Organic dairy farms generally demonstrated lower AMU than conventional systems across three European studies (Alliance to Save Our Antibiotics, 2021; Bennedsgaard et al., 2010; Antillon et al., 2020), with data from the UK showing evidence of a right-skewed distribution of AMU in organic farms similar in shape to that seen in available data for the dairy sector nationally (Veterinary Medicines Directorate (VMD), 2020), with mean AMU approximately 1.5 times the median value (Alliance to Save Our Antibiotics, 2021).

Findings typically reflected more selective use of DCT on organic farms (Brunton et al., 2012; Poizat et al., 2017), but these differences were not always significant (Firth et al., 2019; Antillon et al., 2020). A 2010 survey of British dairy farmers reported that organic farmers were less likely to use blanket DCT, antimicrobial treatment of all cows at the end of lactation, although no statistical analysis was reported. The study suggested similar use of highest priority critically important antibiotics (HP-CIAs) by the two groups (Brunton et al., 2012). However, a 2016 UK survey found organic dairy farms less likely than conventional farms to include HP-CIAs in their three most frequently used veterinary medicines (Higham et al., 2018). Where they were measured, health outcomes such as udder health were not negatively affected by lower AMU (Hamilton et al., 2006; Valle et al., 2007).

Impacts of organic farming on AMR are uncertain. No clear differences in AMR were detected between streptococcal and staphylococcal isolates from organic and conventional milk in studies in Switzerland (Busato et al., 2000; Roesch et al., 2006) or Sweden (Sjöström et al., 2020). In a Swiss study, calves from organic farms had significantly higher odds of carrying commensal E. coli resistant to several antimicrobials including kanamycin and ampicillin, but extended-spectrum beta-lactamase (ESBL) E. coli was found only in calves from conventional farms (Nüesch-Inderbinen et al., 2022).

A UK industry analysis of a convenience sample representing approximately 9% of UK herds found weak positive or insignificant correlations between herd size and/or milk yield and AMU (Kingshay Independent Dairy Specialists, 2021). In Finland, blanket DCT was found to be a minority practice significantly associated with larger herds as well as higher milk production (Vilar et al., 2018). Use of automatic milking systems (AMS) was significantly associated with blanket DCT in Finnish farms (Vilar et al., 2018), although there was no significant difference in AMU between AMS and conventional milking systems in a Dutch study (Deng et al., 2020). In a UK study of milk microbiology, use of AMS rather than conventional milking systems was a significant predictor of resistance (McLaughlin et al., 2022). AMR prevalence in fecal or environmental E. coli was found to be significantly lower in smaller herds in Sweden (De Verdier et al., 2012) and, in Germany, in farms with traits, traits associated with lower intensity farming, such as longer fattening periods and less farm mechanization (Hille, 2017).

Veterinary treatment frequency for mastitis was significantly higher in Swedish herds with an increased proportion of first-parity cows (Nyman et al., 2007), while cows in a UK study were significantly more likely to be treated with antimicrobials in the first 30 days of lactation if they were third or later parity (Cook, 2020). In a study of a single UK farm, ESBL E. coli was significantly more prevalent in feces of lactating cows compared to the herd overall, and around five times more prevalent in multiparous compared to primiparous cows (Watson et al., 2012). A significantly higher prevalence of antibiotic-resistant E. coli was found in calves compared to adult dairy cows in this study as well as in others in Sweden, Germany, the Netherlands and the UK (Watson et al., 2012; Duse et al., 2015; Heuvelink et al., 2019; Schubert et al., 2021; Weber et al., 2021). Calves were found to shed resistant E. coli at a much higher level than young stock and adults, with possible explanations including differing selection pressures within the calf enteric environment compared to later stages (Heuvelink et al., 2019).

Breed associations with AMU are uncertain. Swedish herds with Swedish red-and-white cattle, a traditional breed associated with less intensive farming, had a significantly lower incidence of mastitis cases treated by a veterinarian (Nyman et al., 2007), while no breed association was seen in a convenience sample of UK herds using a dairy consultancy service (Kingshay Independent Dairy Specialists, 2021).

Odds of treatment with antimicrobials in the first 30 days after calving were significantly higher in UK cows calving in summer or winter compared to autumn, while those treated prophylactically with pegbovigrastim (bovine granulocyte colony stimulating factor, G-CSF) had significantly lower odds of requiring treatment (Cook, 2020). Disinfection of the calving area was positively associated with ESBL E. coli prevalence in recently calved cows on German dairies (Weber et al., 2021). Possible explanations for positive associations between disinfection and AMR suggested by these and other authors include poor implementation of disinfection protocols, co-selection for resistance, and increased likelihood of disinfectants being introduced on farms with infectious disease problems (Taylor et al., 2009; AbuOun et al., 2020; Luiken et al., 2022).

Risk factors identified for high AMU in young calves in the Netherlands included keeping calves under eight weeks old on slatted floors rather than non-slatted floors and farmer beliefs that young stock do not require any specific management (Holstege et al., 2018). High-AMU farms in this study had significantly higher rates of respiratory disease and higher probability of a history of Salmonella on-farm (Holstege et al., 2018). Use of non-antimicrobial prophylactic treatments, such as those for ketosis, in periparturient cows was associated with reduced risk of AMR in calves in Germany, which may be due to the effects of these treatments or could reflect good farm management (Weber et al., 2021).

