Veterinary Medicine Hannover, Büscheler Straße 9, D-49456, Bakum, Germany g INRA, UMR1300 Biology, Epidemiology and Risk Analysis in animal health, CS 40706, F-44307 Nantes, France a r t i c l e i n f o Article history: Received 25 January 2016 Received in revised form 13 May 2016 Accepted 1 June 2016 Keywords: Pig Antimicrobial consumption Use Treatment incidence Antibiotic a b s t r a c t Data on sales of antimicrobials using a standardised methodology have shown that there are vast differences between countries in amounts of antimicrobials sold for food-producing animals, but these data do not provide insight on how sales are distributed by species and age groups. The aim of this study was to compare herd level antimicrobial usage for pigs by age category, antimicrobial class and administration route for farrow-to-finish herds in four EU countries. A cross-sectional study was conducted among 227 farrow-to-finish pig herds with at least 100 sows and 500 finishing pigs in Belgium (n = 47), France (n = 60), Germany (n = 60) and Sweden (n = 60). Detailed information about the antimicrobial consumption for breeding and growing pigs was collected. Antimicrobial usage was quantified as active substance expressed as mg and then converted to treatment incidence (TI) based on Defined Daily Doses Animal per 1000 pig-days at risk. TIs varied between and within countries, herds and age groups. The Swedish herds had the lowest and the German herds the highest overall use. Most treatments were applied to weaned piglets except in the Swedish herds where treatments of suckling piglets were most frequent. Antimicrobials were most often applied through feed or water except in the Swedish herds where parenteral treatments were most frequent. Aminopenicillins was the antimicrobial class most commonly used. Use of third and fourth generation cephalosporins constituted 11% of use for the Belgian herds, which was higher compared to the other countries. There was a significant (p < 0.01) association between the within-herd antimicrobial use across different age categories. This study has shown that there were large differences in antimicrobial use for pigs between countries, herds and age groups in farrow-to-finish herds of similar size when actual consumption data were compared. Collecting detailed usage data can be used to efficiently target high users in order to reduce antimicrobial consumption. 2016 Elsevier B.V. All rights reserved. 1. Introduction Antimicrobials are crucial to both veterinary and human medicine but its use is hampered due to the risk of selection for and spread of resistance. Levels of resistance in pathogens varies Corresponding author at: Department of Animal Health and Antimicrobial Strategies, National Veterinary Institute, SVA, SE-751 89 Uppsala, Sweden. E-mail address: marie.sjolund@sva.se (M. Sjölund). between countries (European Food Safety Authority (EFSA, 2015)) and when data on sales of antimicrobials for food-producing animals were compared to data on antimicrobial resistance, a clear association between use and resistance was found (Chantziaras et al., 2014; European Centre for Disease Prevention and Control (ECDC, 2015)) To retain a multitude of treatment options, it is therefore of utmost importance to reduce antimicrobial use to a minimum but without jeopardising animal health. Further, whenever antimicrobials are required to treat infectious diseases in animals, they should be applied prudently in line with the recently http://dx.doi.org/10.1016/j.prevetmed.2016.06.003 0167-5877/ 2016 Elsevier B.V. All rights reserved.

42 M. Sjölund et al. / Preventive Veterinary Medicine 130 (2016) 41 50 issued Guidelines for the prudent use of antimicrobials in veterinary medicine by the European Commission (Commission Notice, 2015). For a number of years, data on sales of antimicrobials using a standardised methodology have been available which has allowed for comparisons between countries and provided valuable information for future work on reducing antimicrobial use (European Surveillance Veterinary Antimicrobial Consumption (ESVAC, 2015)). The ESVAC reports have shown that there are vast differences between countries in amounts of antimicrobials sold for food-producing animals (Grave et al., 2014; ESVAC, 2015), but these data do not provide insight on how sales are distributed by species and age groups. Other sources however have shown that the use of antimicrobials for pigs is higher than for other species (Merle et al., 2012; Bondt et al., 2013; DANMAP, 2013; Hosoi et al., 2013) and use may differ between age groups (Dunlop et al., 1998; Hybschmann et al., 2011; Callens et al., 2012; Jensen et al., 2012; Moreno, 2014; van Rennings et al., 2015). Comparing the sales on species level may change the relation between countries with regards to sales, as size and structure of the pig producing sector varies considerably between the European countries (Eurostat, 2014). Further, sales figures (expressed in kg active substance in relation to a certain measure of biomass produced) are only a crude measurement of actual consumption and do not provide insight into actual use and how use with regards to antimicrobial substance is distributed between different animal species and age categories. Moreover they do not take into account the huge differences in dosage between different antimicrobial compounds. As also suggested by ESVAC (ESVAC, 2015) and others (Bondt et al., 2013), a more detailed data collection of usage on species level would allow for a more in-depth analysis of antimicrobial consumption in different countries with different animal populations and production types. If antimicrobial usage data can be collected on herd level in a standardised manner, benchmarking could be applied across nations to stimulate reduction of antimicrobial use. Benchmarking has previously been shown to be an effective tool in Denmark (Jensen et al., 2014) and the Netherlands (Bos et al., 2013; Autoriteit Diergeneesmiddelen SDa, 2014; Speksnijder et al., 2014), and a similar farm level approach has recently been introduced in Belgium (Antimicrobial Consumption and Resistance in Animals (AMCRA, 2014)). The present study was initiated as a first step for comparing antimicrobial usage on herd level in different countries using a standardised methodology. The specific aim of the study was to compare herd level antimicrobial usage by age category, antimicrobial substance and administration route for farrow-to-finish herds in four different EU countries with different intensities of pig production. Our hypothesis was that usage differs between countries and age groups as well as usage patterns with respect to administration route and choice of active substances. This study was conducted within the European research project MINAPIG (Evaluation of alternative strategies for raising pigs with minimal antimicrobial usage: Opportunities and constraints, www.minapig. eu). 