Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Antibiotic use on German pig farms - A longitudinal analysis for 2011, 2013 and 2014

  • Malin Hemme ,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing – original draft, Writing – review & editing

    malin.hemme@tiho-hannover.de

    Affiliation Department of Biometry, Epidemiology and Information Processing, WHO Collaborating Centre for Research and Training for Health in the Human-Animal-Environment Interface, University for Veterinary Medicine, Hannover, Germany

  • Inga Ruddat,

    Roles Formal analysis, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Department of Biometry, Epidemiology and Information Processing, WHO Collaborating Centre for Research and Training for Health in the Human-Animal-Environment Interface, University for Veterinary Medicine, Hannover, Germany

  • Maria Hartmann,

    Roles Data curation, Methodology, Software, Validation, Visualization

    Affiliation Department of Biometry, Epidemiology and Information Processing, WHO Collaborating Centre for Research and Training for Health in the Human-Animal-Environment Interface, University for Veterinary Medicine, Hannover, Germany

  • Nicole Werner,

    Roles Writing – review & editing

    Affiliation Department of Biometry, Epidemiology and Information Processing, WHO Collaborating Centre for Research and Training for Health in the Human-Animal-Environment Interface, University for Veterinary Medicine, Hannover, Germany

  • Lisa van Rennings,

    Roles Investigation, Validation

    Affiliation Department of Biometry, Epidemiology and Information Processing, WHO Collaborating Centre for Research and Training for Health in the Human-Animal-Environment Interface, University for Veterinary Medicine, Hannover, Germany

  • Annemarie Käsbohrer,

    Roles Writing – review & editing

    Affiliations Federal Institute for Risk Assessment, Berlin, Germany, Institute for Veterinary Public Health, Veterinary University Vienna, Vienna, Austria

  • Lothar Kreienbrock

    Roles Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Supervision, Writing – original draft, Writing – review & editing

    Affiliation Department of Biometry, Epidemiology and Information Processing, WHO Collaborating Centre for Research and Training for Health in the Human-Animal-Environment Interface, University for Veterinary Medicine, Hannover, Germany

Antibiotic use on German pig farms - A longitudinal analysis for 2011, 2013 and 2014

  • Malin Hemme, 
  • Inga Ruddat, 
  • Maria Hartmann, 
  • Nicole Werner, 
  • Lisa van Rennings, 
  • Annemarie Käsbohrer, 
  • Lothar Kreienbrock
PLOS
x

Abstract

To study antibiotic use in livestock in a temporal context with the development of antimicrobial resistance, long-term changes in antibiotic use must be mapped and their possible causes must be explored. Therefore, the present work assesses the changes in antibiotic use over time in German livestock husbandry. In addition, factors associated with antibiotic use were analyzed to identify possible strategies for further reducing antimicrobial usage. For 2011, 2013 and 2014, antibiotic usage data were collected and examined within the VetCAb project. Three hundred participating pig holdings provided information on their antibiotic use based on obligatory application and delivery forms (ADFs) filled in by their veterinarian as well as information on their current stabling capacities for each production type held. Data on sow, piglet, weaner and fattening pig holdings were described separately, using the semi-annual treatment frequency (TF) to measure antibiotic consumption. Multiple linear mixed models were used to investigate the effects of time, farm size, region and farm management category on the treatment frequency. The study yielded significant time changes with p-values below 0.001 in antibiotic administration with a decreasing median TF in piglets from 3.8 in the first half of 2011 (IQR = 1.1–10.6) to 1.7 in the second half of 2014 (IQR = 0.2–4.5) and in fattening pigs from 5.1 in the first half of 2011 (IQR = 0.2–15.4) to 0.7 in the second half of 2014 (IQR = 0.1–6.7). Meanwhile the TF fluctuated between 8.2 and 12.2 in weaners during the observational period (IQRs between zero (lower quartile) and 37.9 (upper quartile)). Piglet, weaner and fattening pig holdings belonging to the upper third of the holdings in size used significantly more antibiotics than the other holdings investigated. Particularly for weaner and fattening pig holdings, a higher TF was noted for farms without breeding units. The region was only a significant factor in weaners. In conclusion, for 2011, 2013 and 2014, the present study shows a clear reduction in antibiotic treatment frequency in German pig holdings. In addition, the association with various factors such as herd size and farm organization on the antibiotic usage frequency is indisputable. Therefore, these factors should be included in monitoring systems and considered when evaluating intervention measures.

Introduction

Knowledge of antibiotic use is necessary for containing of antimicrobial resistance. In 1998, the EU invitational conference, "The Microbial Threat", issued recommendations on how to combat the increasing threat of developing antimicrobial resistance [13]. In 2000, the WHO emphasized the key role of antibiotic and resistance monitoring system data in controlling antimicrobial resistance in their "Global Principles for the Containment of Antimicrobial Resistance in Animals intended for Food" [4]. Recently, the European Medicines Agency (EMA) and the European Food Safety Authority (EFSA) stressed that collecting antimicrobial resistance and consumption data is key to establishing effective measures to control antimicrobial resistance [5]. The World Health Organization (WHO), the Food and Agriculture Organization of the United Nations (FAO), and the World Organization for Animal Health (OIE) presented a global action plan on antimicrobial resistance considering different integrative measures [68].

In recent years, various systems have been developed worldwide to map the quantity and frequency of antibiotic use in veterinary medicine. In Germany, one of these monitoring systems was developed within the VetCAb (Veterinary Consumption of Antibiotics) project, which was the first project to publish representative data on antibiotic use in German livestock husbandry. In 2007 and 2008, the VetCAb feasibility study was conducted to determine whether representative monitoring of antibiotic consumption in Germany is feasible, which data can be used as a basis and how the data collection can be organized. This study’s results were published by Merle et al. [9, 10]. The VetCAb pilot study followed, in which representative data on antibiotic use were collected throughout Germany for 2011 [11]. Thereafter, the VetCAb project was refunded, and the data have been collected from 2013 to present as the longitudinal VetCAb sentinel study.