Feeding of waste milk to calves, found to be a majority practice in a 2010 UK study (Brunton et al., 2012), is no longer recommended following updated guidelines in 2018 (Lloyd, 2018). 46% of UK dairy farmers reported practicing it in 2016 (Higham et al., 2018), comparable to the 48.3% of Swiss dairy farmers in 2020 (Bernier Gosselin et al., 2022). Waste milk is sometimes pasteurized to kill microbes, but this does not solve the problem of antimicrobial residue persistence, and recontamination of pasteurized milk has been demonstrated (Aust et al., 2013). Waste milk feeding of calves was associated with significantly higher prevalence of AMR in multiple studies (Aust et al., 2013; Brunton et al., 2014; Duse et al., 2014; Maynou et al., 2017a; Maynou et al., 2017b; Weber et al., 2021).

Use of DCT, especially blanket rather than selective treatment, is associated with higher overall AMU (Nyman et al., 2007; Scherpenzeel et al., 2016; Stevens et al., 2016; Firth et al., 2019). Selective use has been identified by farmers as a key approach to reducing their AMU (Higham et al., 2018). In a Dutch study, selective DCT was found to reduce AMU compared to blanket application, but at the cost of higher somatic cell count (SCC) at the following calving for cows dried off without antimicrobials (Scherpenzeel et al., 2014). Non-antimicrobial teat sealants, a suggested alternative, are reportedly used mainly alongside, rather than instead of, antimicrobial DCT (More et al., 2012). Teat sealant use has also been reported in association with selective application of DCT (Scherpenzeel et al., 2016). Positive associations between antimicrobial DCT and AMR prevalence have been demonstrated in Sweden and the UK (Duse et al., 2014; Schubert et al., 2021; McLaughlin et al., 2022), highlighting antimicrobial DCT use as an area of concern.

Several approaches to mastitis management were identified for reducing AMU. Identifying mild or subclinical cases and in the first instance managing these with supportive care such as massage or NSAIDs in place of antimicrobials was associated with reduced AMU in multiple studies (Hamilton et al., 2006; Nyman et al., 2007; Stevens et al., 2016; Holstege et al., 2018). A targeted lactating cow treatment protocol piloted in dairy farms in Germany demonstrated reduced AMU associated with the intervention without any negative effect on any of the measured clinical parameters (Schmenger et al., 2020). The protocol included measures such as use of NSAIDs as first-line treatments for mild mastitis and on-farm culture to test for presence of Gram-positive causative agents, for which antibiotic treatment is indicated. Some farms in a Danish study were also reported to use “blinding,” drying-off of affected quarters, to control mastitis, although the impacts of this specific measure were not assessed (Bennedsgaard et al., 2010).

Group treatment with antimicrobials, for example metaphylaxis, is not always indicated, and use of oral and footbath antibiotics on UK farms were found to be significantly associated with the farm being in the top quartile for AMU (Hyde et al., 2017). Disease prevention is an important aspect of reducing AMU and AMR, and Swedish herds free of bovine respiratory syncytial virus (BRSV) had a significantly lower proportion of quinolone-resistant E. coli than recently infected herds or those with a long-term steady infection rate (Duse et al., 2021).

Social norms among farmers and concerns about clinical symptoms persisting or recurring have been associated with antimicrobial course duration as administered by farmers (Swinkels et al., 2015). Dosing rates may also differ from recommended doses on specific product characteristics (SPCs) (Merle et al., 2014), for example due to difficulty accurately estimating body mass of cows (Van Dijk et al., 2015). Veterinary attitudes, beliefs and social pressure from farmers have been shown to be associated with readiness to prescribe antimicrobials, and significant differences in farmer thresholds for seeking veterinary treatment have been observed between nations with similar policy approaches to AMU and AMR (Espetvedt et al., 2013). In Germany a longitudinal observational study of dairy cattle found that the prescribing veterinarian was a significant predictor of farm-level AMU (Hommerich et al., 2019), while a survey of cattle veterinarians in which the majority of respondents practiced in a country in Europe found that the longer a veterinarian had worked in the cattle industry, the less likely they were to be worried about AMR (Llanos-Soto et al., 2021).

Farmer attitudes and personality traits affecting AMU included: beliefs about AMU and AMR (Swinkels et al., 2015; Scherpenzeel et al., 2016), past experiences (Vilar et al., 2018), social norms (Swinkels et al., 2015), opinions of best management practices (Bennedsgaard et al., 2010; Gussmann et al., 2018; Holstege et al., 2018), and individual “treatment threshold” which is defined as the point at which a farmer will call the veterinarian rather than managing a condition themselves (Nyman et al., 2007; Valle et al., 2007; Holstege et al., 2018; Deng et al., 2020). In a study of Norwegian dairy farms, organic production systems were associated with a higher treatment threshold, with organic farmers taking non-pharmaceutical approaches to mild mastitis in the first instance (Valle et al., 2007).

The role of the veterinarian was highlighted in multiple studies, with veterinarian-farmer relationships impacting farm management and disease prevention programs as well as prescribing treatments (Poizat et al., 2017; Speksnijder et al., 2017). More frequent veterinarian contact was associated with greater knowledge of AMR in the UK (Higham et al., 2018) and herd health programs facilitated by veterinarians working in collaboration with farmers were identified as an important tool to help improve herd health and welfare along with decreasing AMU (Stevens et al., 2016; More et al., 2017; Speksnijder et al., 2017; Higham et al., 2018). A trial of farm-specific management changes developed in collaboration with the farmers themselves demonstrated a significantly reduced AMU in intervention farms (Bennedsgaard et al., 2010).