2. Material and methods 2.1. Selection of herds A cross-sectional study was conducted among 227 farrowto-finish pig herds located in Belgium (n = 47), France (n = 60), Germany (n = 60) and Sweden (n = 60) between December 2012 and December 2013. Inclusion criterion was the presence of at least 100 sows and 500 finishing pigs. In Belgium, herds located in the Flanders region (representing 90% of the national pig production) were recruited among those subscribing to a professional newsletter regularly issued by the University of Ghent. The newsletter is distributed to 609 subscribing farmers, veterinarians and other herd advisors with a known interest in biosecurity in pig production. In France, a simple random sample of 110 herds was drawn from a technical database maintained by the French Institute for pig and pork industry (IFIP) which on average covers 46% of the French pig herds with more than 50 sows. Inclusion criteria included localization in the North western region (representing 75% of the French pig production), as well as the presence of at least 100 sows and 500 finishing pigs. Eleven herds were excluded from the original sample mostly due to previous refusal to participate in studies. From the remaining sample, 79 herds were contacted to get the sample of 60 herds. German herds were recruited via consultancy circles together with contacts provided by several veterinary practices in the three regions, Mecklenburg-Vorpommern, Niedersachsen and Nordrhein-Westfalen, with the largest pig production constituting 64% of the total German pig production (Statistisches Bundesamt, 2014). Consultancy circles are regional farmers organisations for pig producers as well as agricultural consulting organisations. A consultancy circle consists of a number of chairpersons together with ordinary members. Several, ten to 15, of the study herds belonged to a consultancy circle in Vechta, Kreislandvolkverband Vechta, which represented 2000 members and 1600 active farms. The office in Vechta employed 50 persons. Swedish herds were selected on a willingness to participate basis among herds either affiliated with the Swedish Animal Health Service (SvDHV) with a herd veterinarian working for SvDHV, or were herds with previous contact with researchers at the National Veterinary Institute as previously described (Sjölund et al., 2015). 2.2. Collection of antimicrobial consumption records Herd visits were undertaken between December 2012 and December 2013. During the herd visits, detailed information about the antimicrobial consumption of breeding and growing pigs was collected. In Belgium, France and Germany, all herds were visited by the same investigator within country. These investigators were researchers within MINAPIG except in France where a veterinary student in the final year conducted all data collection for the students graduation project. In Sweden, recruited herds were visited either by the herd veterinarian from SvDHV (46 herds and 15 veterinarians) or by a researcher within MINAPIG (14 herds and 2 veterinarians) if the herd was not affiliated with SvDHV (Sjölund et al., 2015). Antimicrobial consumption data for the participating herds were collected for one year preceding the visit in Belgium, Germany and Sweden and for the last batch in France. Invoices from veterinarians and feed companies combined with information from the farmer were used in Belgium. Dispensing and application forms from the prescribing veterinarian used as treatment records were used for the German herds. In Sweden, antimicrobial consumption data were retrieved from the farmers treatment records. In France, antimicrobial consumption data were retrieved from the farmers treatment records together with farmers directed interview. Data collection was based on a form developed within the MINAPIG Consortium which can be provided by the authors of the article on request. Antimicrobial consumption, expressed as volume or mass, was recorded by product, strength of product, administration route and age category (breeding pigs, piglets, weaners and finishers). 2.3. Quantification of antimicrobial consumption Antimicrobial usage was quantified using the ABcheck.UGent TM online tool developed by the Unit for Veterinary Epidemiology of the Faculty of Veterinary Medicine, University of Ghent (http:// www.abcheck.ugent.be/). The ABcheck converts recorded antimi-

M. Sjölund et al. / Preventive Veterinary Medicine 130 (2016) 41 50 43 crobial use to active substance expressed as mg and then to treatment incidence (TI) based on Defined Daily Doses Animal (DDDA). The TI is given as the number of DDDAs per 1000 pig-days at risk which is equivalent to how many pigs per 1000 pigs receiving a dose of antimicrobials each day (Timmerman et al., 2006). The TI was calculated for each herd and age group in accordance with Timmerman and co-workers (Timmerman et al., 2006) as follows: Total amount of antimicrobials administered (mg) Total amount of antimicrobials administered (mg) TI = DDDA (mg/kg/day) number of days at risk kg animal at risk 1000 pigs at risk Previously established consensus DDDAs, as described by Postma and co-workers (Postma et al., 2015b), were used. Briefly, DDDAs were established by taking the mean of the recommended dose per kg for all products authorised for pigs with the same active substance and the same administration route. All antimicrobials were classified according to the anatomical therapeutic chemical veterinary (ATCvet) classification system (World Health Organization WHO, 2013 Postma et al., 2015b). TI calculations were performed for each administered antimicrobial product per age category. All TIs for the different products used were then added to each other to reach the overall TI for the particular age category. The number of days at risk for each age category was equal to the rearing period for that category. This is the time period when a pig could receive treatment with antimicrobials. The actual length in days of the rearing periods for the different age categories, i.e. the number of days for the suckling, weaning and fattening periods