Currently, the results from various monitoring systems show substantially reduced antibiotic use in German livestock husbandry [1215], as well as in other European countries (for example MARAN [16], SVARM [17] and DANMAP [18]). Systems in other countries are often based on quantities determined by sales data [19], or the amount prescribed by animal species. In Germany, several monitoring systems present their results based on antibiotic usage frequency, such as the QS monitoring system [20], the nationwide official database implemented by the German Medicinal Products Act (the Federal Office of Consumer Protection and Food Safety; [21]) and the VetCAb project.

To analyze the temporal trend in more detail, the present work focused on the change in antibiotic usage over time in German livestock husbandry, investigating the years 2011, 2013 and 2014. Since some publications have shown that various factors influence both disease incidence [2224] and antibiotic use magnitude [25], the question arises, weather specific factors associated with antibiotic use can be identified. To develop strategies for further reducing antimicrobial usage and to correct the results of the antimicrobial use frequency for other effects, these factors must be analyzed.

Thus, in the present work, the influences of temporal trends, farm size, farm category, region and the veterinarian supervising the farms on the treatment frequency are considered via regression models.

Materials and methods

Study design

The VetCAb sentinel study is a longitudinal extension of the cross-sectional VetCAb pilot study [11] and data for 2011, 2013 and 2014 from Germany are presented. The study population is an open cohort with ongoing recruitment of farms and veterinarians to compensate for possible withdrawals and stabilize the study size and representativeness over time. Therefore, the study population consists of farms that previously participated in 2011, and farms that were recruited later.

Every participant (farmer or veterinarian) provided information on antimicrobial substance application and/or delivery, including information on the delivery/application date, number of animals treated, name and amount of the antibiotic drug used, the medical indication and the treatment duration. Because documenting this information is mandatory for farmers and veterinarians in Germany within a five-year liability period, this information can be obtained retrospectively from the application and delivery forms (ADFs) in farming practice. Data were exported from the veterinarian’s software system, the database of the company "QS Qualität und Sicherheit GmbH", who run a private monitoring system in Germany, and from manual data entry within the study system. Furthermore, farmers provided the number of livestock places for every production type kept.

As in the pilot study, all participants provided consent to use their data for this study, given that all personal data on the pig holding farmers and veterinarians were pseudonymized, as per the privacy statement given to each participant.

All data are transferred manually or imported into a database. Interfaces to several official and private computer systems containing antibiotic use data are integrated in the system. The database is based on the open-source relational database system, MySQL and was designed exclusively for this project. Plausibility checks for completeness and pharmacological plausibility are performed during data input and obvious incorrect entries are intercepted. Subsequently, the dataset is also checked in detail regarding different focuses: Farms without antibiotic use and farms with unusually high treatment frequencies (TFs) are checked for correct information. Missing or incorrect information is revised and, if obtainable, corrected or excluded from the analyses.

In this survey, four production type groups are considered: sows (averaging 200 kg), suckling piglets (averaging 4 kg), weaners (averaging 15 kg) and fattening pigs (averaging 50 kg). The average weights were used per the pilot study [11]. Group allocation is based on the ADF sheet category. Each participating farm could hold one or more production type. Each production type group held on a single farm is defined as a holding in the analysis.

Measuring antibiotic usage

The evaluation focuses on calculating the number of drug applications (treatment units; number of used daily doses; nUDD): as well as estimating the average number of treatments per animal (treatment frequency; TF): [2629]. The measurements were calculated for all applications in a holding within a six-month period (half-year).

This calculation is consistent with those used by other authors [30, 31]; however, as demonstrated in Schaekel et al. [29], most antibiotic consumption calculations use an average body weight and DDDA to calculate the nUDD. Here, ADFs include this information directly.

In this analysis, the number of livestock places is used as the population at risk to calculate the TF, i.e., the population size refers to the possibility of keeping animals rather than the number of animals kept [11, 26]. As livestock places for piglets are not observed directly, the number of livestock places for sows is multiplied by 10.25, the average number of piglets per litter in Germany per Frisch et al. [32].

Statistical analysis

To study the association of various factors with antibiotic use, a multiple linear regression model is performed separately for each defined production type group separately using TF as the outcome. For this purpose, a right-trimmed data set was used to guarantee robust model estimators, where the top 1% of TFs are excluded [33].

The explanatory variable set in each model contains "Time", "Farm size", "Farm category" and "Region" as fixed factors. Due to the hierarchical structure, the variable "Veterinarian" is included in the model as a random factor. To describe "farm size", holdings are categorized into three groups by means of the 33%- and 67%-percentile of the number of livestock places per holding for 2011. The factor "Farm category" is based on the type of production type groups held per farm. Category "Breeding" comprises holdings with sows and piglets only, "Fattening" comprises holdings with weaners and/or fattening pigs only, and the category "Combined" holdings comprises weaners and/or fattening pigs combined with sows and piglets. The category "Changer" represents the holdings that stops keeping one group, kept an additional group or changes the production type group kept over time during the observational period (2011, 2013 and 2014). For the factor, "Region", the examined animal husbandry collective is divided into geographical areas based on agricultural structures in Germany [34].

Antibiotic use is measured twice yearly, resulting in up to six repeated measurements per holding. Mixed models are used for the analyses to account for the hierarchical data structure [35, 36]. Because of the non-equidistant time points, a flexible correlational structure between measurements of one holding is chosen. In addition, a compound symmetry structure is used to model the random veterinarian effect. Since the TF is not normally distributed, various regression models were adapted, and a model is selected based on the distribution of residuals. Due to its non-negative right-skewed distribution, we compare the following models: a negative binomial regression model and mixed models with three different transformations of TF (square root transformation, logarithm transformation after adding 0.1, and logarithm transformation after adding 1). For each production type group, different decisions are made within the model selection process. The distributions of the residuals of multi-factorial models with different transformations for the treatment frequency are provided as supplementary data (S1S12 Figs).