Figure 3 summarizes the evidence regarding factors associated with AMU and AMR at different points in the dairy cattle production cycle, including points in the cycle associated with increased AMU.

A pilot study of 119 beef farms suggests that in the UK, organic beef farms have lower AMU than conventional farms. As observed with broader UK beef sector data (Responsible Use of Medicines in Agriculture Alliance (RUMA), 2021), a small number of farms have disproportionately high AMU among this sample of organic farms (Alliance to Save Our Antibiotics, 2021).

In Germany, larger herds had significantly higher odds of using any antimicrobials than smaller herds, potentially associated with buying in more stock (Hommerich et al., 2019). As well as overall numbers, two Italian studies suggest that cattle arriving on fattening farms in autumn or winter had a significantly higher AMU than those arriving in spring or summer (Diana et al., 2021; Santinello et al., 2022). Treatment incidence was significantly higher in male cattle (Diana et al., 2021), with sex-based differences in susceptibility to respiratory disease presented as a possible cause by the authors. This is consistent with the finding that a reduction in AMU associated with quarantine on arrival was significant for male but not female cattle (Santinello et al., 2022). Non-significant trends were also observed for higher AMU in Blonde d’Aquitaine and Limousin breed cattle (Diana et al., 2020; Diana et al., 2021). Factors characteristic of less intensive farming, including use of dual-purpose breeds, were significantly associated with lower prevalence of cefotaxime-resistant E. coli (Hille, 2017).

Beef production systems include calves from beef-only herds, and “dairy-to-beef” calves, which originate from dairy farms and may be dairy breeds or dairy crossed with beef. An analysis of AMU in Irish beef and dairy-to-beef calves across n=79 suckler beef farms and n=44 dairy farms found no significant difference in overall AMU between the two types of calves over the first 6 months of life but significant differences in indications for treatment. Beef-only farms showed significantly higher incidence of navel ill, joint ill and respiratory disease, while dairy-to-beef calves had a significantly higher incidence of diarrhea, which accounted for 58.3% of group treatments with antimicrobials (Earley et al., 2019). Beef cattle originating from dairy rather than beef-only farms had a significantly higher risk of harboring AMR E. coli in studies in Switzerland (Reist et al., 2013) and the UK (Velasova et al., 2019).

In the UK beef industry, calf-rearing units were associated with higher AMU than suckler or finisher herds (Responsible Use of Medicines in Agriculture Alliance (RUMA), 2021; Doidge et al., 2020). Mixing of calves from different farms after transport to a calf-rearing unit has been highlighted as a likely contributor due to the increased risk of pneumonia (Doidge et al., 2020) and the number of introduced calves was found to be the most important predictor of AMU in Danish veal and young bull production (Fertner et al., 2016). As well as higher AMU in younger age groups, beef calves demonstrated a higher risk of antimicrobial-resistant commensal bacteria than adult cattle (Reist et al., 2013), mirroring findings in dairy cattle (Reist et al., 2013).

Biosecurity was positively correlated with a higher composite management and welfare score based on farm management, staff training, housing, and animal-based measures in an Italian study. Higher composite scores were significantly associated with lower AMU, but biosecurity in itself was not; the authors attribute this to low biosecurity scores across their study population (Diana et al., 2020).

As noted above, significantly lower AMU was observed on farms with better composite animal management and welfare scores in beef fattening farms in Italy, but no conclusions could be made regarding the relative contributions of different components of this (Diana et al., 2020).

In Dutch veal systems, ambient temperatures above 10°C and provision of pelleted feed rather than other starter feeds were significantly associated with lower AMU, while straw bedding was associated with higher AMU (Mallioris et al., 2021). Multiblock analysis of the results of this study showed that housing microclimate explained the greatest part of the observed variation in AMU (Mallioris et al., 2021). In a UK study, beef cattle fed a majority-concentrate diet had a significantly higher diversity of antimicrobial genes and a broader spectrum of AMR mechanisms in the gut flora compared to those fed a diet consisting of equal amounts of forage and concentrate (Auffret et al., 2017).

Two UK studies of veterinarian and farmer decision-making found that the relationship between the farmer and their veterinarian influenced their AMU choices (Doidge et al., 2019; Doidge et al., 2020). Farmers tended to view veterinary advice very positively and articulated a desire for more veterinary guidance on disease prevention to facilitate reduced AMU (Doidge et al., 2020). The characteristics of the veterinarians themselves also predicted AMU decisions: veterinarians working in purely large animal practice, younger veterinarians and veterinarians in situations where practice colleagues had prescribed antimicrobials previously were all more likely to prescribe antimicrobials without a farm visit (Doidge et al., 2019). A 2017 survey of UK beef farmers identified knowledge gaps regarding HP-CIAs and reported that some respondents were unaware which of the veterinary medicines they were using were antimicrobials (Doidge et al., 2020). Use of digital farm management tools and movement records have been associated with lower AMU (Doidge et al., 2020) and lower AMR (Hille, 2017), potentially reflecting on broader farmer attitudes to farm management. As with dairy cattle, actual antimicrobial dosing in beef cattle has been observed to differ from recommended doses (Merle et al., 2014).

Figure 4 summarizes the evidence regarding factors associated with AMU and AMR at different points in the beef production cycle, including points in the cycle associated with increased AMU. The differing boundaries of production units reflect the production stages included in each type of unit.