as defined below, of the individual herds as reported by the farmers were used in the calculations. For breeding pigs (boars, sows, gilts), the time period was set to 365 days. The following age categories were used for the calculation: suckling piglets (birth to weaning); weaners (weaning to an approximate weight of 20 30 kg), fatteners ( 30 kg to slaughter) and breeding pigs including gilts. The kg animal at risk is the total weight of pigs for that particular age category (in kilograms). Uniform weights agreed on by the MINAPIG consortium were used. These weights represent the assumed average weight at treatment and were: two kg for suckling piglets, seven kg for weaners, 35 kg for fatteners, 60 kg for gilts and 220 kg for adult pigs (boars and sows). They were subsequently multiplied with the average number of pigs in a batch calculated from the total number of pigs produced per year for the respective age groups and then divided by the number of batches in a year. For adult pigs, the average number of gilts, boars and sows in the herd over a one-year period was used. If antimicrobial consumption data were provided as the total amount used in a year, this amount was divided by the number of batches produced so that data for growing pigs could be entered per batch which is the required format of the online tool. The TI of suckling piglets, weaned piglets and fatteners were finally combined and recalculated into a standardised lifespan of 200 days to correct for possible differences in ages at slaughter between herds. The TI200 was calculated by dividing the sum of all TIs from birth to slaughter with the number of days for the actual rearing period and then multiplying the TI per day with 200 which was the agreed-on standardised life span of 200 days used in the equation. The standardised TI is hereafter referred to as TI200. 2.4. Statistical analysis After performing descriptive statistics for the relevant variables (TI suckling piglets, TI weaners, TI finishers, TI200, TI breeding pigs), all TI variables were LOG transformed to overcome the problem of a skewed distribution. To overcome the problem of the observations where the TI for a particular age category was 0 (which cannot be log transformed), first a value of 1 was added to all observations and then this data was log transformed. In that way the herds with no reported antimicrobial consumption received a value of 1 in the original scale which then becomes 0 again after log transformation. Differences in TIs between countries were tested by means of ANOVA. Scheffé s method was used for post hoc comparison. To compare TIs for the different age categories, multivariable linear regression was performed including country as independent fixed variable to account for country differences. Normal probability tests and plots were examined to check whether assumptions of normality and homoscedasticity were fulfilled. Furthermore, cross tabulation was used and graphs were prepared to visualize the results. All statistical analyses were performed using SPSS statistics 22 (IBM Corporation, 2013, Armonk, New-York). 3. Results 3.1. Herd characteristics The number of sows enrolled in the study were 15 816 for the Belgian herds, 12 015 for the French herds, 23 770 for the German herds and 14 279 for the Swedish herds. On a national level, the numbers of sows were 472 000, 1 043 00, 2 054 000 and 148 000 for Belgium, France, Germany and Sweden, respectively (Eurostat, 2014). The corresponding figures for the number of fattening pigs produced 1 were 394 604 for the Belgian herds, 273 464 for the French herds, 318 369 for the German herds and 302 748 for the Swedish herds. On a national level, the numbers of slaughtered pigs were 11 915 000, 23 747 000, 58 628 000 and 2 551 000 for Belgium, France, Germany and Sweden, respectively (Eurostat, 2014). The median herd sizes were 290, 174, 300 and 188 sows for the Belgian, French, German and Swedish herds, respectively. The median weaning age in the studied herds were 24, 22, 25 and 35 days for the Belgian, French, German and Swedish herds, respectively. The median rearing periods from weaning until allocation in the fattening unit were 49, 49, 56 and 49 days for the Belgian, French, German and Swedish herds, respectively. The median time pigs spent in fattening units before being slaughtered were 125, 130, 112 and 105 days for the Belgian, French, German and Swedish herds respectively. Additional information on herd characteristics can be found in previously published articles by Backhans et al. (2015), Postma et al. (2015a) and Sjölund et al. (2015). 3.2. Treatment incidences Treatment incidences over a standardised life span of 200 days (TI200), varied considerably between countries as shown in Fig. 1 and Table 1. The lowest average TI200 was observed for the Swedish (SE) herds followed by the French (FR) herds, the Belgian (BE) herds and lastly German (DE) herds which had the highest TI200 (Table 1). French and Belgian herds did not differ from each other (p = 0.1) whereas significant differences were observed between the other countries (p < 0.05). In all countries considerable variations between herds were observed (Fig. 1;Table 1). Frequency distributions of TI200 for the different countries showed a righttailed distribution (Fig. 2). When TIs for growing pigs were split for different age categories, it was observed that weaned piglets received most treatments followed by suckling piglets (Fig. 1; Table 1). However, the reverse situation was observed for the Swedish herds where most treatments were applied to suckling piglets. Treatment incidences for 1 Some herds sold pigs at the end of the weaning period before pigs were allocated to the fattening units. These pigs are not included in the numbers presented.