The analyses were performed using the procedures GLIMMIX and MIXED in SAS, version 9.3, TS level 1M2 (SAS Institute Inc., Cary, NC, United States). To describe the influencing factors’ effects on the TF, the back-transformed estimators of the least-squares means with their associated 95% confidence intervals are considered for each category of factors. F-tests are used to assess the statistical significance of the fixed effects, considering p-values below 5% as statistically significant. The impact of the veterinarian random effect is analyzed using a likelihood ratio chi-square test comparing the full model with the reduced model, thus omitting the hierarchical level.

Results

Description

Study population.

In total, 51 311 ADFs from participating pig farms were analyzed. For weaners and fattening pigs, the median was 4 antibiotic substance prescriptions per holding biannually (IQR = 1–13 prescriptions per holding for weaners and IQR = 1–10 prescriptions per holding for fattening pigs), while for sows and piglets the median were 6 and 7 prescriptions per holding (IQR = 2–14 prescriptions per holding for sows and IQR = 2–16 prescriptions per holding for piglets), respectively.

The number of participating holdings within the analyses increased over time. In the first half of 2011, 755 holdings were integrated into the evaluation, and 1 102 holdings were analyzed in the second half of 2014. The population composition changed only slightly relative to "Farm size", "Farm category" and "Region". Sow, piglet and weaner holdings comprised approximately 20% of these holdings each, while fattening pig holdings comprised approximately 40%. The increased number of evaluable fattening pig holdings per half-year was slightly more than that of other groups (see Table 1).

thumbnail
Table 1. Biannual treatment frequency distribution for sows, piglets, weaners and fattening pigs.

https://doi.org/10.1371/journal.pone.0199592.t001

Antibiotic usage and treatment frequency.

To describe the data basis of further statistical analyses, Table 1 shows the distribution of the biannual TF for the four production type groups. Consequently, following statements refer to a mere description of the data. In general, the median semi-annual TF constantly decreased from the first half of 2011 until the second half of 2014. This reduction is most noticeable in the fattening pigs, where the median TF decreased from 5.1 (in the first half of 2011) to 0.7 (in the second half of 2014) and furthermore in the piglets, where the median semi-annual TF decreased from 3.8 in the first half of 2011 to 1.7 in the second half of 2014. This is a TF reduction of more than 55% in piglets and 86% in fattening pigs within four years. The descriptive results of sow and weaner TF deviated. The median of both groups varied only slightly over time. For sows, the median TF remained between 1.0 and 1.2. For weaners, the values vary at higher levels between 8.2 and 12.2.

Fig 1 illustrates descriptive results about the change over time in the proportions of participants without antibiotic usage among the groups. The most obvious change occurred in the piglets. From less than 10% of holdings without antibiotic usage, their proportion increased constantly up to 21.5% in the second half of 2014. In the remaining groups, proportions varied: in 2013 and 2014, weaners and fattening pigs showed higher percentages of holdings without antibiotic use than in 2011. However, in 2013 and 2014 values varied between 18% and 26% for weaners and between 19% and 23% for fattening pigs. The proportion of sow holdings without antibiotic use always remained under 15%.

thumbnail
Fig 1. Percentage and trend line for participating farms with no antibiotic use by half-year for each production type group.

https://doi.org/10.1371/journal.pone.0199592.g001

Regression models

To improve the statistical significance of the data Tables 25 show the regression model results for each production type group. The estimates of fixed effects regression coefficients and random effects covariance parameters are provided in the supplementary material (S1 and S2 Tables). Within the model selection process, for groups with lower TFs, the highest goodness of fit was seen using the logarithm transformation after adding 0.1 (S1S12 Figs). This occurred in the sow data with medians between 1.0 and 1.2 (Table 1). For distributions with a slightly higher average antibiotic use, the best model fit is obtained using the square root transformation. This applied to the distribution of the residuals of the TF in piglets (median from 1.7 to 4.7), weaners (median from 8.2 to 12.2) and fattening pigs (median from 0.7 to 5.3).

thumbnail
Table 2. Multi-factorial model results with logarithm transformation for the treatment frequency in sows.

https://doi.org/10.1371/journal.pone.0199592.t002

thumbnail
Table 3. Multi-factorial model results with square root transformation for the treatment frequency in piglets.

https://doi.org/10.1371/journal.pone.0199592.t003

thumbnail
Table 4. Multi-factorial model results with square root transformation for the treatment frequency in weaners.

https://doi.org/10.1371/journal.pone.0199592.t004

thumbnail
Table 5. Multi-factorial model results with square root transformation for the treatment frequency in fattening pigs.

https://doi.org/10.1371/journal.pone.0199592.t005

During the observational period, significant changes occurred in the TF among all groups; however, for sows, the mean TF values estimated via the model remained constant between 0.7 and 1.1. For weaners, the mean values are estimated between 14.7 and 23.1. Although the p-value shows a global statistical significance, no general trend is observed. In contrast, the estimated means are significantly reduced in the piglets and fattening pigs.

"Farm size" has a significant impact on the TF in all models. The estimated TF means increased with increasing size. For all groups in the category, "Upper third", the estimated mean is at least twice as high as for the "Lower third".

The factor "Region" has no significant impact on any group’s TF, except in the weaner model, since the category "East" has a nearly four-fold higher estimated mean (39.0) compared with the "Northwest" and "Middle" regions (11.4).

The factor "Farm category" significantly affects the TF in piglets, weaners and fattening pigs, while the sow model is unaffected. In piglets, the model estimated similar means in the categories, "Breeding" and "Combined", at 2.6 and 2.5. The category "Changer" has a significantly higher estimated mean of 4.2. The estimated means of the weaners and fattening pigs in the category "Fattening" are significantly higher (25.8 and 4.6) than in the categories "Combined" (13.7 and 3.1) and "Changer" (17.7 and 33.1).

In investigating the effect of veterinarians, values below 0.05 were observed for all p-values in the likelihood ratio test for each group model. This indicates that models addressing the veterinarian clusters fit the data better than models omitting the random veterinarian effect.