A UK analysis found that lowland flocks had the highest AMU, followed by upland, then hill flocks (Davies et al., 2017), the reasons for which require further investigation. The same study identified a non-significant trend for lower AMU in organic flocks, consistent with the limited available industry data (Alliance to Save Our Antibiotics, 2021), although estimates for industry-wide AMU in sheep vary markedly. A study conducted in Greece reported a significantly lower prevalence of AMR in organic compared to conventionally raised dairy sheep (Malissiova et al., 2017), while another in Greece found a positive association between intensive production and presence of resistant Staphylococcus isolates in bulk milk tanks, although the criteria for intensive production were not specified (Lianou et al., 2021).

A UK study identified a possible association between the proportion of ewes lambing indoors and farm-level AMU, but this was not statistically significant (Lovatt and Davies, 2019).

The only study of biosecurity and AMU or AMR was an investigation of AMR in sheepdog puppies and lambs in Greece. This identified identical phenotypic resistance profiles in E. coli isolates from two pairs of animals (one puppy and one lamb) on the same farm (Chatzopoulos et al., 2016). Phylogenetic analysis would have been beneficial for further investigation.

No significant link was found between vaccination against footrot and AMU in a study of 152 UK flocks (Lovatt and Davies, 2019). In UK flocks, lameness accounted for 65.5% of the AMU in one study, and incidence varied markedly between farms (Davies et al., 2017), although the etiologies of the lameness cases were not differentiated.

A UK survey investigating uptake of contagious ovine digital dermatitis (CODD) management guidelines reported that 45% of veterinarians said that they decreased their use of whole-flock antimicrobial treatments and 57% were recommending reduction in use of antimicrobial footbaths on farms because of evidence-based guidelines. Farmers also reported changing their practices because of the updated advice, with 46% updating biosecurity measures and 52% reconsidering their choice of antibiotic (Duncan et al., 2022). As well as access to up-to-date guidelines, individual farmer characteristics, in particular attitudes to change, were identified as predictors of AMU (Doidge et al., 2021a).

Quantitative analysis of AMU on UK sheep farms reported that 21% of unexplained variation in farm AMU was between veterinary practices (Davies et al., 2017), suggesting differences in prescribing behaviors. A survey of veterinarians investigating prescribing decisions in various scenarios identified predictors including veterinarian characteristics as well as farmer-veterinarian relationships. Younger veterinarians and those working in farm-only veterinary practices were more likely to prescribe antimicrobials to a sheep farmer without a visit, while being confident in the farmer’s ability to identify the disease and feeling that the farmer was unwilling to pay for a visit also increased the likelihood of prescribing (Doidge et al., 2019).

Figure 5 summarizes the evidence regarding factors associated with AMU and AMR at different points in the sheep production cycle, including points in the cycle associated with increased AMU. Whilst sheep production in the UK is characterized by the movement of sheep from hill farms to upland to lowland farms, the production cycle is presented in a simplified manner here for clarity.

A UK study carried out prior to the EU ban on AMU for growth promotion, which came into effect on 1 January 2006 (Casewell, 2003), observed markedly lower AMU on organic pig farms compared to conventional farms, although only 12 pig farms were included in the study (Veterinary Laboratories Agency (VLA), 2006). Recent data from Switzerland and the UK suggest this association between organic pig farming and lower AMU still exists, with a pilot study of 18 organic UK pig farms reporting mean AMU of 1.42mg/kg, compared to 110mg/kg reported in national monitoring (Alliance to Save Our Antibiotics, 2021; Arnold et al., 2016; Veterinary Medicines Directorate (VMD), 2020). In Denmark, AMU was significantly lower in the free-range herds than in the indoor herds, but there was no significant difference between organic free range and conventional free-range pigs (Nielsen et al., 2021). An association between membership of two farm assurance schemes (Scottish Pig Industry Initiative and Freedom Foods) and lower self-reported AMU was also observed in an older UK study, although AMU regulations and norms have changed significantly since the data for this study were collected in 2001 (Stevens et al., 2007). Taken together, these findings suggest that there is lower AMU in organic systems and free-range systems but that the differences between organic free-range and conventional free-range may not be as marked.

A cross-sectional study of conventional and organic pig farms in Denmark, France, Italy and Sweden, found a significantly higher prevalence of AMR amongst E. coli isolates from conventional rather than organically raised pigs in all four countries (Österberg et al., 2016). However, a fecal analysis covering the same four countries reported that organic status had little impact on antimicrobial resistance gene (ARG) abundance for the six ARGs measured, with country of origin the main predictor (Gerzova et al., 2015). Studies in the UK and the Netherlands have reported lower AMR prevalence on organic compared to conventional pig farms (Veterinary Laboratories Agency (VLA), 2006; Hoogenboom et al., 2008), although neither calculated p-values. Two Spanish studies found significant positive associations between intensive production and AMR when compared with organic and conventional free-range herds (Galan-Relano et al., 2019; Mencía-Ares et al., 2021).

AMU has been identified as a risk factor for AMR (Docic and Bilkei, 2003; Taylor et al., 2009; Burow et al., 2019; AbuOun et al., 2020; Ceccarelli et al., 2020; Luiken et al., 2020; Mencía-Ares et al., 2021; Luiken et al., 2022), although resistance is also observed in pigs with no history of antimicrobial treatment (Taylor et al., 2009; Burow et al., 2019) and one study noted a lack of correlation between AMU and AMR on 60 pig farms (De Koster et al., 2021). Specific aspects of treatment such as choice of antimicrobial or use of group treatments have been found to be significantly associated with AMR (Dohmen et al., 2017). Evidence that antimicrobial dosing rates used on farms frequently deviate from licensed or recommended doses (Merle et al., 2014; Sarrazin et al., 2019) highlights factors other than just quantity of antimicrobials used.