44 M. Sjölund et al. / Preventive Veterinary Medicine 130 (2016) 41 50 Fig. 1. Antimicrobial treatment incidences (TI), expressed as treatment incidence per 1000 pig-days at risk, for pigs in farrow-to-finish herds in Belgium (n = 47), France (n = 60), Germany (n = 60) and Sweden (n = 60) shown by age category (suckling piglets, weaned piglets (nursery), finishers, growing pigs (TI200) and breeding pigs). TI200 denotes the overall TI for growing pigs with a standardised lifespan of 200 days. Different letters denote significant differences between countries within the respective age categories. Table 1 Antimicrobial treatment incidences (TI) for farrow-to finish pig herds expressed as the TI per 1000 pigs-days at risk shown for different age categories, including a calculated TI for growing pigs from birth to slaughter with a standardised life span of 200 days. Differences in TIs between countries were tested by means of ANOVA. Scheffé s method was used for post hoc comparison. Belgium (n = 47) France (n = 60) Germany (n = 60) Sweden (n = 60) Median Mean Median Mean Median Mean Median Mean Age category (Min Max) (Min Max) (Min Max) (Min Max) Suckling piglets 151.8 175.6 a,c 12.9 59.1 b 138.9 245.0 a 54.7 76.0 c (0.0 883.6) (0.0 637.7) (8.1 1496.4) (1.6 367.9) Weaned piglets 339.5 407.1 a 320.1 374.3 a 487.6 633.4 b 6.1 21.4 c (0.0 1542.4) (0.0 1794.6) (0.0 1965.8) (0.0 260.5) Fattening pigs 8.3 33.0 a,c 0.0 7.3 b 20.9 52.9 a 2.8 6.1 c (0.0 231.9) (0.0 231.8) (0.0 399.2) (0.0 64.9) Entire growing period (TI200) 107.8 142.9 a 94.7 108.0 a 189.0 242.8 b 14.3 22.7 c (0.0 531.1) (0.0 539.0) (3.8 673.9) (1.6 116.0) Breeding pigs 6.1 15.6 a 0.7 22.1 a 21.1 42.0 b 8.4 10.9 a (Boars, sows, gilts) (0.0 134.9) (0.0 382.5) (0.2 204.5) (0.0 45.0) * Different superscripts within a row indicate significant differences. suckling piglets was lowest among the French herds and was significantly lower than in the other countries (p < 0.01). The Swedish herds had the second lowest TI for suckling piglets followed by the Belgian herds and then the German herds. In addition to the French herds, the Swedish herds differed from the German herds (p < 0.01) but no differences were observed between Belgian and Swedish herds (p = 0.1) and Belgian and German herds (p = 0.4) in TIs for suckling piglets. Average TIs for weaned piglets was lowest for the Swedish herds with a mean TI of 21.4, which differed significantly (p < 0.01) to TIs for weaners in the three other countries (Fig. 1; Table 1). The TIs for the Belgian and French herds were similar to each other (p = 0.9), while the mean TI for weaned piglets of the German herds differed to those for the Belgian and French herds (p < 0.05). For both suckling and weaned piglets there were large variations in TIs between herds in all four countries. For suckling piglets, two herds in Belgium (4%) and four in France (7%) reported that no treatments had been applied to suckling piglets while all herds in Germany and Sweden reported treatment of suckling piglets. For weaned piglets, only Germany reported treatment of weaned piglets in all herds while four of the Belgian herds (8%), six of the French (10%) and five of the Swedish (9%) herds did not report any treatments for this age category. Overall, treatments of finishing pigs were low but also varied between herds and countries with the lowest average TI for the Swedish herds, followed by that of the French herds which was still significantly different to that of the Swedish herds (p < 0.01) (Fig. 1; Table 1). The average TI for the Belgian herds for finishers was only significantly different to the French herds (p < 0.01), whereas the average for the German herds differed to the French and Swedish herds (p < 0.01). Forty-two (70%) of the French herds did not report any use of antimicrobials for their fattening pigs. The corresponding figures for the other three countries were 11 for the Belgian herds (23%), 7 for the German herds (12%) and 3 for the Swedish herds (5%). Overall, it was observed that from birth till slaughter (total TI = 117 174), 26% of all treatments were performed during the suckling phase, 69% during the weaning phase, whereas only 5% during the fattening phase. Also for breeding animals, mainly sows, large variations between herds and also between countries were found (Fig. 1;

M. Sjölund et al. / Preventive Veterinary Medicine 130 (2016) 41 50 45 Fig. 2. a-da Within-herd associations of log-transformed treatment incidences (TI) per 1000 pig-days at risk for different age categories (suckling piglets sucklers ; weaned piglets nursery ; fattening pigs finishers ; gilts, sows, boars breeding pigs ) for 227 farrow-to-finish pig herds in Belgium (n = 47), France (n = 60), Germany (n = 60) and Sweden (n = 60). Table 1). Swedish herds had the lowest average TI for sows followed by the Belgian herds and not significantly different from the Swedish herds (p = 0.6). The French herds differed from those of Germany and Sweden (p < 0.05). Further exploration of the distribution of TIs for breeding animals showed that TIs for the 10% of the herds which used most were 40.7 (BE), 69.6 (FR), 122.9 (DE) and 21.4 (SE). For growing pigs, the corresponding TI200 for the 10% of the herds which used most were 315.8, 264.9, 483.7 and 54.9 for Belgium, France, Germany and Sweden, respectively. 3.3. Associations of treatment incidences between age groups When within-herd TIs were investigated for the different age categories, an association (p < 0.01) was found (Fig. 2a d). This association was observed also when corrected for country effect. The association was strongest for the Swedish herds when TIs for breeding animals were compared to suckling piglets and suckling piglets to fattening pigs, i.e. herds that treated their breeding animals also tended to treat their suckling piglets and fatteners. The association was on the other hand strongest for the German herds when TIs for suckling piglets were compared to weaned piglets and weaned piglets to finishing pigs, i.e. herds that treated their suckling piglets also tended to treat their weaned piglets and fatteners. 3.4. Antimicrobial usage by active substance and administration routes The most commonly used antimicrobials were aminopenicillins followed by polymixins as shown in Table 2. Macrolides were the third most commonly used antimicrobials followed by tetracyclines. When overall use within a country was investigated, different usage patterns were seen. Use in the Belgian herds was similar to that of the overall pattern with aminopenicillins being most common followed by polymixins and macrolides. Third and fourth generation cephalosporins were the fourth most commonly used antimicrobials in the Belgian herds constituting 11% of the overall use. In the French herds, polymixins were most commonly used, tetracyclines were second most commonly used, followed by the aminopenicillins and the macrolides. Aminopenicillins also dominated use in the German herds which was followed by macrolides, tetracyclines and polymixins. The pattern for the Swedish herds was different with benzylpenicillins dominating, followed by the trimethoprim-sulphonamides, macrolides and aminopenicillins only being fourth. The proportion of the total number of antimicrobial treatment incidences per 1000 pig-days at risk shown by antimicrobial class is shown in Fig. 3. Additional information on which antimicrobials and combinations of antimi-