A sensitivity analysis to investigate the results’ stability was performed by omitting the cases where one veterinarian ID is connected with only one farm ID (data not shown). The analysis showed no changes in interpreting of the p-values that assessed the fixed effects and the veterinarian random effect for all group models.

Discussion

This longitudinal study used data from mandatory ADF documentation in Germany. Veterinarians and farmers participated voluntarily in a panel study and provided data from 2011, 2013, and 2014. Based on the calculated semi-annual TF for each pig holding, the influence of various factors on antibiotic consumption on pig farms was analyzed, using regression models. This study’s main achievement is observing a cohort of pig holdings over several years.

This study is a longitudinal extension of the cross-sectional VetCAb project. Van Rennings et al. [11] showed the successful implementation of a monitoring system for antibiotic use by presenting the first results representative for German pig holdings. To transfer this knowledge, the study design concerning data collection and analysis as well as the general measurement of antibiotic usage was adopted. Furthermore, recruitment focused on maintaining the pilot collective to minimize a possible selection and migration bias.

The data presented here are based on voluntary participation, thus carrying the risk of selection bias, or migration bias in longitudinal studies. The true antimicrobial use may have been higher and the reduction less pronounced than that in the presented data. However, the TF range results suggest that both selection bias [11] and migration bias are unlikely.

Additionally, a bias due to ADF misallocation must be discussed. Using ADFs as data sources risks assigning the ADFs and antibiotic treatments to the incorrect production type groups. The group designations were not standardized, and were often used in colloquial speech for the ADFs, resulting in the allocations being ambiguous. Specifically, the term "weaner" has various meanings in veterinary and farming practice in Germany. Therefore, an exalted misallocation frequency is possible; however, we assume that veterinarians will define production types homogeneously by adhering to the Medicinal Products Act, so that this effect will decrease over the years.

In this study, antibiotic use is measured by the treatment frequency calculated based on the nUDD. Most other studies on antibiotic consumption in veterinary medicine are based on sales data (viz. amounts). Merle et al. [9], van Rennings et al. [27] and Schaekel et al. [29] described the differences between UDD and DDDA in detail. Here, documenting data for calculating used daily doses (UDD) offers two practical advantages. First, in contrast to the theoretical DDDA (defined daily dose for animals), which estimates the use based on the sales data or amounts applied, nUDD gives direct insight into the antibiotic consumption on site. The key differences are that the TF (nUDD) provides information on the actual number of animals treated, and, if the total amount of antimicrobials used is recorded, allows assessing the actual dose of the active ingredient used. If DDDA is applied, standard dosages and standard animal weights are used instead, which will thus provide inaccurate information when animals are treated during different live periods. Consequently, the results are observed with additional variability and therefore may lead to diluted effects when comparing TFs. Thus, calculations based on UDD, especially for running risk factor models, may achieve less biased results. Because the number of treated animals and treatment days are documented on the ADFs in Germany, the nUDD can be calculated directly, without the quantity of active substances or standardized animal weights, so that these possible sources of error remain accurate. Second, the production type of the treated animal (group) is usually indicated on the ADFs in Germany; therefore, using the ADFs facilitates associating the antibiotic consumption with the group treated.

Although TF is similar to amount-based calculations for antibiotic use, its comparison to other studies is restricted due to systematically different approaches in defining standard weights and DDDA. In addition, this comparison is hindered by varying group definitions. In the present VetCAb study, the observed animal holdings are divided into the groups piglet, weaner, fattening pig and sow. In other surveys, production type groups were summarized, such as those of Bos et al. [37] and Jensen et al. [38], who analyzed piglets and sows together in one group. Conversely, some document no group separation at all, such as in Hosoi et al. [39]. Even if production types are separated into groups, the definitions of these groups may differ, such as in the surveys of Callens et al. [40], Trauffler et al. [28] and Sjölund et al. [41]. Hence, it is crucial to differentiate antibiotic use data and standardize definitions when comparing data between studies, regions, and countries.

Overall, internationally and even nationally, no harmonized approach exists for assessing antibiotic consumption; therefore, increased attention should be paid to standardizing definitions and calculation methods in antibiotic monitoring to compare antibiotic use.

Despite the many differences described above, temporal trends within the systems can be compared. In this study, sufficient data were available to assess the temporal development of antibiotic usage in all animal groups. While the median TF for sows remained unchanged and constantly low over the study period, piglets and fattening pigs showed remarkably reduced antimicrobial consumption. In contrast, antibiotic use in weaners fluctuated significantly, which requires further investigation.

In general, the marked differences in the TFs between groups may be explained by animals’ different exposures to infectious hazards. Since fattening animals (piglets, weaners and fattening pigs) receive the most antibiotics in pig farming, this area should be the focus for reduction. The fluctuating TF medians in the weaner collective were striking, which could be explained by the possible misallocation described above.

In total, a statistically significant reduction in antibiotic use over time is evident in this study. This is consistent with reported sales data in Germany, where since 2011, a reduction by 468 tons to approximately 1 238 tons was observed in 2014 [12]. This trend appears to continue in the following years. In the official monitoring system, the median and the upper quartile of the TF decreased constantly. For weaners (up to 30 kg bodyweight (BW)) and fattening pigs (from 30 kg BW), the medians decreased from 4.8 and 1.2, respectively, in the second half of 2014 to 3.4 and 0.4, respectively, in the first half of 2016 [14, 15]. In the QS system, the therapy index medians decreased from 10.71 (weaners, up to 30 kg BW) and 1.76 (fattening pigs, from 30 kg BW), respectively, in the second half of 2014 to 3.53 and 0.37, respectively, in the first half of 2016 [13].