Conflicting results were found regarding farm size as a determinant of AMU. Several studies found lower AMU in larger herds (Casal et al., 2007; Scali et al., 2020; Mallioris et al., 2021), some found the relationship with herd size non-significant (Van Rennings et al., 2015; Arnold et al., 2016; Backhans et al., 2016; Collineau et al., 2017a; Kruse et al., 2020), whilst others observed higher AMU in larger herds (Stevens et al., 2007; Van Der Fels-Klerx et al., 2011; Visschers et al., 2016; Collineau et al., 2018; Raasch et al., 2018; Stygar et al., 2020; Matheson et al., 2022). One possible reason that larger farms might have lower AMU is a positive association between farm size and level of biosecurity identified in several studies (Laanen et al., 2013; Mencía-Ares et al., 2021; Yun et al., 2021).

The production stages included in the farm appear to impact AMU and AMR, but the trends identified differ between studies. AMU was significantly higher in farrow-to-finish and piglet producers compared with finishing farms in Switzerland and England (Echtermann et al., 2019; Matheson et al., 2022), but the opposite trend was observed for antimicrobial prophylaxis in Spain (Casal et al., 2007). Other papers reported AMU to be higher in mixed-species farms compared with farrow-to-finish or specialized fattening farms (Van Der Fels-Klerx et al., 2011) and higher in farrow-to-finish than purely piglet producers (Van Der Fels-Klerx et al., 2011; Mallioris et al., 2021). In a UK study, finisher-only farms were at significantly higher risk of multi-drug resistant (MDR) E. coli presence than breeder-finisher farms, possibly due to the practice of buying pigs in from multiple sources (AbuOun et al., 2020).

Peaks in AMU were detected at weeks 1, 4 and 9, in pig farms in nine European countries, although relative contributions of therapeutic and prophylactic AMU to these peaks were not investigated (Sarrazin et al., 2019). These peaks coincide with the start of each production phase (piglet, weaner, finisher), which can involve stressful management changes and husbandry interventions.

Suckling pigs have been identified as a high-AMU age group (Van Rennings et al., 2015; Backhans et al., 2016; Raasch et al., 2018; Echtermann et al., 2019; Yun et al., 2021) but risk factors affecting relative AMU among piglets in this age group are uncertain. Surgical castration was found to have no effect on AMU in piglets in Spain despite impacts on productive performance (Morales et al., 2017), while tooth clipping in piglets showed a possible association with AMR in two studies but was not statistically significant in the multivariate models for either study (Dorado-García et al., 2015; Dohmen et al., 2017). Weaning is a period of high stress for piglets and is a period of particularly high risk for antimicrobial use (Sjölund et al., 2016; Martin et al., 2020; O'Neill et al., 2020), although relative AMU compared with suckling piglets varies between countries (Moreno, 2013; Sjölund et al., 2015). Later weaning age was significantly associated with lower AMU in one study of Dutch and Belgian pig farms (Caekebeke et al., 2020), while other studies either showed non-significant associations (Postma et al., 2016a), significance in only some of the countries covered (Collineau et al., 2018) or no association (Collineau et al., 2017a).

Whilst sows have been observed to have a lower prevalence of AMR commensals than piglets (Crombe et al., 2012) and other age groups (Nollet et al., 2006), similarities in resistance profiles of E. coli isolates from young piglets and their dams in Germany suggest colonization of piglets with maternal flora (Burow et al., 2019). In Spain, shedding of cephalosporin-resistant E. coli (CR-EC) by the dam was found to be the most important predictor for isolation of CR-EC in young pigs at slaughter, regardless of AMU history (Cameron-Veas et al., 2016).

Better overall biosecurity has been found to be significantly associated with decreased AMU (Laanen et al., 2013; Collineau et al., 2017a), but also with increased risk of AMR (Luiken et al., 2022), with these AMU and AMR findings also observed in studies addressing internal biosecurity specifically.

Use of homeopathic treatments had a significant negative association with AMU on Swiss fattening farms (Arnold et al., 2016), while use of group treatments for antimicrobials was associated with a higher overall AMU (Casal et al., 2007; Merle et al., 2014; Martin et al., 2020). General good health of the animals on-farm and use of anthelmintics were negatively associated with AMR in Swiss finishing farms (Schuppers et al., 2005).

In Swedish pig herds, weaners from specific pathogen-free herds had lower AMU than those from other herds (Sjölund et al., 2015). In Denmark, AMU to treat respiratory signs was found to be closely associated with seroconversion to porcine respiratory disease complex (PRDC) (Antunes et al., 2019), suggesting that PRDC control may contribute to minimizing AMU.

Several studies found significant associations between vaccination and lower AMU, although these associations were limited either to use of multiple vaccinations (Collineau et al., 2017a) or introduction of vaccines to address existing problems (Fricke et al., 2015; Rojo-Gimeno et al., 2016). Several other papers reported no significant changes in AMU with various vaccination protocols (Sjölund et al., 2015; Kruse et al., 2017; Mallioris et al., 2021). Some vaccinations were associated with higher AMU (Stevens et al., 2007; Postma et al., 2016a; Collineau et al., 2018; O'Neill et al., 2020) and one had a significant positive association with MRSA risk (Dorado-García et al., 2015).