46 M. Sjölund et al. / Preventive Veterinary Medicine 130 (2016) 41 50 Table 2 Proportion of amount of antimicrobials used in 227 farrow-to-finish pig herds in Belgium (n = 47). France (n = 60). Germany (n = 60) and Sweden (n = 60) shown by country, total use and use by age category. Superscript numbers within column denotes rank of amount used. Country Overall Age category Antimicrobial class Belgium France Germany Sweden Total % LA b of total Suckling piglets Weaned piglets Fattening pigs Breeding pigs Aminoglycosides <0.1% 7.9% 1.2% 0.2% 2.4% a 9.8% 89.7% 0.4% <0.0% Aminopenicillins 37.7% 1 15.7% 3 35.7% 1 6.2% 4 30.0% 1 17.7% 15.9% 78.6% 4.7% 0.9% Amphenicols 0.1% 0.5% <0.1% a 0.1% 100.0% 69.1% 10.7% 6.7% 13.5% Benzylpenicillin 0.4% a 1.0% 61.2% 1 4.0% 48.0% 73.8% 8.5% 6.1% 11.6% Benzylpenicillin in combination <0.1% 2.1% 4.6% 0.9% 2.7% a 90.2% 8.6% <0.0% 1.2% 3rd & 4th generation Cefalosporins 10.8% 4 1.2% 1.8% a 3.7% 88.3% 94.2% 4.9% 0.1% 0.7% Colistin c 17.5% 2 30.1% 1 13.6% 4 4.3% 17.7% 2 a 1.6% 98.1% 0.3% 0.1% Fluoroquinolones 5.3% 0.4% 1.3% 1.3% 2.1% 4.3% 58.2% 28.5% 4.4% 9.0% Lincosamides <0.1% 1.5% 0.4% a 0.5% a <0.1% <0.1% 0.1% <0.1% Lincosamides and spectinomycin 1.6% 2.9% 0.5% a 1.2% a 41.2% 37.9% 14.3% 6.6% Macrolides 14.7% 3 11.1% 4 17.7% 2 9.0% 3 15.0% 3 60.9% 56.7% 34.2% 4.7% 4.5% Pleuromutilins 0.1% 0.5% 1.4% 1.0% 0.9% a a 70.9% 15.9% 13.1% Sulfonamides and trimethoprim 5.1% 8.0% 3.4% 13.1% 2 5.4% a 11.2% 74.1% 2.8% 11.9% Tetracyclines 6.8% 18.2% 2 17.3% 3 2.9% 14.1% 4 2.3% 1.8% 77.6% 9.2% 11.5% a The antimicrobial or formulation of the antimicrobial was not used at all. b LA long acting formulation. c Also includes colistin in combination with aminopenicillins constituting less than 0.1% of the overall colistin use. Fig. 3. Proportion of antimicrobial classes of the total number of antimicrobial treatment incidences per 1000 pig-days at risk for 227 farrow-to-finish pig herds in Belgium (n = 47), France (n = 60), Germany (n = 60) and Sweden (n = 60). Colistin combinations constituted less than 0.1% of the overall colistin use and only contained aminopenicillins in addition to colistin. crobials that were used in the herds in the four countries is found in Supplementary Table 1 (Table S1). Further analysis on the use by substance for the different age categories showed that the aminopenicillins were mostly applied to weaned piglets followed by the suckling piglets (Table 2). Also the largest proportion of the polymixins were applied to weaned piglets followed by the suckling piglets. Macrolides were on the other hand most often applied to suckling piglets followed by weaned piglets. Tetracyclines were mostly used for weaned piglets followed by use for breeding animals. Third and fourth generation cephalosporins was mainly applied to suckling piglets followed by use for weaned piglets. Fluoroquinolones was also mainly used for suckling piglets and weaned piglets. On country level, overall use of third and fourth generation cephalosporins constituted 11% (BE), 1% (FR), 2% (DE) and 0% (SE) of total use (Table 2). For fluoroquinolones the corresponding figures were 5% (BE), <1% (FR), 1% (DE) and 1% (SE). Long acting (LA) formulations of injectable products were used to a varying degree for the parenteral treatments in the different countries (Table 2). The majority (88%) of the third and fourth generation cephalosporins used were LA formulations whereas only 4% of the fluoroquinolones used were LA formulations (Table 2). Of the aminopenicillins, 18% were LA formulations and 48% of the benzylpenicillins applied were LA formulations, while 61% of the macrolides used were LA formulations. Antimicrobials were most commonly applied orally through water or feed (Fig. 4). Overall, 71% of the antimicrobials were

M. Sjölund et al. / Preventive Veterinary Medicine 130 (2016) 41 50 47 Fig. 4. Proportion of the total number of antimicrobial treatment incidences per 1000 pig-days at risk for 227 farrow-to-finish pig herds in Belgium (n = 47), France (n = 60), Germany (n = 60) and Sweden (n = 60) shown by administration route. *Topical treatments constituted <0.05% and <0.01% of the total number of treatment incidences for Belgium and Germany respectively. Topical treatments were not reported for the French and Swedish herds. applied orally and 29% applied parentally also including a few topical (TI = 10 in total) treatments. Differences were however observed between the countries, where Swedish herds applied 13% of the antimicrobials orally whereas the French herds applied 87% orally. Oral administration for the Belgian and German herds was similar, 70% and 71%, respectively. 4. Discussion To our knowledge, this is the first comprehensive investigation of actual antimicrobial use in farrow-to-finish pig herds comprising several countries and using a standardised methodology. We are now able to confirm that antimicrobial usage actually differs a lot between countries, herds and age groups. Previous studies have mostly been conducted at national level and different methods for quantifying antimicrobial use have been used (Stege et al., 2003; Timmerman et al., 2006; Callens et al., 2012; Bos et al., 2013; Moreno, 2014; van Rennings et al., 2015), disabling direct comparisons. These studies have on the other hand shown that there are differences in which antimicrobials (Timmerman et al., 2006; Callens et al., 2012; Bondt et al., 2013) and to which age groups (Dunlop et al., 1998; Hybschmann et al., 2011; Callens et al., 2012; Jensen et al., 2012; Moreno, 2014; van Rennings et al., 2015) antimicrobials have been applied. However, in these national studies, data were collected from different types of herds and production stages which could have influenced consumption and consequently hampers comparability. The highest average overall antimicrobial consumption was found among the German herds and the lowest in Swedish herds when TIs for the entire growing period, TI200, were compared which corresponds with the European sales figures (ESVAC, 2015). There are a number of reasons which may account for these differences. Firstly, the prevalence of different pathogens and consequently presence of subclinical and clinical disease may vary. Farmers and veterinarians enrolled in a British focus group study on antimicrobial use and prescribing behaviours considered health status