The reduced antibiotic usage can also be seen in other European countries, as shown in Sjölund et al.’s study [42]. The sixth European Surveillance of Veterinary Antimicrobial Consumption (ESVAC) report [19] showed lower sales of veterinary antimicrobial agents for food-producing species per population corrected unit (in mg/PCU) in 2014 for all animals in countries including Belgium, the Netherlands, Sweden and Germany compared to former years. Austria, Denmark, Poland and the UK also reduced their veterinary antimicrobial agent sales for food-producing species per population corrected unit, but in these countries, the reductions began later (in 2012 and 2013, respectively) [19]. In Denmark, the total amounts of antibiotic substances as well as the DAPD (DDDA per 1 000 animals per day), increased from 2011 to 2013 in all groups, until the DAPD decreased in 2014 [43]. However, antibiotic use in Denmark was significantly reduced a few years prior; compared with 2009, the DAPD decreased in 2011 in all groups. This was related to the yellow card initiative, a benchmarking system for pig farms established by the Danish Veterinary and Food Administration (DVFA) in 2010 [44]. In the Netherlands, the reduction in antimicrobial use began in 2007 and has continued since [16].

In summary, nearly all countries with antibiotic use monitoring systems in livestock husbandry document success in reducing antibiotic consumption for many reasons. One of the main reasons is that awareness in both society and the agricultural sector has improved. This increase developed due to implementing monitoring systems, publishing scientific studies from across Europe (e.g., [28, 37, 4550]) and various measures taken by international and national policy makers [19, 5153].

International surveys show that antibiotic use in animal husbandry cannot be reduced indefinitely. In 2015, the National Veterinary Institute of Sweden (SVA) published the yearly Swedish Veterinary Antibiotic Resistance Monitoring (SVARM) results, which showed that the temporal trend of veterinary antimicrobial consumption in Sweden has reduced over several years. In contrast, antibiotic sales for pigs remained stagnate over the past five years. However, another change occurred instead. Product sales for individual medications increased, while sales of group medications decreased [17]; thus, it appears that more individual and less metaphylactic treatments have occurred.

Diseases requiring antibiotic treatment can occur in any animal holding type. For this reason, the aim cannot be to ban antibiotic uses, but to change its application. Responsible antibiotic handling does not mean non-use, but prudent use. It must be assumed that in the future, a plateau will be reached, and the reduction will stagnate at the necessary level. Further studies are needed to assess the antimicrobial level needed to treat diseased animals without conflicting with animal welfare through the legal compulsion for further reductions.

Furthermore, all measures must be evaluated regularly and in detail to adjust the monitoring and benchmarking system if necessary. As a first step, multiple linear regression for each defined production type group was performed to investigate weather monitoring antibiotic use should be accompanied by evaluating secondary data.

In this study, the estimated means of TF rose with increasing farm size in all production type groups. In contrast, Vieira et al. [54] identified higher TF on small farms and mentioned worse hygiene management as a possible reason. Van Rennings et al. [11] also obtained different results. The authors determined that farm size did not significantly impact the TF, except for a slight influence in the weaner model. Considering that part of van Rennings et al.’s study population [11] was also analyzed in this study, this influence appears to have changed over time, as the extended study population and more detailed model may have influenced the recent study. Van der Fels-Klerx et al. [25] noted an influence of farm size on antibiotic use in their study population in 2011. They suggested that a rising probability of infection with an increased number of animals could be responsible for this and refer to Österberg et al. [22], Hautekiet et al. [23] und García-Feliz et al. [24]. Regarding the link between herd size and antimicrobial resistance, various patterns are also described [55].

In our study, spatial factors did not significantly impact the TF, except for in weaners. This result is consistent with those of van Rennings et al. [11]. Although a high density of pig holdings prevails in some regions, and the presumption suggests higher infection pressure due to the proximity of neighboring stables [56], high farm density and high antibiotic use were not significantly associated in this study. Even in the weaner model, a statistically significant higher TF was determined in the "East" region, which comprises few but large farms [34].

The factor "Farm category" has a significant effect on the TF in piglets, weaners and fattening pigs, while the sow model was unaffected. Fattening pigs show similar results to those of van der Fels-Klerx et al. [25]; on specialized fattening farms, significantly more antibiotics were used than on combined farms / farrow-to-finish farms. This finding corresponds to the assumption that increased animal movement and pooling of animals from different stables, and thus of different farm-specific germ spectra, results in an increased risk of infection. These results concur with those published by Casal et al. [57] and Moreno [58]. The same applies to the weaners. In sow holdings, the farm category has no significant influence on the TF, whereas in piglet holdings, a significant impact could be seen, but neither "Combined" nor specialized "Breeding" farms used considerably more antibiotics. (In the piglet group, the significance of the factor seems to be due to the high-level antibiotic usage in the "Changers" group. In this group, production system conversion leading to unstable production conditions may have influenced the disease incidence, leading to increased antibiotic use.) These findings support the above hypothesis; sows and piglets were not moved to other stables, regardless of farm category. The increased infection pressure is only apparent after the first regrouping. Van der Fels-Klerx et al. [25] has received other results in this regard; specialized sow farms used fewer antibiotics for piglets than did farrow-to-finish farms. They hypothesized that this result may be due to the term "piglet" is being used longer on farrow-to-finish farms than on specialized farms.

According to van Rennings et al. [11], the influence of veterinarian is taken into account within the adjusted analyses and has again a significant effect on the TF. This could be explained by the veterinarian’s different specializations or typical antimicrobial treatment routines (e.g., using combined products).

Overall, the factors, "Farm size", "Veterinarian" and "Farm category", appears to impact the TF, while the "Region" factor shows no effect when the data are adjusted for confounding. However, it has to be assumed that these three identified factors are not solely influential; thus, to improve existing systems for antibiotic monitoring or establish new intervention measures, additional studies are needed. Evaluating the secondary data may greatly benefit this.

Conclusions

In the longitudinal VetCAb sentinel study, antibiotic administration decreased significantly over time in the study population. The factors, "Farm size", "Veterinarian" and "Farm category", appeared to impact the treatment frequency. Considering these effects, these factors should complement antibiotic use monitoring.

