The use of antimicrobials in global pig production: a systematic review of methods for quantification

BACKGROUND
Overuse of antimicrobials in both humans and animals is recognized as one of the main drivers of Antimicrobial Resistance (AMR); and the optimisation of their use has been advocated as a key strategy for dealing with AMR. The measurement of antimicrobial use is vital for the design, monitoring and evaluation of such strategies. This systematic review describes and compares methods and measurements used to quantify antimicrobial use in pigs in order to inform efforts to standardize measurement.


METHODS
The peer-reviewed literature was systematically searched using four online databases: MEDLINE, ScienceDirect, Scopus and Web of Science. Eligibility criteria for inclusion in the review included: articles published in English, involving pigs of any age and types of production, providing quantitative data on antimicrobial use, containing a clear description of the methodology, and having moderate to high rank in the quality assessment.


RESULTS
Of 2,362 abstracts reviewed, a total of 25 studies were included based on the eligibility criteria. All studies were published between 2001 and 2017. Twenty of the studies were conducted in eight European countries. Twelve studies estimated antimicrobial use and eight studies were primarily methodological papers comparing different methods or variables, or developing new methods. The two main sources of antimicrobial use data were farm surveys and national sales data. A large variety of units of measurement was found. In this review, the ten measurements identified were categorized into four groups: 1) antimicrobials use measured by milligrams of active substance per animal weight; 2) antimicrobials use measured by daily dose per weight at treatment; 3) antimicrobial use measured by daily dose per treatment period; and 4) antimicrobials use measured by daily dose per period at risk of treatment.


CONCLUSION
There is no global standardized measurement of antimicrobial use in pigs. Given the importance of monitoring the use antimicrobials, we recommend that at a minimum, all countries should develop macro-level monitoring using national sales data and report use by milligram of active ingredients per Population Correcting Unit. Monitoring in specific animal species requires the development of systems to capture prescription at national or farm level. Findings from monitoring antimicrobial use may help to guide effective interventions for optimising use of antimicrobials, as recommended by the WHO Global Action Plan on AMR.


Background
Antimicrobial Resistance (AMR) is an increasingly serious threat to global public health. Overuse of antimicrobials can accelerate the emergence of antimicrobial resistance (World Health Organization, 2015b). In livestock industries, large amounts of antimicrobials are used for both therapeutic and non-therapeutic purposes including growth promotion (Aarestrup, 2005). In response to global concerns about AMR, in 2008, the World Organization for Animal Health (OIE) launched guidelines on the prudent use of antimicrobials in veterinary medicines, which describes the respective responsibilities of relevant stakeholders such as veterinarians, regulators, pharmaceutical industries, animal producers and consumers (World Organisation for Animal Health (OIE), 2008).
Measuring antimicrobial use is critical to understanding the magnitude and profile of antimicrobial resistance in countries. Measurement is the first step to detecting whether there is excessive and A C C E P T E D M A N U S C R I P T inappropriate use and monitoring whether policies aimed at optimising use are successful.
Recognising this, international organizations such as FAO, OIE and WHO, have recommended that countries develop systems for monitoring antimicrobial consumption (World Health Organization, 2015a, OIE, 2016, FAO, 2016. The World Health Organization (WHO) guidelines defines antimicrobial "consumption" data captured from aggregate sales data such as form importer, local manufacturer or wholesales, whilst data on antimicrobial "use" are collected from patient-level data such as medical records and prescriptions (World Health Organization, 2017). Whilst there has been significant progress in the monitoring of antimicrobial use and consumption in the human health sector, action in the animal health sector has lagged behind (Schar et al., 2018). Some European countries established national programs for the surveillance of antimicrobial consumption in animals for more than 20 years ago, specifically DANMAP in Denmark in 1995(Statens Serum Institut, 2012, MARAN in Netherlands in 1998(Anonymous, 2012 and SWEDRES-SVARM (SWEDRES and SVARM, 2014). The European Medicines Agency established the European Surveillance of Veterinary Antimicrobial Consumption (ESVAC) project in 2009 (Agency, 2017) . ESVAC compiles, verifies and reports on antimicrobial consumption of veterinary antimicrobial agents in 29 European countries. Data are collected through a network of national focal points. Furthermore, ESVAC has been striving to set up a standardised methodology to allow for cross country comparisons. The monitoring of antimicrobial consumption serves various objectives. It monitors time trends of antimicrobial use, compares use by different antimicrobial classes, identifies high users and promotes more prudent use, and studies the association between level of usage and bacterial resistance (Collineau et al., 2017).
Currently, there is a wide variation in the availability and type of data, methods and use measurement across countries. The lack of uniformity hampers cross-country comparisons (Collineau et al., 2017). In order to guide the strengthening of existing monitoring systems and the development of new ones to facilitate cross-country comparisons, it is essential to understand the different existing methods, their strengths, limitations and operational feasibility. However, in this study, for simplicity the term "use" is applied to refer to both use at farm level and consumption at aggregate national or sub-national level.