Use of a private water source rather than mains has been identified as a risk factor for increased AMU on Dutch farms, potentially due to inconsistent water quality (Mallioris et al., 2021). The practice of farms milling their own feed was also significantly associated with reduced AMU in a study of Irish pig farms, although the authors point out that this may be a biosecurity indicator (O'Neill et al., 2020). Links between feeding practices and AMR were identified, with feeding of whey and limited (rather than ad libitum) feeding significantly negatively associated with AMR in Swiss finishers (Schuppers et al., 2005).

Feed composition may influence the livestock resistome, mediated by nutritional effects on the gut microbiome. Taxonomic variation in the gut microbiome was found to explain resistome variation in pigs on farms across Europe, suggesting that by influencing the gut microbiota, it may be possible to limit colonization by resistant organisms (Munk et al., 2018).

Marketing authorization for medicinal products containing zinc oxide at therapeutic levels (≥1500ppm) was withdrawn in the EU in June 2022, following a 5-year phasing out period, in order to reduce selection pressure for AMR associated with zinc supplementation. One study found no significant association between in-feed zinc and AMU (Sjölund et al., 2015), while another found a non-significant trend of positive correlation between AMU and zinc use (Nielsen et al., 2021). In a controlled trial, in-feed zinc supplementation at 2500 ppm was significantly associated with MDR E. coli carriage (Bednorz et al., 2013).

An experimental comparison of fully-slatted and straw-bedded floors reported overall reduced AMU on fully-slatted rather than straw-bedded floors (Scott et al., 2006). On UK pig farms, having finisher pens with outdoor space and use of automatically controlled natural ventilation were both associated with lower AMU (Matheson et al., 2022). Finnish fattening farms with average-to-poor air quality or problems with pen cleanliness and condition were found to have higher AMU, and a significant positive interaction effect was observed between high stocking density and poor pen condition (Stygar et al., 2020). A study covering nine European countries found high pig stocking density to be positively associated with AMR genes in the farm environment (Luiken et al., 2022).

Evidence of the role of socio-demographic characteristics as predictors of AMU appears inconsistent. One study found higher AMU in farms run by older farmers (Backhans et al., 2016), while others found no significant association between age and AMU (Postma et al., 2016a; Visschers et al., 2016). Farmer gender was not significant in most studies (Postma et al., 2016a; Visschers et al., 2016; Collineau et al., 2017a), although one study found female farmers significantly associated with higher AMU in suckling piglets (Backhans et al., 2016). The same study suggested university-educated farmers reported significantly higher AMU in suckling piglets, but the association between education and AMU was not identified in another study the same year (Postma et al., 2016b). In a study of veterinary prescribing on German pig farms, there was significant variation between individual veterinarians’ prescribing practices but predictors were not explored (Van Rennings et al., 2015).

Use of external farm consultant services was associated with lower AMU in Swiss farms, potentially an indicator of farmer attitudes or consultant advice (Arnold et al., 2016). General views regarding antimicrobials were found to be a significant predictor of AMU among Swiss (Visschers et al., 2014) but not Swedish farmers (Backhans et al., 2016). Perceiving an expectation from society to reduce AMU and scoring higher for optimistic views regarding the risks and benefits of reducing AMU were both associated with lower AMU by farmers (Van Asseldonk et al., 2020).

While financial concerns were reported as a key barrier to reducing AMU (Visschers et al., 2014; Van Asseldonk et al., 2020), early trials of farm-specific AMU reduction plans showed a net economic gain for farmers (Rojo-Gimeno et al., 2016) and have been demonstrated to be successful in reducing AMU without compromising production parameters (Rojo-Gimeno et al., 2016; Collineau et al., 2017b). Voluntary schemes and mandatory national programs involving external monitoring of farm AMU were associated with AMU reductions in Switzerland (Echtermann et al., 2020) and Denmark (Jensen et al., 2014) respectively.

Figure 6 summarizes the evidence regarding factors associated with AMU and AMR at different points in the pig production cycle, including points in the cycle associated with increased AMU. In Europe, unit types include multiple combinations of production stage. Breeder farms breed piglets to sell as piglets or weaners or to rear to slaughter. Nursery units buy in piglets to sell on as weaners or to rear to slaughter, while finisher farms buy in weaners to rear to slaughter.

A UK study reported that of the seven organic broiler farms visited, only one had used antimicrobials in the two-year period examined, compared to all of the six conventional farms, although this was prior to the 2006 EU ban on AMU for growth promotion (Veterinary Laboratories Agency (VLA), 2006). A more recent UK report using a sample of six organic broiler farms gave a substantially lower mean AMU, at 2.95mg/kg (Alliance to Save Our Antibiotics, 2021) than the industry average of 17mg/kg PCU (Veterinary Medicines Directorate (VMD), 2020).

AMR patterns among organic and conventional farms have been much more widely researched and consistently reported lower AMR in chickens from organic farms (Soonthornchaikul et al., 2006; Hoogenboom et al., 2008; Miranda et al., 2008; Alvarez-Fernandez et al., 2013; Much et al., 2019; Hansson et al., 2021). However, some papers also identified a higher bacterial load in organic chickens (Hoogenboom et al., 2008; Miranda et al., 2008; Alvarez-Fernandez et al., 2013) and resistance patterns for individual antimicrobials were sometimes less clear (Miranda et al., 2008; Much et al., 2019). In Italy, antibiotic-free intensive broiler production was associated with significantly lower AMR in commensal flora than conventional production (De Cesare et al., 2022), although in this and other studies, AMR bacteria were detected despite absence of AMU (Veterinary Laboratories Agency (VLA), 2006).