Supporting information

S1 Fig. Distribution of the residuals of the multi-factorial model with logarithm transformation after adding 0.1 for the treatment frequency in piglets.

https://doi.org/10.1371/journal.pone.0199592.s001

(TIF)

S2 Fig. Distribution of the residuals of the multi-factorial model with logarithm transformation after adding 1 for the treatment frequency in piglets.

https://doi.org/10.1371/journal.pone.0199592.s002

(TIF)

S3 Fig. Distribution of the residuals of the multi-factorial model with square root transformation for the treatment frequency in piglets.

https://doi.org/10.1371/journal.pone.0199592.s003

(TIF)

S4 Fig. Distribution of the residuals of the multi-factorial model with logarithm transformation after adding 0.1 for the treatment frequency in sows.

https://doi.org/10.1371/journal.pone.0199592.s004

(TIF)

S5 Fig. Distribution of the residuals of the multi-factorial model with logarithm transformation after adding 1 for the treatment frequency in sows.

https://doi.org/10.1371/journal.pone.0199592.s005

(TIF)

S6 Fig. Distribution of the residuals of the multi-factorial model with square root transformation for the treatment frequency in sows.

https://doi.org/10.1371/journal.pone.0199592.s006

(TIF)

S7 Fig. Distribution of the residuals of the multi-factorial model with logarithm transformation after adding 0.1 for the treatment frequency in weaners.

https://doi.org/10.1371/journal.pone.0199592.s007

(TIF)

S8 Fig. Distribution of the residuals of the multi-factorial model with logarithm transformation after adding 1 for the treatment frequency in weaners.

https://doi.org/10.1371/journal.pone.0199592.s008

(TIF)

S9 Fig. Distribution of the residuals of the multi-factorial model with square root transformation for the treatment frequency in weaners.

https://doi.org/10.1371/journal.pone.0199592.s009

(TIF)

S10 Fig. Distribution of the residuals of the multi-factorial model with logarithm transformation after adding 0.1 for the treatment frequency in fattening pigs.

https://doi.org/10.1371/journal.pone.0199592.s010

(TIF)

S11 Fig. Distribution of the residuals of the multi-factorial model with logarithm transformation after adding 1 for the treatment frequency in fattening pigs.

https://doi.org/10.1371/journal.pone.0199592.s011

(TIF)

S12 Fig. Distribution of the residuals of the multi-factorial model with square root transformation for the treatment frequency in fattening pigs.

https://doi.org/10.1371/journal.pone.0199592.s012

(TIF)

S1 Table. Estimates of regression coefficients for fixed effects in the multi-factorial models per production type.

https://doi.org/10.1371/journal.pone.0199592.s013

(DOCX)

S2 Table. Estimates of covariance parameters for random effects in the multi-factorial models per production type.

The coloured lines show the estimated variances for observations within one veterinarian or within one of the six time-points.

https://doi.org/10.1371/journal.pone.0199592.s014

(DOCX)

Acknowledgments

The VetCAb project is commissioned and funded by the Federal Institute for Risk Assessment (BfR). We would like to thank the colleagues of the Institute of Pharmacology, Pharmacy and Toxicology of the Faculty of Veterinary Medicine of the University of Leipzig, who supported the part of the study conducted in 2011.