Biomass
The weight or total quantity of living organisms of one animal species or of all the species in the community. Using biomass for antimicrobial consumption aims to compare the weight of animals between different species and between human and animals.
This review covers use of antimicrobials in pigs, with the following research question: "What methods and measurements are used to quantify the use of antimicrobials?"

Search strategy
o SPIDER tool A "SPIDER" tool was applied in order to specifically identify relevant quantitative and mixedmethod studies. It covers the Sample, Phenomenon of interest, Design, Evaluation and Research type) (Cooke et al., 2012). o Eligibility assessment of studies and inclusion criteria The following inclusion criteria were considered: (i) the paper was published in, or translated into, the English language, A C C E P T E D M A N U S C R I P T (ii) the study involved pigs of any age and type of production, (iii) the study provided quantitative data on antimicrobial use with a focus or clear explanation of the methodology in pigs or other food producing animals including pigs,

Quality assessment
The methodological quality of the studies was assessed using an instrument adapted from the Critical Appraisal Skills Programme (CASP) (Critical Appraisal Skills Programme (CASP), 2014). The four criteria of quality assessment were a) aim, b) method, c) result and d) application. The answer to the four criteria are either 'yes', or 'no' or 'cannot tell'. Each criterion has certain a number of sub-criteria, there were in total eleven sub-criteria for quality assessment; see Table A2 (annex). If the assessment by the two independent reviewers (AS and VT) was 'no' or 'cannot tell', the score for that question was zero; the score for yes was one. When there were conflicting views, the reviewers discussed and sought consensus. In this review, the studies were ranked by quality criteria. The quality ranking was classified into three groups: High meant >75% of all eleven sub-criteria were met, moderate meats 50-75% were met, weak meant <50% were met.

Data extraction and synthesis
The full text of all relevant articles was reviewed and summarised using a standardised data extraction

Search processes
The search from the four database and hand search identified 2,362 articles. After screening and removal of duplications, 90 manuscripts remained for further screening. Of these 90 manuscripts, 37 manuscripts were selected on the basis of the inclusion criteria. Of these 37 manuscripts, seven articles described antimicrobials without essential information on the pattern or volume of antimicrobials use; these were excluded. Two articles were not included, because they were review articles. Another three articles were excluded as they only focused on the association between specific groups of antimicrobial and AMR. No studies were excluded due to low rank of quality assessment (<50%). In summary, a total of 12 studies were excluded from the set of 37 studies, leaving 25 manuscripts that met the inclusion criteria and were included in this systematic review.
< Figure 1> Description of the studies

<Table 2>
The quality assessment is reported in table A2 of the Annex. In general, the hypotheses and the objectives of the study were clearly described. Fifteen (60%) studies were ranked as high quality (meeting more than 75% of all eleven sub-criteria). Ten remaining studies were of moderate quality.
None had low quality assessment.

Methods for measuring antimicrobial use
A large variation in terms of the methodological approaches and units of measurement of antimicrobial use was found.

Types of studies and data sources
As shown in table 2, eight studies were primarily methodological, for example comparing antimicrobial use by using different methods or variables (Carmo et al., 2017, Dupont et al., 2016, Bondt et al., 2013 or developing new methodologies . Twelve studies aimed to estimate antimicrobial use , Mitema et al., 2001b al., 2014, Eagar et al., 2012. One study examined both improving the national surveillance and measuring the antimicrobial use (Filippitzi et al., 2014). The remainder of studies assessed the association between the use of antimicrobials and farm management practice .
Data on antimicrobial use were collected from various sources. Of 25 studies, seven collected data through farm surveys , six compiled national data from the surveillance of antimicrobial consumption (Carmo et al., 2017, Dupont et al., 2016, Bondt et al., 2013, four collected data through veterinary prescriptions , and four from a review of sales of pharmaceutical products (Carmo et al., 2017, Eagar et al., 2012, Mitema et al., 2001b.
Three studies drew information from more than one data source  and one study used data on food animal antimicrobial utilization from the US, estimating the quantity of antimicrobials used in China .
Twenty-two studies (88%) reported antimicrobial use by major classes, while three studies (12%) reported in aggregation all classes of antimicrobial . Twenty studies (80%) reported the use of antimicrobials specific to pigs or other animal species but five studies (20%) only reported total use in all animal species , Eagar et al., 2012, Mitema et al., 2001a.