Dutch national statistics reported higher AMU in breeding flocks than in fattening flocks (Van Geijlswijk et al., 2019). In Irish farms, nearly half of all AMU in broilers was reported to be in the first week of life (Martin et al., 2020). Fattening farm traits associated with higher AMU included proximity to other farms (Caucci et al., 2019), higher broiler weight at slaughter (Hughes et al., 2008), and, possibly, larger farms (Bos et al., 2013). An analysis of different stages in the broiler production chain found that variations in AMU on broiler fattening farms were not affected by the breeder from which the chicks were sourced. AMU differences were largely associated with the different fattening farms as well as unexplained variation between chick batches (de Jong and van Riel, 2020).

Although Dutch national statistics reported lower AMU in farms with “alternative” chicken breeds associated with less intensive systems (Van Geijlswijk et al., 2019), an experimental study of fast- and slow-growing broiler breeds with no antimicrobial treatment showed a similarly high rate of AMR in both groups (Montoro-Dasi et al., 2020). The same authors observed rapid development of resistance without any AMU in another experiment, and found no significant differences in AMR between groups with high and low stocking density or groups with different levels of ventilation (Montoro-Dasi et al., 2021).

One study of intensive, semi-intensive and organic free-range systems in Portugal found significantly lower AMR in broilers from semi-intensive farms (described as “indoor extensive” in the study) than from either intensive or organic systems (Fraqueza et al., 2014). Another comparing intensive and extensive systems found no significant differences in AMR profiles between systems (Oliveira et al., 2010). Similarly, no significant difference was found between AMR prevalence in free-range and housed chickens in Greece (Economou et al., 2015). A significantly higher prevalence of tetracycline resistance was found in indoor than free-range chickens in a French study, but this could be due to the fact that in this study, carried out prior to the phasing out of AGPs in Europe, around half of the indoor farms and none of the free-range farms reported using AGPs (Avrain et al., 2003).

Possible links between lower AMU and better overall biosecurity were identified in a Belgian-Dutch study (Caekebeke et al., 2020). Multinational metagenomic studies gave mixed results: one found little evidence for any impact of biosecurity on broiler fecal resistomes (Luiken et al., 2019), but another found a significant negative association between better biosecurity and relative AMR gene (ARG) abundance in dust from broiler houses (Luiken et al., 2022).

A study of French free-range broilers identified the use of chicken paper topped with feed to be a protective factor for AMU in hatchlings (Adam et al., 2019). No significant effect of stocking density on AMU was found on conventional farms in Italy (Tarakdjian et al., 2020). The impacts of bedding materials are unclear. Thinner litter was associated with lower AMU in French broilers, although the reasons for this were not clear (Adam et al., 2019). A metagenomic study found mixed results regarding the effects of bedding on ARGs in the environment. Shredded straw bedding was associated with lower absolute abundance of ARGs but higher relative abundance of ARGs compared to the environmental bacterial population. The same study found lower absolute ARG abundance during summer, which could be suggestive of climatic influences (Luiken et al., 2022).

In the UK, controlled (rather than ad libitum) feeding and use of competitive exclusion (CE) products were both negatively associated with prophylactic AMU (Hughes et al., 2008). As well as associations with AMU, laboratory trials have shown some promise for CE products to mitigate AMR (Ceccarelli et al., 2017; Dame-Korevaar et al., 2020; Methner and Rösler, 2020). A genomic study across nine European countries found that taxonomic variation in gut microbiome, potentially influenced by diet, explained resistome variation in broilers (Munk et al., 2018).

Therapeutic AMU on UK broiler farms was reported mainly in response to mortality, respiratory or enteric disease (especially necrotic enteritis) (Hughes et al., 2008), similar to other countries in Europe, although AMU for respiratory disease varies greatly between countries (Joosten et al., 2019). In UK broilers, vaccination against infectious bursal disease was positively associated with higher therapeutic AMU (Hughes et al., 2008), potentially due to temporary immunosuppression caused by this vaccine (Prandini et al., 2016). In free-range broilers in France, the use of essential oils as prophylactic treatment for any condition was significantly associated with lower AMU (Adam et al., 2019).

Farmer and veterinarian attitudes have not been as thoroughly researched for broiler chickens as for some other livestock species. As in other species, dosing accuracy may vary (Persoons et al., 2012). In France, AMU in broilers was significantly associated with the farmer’s perception of their flock’s health (Adam et al., 2019). A European trial of interventions targeting farmer behavior, along with general farm management and disease prevention, reported successes in reducing broiler farm AMU without negatively impacting economic performance (Roskam et al., 2019).

Figure 7 summarizes the evidence regarding factors associated with AMU and AMR at different points in the broiler production cycle, including points in the cycle associated with increased AMU. In broiler chickens, grandparent, parent and fattening units in the same production chain may be spread across different countries.

No significant difference in probability of AMU was found between organic free-range and battery hens across Belgium, Germany, Italy, and Switzerland (Van Hoorebeke et al., 2011). Although a sample of 14 UK organic laying hen farms reported lower AMU than the industry mean in 2018-2019 (Alliance to Save Our Antibiotics, 2021; Veterinary Medicines Directorate (VMD), 2020), further studies are required to substantiate this.