to be a key factor in antimicrobial consumption (Coyne et al., 2014). This may in part explain the low use in the Swedish herds since Sweden is declared free from PRRS (Carlsson et al., 2009), and in between-herd variations in the other countries since, for example, more severe disease has been observed in pigs concomitantly infected with PRRS and Streptococcus suis (Thanawongnuwech et al., 2000). Secondly, pig and herd density may influence disease transmission between herds within a region and pig density is lower in Sweden compared to the other countries in this study and all of the Belgian, French and German herds originated from pig dense regions (Eurostat, 2014). However, a recent Danish study, which investigated clustering of herds with large amounts of prescribed antimicrobials, found that not only farm density influenced antimicrobial use but other factors also played a role (Fertner et al., 2015b), but further research is required to identify these factors. In addition to the variation between countries, there were also large between-herd variations within the different countries. Biosecurity measures such as all-in all-out production and a single supplier of weaners, all related to reducing disease transmission, have been identified as common practice in herds with low antimicrobial use (Fertner et al., 2015a). The correlation of use between different age categories within herds irrespective of country in our study also support that presence and spread of infections within a herd can influence consumption without ruling out other factors. Another clue to the large differences in antimicrobial use may also lie in how antimicrobials were applied. Group treatment through feed or water was the most common way to treat pigs in the herds in Belgium, France and Germany. However, in the Swedish herds, with the lowest overall use, parenteral treatment was most common. This discrepancy is also seen in sales of antimicrobials (ESVAC, 2015) and could account for some of the differences in consumption. Moreover, it was common practice in many participating herds in Belgium, France and Germany to apply treatments to entire batches of pigs, mainly suckling and weaned piglets, at strategic time points when pigs were judged most likely to contract disease (data not shown). Treatment of entire batches will inherently increase consumption compared to treatment of single individuals within a batch. In addition to the variations between herds and countries, there were also variations to which age group most treatments were applied. Weaned piglets was the age group that received most treatments which is in accordance with several other studies (Hybschmann et al., 2011; Callens et al., 2012; Jensen et al., 2012; Moreno, 2014). Yet again, the Swedish herds differed in that treatment of suckling piglets was more common than treatment of weaned piglets, similar to what was described in Canada (Dunlop et al., 1998) and Germany in another study (van Rennings et al.,

48 M. Sjölund et al. / Preventive Veterinary Medicine 130 (2016) 41 50 2015). To determine the cause of these differences requires further investigations. The treatment incidence for fattening and breeding pigs was in general lower in comparison to the younger age categories, but also for these age categories there were apparent differences between the participating countries as well as between herds. Quite many herds reported no treatments at all for fattening pigs and as many as 70% of the French herds did not report any treatment of fatteners. It is possible that treatment incidences could be under- or overestimated since data for the French herds were only collected for one batch and not during an entire year as was done in the other three countries. Thus any seasonal differences in occurrence of clinical disease and unexpected outbreaks of acute disease requiring medication influencing overall treatment incidence could be lacking. This may explain the low reported antimicrobial use in some herds but on the other hand, low use in fatteners and sows has previously been reported for French herds (Hémonic et al., 2013). Further, therapeutic treatments of individual pigs are not always recorded (Callens et al., 2012) and five Belgian herds of the 52 originally enrolled herds were excluded from the analysis due to lacking information on individual therapeutic treatments although these treatments most certainly did not have a large impact on the overall use. For the Swedish herds, where parenteral therapeutic treatments of individual animals were most common, the risk of underestimation of overall use was considered low (Sjölund et al., 2015). However, Swedish data for sows may be underestimated since 19 of the participating herds were satellites to sow pools 2 and, consequently, treatments performed during gestation were not recorded for these herds (Sjölund et al., 2015). Concerning within herd use for different age categories, a positive association in antimicrobial use between the different age categories as well as a positive association between the use for breeding animals and the use for pigs from birth to slaughter was found. This is in contrast to claims that a high antimicrobial use of young animals prevents disease problems further on. The observed associations may reflect a certain attitude/behaviour of the farmer towards regular use of medicines (Visschers et al., 2015). It may also be caused by an overall higher disease pressure in these herds. A final potential explanation might be the deteriorating effect of antimicrobial use at a young age on the bacterial composition of the gut resulting in higher disease susceptibility and treatment needs in later stages of their lives (Callens et al., 2014). Differences were also found regarding the antimicrobial classes that were most commonly used in herds in the four countries in our study. These differences could depend on a number of factors. First of all, there are differences in levels of resistance in bacteria between countries (EFSA, 2015) which may influence treatment choices. Availability of antimicrobials could also influence treatment choices. The number of products available for use for pigs in the four countries has previously been described in detail where Sweden had the fewest number of available products and Germany the most which could in part also explain the differences (Postma et al., 2015b). Tetracyclines have often been shown to be the most common drug for use in pigs (Timmerman et al., 2006; Bondt et al., 