References

  1. 1. The Copenhagen Recommendations. The Copenhagen Recommendations—Report from the Invitational EU Conference on The Microbial Threat. Copenhagen, Denmark. 1998. 45 p.
  2. 2. Stege H, Bager F, Jacobsen E, Thougaard A. VETSTAT—the Danish system for surveillance of the veterinary use of drugs for production animals. Prev Vet Med 2003; 57: 105–115. pmid:12581594
  3. 3. Frimodt-Møller N. Microbial Threat–the Copenhagen Recommendations Initiative of the EU. J Vet Med 2004; 51: 400–402. pmid:15525373
  4. 4. WHO. Global Principles for the Containment of Antimicrobial Resistance in Animals for Food. Geneva, Switzerland: WHO. 2000. 27 p.
  5. 5. EMA (European Medicines Agency) and EFSA (European Food Safety Authority). EMA and EFSA Joint Scientific Opinion on measures to reduce the need to use antimicrobial agents in animal husbandry in the European Union, and the resulting impacts on food safety (RONAFA). EFSA Journal 2017; 15: e04666–n/a.
  6. 6. WHO. Global action plan on antimicrobial resistance. 2015. 28 p.
  7. 7. FAO. The FAO Action Plan on Antimicrobial Resistance 2016–2020. Rome. 2016. 25 p. 978-92-5-109392-4.
  8. 8. OIE. The OIE Strategy on Antimicrobial Resistance and the Prudent Use of Antimicobials. 2016. 12 p.
  9. 9. Merle R, Robanus M, Hegger-Gravenhorst C, Mollenhauer Y, Hajek P, Kasbohrer A, et al. Feasibility study of veterinary antibiotic consumption in Germany—comparison of ADDs and UDDs by animal production type, antimicrobial class and indication. BMC Vet Res 2014; 10: 7. pmid:24401194
  10. 10. Merle R, Hajek P, Käsbohrer A, Hegger-Gravenhorst C, Mollenhauer Y, Robanus M, et al. Monitoring of antibiotic consumption in livestock: A German feasibility study. Prev Vet Med 2012; 104: 34–43. pmid:22115924
  11. 11. van Rennings L, von Münchhausen C, Ottilie H, Hartmann M, Merle R, Honscha W, et al. Cross-Sectional Study on Antibiotic Usage in Pigs in Germany. PLoS One 2015; 10: e0119114. pmid:25785688
  12. 12. GERMAP. GERMAP 2015—Bericht über den Antibiotikaverbrauch und die Verbreitung von Antibiotikaresistenzen in der Human- und Veterinärmedizin in Deutschland. Rheinbach: Bundesamt für Verbraucherschutz und Lebensmittelsicherheit, Paul-Ehrlich-Gesellschaft für Chemotherapie e.V. 2016. 196 p.
  13. 13. QS Qualität und Sicherheit GmbH. Übersicht: Kennzahlen QS-Antibiotikamonitoring und HIT-Datenbank (Stand: November 2016). 2016. Available from: https://www.q-s.de/qs-system/monitoringprogramme-antibiotikamonitoring.html.
  14. 14. BVL. Bekanntmachung des Medians und des dritten Quartils der vom 1. Juli bis 31. Dezember 2014 erfassten bundesweiten betrieblichen Therapiehäufigkeiten für Mastrinder, Mastschweine, Masthühner und Mastputen nach § 58c Absatz 4des Arzneimittelgesetzes. 2015. Available from: https://www.bvl.bund.de/SharedDocs/Downloads/05_Tierarzneimittel/bekanntmachungen/2015_03_31_Bekanntmachung_BAnz.pdf?__blob=publicationFile&v=4.
  15. 15. BVL. Bekanntmachung des Medians und des dritten Quartils der vom 1. Januar 2016 bis 30. Juni 2016 erfassten bundesweiten betrieblichen Therapiehäufigkeiten für Mastrinder, Mastschweine, Masthühner und Mastputen nach § 58c Absatz 4des Arzneimittelgesetzes. 2016. Available from: https://www.bvl.bund.de/SharedDocs/Downloads/05_Tierarzneimittel/Fachmeldungen/BAnz_Therapiehaufigkeit_30092015.pdf?__blob=publicationFile&v=5.
  16. 16. MARAN. MARAN: Monitoring of antimicrobial resistance and antibiotic usage in animals in the Netherlands in 2014. Lelystad, Netherlands: Central Veterinary Institute, part of Wageningen University and research Centre (CVI). 2015. 72 p.
  17. 17. SVARM. SWEDRES/SVARM 2015—Consumption of antibiotics and occurence of antibiotic resistence in Sweden. Solna and Uppsala, Sweden: Public Health Agency of Sweden and National Veterinary Institute. 2015. 123 p.
  18. 18. DANMAP. DANMAP 2015—Use of antimicrobial agents and occurrence of antimicrobial resistance in bacteria from food animals, food and humans in Denmark. Denmark: Microbiology and Infection Control, Statens Serum Institute; National Food Institute, Technical University of Denmark. 2016. 144 p. ISSN 1600-2032.
  19. 19. European Medicines Agency ESoVAC. Sales of veterinary antimicrobial agents in 29 European countries in 2014. Trend from 2011 to 2014. 2016. 176 p.
  20. 20. QS Qualität und Sicherheit GmbH. Leitfaden Antibiotikamonitoring Schwein. 2015. Available from: https://www.q-s.de/dokumentencenter/dc-antibiotikamonitoring-schwein.html.
  21. 21. Anonym (2014) Verordnung über die Durchführung von Mitteilungen nach §§ 58a und 58b des Arzneimittelgesetzes (TAMMitDurchfV).
  22. 22. Österberg J, Vågsholm I, Boqvist S, Lewerin SS. Feed-borne Outbreak of Salmonella Cubana in Swedish Pig Farms: Risk Factors and Factors Affecting the Restriction Period in Infected Farms. Acta Vet Scand 2006; 47: 13–22. pmid:16722302
  23. 23. Hautekiet V, Geert V, Marc V, Rony G. Development of a sanitary risk index for Salmonella seroprevalence in Belgian pig farms. Prev Vet Med 2008; 86: 75–92. pmid:18453017
  24. 24. García-Feliz C, Carvajal A, Collazos JÁ, Rubio P. Herd-level risk factors for faecal shedding of Salmonella enterica in Spanish fattening pigs. Prev Vet Med 2009; 91: 130–136. pmid:19539388
  25. 25. van der Fels-Klerx HJ, Puister-Jansen LF, van Asselt ED, Burgers SLGE. Farm factors associated with the use of antibiotics in pig production. J Anim Sci 2011; 89: 1922–1929. pmid:21606448
  26. 26. Hemme M, Käsbohrer A, von Münchhausen C, Hartmann M, Merle R, Kreienbrock L. Unterschiede in der Berechnung des betriebsbezogenen Antibiotika-Einsatzes in Monitoringsystemen in Deutschland–eine Übersicht. Berl Münch Tierärztl Wschr 2017; 130: 93–101.
  27. 27. van Rennings L, Merle R, von Münchhausen C, Stahl J, Honscha W, Käsbohrer A, et al. Variablen zur Beschreibung des Antibiotikaeinsatzes beim Lebensmittel liefernden Tier. Berl Münch Tierärztl Wschr 2013; 126: 297–309.
  28. 28. Trauffler M, Griesbacher A, Fuchs K, Köfer J. Antimicrobial drug use in Austrian pig farms: plausibility check of electronic on-farm records and estimation of consumption. Vet Rec 2014; 175: 402. pmid:25053268