A pair of analyses comparing organic and conventionally reared laying hens in Germany identified significantly lower AMR prevalence in E. coli and Enterococcus species in organic chickens although a less clear-cut pattern was observed for Campylobacter jejuni (Schwaiger et al., 2008; Schwaiger et al., 2010) and a further study found no significant difference in AMR risk between organic free-range and conventional battery hens (Van Hoorebeke et al., 2011).

National level data from the Netherlands reported AMU in rearing farms for laying hens to be substantially higher than laying hen farms (Van Geijlswijk et al., 2019), but no studies were found investigating AMR at different production stages. Contact with other species may be an AMR risk: in Belgian mixed-species farms with MRSA-positive pigs, MRSA was also isolated from chickens, albeit at a lower prevalence (Verhegghe et al., 2013).

In a survey of Scottish backyard poultry keepers, over 60% of respondents reported no AMU, and those who did report using any antimicrobials administered them less than once a year, although equivalent statistics are not available for commercial flocks to aid comparison. Inconsistent biosecurity and infrequent veterinary attention were highlighted by the authors as a risk for spread of AMR and infectious disease in general (Correia-Gomes and Sparks, 2020). Another UK study found that backyard poultry treated in companion animal veterinary practices often had advanced disease when presented at the practice. In 33.0% of chicken consultations, antimicrobials were prescribed, and 43.8% of these were HP-CIAs (Singleton et al., 2021).

A UK supermarket publishing AMU figures in 2018 reported higher AMU on its free-range egg supplier farms compared with cage or colony egg suppliers, but did not provide methodology or analysis (ASDA, 2018). No significant difference in probability of using any antimicrobials was found between different housing types in four European countries. This study did find some effects of housing type on AMR, with resistance in E. faecalis lower in free-range conventional farms compared to caged battery farms, but higher in E. coli in farms using raised-floor housing compared to battery cages (Van Hoorebeke et al., 2011). An observational study of laying hen farms in Switzerland found no consistent association between housing and management factors and AMR (Harisberger et al., 2011).

No publications were identified investigating diet or supplements, although laboratory-based studies have suggested that administration of CE flora in early life may protect against colonization with AMR E. coli (Methner et al., 2019; Methner and Rösler, 2020).

In a survey of UK free-range egg farmers, sourcing medications from veterinarians rather than agricultural merchants was significantly associated with introducing AMU reduction measures on their farms and increased contact with veterinarians was associated with a higher level of optimism about the scope for AMU reduction (Rayner et al., 2019). It is worth noting that in the UK antibiotics require a veterinary prescription, while some non-antibiotic medications can be bought without prescriptions.

Figure 8 summarizes the available evidence regarding factors associated with AMU and AMR at different points in the laying chicken production cycle.

Apparently no studies have compared AMU between representative samples of organic and conventional turkey farms. In a report on UK organic farms covering different livestock types, only one organic turkey farm (which recorded no antimicrobial usage) had responded to the request for data (Alliance to Save Our Antibiotics, 2021). An Italian study comparing AMR and natural immunity in organic and conventional turkey farms reported a general tendency for lower AMU in organic farms but reported no statistical analysis of the trend. In this study organic farms, which had stricter requirements for welfare standards such as stocking density, showed a non-significant trend for lower AMR. Organic status was also significantly associated with higher serum lysozyme concentration and serum bactericidal activity, indicators of natural immunity, which the authors attribute to better welfare, and which they suggest may contribute to lower AMR (Mughini-Gras et al., 2020). Comparison of organically and conventionally produced turkey meat in Germany found significantly lower AMR prevalence in organically produced meat but higher prevalence of Campylobacter (Tenhagen et al., 2020). As with other livestock species, AMR has been identified in turkeys that have not been treated with antimicrobials at all (Horie et al., 2021). Two studies, one a prevalence survey of phenotypic resistance in turkey commensal E. coli and the other a metagenomic analysis of turkey fecal resistomes, found no significant association between AMU and AMR in three European countries (Horie et al., 2021; Ceccarelli et al., 2020).

Among Italian turkey farms, being located in an area with a higher density of turkey farms was a significant risk factor for higher AMU (Caucci et al., 2019). In the UK, turkey breeding farms with over 10,000 birds had a significantly higher risk of AMR presence than farms with fewer birds, but this pattern was not seen in fattening turkey farms (Jones et al., 2013).

Compliance with biosecurity protocols on French turkey farms, such as changing clothes and shoes before entering the facility, were found to be a significant protective factor for AMU (Chauvin et al., 2005). In UK turkey farms, both internal and external biosecurity factors were significantly associated with risk of AMR. Factors indicating contact with, or proximity to, other animal species were significant risk factors for AMR, while good hygiene and disinfection practices were significant protective factors (Jones et al., 2013). In contrast, a study of turkey farms in Germany, France and Spain found no significant associations with biosecurity but a significant negative association between proximity to another turkey farm and presence of AMR E. coli (Horie et al., 2021). Biosecurity at hatchery level was identified as an important control point by a UK study which found resistant E. coli and Salmonella in multiple areas of the hatchery environment, and evidence of pseudo-vertical transmission even in hatcheries using the recommended egg collection and disinfection protocols (Mueller-Doblies et al., 2013).

In fattening turkeys in France, use of CE flora was significantly associated with lower AMU. In this study, farmers having expectations of veterinary antimicrobial prescriptions and these expectations being met were significantly associated with higher overall AMU compared to farms in which staff had no particular expectations of antimicrobial prescriptions (Chauvin et al., 2005).

Figure 9 summarizes the available evidence regarding factors associated with AMU and AMR at different points in the broiler turkey production cycle.