2013) but in our study, tetracylines was only the fourth most commonly used antimicrobial class. However, also for tetracyclines we found differences between countries as it was the secondly most common used class in the French herds but not very commonly used in the Belgian and Swedish herds. The introduction of more novel products and long-acting formulations alongside with the develop- 2 Sow pool is a system where pregnant sows are leased from a central herd unit to other herds ( satellite herds). The sows are transported to the satellite herds three weeks before farrowing, and return to the central unit after weaning. ment of resistance may have shifted use away from tetracyclines. Yet the differences may also be due to the applied unit of measurement. When the use is expressed in weight of used active substance rather than in treatment days (as in the current study) the use of tetracycline will likely be over-represented. The frequent use of macrolides and polymixins, mainly colistin and to a much lesser extent colistin in combination with aminopenicillins, indicate that diarrhoea, especially in weaned pigs, continues to be a major challenge in pig production. The frequent use of these antimicrobials is rather worrying since in the recently issued Guidelines for the prudent use of antimicrobials in veterinary medicine (Commission Notice, 2015), the recommendation is to reduce use of both macrolides and colistin. Moreover, these guidelines explicitly state that colistin is a last-resort drug in human medicine which should not be used as a substitute for good management practices. The recent finding of plasmid-mediated polymixin resistance even more so calls for a restrictive use of colistin (Liu et al., 2015). Further, regarding antimicrobials considered to be especially critical to human medicine, we found a relatively frequent use of third and fourth generation cephalosporins, mainly applied to suckling piglets, for the Belgian herds as also previously shown by Callens and co-workers (Callens et al., 2012). Use of third and fourth generation cephalosporins increases the prevalence of extended spectrum -lactamase (ESBL) producing Escherichia coli (E. coli) (Agersø et al., 2012). Within this context, herd veterinarians with the aid of the Center of expertise on antimicrobial consumption and resistance in animals (AMCRA) needs to continue to enforce its work to reduce use of cephalosporins (AMCRA, 2014) in Belgium. The use of fluoroquinolones was more modest but the same pattern was seen where the Belgian herds had the highest use and the French herds used the least (<1%), while the overall use for the four countries was 4%. The low use in the French herds is most likely a result of the voluntary restriction on the use of third and fourth generation cephalosporins adopted by the pig sector in 2010 (ESVAC, 2015). Since 2013, the use of fluoroquinolones and third and fourth generation cephalosporins has also been restricted by law in Sweden (SJVFS 2013:42, 2013). As of yet, there are not restrictions on the use of fluoroquinolones and cephalosporins in Germany. The overall high antimicrobial use for weaners, in many instances consisting of antimicrobials which should be used sparingly, once again points to that improving health in weaned pigs may be most efficient in reducing overall use of antimicrobials. There are also some aspects on the methodology used in this study which need to be discussed. Although much effort was put in the planning of data collection to as far as possible ensure a uniform way for collecting data irrespective of investigator and country, data collection procedures varied. This could have had an impact on the results but as ways of recording treatments varied between countries, it was necessary to adopt to national conditions to retrieve as much data as possible. In Sweden, where several investigators were involved in data collection, all investigators received training by the two principal Swedish investigators in order to harmonise data collection as much as possible. By involving herd veterinarians from SvDHV for data collection, we got access to antimicrobial consumption data retrieved from their central data base used for the mandatory reporting of antimicrobial treatments to the Swedish Board of Agriculture (SJVFS 2013:42, 2013) which would vouch for completeness and accuracy of data. In general, there may also have been a selection bias towards herds where the farmers had a general interest in questions regarding antimicrobials. Moreover, farmers suspecting a high level of antimicrobial use in their herd may have been more reluctant to participate in the study. Thus, overall antimicrobial use in the participating herds may be lower. In fact, the Swedish herds used less when compared to national data on prescriptions for pigs as previously shown

M. Sjölund et al. / Preventive Veterinary Medicine 130 (2016) 41 50 49 (Sjölund et al., 2015). Whether this is the case for the other participating countries is not known but as only farrow-to-finish herds were enrolled, it is reasonable to assume that trade of pigs was less in these herds compared to more specialized production systems purchasing or selling pigs where the risk of introduction of disease must be considered higher. A clear association between improved biosecurity and reduction in treatments has been previously shown for Belgian farrow-to-finish herds (Laanen et al., 2013). Nevertheless, there were large in-between herd variations in terms of usage with some herds accounting for most of the use. This shows that herds with different levels of use were included in the study sample. Extrapolation to farrow-to-finish herds in general should however be made with caution. It must also be noted that TIs were calculated based on consensus DDDAs adopted for the MINAPIG project (Postma et al., 2015b). DDDA is a technical unit of measurement and should not be assumed to reflect the prescribed or used daily doses. The TIs reported in this study may therefore not accurately reflect the true TI for a particular age group or herd. On the other hand, comparability is not affected and the method allows for comparisons on usage between countries using the same unit of measurement. Comparisons to other studies using TIs based on other DDDAs should however be made with caution as shown by Taverne and co-workers (Taverne et al., 2015). 