  29. 29. Schaekel F, May T, Seiler J, Hartmann M, Kreienbrock L. Antibiotic drug usage in pigs in Germany—Are the class profiles changing? PLoS One 2017; 12: e0182661. pmid:28841685
  30. 30. Timmerman T, Dewulf J, Catry B, Feyen B, Opsomer G, Kruif Ad, et al. Quantification and evaluation of antimicrobial drug use in group treatments for fattening pigs in Belgium. Prev Vet Med 2006; 74: 251–263. pmid:16675051
  31. 31. Persoons D, Dewulf J, Smet A, Herman L, Heyndrickx M, Martel A, et al. Antimicrobial use in Belgian broiler production. Prev Vet Med 2012; 105: 320–325. pmid:22459488
  32. 32. Frisch J, Fritzsche S, Fröba N, Funk M, Gaio C, Grimm E, et al. Betriebsplanung Landwirtschaft 2012/13 KTBL-Datensammlung. Darmstadt: Kuratorium für Technik und Bauwesen in der Landwirtschaft 2012.
  33. 33. Huber PJ. Robust Statistics: Wiley; 2004.
  34. 34. Merle R, Busse M, Rechter G, Meer U. Regionalisierung Deutschlands anhand landwirtschaftlicher Strukturdaten. Berl Münch Tierärztl Wschr 2012; 125: 52–59.
  35. 35. Vonesh EF. Generalized linear and nonlinear models for correlated data: theory and applications using SAS: SAS Institute; 2012.
  36. 36. Dohoo I, Martin W, Stryhn H. Veterinary Epidemiologic Research - 2nd Edition; MacPike S, editor. Canada: VER Inc.; 2009.
  37. 37. Bos MEH, Taverne FJ, van Geijlswijk IM, Mouton JW, Mevius DJ, Heederik DJJ, et al. Consumption of Antimicrobials in Pigs, Veal Calves, and Broilers in The Netherlands: Quantitative Results of Nationwide Collection of Data in 2011. PLoS One 2013; 8: e77525. pmid:24204857
  38. 38. Jensen VF, Jacobsen E, Bager F. Veterinary antimicrobial-usage statistics based on standardized measures of dosage. Prev Vet Med 2004; 64: 201–215. pmid:15325773
  39. 39. Hosoi Y, Asai T, Koike R, Tsuyuki M, Sugiura K. Use of veterinary antimicrobial agents from 2005 to 2010 in Japan. Int J Antimicrob Agents 2013; 41: 489–490. pmid:23434535
  40. 40. Callens B, Persoons D, Maes D, Laanen M, Postma M, Boyen F, et al. Prophylactic and metaphylactic antimicrobial use in Belgian fattening pig herds. Prev Vet Med 2012; 106: 53–62. pmid:22494905
  41. 41. Sjölund M, Backhans A, Greko C, Emanuelson U, Lindberg A. Antimicrobial usage in 60 Swedish farrow-to-finish pig herds. Prev Vet Med 2015; 121: 257–264. pmid:26211838
  42. 42. Sjölund M, Postma M, Collineau L, Lösken S, Backhans A, Belloc C, et al. Quantitative and qualitative antimicrobial usage patterns in farrow-to-finish pig herds in Belgium, France, Germany and Sweden. Prev Vet Med 2016; 130: 41–50. pmid:27435645
  43. 43. DANMAP. DANMAP 2014—Use of antimicrobial agents and occurrence of antimicrobial resistance in bacteria from food animals, food and humans in Denmark. Denmark: Statens Serum Institute; National Veterinary Institute, Technical University of Denmark; National Food Institute, Technical University of Denmark. 2015. 112 p. ISSN 1600-2032.
  44. 44. Ministry of Food Agriculture and Fisheries. Special provisions for the reduction of the consumption of antibiotics in pig holdings (the yellow card initiative). 2015. Available from: https://www.foedevarestyrelsen.dk/english/SiteCollectionDocuments/25_PDF_word_filer%20til%20download/Yellow%20Card%20Initiative.pdf.
  45. 45. Aarestrup FM. Association between the consumption of antimicrobial agents in animal husbandry and the occurrence of resistant bacteria among food animals. Int J Antimicrob Agents 1999; 12: 279–285. pmid:10493603
  46. 46. Aarestrup FM. Veterinary Drug Usage and Antimicrobial Resistance in Bacteria of Animal Origin. Basic Clin Pharmacol Toxicol 2005; 96: 271–281. pmid:15755309
  47. 47. Chauvin C, Beloeil P-A, Orand J-P, Sanders P, Madec F. A survey of group-level antibiotic prescriptions in pig production in France. Prev Vet Med 2002; 55: 109–120. pmid:12350315
  48. 48. Grave K, Jensen VF, Odensvik K, Wierup M, Bangen M. Usage of veterinary therapeutic antimicrobials in Denmark, Norway and Sweden following termination of antimicrobial growth promoter use. Prev Vet Med 2006; 75: 123–132. pmid:16580756
  49. 49. Jensen VF, Emborg HD, Aarestrup FM. Indications and patterns of therapeutic use of antimicrobial agents in the Danish pig production from 2002 to 2008. J Vet Pharmacol Ther 2012; 35: 33–46. pmid:21564137
  50. 50. Merle R, Mollenhauer Y, Hajek P, Robanus M, Hegger-Gravenhorst C, Honscha W, et al. Verbrauchsmengenerfassung von Antibiotika beim Schwein in landwirtschaftlichen Betrieben Berl Münch Tierärztl Wschr 2013; 126: 326–332.
  51. 51. DIMDI-AMV (2010) Verordnung über das datenbankgestützte Informationssystem über Arzneimittel des Deutschen Instituts für Medizinische Dokumentation und Information (DIMDI-Arzneimittelverordnung—DIMDI-AMV). In: Bundesministerium für Gesundheit, editor: Bundesministerium für Ernährung, Landwirtschaft und Verbraucherschutz, BGBl. I S. 140. pp. 4.
  52. 52. Anonymous (2013) Sechzehntes Gesetz zur Änderung des Arzneimittelgesetzes, Bundesgesetzblatt Jahrgang 2013 Teil I Nr. 62. Bonn.
  53. 53. DART. DART 2020 Deutsche Antibiotika-Resistenzstrategie. Berlin: Bundesministerium für Gesundheit, Bundesministerium für Ernährung, Landwirtschaft, Bundesministerium für Bildung und Forschung. 2015. 1–32 p.
  54. 54. Vieira A, Pires S, Houe H, Emborg H-D. Trends in slaughter pig produktion and antimicrobial consumption in Danish slaughter pig herds, 2002–2008. Epidemiol Infect 2011; 139: 1601–1609. pmid:21134323
  55. 55. Burow E, Käsbohrer A. Risk Factors for Antimicrobial Resistance in Escherichia coli in Pigs Receiving Oral Antimicrobial Treatment: A Systematic Review. Microb Drug Resist 2016; 23: 194–205. pmid:27249658
  56. 56. Elbers ARW, Stegeman A, Moser H, Ekker HM, Smak JA, Pluimers FH. The classical swine fever epidemic 1997–1998 in the Netherlands: descriptive epidemiology. Prev Vet Med 1999; 42: 157–184. pmid:10619154
  57. 57. Casal J, Mateu E, Mejía W, Martín M. Factors associated with routine mass antimicrobial usage in fattening pig units in a high pig-density area. Vet Res 2007; 38: 481–492. pmid:17425935
  58. 58. Moreno MA. Survey of quantitative antimicrobial consumption in two different pig finishing systems. Vet Rec 2012; 171: 325. pmid:22915683