Numerators: the amount of antimicrobial use
Measuring numerators varied greatly, for example, by milligrams or kilograms of active ingredient and other more sophisticated adjustments such as defined daily dose, daily product dose, animal daily dose, used daily dose, prescribed daily dose, (see detail in Figure 2).

Denominators: the number or mass of animals
For denominator data, eight studies used national level animal population which was retrieved from government agencies such as National Statistics, Central registry for livestock (Carmo et al., 2017, Dupont et al., 2016, Bondt et al., 2013. Two studies applied data from the Food and Agriculture Organization (FAOSTAT) . For the twelve studies at farm level, the number of animals reported by a certain production type and the time period during the study period .

Unit of measurement: indicators used
Of the total 25 studies, there were ten different units of measurement. Nine studies calculated the total volume of antimicrobials used in the country per year (Carmo et al., 2017, Eagar et al., 2012, Mitema et al., 2001b. Five of these studies , Merle et al., 2012, Mitema et al., 2001b, Carmo et al., 2017  ,  used treatment data at the farms. Only one study attempted to estimate non-therapeutic antimicrobial use in livestock. This was done by multiplying the number of animals in different phases of production by the estimated feed consumed per day and the duration in days in each phase that the swine received antimicrobials through feed and doses of antimicrobials in the feed .

Antimicrobials use measured by milligrams of active substance per animal weight
Six studies used some measure of the biomass of animals in order to indicate the intensity of antimicrobial use (Carmo et al., 2017. Biomass is the total weight of live animals. Two studies ) calculated biomass at farms by multiplying the number of animals and the average weight. One study estimated biomass by using the carcass weight, which is the whole-body weight of a slaughtered animal after blood is drained, evisceration and skinning .
Biomass can be calculated by using a population correction units (PCU). The PCU provides a better measurement of animal weight exposed to antimicrobial treatment: one PCU is equivalent to one kilogram of biomass of live animal or slaughtered animals where the animal had been exposed to antimicrobials throughout their lifecycle. For example, gross weight at slaughter was 150 kilograms, but the PCU was 65 kilograms and 25 kilograms for slaughtered and fattening pigs (Agency, 2013).
Two studies (Carmo et al., 2017 calculated the total national PCU, with reference to the guidelines produced by ESVAC, by multiplying the numbers of livestock animals and slaughtered animals by the theoretical weight at the time they were exposed to antimicrobial treatment. Another study estimated the PCU by multiplying the numbers of live animals in a production period and a ratio of carcass weight to live weight of animals .

Antimicrobials use measured by daily dose per weight at treatment
The daily dosage is a measure of the amount of a specific active pharmaceutical ingredient (e.g. in milligrams) required to treat one kilogram of animal in one day with that antimicrobial preparation, and is based on the average dosage of a medicine per kilogram per day for a specific type of animal.
Defined Daily Dose (DDD) is a technical unit of measurement of antimicrobial consumption in humans, calculated by standard DDD-value. In animals, measuring antimicrobial by defined daily dosage is calculated by using a specified dose of medicine (Animal Daily Dose value (ADD-value)), so called Animal Daily Dose (ADD) (Dupont et al., 2016, Bondt et al., 2013 or by using the mean authorised dosage  so called Daily Doses Animal (DDDA).
The ADD-value is specifically defined as the average maintenance dose per day for a drug used for its main indication for each animal species. The ADD-value was used in Denmark and Austria. They were based on the dose recommendations of each medicinal product registered in a country for each antimicrobial agent, administration route and animal species and when appropriate, also age group (Dupont et al., 2016, Trauffler et al., 2014a, Bondt et al., 2013. Product-related Daily Doses (PrDD) or Daily Product Dose (DPD) calculated the daily dose to an assumed factor of 0.8, correcting for the fact that the maximum doses are not used in every treatment ; this means only 80% of the maximal dosage of the active substances were administered per day per kilogram biomass .