5. Conclusions The in-between herd differences with respect to antimicrobial consumption both in terms of amounts used and how pharmaceuticals were applied as well as the distribution between age groups show the importance of collecting usage data by country, species, herd, and age group using a standardised methodology. This allows for in-depth comparisons with the aim to improve health and reduce the need for antimicrobials. This study has identified that interventions directed at management of weaned piglets may be of highest priority since most treatments were applied to weaned piglets but further studies are required to identify specific target areas which may reduce the need for antimicrobials in this particular age group. For future studies, it would be interesting to identify common denominators of herds with a low antimicrobial use irrespective of nationality which can be used to provide advice on how to improve health, productivity and subsequently the need for antimicrobials. Conflict of interest None Funding This work was supported by EMIDA ERA-NET. Acknowledgements This project was part of the European MINAPIG project (Evaluation of alternative strategies for raising pigs with minimal antimicrobial usage: Opportunities and constraints, www.minapig.eu), funded by the ERA-NET programme EMIDA (EMIDA19). We would like to thank the MINAPIG consortium for designing the study and for collaborations on the establishment of consensus DDDAs. The MINAPIG consortium consists of the following members, in alphabetical order: Annette Backhans, SLU, Sweden; Catherine Belloc, ONIRIS, France; Lucie Collineau, SAFOSO, Switzerland; Jeroen Dewulf, Ghent University Belgium; Ulf Emanuelson, SLU, Sweden; Elisabeth Grosse Beilage, TiHo Hannover, Germany; Bernd Grosse Liesner, Boehringer Ingelheim, Germany; Christian Alexander Körk, Boehringer Ingelheim, Germany; Ann Lindberg, SVA, Sweden; Svenja Lösken, TiHo, Hannover Germany; Merel Postma, Ghent University, Belgium; Hugo Seemer, Boehringer Ingelheim, Germany; Marie Sjölund, SVA, Sweden; Katharina Stärk, SAFOSO, Switzerland and Vivianne Visschers, ETHZ, Switzerland. We would also like to thank all farmers and their herd veterinarians who participated in this study. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.prevetmed.2016. 06.003. References Antimicrobial Consumption and Resistance in Animals (AMCRA), 2014. Belpork Pioneers Antibiotic Registration. Pig Farmers Will Soon Be Able to Make Comparisons. http://amcra.be/en/news/antibiotic-awareness-day-18- november-2014-press-releases-amcra-belpork. Agersø, Y., Aarestrup, F.M., Pedersen, K., Seyfarth, A.M., Struve, T., Hasman, H., 2012. Prevalence of extended-spectrum cephalosporinase (ESC)-producing Escherichia coli in Danish slaughter pigs and retail meat identified by selective enrichment and association with cephalosporin usage. J. Antimicrob. Chemother. 67, 582 588. SDa, Autoriteit Diergeneesmiddelen, 2014. In: Werner, F.J.M. (Ed.), Usage of Anitbiotics in Agricultural Livestock in the Netherlands in 2013. SDa, Autoriteit Diergeneesmiddelen, Utrecht the Netherlands. Backhans, A., Sjölund, M., Lindberg, A., Emanuelson, U., 2015. Biosecurity level and health management practices in 60 Swedish farrow-to-finish herds. Acta Vet. Scand. 57 (14). Bondt, N., Jensen, V.F., Puister-Jansen, L.F., van Geijlswijk, I.M., 2013. Comparing antimicrobial exposure based on sales data. Prev. Vet. Med. 108, 10 20. Bos, M.E., Taverne, F.J., van Geijlswijk, I.M., Mouton, J.W., Mevius, D.J., Heederik, D.J., Netherlands Veterinary Medicines Authority, S., 2013. Consumption of antimicrobials in pigs, veal calves, and broilers in the Netherlands: quantitative results of nationwide collection of data in 2011. PLoS One 8, e77525. Callens, B., Persoons, D., Maes, D., Laanen, M., Postma, M., Boyen, F., Haesebrouck, F., Butaye, P., Catry, B., Dewulf, J., 2012. Prophylactic and metaphylactic antimicrobial use in Belgian fattening pig herds. Prev. Vet. Med. 106, 53 62. Callens, B., Faes, C., Maes, D., Catry, B., Boyen, F., Francoys, D., de Jong, E., Haesebrouck, F., Dewulf, J., 2014. Presence of antimicrobial resistance and antimicrobial use in sows are risk factors for antimicrobial resistance in their offspring. Microb. Drug Resist. 21 (1), 50 58. Carlsson, U., Wallgren, P., Renstrom, L.H., Lindberg, A., Eriksson, H., Thoren, P., Eliasson-Selling, L., Lundeheim, N., Norregard, E., Thorn, C., Elvander, M., 2009. Emergence of porcine reproductive and respiratory syndrome in Sweden: detection, response and eradication. Transbound. Emerg. Dis. 56, 121 131. Chantziaras, I., Boyen, F., Callens, B., Dewulf, J., 2014. Correlation between veterinary antimicrobial use and antimicrobial resistance in food-producing animals: a report on seven countries. The J. Antimicrob. Chemother. 69, 827 834. Commission Notice, 2015. Guidelines for the prudent use of antimicrobials in veterinary medicine. Off. J. Eur. Union 58 (C299/04). Coyne, L.A., Pinchbeck, G.L., Williams, N.J., Smith, R.F., Dawson, S., Pearson, R.B., Latham, S.M., 2014. Understanding antimicrobial use and prescribing behaviours by pig veterinary surgeons and farmers: a qualitative study. Vet. Rec. 175, 593. DANMAP, 2013. Use of Antimicrobial Agents and Occurrence of Antimicrobial Resistance in Bacteria from Food Animals, Food and Humans in Denmark. National Food institute, Statens Serum Institut, Copenhagen http://www. danmap.org/ /media/projekt%20sites/danmap/danmap%20reports/ DANMAP%202013/DANMAP%202013. ashx. Dunlop, R.H., McEwen, S.A., Meek, A.H., Black, W.D., Clarke, R.C., Friendship, R.M., 1998. Individual and group antimicrobial usage rates on 34 farrow-to-finish swine farms in Ontario, Canada. Prev. Vet. Med. 34, 247 264. European Centre for Disease Prevention and Control (ECDC), EFSA (European Food Safety Authority), EMA (European Medicines Agency), 2015. ECDC/EFSA/EMA first joint report on the integrated analysis of the consumption of antimicrobial agents and occurrence of antimicrobial resistance in bacteria from humans and foodproducing animals. Stockholm/Parma/London: ECDC/EFSA/EMA, 2015. EFSA J. 13 (1), 114, http://dx.doi.org/10.2903/j.efsa.2015.4006 (4006). European Food Safety Authority (EFSA), 2015. EU Summary Report on antimicrobial resistance in zoonotic and indicator bacteria from humans, animals and food in 2013. EFSA J. 13 (2), 178, http://dx.doi.org/10.2903/j.efsa. 2015.4036. European Surveillance of Veterinary Antimicrobial Consumption (ESVAC), 2015. Sales of Veterinary Antimicrobial Agents in 26 EU/EEA Countries in 2013 European Medicines Agency. http://www.ema.europa.eu/docs/en GB/ document library/report/2015/10/wc500195687.pdf.