Antimicrobial use measured by daily dose per treatment period
The Used Daily Dose (UDD) is the actual administered daily dose per kilogram biomass of a drug based on administered data reported by the farmer at farm level by a specific study. The formula for the UDD calculation is the weight of active substance divided by the number of treated animals, multiplied by the average weight of animals and treatment duration. Three studies applied UDD (Carmo et al., 2017. One study quantified antimicrobial use as a Prescribed Daily Dose (PDD). This was calculated for each active pharmaceutical ingredient and for each prescription according to the amount of active pharmaceutical ingredient per prescription (mg) divided by the average weight of the animals multiplied by the number of animals and treatment period .

Antimicrobials use measured by daily dose per period at risk of treatment
To compare each administered antimicrobial in specific individual species, the treatment incidence was used in five studies . It was defined as the number of pigs per 1,000 pigs that are treated daily with one ADD or UDD, which is equivalent to how many pigs per 1,000 pigs receive a dose of antimicrobials each day. In order to calculate the treatment incidence, the total UDD or ADD is divided by the treatment period, standard weight and population, then multiplied by 1,000. One study applied 'treatment incidence' rate for slaughtered pigs by dividing the number of ADD by 100 slaughtered pigs at risk .
One study calculated 'treatment frequency' by using the sum of all UDD divided by population size.
It identified how many days, on average, an animal in a herd is treated with one active pharmaceutical ingredient . <Figure 2>

Volume of antimicrobial use
As described above, this review uncovered a large variation in how antimicrobial use was measured, and the actual magnitudes of use. The annual antimicrobial use in pigs ranged from 20,000 kilograms ,300 kilograms at different farm and country levels. One study estimated 34 million kilograms of antimicrobials was found in medicated feed in pigs in China due to the massive number of livestock . However, more than one million kilograms were quantified in the studies in food animals in Germany  and South Africa (Eagar et al., 2012) and about 63 million kilograms globally . On the other hand, lower use was documented in Kenya where only 15,000 kilograms of antimicrobials were used in one year in all animal species (Mitema et al., 2001b). A wide range of volume per biomass was reported, ranged from 33.9 mg per biomass in Austria  with about 400 mg per biomass in Japan .
The ADD varied from lower than one  to 16 ADD  in different phases of pig production and countries. Treatment incidence per 1,000 pigs at risk per day ranged from lower than 10 (Carmo et al., 2017 to more than 200 treatment incidences .
However, careful interpretation across countries is needed as these measurements are not standardized. Also, the magnitudes of use are determined by the type of pig farms, animal demographic and the socio-economic context of a country. See details in table 3.

Data sources
Two main sources of data emerge from this review: national sales data and primary data collected through pig farm surveys. In many European countries, the national monitoring of antimicrobial consumption relies on national sales data of pharmaceutical products, the disadvantage of sales data is the lack of information on which species they are being used for, the indication, dose and duration of treatment. Farm or pharmaceutical company surveys apply prospective longitudinal or cross-sectional studies which provide additional detailed use by species and production types (European Medicines Agency, 2013). One study applies bottom up approach for national consumption data estimate, it collects data from some herds and extrapolates to the national level . However, this approach could be inaccurate as the sampled farms are not designed as national representative samples.
Data sources for animal populations can be retrieved from total national data collection by government agencies such as slaughter house and production information, or it can be obtained from other sources such as the Association of Pig Farmers. Data from international organizations such as the FAOSTAT database hosted by the Food and Agriculture Organization is another source of the size of animal populations . Even though, FAOSTAT information is limited such as estimates for non-responses and incomplete report, and the lack of granularity on number of animal of species; it can be applied when data at the country is not available. Using different weights of animals at treatment across studies resulted in substantial differences in use and hinders comparability (Carmo et al., 2017, Dupont et al., 2016. Several studies reported the quantity of use in kilograms of active ingredient without denominator data. Though simple, its main limitation is that it does not give any indication of intensity of use. To address this deficiency, measurements of use per weight have been widely used. However, using kilogram of active ingredients does not take into account the differences in drug strengths, doses administered and pharmacokinetics. The use of higher strengths, dosage and more treatment days led to higher antimicrobial use than those which were applied at lower strengths and dosage .

Methods and units of measurement
There is also a large variation in strengths and dosages of antimicrobials use in human health. In order to standardise the measurement, the DDD was developed and is now used globally to measure antimicrobial consumption in humans with standardised reporting by DDD per 1000 inhabitant-days.
This facilitates international comparison on antimicrobial use (Natsch et al., 1998) Moreover, there are not only different units of measurement, but countries also name their measurement differently, such as ADD in Denmark  and ADDD in Netherlands (NETHMAP and MARAN, 2013).

A C C E P T E D M A N U S C R I P T
There has been an attempt to establish a consensus on DDDA for each active substance and administration route for veterinary antimicrobial products authorized in four European countries (Postma et al., 2015); this effort has yet to scale up to all European countries. Another approach to calculate the daily dose is by using an actual dose administered to animal. Instead of using ADDvalue, a DPD is proposed to by adjusting the recommended maximum daily dose by a factor of 0.8 of maximal dose for specific medicinal products; assuming that the maximum doses are not used in every treatment .
To differentiate antimicrobial use between herds, antimicrobial per treatment periods were calculated based on real use data at farm level. In 2006, a measurement called UDD was introduced firstly in a study in pig farms . The UDD was calculated based on the definite number of treated animals in a treatment period and the dosages of antimicrobials to animals in farms; the UDD avoids differences between ADD-values and supports comparison of use across countries and across studies. Moreover, the ratio between UDD/ADD reflects the appropriateness of dosing where the higher the ratio, the more excessive the use. Another measurement of antimicrobial use that takes into account the treatment period was PDD; it reports antimicrobial use by antimicrobial prescription. PDD also shows the veterinarian's prescribing pattern. However, antimicrobial prescription is not always equal to the actual antimicrobial administration (Chauvin et al., 2001).
There are several methods that relate to the association between the actual volume of specific antimicrobials used in a specific time period such as 'treatment frequency'  and 'treatment incidence rate' . Furthermore, the 'treatment incidence' has been introduced for a comparison of data between farms, considering the period at risk of treatment . The treatment incidence rate can compare the antimicrobial use per can be compared between herd and production types. It can be calculated based on both ADDvalues or UDD. However, comparison of 'treatment incidence' to other studies should be done with caution when ADD-value is used .
The wide variation in methods and indicators across the studies, and the relative lack of swinespecific data prevent this review from making valid comparisons of antimicrobial use in swine production or documenting trends.

European experiences and international recommendations
In DCDvet of active substance, which take into account differences in dosing, pharmaceutical forms and routes of administration used by these three species (European Medicines Agency, 2015).

A C C E P T E D M A N U S C R I P T
To rectify the weakness of national sales data, in 2013, the EMA recommended that countries conduct farm surveys of veterinary prescriptions or antimicrobial administration records in the logbooks kept by farmers, specific for different species (see ESVAC guidelines of data collection at farm level) (European Medicines Agency, 2013). Though this additional data collection from farms demands substantial resources, infrastructure development and enforcement of veterinary prescriptions at farm level, the benefit is high as it provides accurate information on antimicrobial use by classes and animal species and indications, and evidence can be used to facilitate the development of specific interventions and improve the specific training and education in veterinarians and farmers.
To date, the OIE has also relied on antimicrobial sales data as indicators of actual use, and also recommends that OIE member countries to collect and report data on quantity of antimicrobial consumption in kilogram of antimicrobial agents for different types of indication (therapeutic use or growth promotion), different animal species group and different routes of administration. In the second OIE annual report in 2017 on the use of antimicrobial agents intended for use in animals, OIE recommended to use animal biomass as a denominator so that the quantitative data on antimicrobial agent can be compared among countries. Animal biomass is calculated as the total weight of the live domestic animals, used as a proxy to represent those likely to have exposed to the quantities of antimicrobial agents reported (World Organisation for Animal Health (OIE), 2017).

Policy utilities
Data on antimicrobial usage is needed for a number of reasons such as monitoring time trends of use and assessing the effectiveness of interventions. Ideally it should be disaggregated by different antimicrobial classes in particular the critically important for human health. It can also be used to investigate the association between the magnitude of use and bacterial resistance (Collineau et al., 2017, Schar et al., 2018.

Conclusion
We systematically reviewed the peer-reviewed literatures on the methods and measurements for antimicrobial use in pigs globally. Ten different units of measurement were identified from 25 studies of high-and medium-quality studies; which vary greatly in term of objectives, data sources and units of measurement both numerators and denominators. The non-homogeneity of the unit of measurement limits the cross-study comparative analysis. Additionally, different levels of data such as from farm surveys and national sales data used by these studies also produce different magnitude of use across studies.

Recommendations
Given the importance of measuring antimicrobial use in monitoring progress of policies in optimizing use, at a minimum, all developing countries should develop macro-level monitoring using national

Conflict of interest
The authors declare that they have no financial and personal relationships with other people or organisations that could inappropriately have influenced the present study and manuscript.

Funding
This study was supported by internal funding.